Identification of earlier predictors of pregnancy complications through wearable technologies in a Brazilian multicentre cohort: Maternal Actigraphy Exploratory Study I (MAES-I) study protocol

Introduction Non-invasive tools capable of identifying predictors of maternal complications would be a step forward for improving maternal and perinatal health. There is an association between modification in physical activity (PA) and sleep–wake patterns and the occurrence of inflammatory, metabolic, pathological conditions related to chronic diseases. The actigraphy device is validated to estimate PA and sleep–wake patterns among pregnant women. In order to extend the window of opportunity to prevent, diagnose and treat specific maternal conditions, would it be possible to use actigraphy data to identify risk factors for the development of adverse maternal outcomes during pregnancy? Methods and analysis A cohort will be held in five centres from the Brazilian Network for Studies on Reproductive and Perinatal Health. Maternal Actigraphy Exploratory Study I (MAES-I) will enrol 400 low-risk nulliparous women who will wear the actigraphy device on their wrists day and night (24 hours/day) uninterruptedly from 19 to 21 weeks until childbirth. Changes in PA and sleep–wake patterns will be analysed throughout pregnancy, considering ranges in gestational age in women with and without maternal complications such as pre-eclampsia, preterm birth (spontaneous or provider-initiated), gestational diabetes, maternal haemorrhage during pregnancy, in addition to perinatal outcomes. The plan is to design a predictive model using actigraphy data for screening pregnant women at risk of developing specific adverse maternal and perinatal outcomes. Ethics and dissemination MAES-I has been reviewed and approved by each institutional review board and also by the National Council for Ethics in Research. Detailed information about the study is provided in the Brazilian Cohort website (www.medscinet.com/samba) and findings will be published in the scientific literature and institutional webpages.


Strengths and limitations of this study
• This multicentre cohort will collect comprehensive data on the main maternal and perinatal complications as pre-eclampsia, small for gestational age/fetal growth restriction, preterm birth and gestational diabetes mellitus.
• Physical activity and sleep patterns will be estimated through an inovative wearable device used in the natural environment of the study subject.
• Physical activity and sleep patterns will be estimated from the beginning of the second half of pregnancy until delivery, covering a wide interval during pregnancy and enabling the study of PA and sleep patterns changes throughout pregnancy.
• One possible limitation refers to the uncovered first half of pregnancy regarding this information.

Background
Reducing the global maternal mortality ratio to less than 70 per 100,000 live births by 2030 is one of the targets of the new United Nations Sustainable Development Goals [1]. Multiple challenges need to be tackled to achieve this target, but the 2016-2030 health and development agenda goes well beyond mortality reduction. The Global Strategy for Women's, Children's and Adolescent's Health aims to ensure that every newborn, women and child not only survive but thrive. This will only be possible if a transformative agenda, with innovation at the central stage, is put into action [2].
One of the major challenges that need to be addressed is optimizing the recognition of early predictors and identifiers of maternal and perinatal complications. Delays in diagnosing and managing maternal complications have been associated with poor outcomes [3]. The reduced self-perception of clinical signs related to maternal complications, difficulties in accessing the health system and poor quality of care may contribute to the late identification of complications and worsened prognosis. The development of a non-invasive Antenatal Care (ANC) tool capable of identifying maternal sub-clinical signs during pregnancy may provide the window of opportunity for early identification of abnormal patterns of physiologic parameters related to pregnancy complications and enable their prevention or early treatment. Shortening the time between the onset of a complication and the initiation of the appropriate management allow for secondary prevention and reduction of maternal morbidity and mortality [3][4][5][6][7].
Pervasive computing (i.e. the trend towards embedding microprocessors in everydaylife objects so they can generate data) and wearable technology (i.e. clothing and accessories incorporating computer and advanced electronic technologies such as wrist and/or waistband sensors) are ubiquitous and able to generate a new dataset that needs to be correlated with pregnancy outcomes. Preterm birth and preeclampsia, for instance, are two important pregnancy complications that have a relatively long subclinical phase before the appearance of signs or symptoms [8,9]. It is plausible that during this subclinical phase of certain conditions the pattern of physical activity (PA) or sleep-wake rhythm is affected in some way and this change could be captured through wearable devices. Although some studies show that PA patterns and and diseases in the general population [10,11], published literature correlating wearable technology data and maternal complications are very scarce.
The human circadian rhythm is ruled by endogenous physiologic mechanisms and environmental stimuli [12]. There is solid evidence showing that modification of circadian rhythm or sleep and PA patterns are an underlying condition related to inflammatory, degenerative and/or metabolic chronic diseases as diabetes, hypertension, and cancer [13]. Circadian misalignment is defined as having inappropriate timed sleep and wake, misplaced feeding periods and modification of activity behavior.
The determination of cause or consequence effect between these modifications and the development of pathological conditions is a complex task. It seems that changes in appetite-stimulating hormones, glucose metabolism, inflammatory markers, and mood are some of the related pathways [13][14][15]. Leproult  Some metabolic, cognitive, cardiovascular and other chronic degenerative diseases have been associated with particular patterns of PA and sleep [10,11,[16][17][18]. A previous observational study assessed various sleep parameters during pregnancy, e.g.
sleep onset latency (SOL), wake after sleep onset (WASO) and total nocturnal sleep time (TST). Difficulty in initiating sleep in early pregnancy was associated with higher body mass index, greater weight gain and higher blood pressure during pregnancy  [17]. Palagini et al. reviewed clinical evidence between chronic sleep loss and pregnancy adverse outcomes, discussing common mechanisms of stress system activation [19]. Low-quality evidence suggests an association between sleep loss and prenatal depression, gestational diabetes, preeclampsia, abnormal length of labour, caesarean delivery, abnormal fetal growth, and preterm birth. Those results corroborate with other findings regarding pregnancy and sleep disorders [20][21][22][23].
The assessment of PA and the sleep patterns can be performed using small wrist (or waist) devices similar to a regular watch (actigraphy technology). The type of sensor, batteries, materials and output data have been substantially developed in recent years, enabling low cost, comfort, discretion and performance [24]. Nowadays there are devices that are portable, lightweight and with a large capacity to storage information, including a software with automatic scoring algorithms packages for the detection of wakefulness, sleep periods and PA [24,25]. The actigraphy estimation of PA and sleep patterns is validated as a proxy for chronobiologic behavior [26][27][28][29] and 7 to 14 days using the actigraph device provides reliable estimates of PA behavior in older adults [30][31][32]. Both hip and wrist devices show reliable and acceptable performance in estimating PA and sleep-wake patterns [33][34][35][36].
The main advantages of using wearable devices for actigraphy is the non-invasiveness, 24/7 monitoring of the circadian pattern and PA, and informing sleep habits and parameters in the user natural environment [24,25,28]. We propose an innovative and strategic approach to monitor PA and sleep-wake patterns during pregnancy, establishing a large database comprised of clinical, epidemiological, PA and sleep-wake variables potentially capable of composing a prediction model for maternal complications during pregnancy. The main goal of this study is to identify early predictors of pregnancy complications by correlating data generated on PA and sleep patterns through wearable devices (wristband sensors) with maternal and perinatal complications and outcomes.

Methods/Design
Study design F o r p e e r r e v i e w o n l y 7 We will conduct a cohort study of 400 low-risk pregnant women using wrist sensor bands able to capture information on daily physical activity and sleep patterns (exposure). This cohort study will be implemented in 5 ANC clinics linked to obstetric units in 3 different regions of Brazil that are already part of the Brazilian Network for Studies on Reproductive and Perinatal Health [37], as shown in Table 1. During a period of eight months, the ANC clinics will identify eligible cases for using the wristband sensors. Wearable technology data will be correlated with the occurrence of pregnancy and childbirth complications and outcomes, thus understood as an effect.
Eligible women will be identified up to 21 weeks of gestation and invited to participate.
A proper consent form will be applied and the women who agree to participate will receive a wristband sensor to be used starting at 19-21 weeks until childbirth, uninterruptedly.

Study setting and population
Brazil is a multi-ethnic mixed-race population of diverse resourced settings [38].
Despite the considered high global overall human development index (HDI 0.727) in 2010, the HDI of Brazilian municipalities ranged from 0,862 to 0,418 [39]. The possibility of considering such mixed population is suitable to explore information regarding maternal patterns of mobility and sleep, maximizing external validity and comparisons to other populations. A few reasons might support the study population focused in low-risk nulliparous women: 1) Previous obstetric history can refer to known risk factors for many maternal complications such as preterm birth, preeclampsia, and diabetes [13,40]. Nulliparous women enable unbiased sampling regarding obstetric history. 2) Women with previous morbidity such as hypertension, diabetes, nephropathy or others chronic/degenerative diseases are more likely to present abnormalities of sleep-wake rhythm or physical activity patterns during pregnancy.

Sampling
The five participating centers are regional referral obstetric units responsible for antenatal care assistance mainly for high-risk pregnant women. Participating centers  Table 1. Nevertheless, there are primary health care units strategically linked with these participating centers, enabling the identification and enrollment of low-risk pregnant women. The recruitment strategies include approaching all eligible women in these participating centers and their linked facilities. An informed consent form will be applied for women who agree to participate.

Eligible women: Low-risk pregnant subjects
There is no international consensus on the criteria for low-risk pregnancies, although there are several known factors associated with maternal and perinatal adverse outcomes. A recent study evaluating complications of "low-risk" pregnancies of US Americans (10 million births from 2011 through 2013) showed that 29% of low-risk women had an unexpected complication requiring no routine obstetric/neonatal care [41]. This shows the difficulty in establishing a "low-risk profile" for maternal/perinatal complications. As an exploratory study, we will exclude potential known confounders of pre-pregnancy conditions related to adverse maternal or perinatal outcomes as shown in Table 2, in order to assess PA and sleep patterns of a mostly "normal" population. Lifestyle habits and body composition (Body mass index, height, etc.) characteristics, and some non-severe chronic diseases as thyroid disorders, non-severe anaemia and/or asthma are not among exclusion criteria but may be part of subgroup analyses (composition of any previous disorder, e.g.). Intra and inter-individual analyses of PA and sleep patterns enable the identification of potential confounders affecting primary outcomes, avoiding potential biases.
Eligible women will be enrolled between 04-21 weeks. The inclusion and exclusion criteria are presented in Table 2.

Data collection methods
Essentially, MAES I study is comprised of 4 key set points -3 clinical visits during pregnancy and a postnatal visit. The clinical visits will be held at 1) between [19][20][21] weeks; 2) between 27-29 weeks; 3) between 37-39 weeks. At first, second, third and postnatal visits, additional information regarding maternal history, details on pregnancy complications, maternal biophysical data (weight, height, skinfolds) and pregnancy adverse outcomes will be collected following a specific Standard Operating Procedure (SOP) specially developed for MAES-I study. Additionally, the Perceived Stress Scale [42] and Resilience Scale [43] will be applied during 27-29w visit. Figure 1 shows the set points of MAES-I study.
Eligible women will be invited to use a small wrist device similar to a regular watch (GENEActiv Original -Activinsights®). The device contains sensors that estimate PA and sleep-awake patterns through a proper software algorithm. At the first set point of MAES-I study (between 19+0 -21+0 weeks of gestation), eligible women who agreed to participate will start using the wrist bracelet device on the non-dominant arm during day and night (24h/day) uninterruptedly until childbirth (including bathing or aquatic activities). The acquisition of actigraphy data can be performed in different frequencies (from 10Hz to 100Hz). Since the frequency of data acquisition impacts on the battery life of the device (inverse relationship), the measurement frequency will be set up according to the participant´s gestational age (Table 3). This information will be registered in the database accordingly. The data accumulated will be downloaded during participant´s antenatal care visits, according to the maximum return periods showed in Table 3. The maximum return periods were calculated taking into consideration the expected battery life.
A leaflet with detailed information and FAQ (Frequently asked questions) on the device will also be provided to the women. They will also have a cell phone number to call whether doubts arise regarding the procedures for using the device, or if any technical or medical concern arises. All actigraphy data collected will be entered into proper software to interpret data and generate an output file. Then, the actigraphy data will be uploaded to an online database platform developed by MedSciNet®, where all clinical data of the study will also be registered. The actigraphy software uses several algorithms to estimate physical activity and sleep patterns. The database in centralized, secure, internetbased and allows several procedures for prospective and retrospective monitoring, hierarchical access (local user, general manager, etc.). The database will be translated into Portuguese and English, facilitating data collection for Portuguese-speaking team and international monitoring. A correspondent paper form will be available for data collection if necessary (e.g. internet connection failure for instance). Decision to start monitoring PA and sleep patterns between 19-21weeks There are various underlying mechanisms involved in the development of the maternal and perinatal adverse outcomes that will be assessed, as preterm birth, preeclampsia, gestational diabetes, fetal growth restriction and small for gestational age. The preclinical phase, stage where there are no clinical signs or symptoms, might be different for each disease and dependent on environmental and individual aspects. The study of adverse maternal and perinatal predictors has been focused in early pregnancy so far (first trimester), aiming to maximize the window of opportunity for preventative interventions. However, we hypothesized that the modification of PA and/or sleep pattern due to maternal underlying changes of biological function might not be evident at a very early stage in pregnancy before the beginning of the pre-clinical phase. Our hypothesis is that it possibly occurs shortly before symptoms.
Additionally, to establish the period between 19-21 weeks as appropriate to start the assessment of PA and sleep patterns taking into consideration that the prevalence of the main maternal complications, as preeclampsia, fetal growth restriction, and preterm birth, are more common in late pregnancy. A recent cross-sectional study conducted in 20 referral centres in Brazil, including the five participating centres of this proposal, showed that the occurrence of preterm birth before 28 weeks comprised less than 1% of all births and less than 8% of all preterm births [44]. In addition, the early onset of preeclampsia (before 34 weeks of gestation) complicates less than 0.4% of all pregnancies, according to a large retrospective cohort of more than 450,000 deliveries in USA [45]. Figure 2 outlines the predicted prevalence of preterm birth and preeclampsia in the second trimester, highlighting its clinical presentation through pregnancy in red. Our hypothesis is that PA and sleep patterns might be altered closely to the clinical presentation, still in preclinical phase. Thus, the start of assessment between 19-21 weeks seems to be very reasonable, providing a wide interval to monitor and predict the main maternal and perinatal adverse outcomes.

Actigraph device
The actigraph device that will be used to monitor PA and sleep-wake patterns is GENEActiv Original (GENEActiv, Activinsights Ltd, Huntingdon, UK). The device has multiple sensors as microelectromechanical (MEMS) accelerometer, temperature

Wrist vs waist wear: advantages and performance
Wrist wear of actigraph devices provides more comfortable use during wake and sleep periods and highest wear time compared to waist monitors [33,46]. A non-systematic review published in 2011 showed that actigraphy is a useful and reliable tool to assess sleep patterns and circadian rhythm disorders, although there are some limitations on diagnosing sleep diseases or measuring sleep stages [25]. Actigraphy showed excellent concordance with polysomnography in assessing sleep parameters in healthy subjects (sensitivity >90% in estimating total sleep time, for instance). A recent study evaluated the concordance of physical activity estimation of wrist device in free-living settings in forty overweight or obese women [34]. They used both wrist and hip devices, and a small camera that captured participant behaviour for 7 days, enabling the monitoring of physical activity behaviour (gold-standard comparison). The hip and wrist machine learning (ML) classifiers used are different due to the different methods/algorithms to estimate physical activity [34]. The sensitivity and specificity of hip and wrist estimations according to Ellis et al are showed in Table 4 [34].
Two years ago, the same author had published a similar evaluation using 40 adults (women and men), showing that the hip and wrist accelerometers obtained an average accuracy of 92.3% and 87.5%, respectively, in predicting PA types [47].
Staudenmayer et al developed an investigation with 20 participants also using two devices (wrist-and hip-worn), and showed that wrist actigraphy can estimate energy expenditure accurately and relatively precisely [48]. Another study evaluating PA patterns in a free-living environment with wrist devices showed that women in the top 40% or bottom 40% of the distribution of daily PA, hip and wrist accelerometers agreed on the classification for about 75% of the women [49]. Additionally, the total activity (counts per day) was moderately correlated (Spearman's r = 0.73) between the wrist and hip worn devices.
At the best of our knowledge, there are no systematic reviews or other high-quality evidence-based recommendation supporting a particular method. Although wrist wear of actigraphy is more conventional and accurate, it might not be the best choice for assessing long periods of PA or sleep patterns, even more considering the similar performance of the wrist wear. The current proposal does not intend to diagnose pathologic behaviours or diseases, but to identify different patterns along pregnancy and in different subgroups of women. Therefore, supported by the evidence that wrist wear of actigraphy devices can accurately and more comfortably estimate PA and sleep patterns, mainly for long periods and in the free-living environment, the MAES-I study group addopted wrist wear devices.

Main variables
The independent variables assessed as potential predictors of maternal complications will be related to sleep-wake cycle and mobility as: The actigraph device collects many pieces of information related to body position and body movements to estimate the described sleep variables. Then, actigraphy software will be used to analyse the data and generates the output variables. according to Count per minute (CPM) cut points, the PA intensity can be categorized [50]. The information is translated by the software using proper algorithms into quantitative variables as following: -MET rates: Metabolic Equivalents (METs) are commonly used to also express the intensity of physical activities. One MET is the energy cost of resting quietly, often defined in terms of oxygen uptake as 3.5 mL·kg -1 ·min -1. MET rate expresses a person's working metabolic rate relative to their resting metabolic rate. Briefly, the triaxial piezoelectric sensors stressed by acceleration forces can estimate the intensity of movements, converted to the oxygen consumption required to perform such movement.
-Step counts/day: estimated steps count per day.

Outcomes
The primary outcomes are late pregnancy complications as: hours of the childbirth.
Secondary outcomes include childbirth variables and neonatal adverse outcomes as fetal death, caesarean section, small for gestational age (defined as birth weight below percentile 10 for gestational age), Apgar score < 7 at 5 minutes, neonatal severe morbidity (Table 5) and neonatal mortality before discharge.  [55].

Analyzes and statistics details
According to these studies, the predicted incidence of these complications, the leading causes of maternal and perinatal adverse outcomes, seems adequate for the current proposal and sample size estimation, although the complications are not cumulative.
Firstly, we will identify PA and sleep-wake patterns of women who did not develop adverse maternal or perinatal outcomes. It will permit the recognition of a normal PA and sleep-wake patterns in low-risk population without complication during pregnancy. Using the same population, we will analyze changes in PA and sleep-wake patterns through pregnancy, allowing for gestational age periods.
Then, we will compare the PA and sleep-wake patterns of women who developed specific adverse maternal or perinatal outcomes with those who did not. The differences between groups might be identified to be used as potential markers for specific pregnancy complications.
After that, we will analyze changes in PA and sleep-wake patterns of women who developed adverse maternal or perinatal outcomes through pregnancy, comparing the patterns before with those after the onset of maternal complications, trying to discover which changes might be related to pregnancy complications.
Finally, we will develop a predictive model for screening pregnant women for risk of specific adverse maternal and perinatal outcomes using PA and sleep-wake data estimated with actigraphy technology.
The analysis will be performed using the actigraph software that translates the collected information into PA and sleep-wake parameters. Additionally, Friedman and Wilcoxon for paired samples, t-test, and ANOVA for repeated measures will be applied to achieve statistical analyses.

Discontinuation of participants
The criteria for discontinuation include: -Withdrawal of consent; -Not regularly using the actigraph device for long periods, above 50% of all planned time. The information that they are not using the device properly will be recorded if women notice the MAES-I team. Otherwise, the low use of the device will be noticed after data discharge during antenatal care visits.
-The loss to follow-up, not allowing the download of actigraphy data.

Data and Sample Quality
All entered data will be prospectively and retrospectively monitored by local research assistants and a global monitor. Internal consistency of variables will be constantly performed by the database and error messages are automatically flagged. A local research assistant will be responsible for checking all forms and actigraphy data before locking forms, assuring good quality of data. Then, the local principal investigator (PI) will be responsible for signing the case, enabling its incorporation to the final database.  [56], are not uncommon, establishing a barrier between early recognition of symptoms and timely interventions capable to successfully treat potentially life-threatening conditions. We believe that women will feel encouraged, empowered and willing to participate in the study that aims to develop a potentially useful prenatal care tool to identify the risk for maternal and perinatal morbidity and mortality. Following national ethical regulations, the participants will not receive any financial compensation.
Women who agree to participate in the study will not have any disadvantages or injury of their prenatal care. On the contrary, they will receive a telephone number to contact the clinical researchers at any time (24/7 service), which enables a closer contact with researchers and providers of care, since the MAES-I team are committed to contacting providers of care if any potential complication is noticed by participants.
The participating women will not be responsible if loss, theft or damage to the wrist device occurs. However, they will be asked to return the device after they finished the participation in the study, in order to use it for new women entering the study. No selfdamage is expected in those who use the device. The identity of all women will be kept confidential.
Participating women will not be able to identify any PA or sleep parameters at any stage of the study. The download of the data is only possible through the own licensed software of the device.  [57].

Patient and Public Involvement
Patients and public were not involved in this study for the development of the research question and outcome measures. However, the choice for a wrist device was based on the preference of users as reported. Participants of the study will have access to the results by its webpage that will open access.

Discussion
The actigraphy is an innovative, non-invasive, non-operator dependent, wearable technology, which enables the estimative under real life conditions of diverse variables related to mobility, physical activity, sleep-wake, and circadian cycle patterns.
Actigraph devices show high sensitivity in sleep-wake parameters detection and are currently highly recommended by the American Sleep Disorder Association for diagnosis and therapy response of circadian rhythm disorders [27,28,58]. Although some studies show that 7 to 14 days using the actigraph device provides reliable estimates of physical activity behavior in older adult, it is not absolutely clear how many days is needed to estimate habitual PA by using the wrist/waist device during pregnancy. In general, it seems to depend mainly on the type of actigraph device, position of wear and target population [30,33]. Nevertheless, MAES-I study will provide sufficient data to assess different patterns along pregnancy.
The use of wearable physical activity monitors has grown enormously due to the interest about the relationship between the pathophysiology of diseases and physical activity and sleep patterns. A recent study on the use of physical activity monitors in human physiology research unravels the current and potential uses of actigraph device as in strategies to promote healthier behaviour or to predict outcomes [59]. and subjective assessments of sleep time [60].
Alterations in sleep patterns, as less deep sleep and more nocturnal awakenings, can be observed in pregnancy as early as 10-12 weeks gestation [61]. Sleep disturbances during pregnancy have been associated with preterm delivery, gestational hypertensive disorders, glucose intolerance and increased risk of caesarean delivery [24]. Shorter night time sleep was also associated with hyperglycemia [60]. Persistent sleep deficiency is correlated with depressive symptoms and stress perception by pregnant women [61]. These studies lay correlation between PA patterns and sleep disturbances determining complications, in a well-established relationship of cause and consequence, although sometimes it could not be adequately determined due to the study design [17].
In a distinct way, our analysis intends to figure out if the maternal complication could be identified by physical activity and/or sleep patterns modifications, even during its  The current clinical and biological predictors   for the main maternal complications as preeclampsia, preterm birth, maternal haemorrhage, and gestational diabetes still lack for effective sensitivity and specificity.
If this is confirmed to be true, an important step will be achieved for a possible introduction of screening non-invasive procedures during prenatal care with the purpose of identifying women at higher risk of developing those conditions. Therefore, they could receive specific orientation on prevention and earlier detection of the onset of condition for taking immediate action to look for professional health care and receiving appropriate interventions, avoiding delays that are the most striking factor for the low quality of care the women usually receive in low and middle-income settings, contributing to the still high burden of maternal morbidity and mortality. If we were successful in identifying such "specific patterns of physical activity and sleep" as predictors for pregnancy complications, further validation studies will necessarily be recommended for assessing its effectiveness for the whole management of such conditions. Additionally, MAES-I will enable further specific studies among high risk population and also will help to identify the best gestational age for monitoring, giving the means to target a specific gestational age interval.   1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57        complications would be a step forward in the improvement of maternal and perinatal 3 health. There is association between modification of physical activity (PA) and sleep- 4 wake patterns and the occurrence of inflammatory, metabolic, pathologic conditions 5 as chronic diseases. The actigraph device is validated to estimate PA and sleep-wake 6 patterns and might be valuable to identify predictors for maternal complications, 7 widening the window of opportunity to prevent, diagnose or treat specific conditions 8 prior to the development of typical symptoms or clinical signs, assessing PA and sleep- 9 wake patterns during pregnancy. 10 Methods and analysis: A cohort will be held in 5 centres from the Brazilian Network 11 for Studies on Reproductive and Perinatal Health. MAES-I will enroll 400 low-risk 12 nulliparous women that will wear the actigraph device on the wrist during day and 13 night (24h/day) uninterruptedly from 19-21 weeks until childbirth. Changes in PA and 14 sleep-wake patterns through pregnancy will be analyzed, considering gestational age 15 ranges, in women with and without maternal complications during pregnancy, such as 16 preeclampsia, preterm birth (spontaneous and provider-initiated), gestational 17 diabetes, maternal haemorrhage and also perinatal outcomes. A predictive model for 18 screening pregnant women for risk of specific adverse maternal and perinatal 19 outcomes is planned to be then developed using the actigraphy data. 20 Ethics and Dissemination: MAES-I study has been reviewed and approved by each 21 Institutional Review Board (IRB) and also by the National Council for Ethics in Research. 22 Detailed information of the study is provided in the Brazilian Cohort website 23 (www.medscinet.com/samba) and findings will be publicized in scientific literature and 24 Institutional webpages.  1 • This multicentre cohort will collect comprehensive data on the main maternal 2 and perinatal complications as pre-eclampsia, small for gestational age/fetal 3 growth restriction, preterm birth and gestational diabetes mellitus. 4 • Physical activity and sleep patterns will be estimated through an innovative 5 wearable device used in the natural environment of the study subject. 6 • Physical activity and sleep patterns will be estimated from the beginning of the 7 second half of pregnancy until delivery, covering a wide interval during 8 pregnancy and enabling the study of PA and sleep patterns changes throughout 9 pregnancy. 10 • One possible limitation refers to the uncovered first half of pregnancy 11 regarding this information.  Reducing the global maternal mortality ratio to less than 70 per 100,000 live births by 2 2030 is one of the targets of the new United Nations Sustainable Development Goals 3 [1]. Multiple challenges need to be tackled to achieve this target, but the 2016-2030 4 health and development agenda goes well beyond mortality reduction. The Global 5 Strategy for Women's, Children's and Adolescent's Health aims to ensure that every 6 newborn, women and child not only survive but thrive. This will only be possible if a 7 transformative agenda, with innovation at the central stage, is put into action [2]. 8 One of the major challenges that need to be addressed is optimizing the recognition of 9 earlier predictors and identifiers of maternal and perinatal complications. Delays in 10 diagnosing and managing maternal complications have been associated with poor 11 outcomes [3]. The reduced self-perception of clinical signs related to maternal 12 complications, difficulties in accessing the health system and poor quality of care may 13 contribute to the late identification of complications and worsened prognosis. The 14 development of a non-invasive Antenatal Care (ANC) tool capable of identifying 15 maternal sub-clinical signs during pregnancy may provide the window of opportunity 16 for the earlier identification of abnormal patterns of physiologic parameters related to 17 pregnancy complications. We consider earlier identification when the recognition 18 could be made before clinical presentation, when standard criteria based on clinical 19 signs, symptoms, and supplementary tests are presented. Shortening the time 20 between the onset of a complication and the initiation of the appropriate 21 management allow for secondary prevention and reduction of maternal morbidity and 22 mortality [3][4][5][6][7]. 23 Pervasive computing (i.e. the trend towards embedding microprocessors in everyday- 24 life objects so they can generate data) and wearable technology (i.e. clothing and 25 accessories incorporating computer and advanced electronic technologies such as 26 wrist and/or waistband sensors) are ubiquitous and able to generate a new dataset 27 that needs to be correlated with pregnancy outcomes. Preterm birth and 28 preeclampsia, for instance, are two important pregnancy complications that have a 29 relatively long subclinical phase before the appearance of signs or symptoms [8,9]. It is 30 plausible that during this subclinical phase of certain conditions the pattern of physical 31  [10,11], published literature correlating wearable technology data 4 and maternal complications are very scarce. 5 The human circadian rhythm is ruled by endogenous physiologic mechanisms and 6 environmental stimuli [12]. There is solid evidence showing that modification of 7 circadian rhythm or sleep and PA patterns are an underlying condition related to 8 inflammatory, degenerative and/or metabolic chronic diseases as diabetes, 9 hypertension, and cancer [13]. Circadian misalignment is defined as having 10 inappropriate timed sleep and wake, misplaced feeding periods and modification of 11 activity behavior. 12 The determination of cause or consequence effect between these modifications and 13 the development of pathological conditions is a complex task. It seems that changes in 14 appetite-stimulating hormones, glucose metabolism, inflammatory markers, and mood 15 are some of the related pathways [13][14][15]. Leproult et al. evaluated the effect of 16 circadian misalignment on metabolic and inflammation markers in cardiovascular 17 disease [15]. The insulin action and release, and also the levels of some inflammatory 18 markers that are predictors for cardiovascular diseases, were abnormal in individuals 19 with circadian misalignment. The mechanisms involved in the association between 20 changes of PA pattern and pathologic conditions seem to be of multiple etiologies. Sani 21 et al. assessed the circadian rithm of more than 2,300 African descendant adults. More 22 than evaluating physical activity itself, the authors aimed to identify chronobiologic 23 patterns of adults from different socioeconomic settings. The study identified that 24 chronobiologic behavior can vary depending on individual BMI, socioeconomic 25 background, work type and time of sunlight exposure. Possibly, many other factors are 26 involved in modifications of chronobiologic behavior, such as pathologic conditions. 27 Some metabolic, cognitive, cardiovascular and other chronic degenerative diseases 28 have been associated with particular patterns of PA and sleep [10,11,[16][17][18]. A 29 previous observational study assessed various sleep parameters during pregnancy, e.g. 30 sleep onset latency (SOL), wake after sleep onset (WASO) and total nocturnal sleep 31  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  pregnancy adverse outcomes, discussing common mechanisms of stress system 4 activation [19]. Low-quality evidence suggests an association between sleep loss and 5 prenatal depression, gestational diabetes, preeclampsia, abnormal length of labour, 6 caesarean delivery, abnormal fetal growth, and preterm birth. Those results 7 corroborate with other findings regarding pregnancy and sleep disorders [20][21][22][23]. 8 The assessment of PA and the sleep patterns can be performed using small wrist (or 9 waist) devices similar to a regular watch (actigraphy technology). The type of sensor, 10 batteries, materials and output data have been substantially developed in recent 11 years, enabling low cost, comfort, discretion and performance [24]. Nowadays there 12 are devices that are portable, lightweight and with a large capacity to storage 13 information, including a software with automatic scoring algorithms packages for the 14 detection of wakefulness, sleep periods and PA [24,25]. The actigraphy estimation of 15 PA and sleep patterns is validated as a proxy for chronobiologic behavior [26][27][28][29] and 7 16 to 14 days using the actigraph device provides reliable estimates of PA behavior in 17 older adults [30][31][32]. Both hip and wrist devices show reliable and acceptable 18 performance in estimating PA and sleep-wake patterns [33][34][35][36]. 19 The main advantages of using wearable devices for actigraphy is the non-invasiveness, 20 24/7 monitoring of the circadian pattern and PA, and informing sleep habits and 21 parameters in the user natural environment [24,25,28]. We propose an innovative 22 and strategic approach to monitor PA and sleep-wake patterns during pregnancy, 23 establishing a large database comprised of clinical, epidemiological, PA and sleep-wake 24 variables potentially capable of composing a prediction model for maternal 25 complications during pregnancy. The main goal of this study is to identify earlier 26 predictors of pregnancy complications by correlating data generated on PA and sleep 27 patterns through wearable devices (wristband sensors) with maternal and perinatal 28 complications and outcomes. 29 1 We will conduct a cohort study of 400 low-risk pregnant women using wrist sensor 2 bands able to capture information on daily physical activity and sleep patterns 3 (exposure). This cohort study will be implemented in 5 ANC clinics linked to obstetric 4 units in 3 different regions of Brazil that are already part of the Brazilian Network for 5 Studies on Reproductive and Perinatal Health [37], as shown in Table 1. During a 6 period of eight months, the ANC clinics will identify eligible cases for using the 7 wristband sensors. Wearable technology data will be correlated with the occurrence 8 of pregnancy and childbirth complications and outcomes, thus understood as an 9 effect. 10 Eligible women will be identified up to 21 weeks of gestation and invited to participate. 11 A proper consent form will be applied and the women who agree to participate will 12 receive a wristband sensor to be used starting at 19-21 weeks until childbirth, 13 uninterruptedly. 14 15 Brazil is a multi-ethnic mixed-race population of diverse resourced settings [38]. 16 Despite the considered high global overall human development index (HDI 0.727) in 17 2010, the HDI of Brazilian municipalities ranged from 0,862 to 0,418 [39]. The 18 possibility of considering such mixed population is suitable to explore information 19 regarding maternal patterns of mobility and sleep, maximizing external validity and 20 comparisons to other populations. A few reasons might support the study population 21 focused in low-risk nulliparous women: 1) Previous obstetric history can refer to 22 known risk factors for many maternal complications such as preterm birth, 23 preeclampsia, and diabetes [13,40]. Nulliparous women enable unbiased sampling 24 regarding obstetric history. 2) Women with previous morbidity such as hypertension, 25 diabetes, nephropathy or others chronic/degenerative diseases are more likely to 26 present abnormalities of sleep-wake rhythm or physical activity patterns during 27 pregnancy.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  The five participating centers are regional referral obstetric units responsible for 1 antenatal care assistance mainly for high-risk pregnant women. Participating centers 2 are listed in Table 1. Nevertheless, there are primary health care units strategically 3 linked with these participating centers, enabling the identification and enrollment of 4 low-risk pregnant women. The recruitment strategies include approaching all eligible 5 women in these participating centers and their linked facilities. An informed consent 6 form will be applied for women who agree to participate. 7 Eligible women: Low-risk pregnant subjects 8 There is no international consensus on the criteria for low-risk pregnancies, although 9 there are several known factors associated with maternal and perinatal adverse 10 outcomes. A recent study evaluating complications of "low-risk" pregnancies of US 11 Americans (10 million births from 2011 through 2013) showed that 29% of low-risk 12 women had an unexpected complication requiring no routine obstetric/neonatal care 13 [41]. This shows the difficulty in establishing a "low-risk profile" for maternal/perinatal 14 complications. As an exploratory study, we tried to exclude potential known 15 confounders of pre-pregnancy conditions related to adverse maternal or perinatal 16 outcomes as shown in Table 2, in order to assess PA and sleep patterns of a mostly 17 "normal" population. Nonetheless, lifestyle habits and body composition (Body mass 18 index, height, etc.) characteristics, and some non-severe chronic diseases as non- 19 severe anaemia and/or asthma are not among exclusion criteria but may be part of 20 subgroup analyses (composition of any previous disorder, e.g.). Intra and inter- 21 individual analyses of PA and sleep patterns enable the identification of potential 22 confounders affecting primary outcomes, avoiding potential biases. It means that 23 comparison of PA and sleep pattern parameters collected in different stages of 24 pregnancy from the same participant (intra-individual analysis) and collected at the 25 same stage of pregnancy from different participants (inter-individual analysis) will be 26 carried out. 27 Eligible women will be enrolled between 04-21 weeks. The inclusion and exclusion 28 criteria are presented in Table 2.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   9 Essentially, MAES I study is comprised of 4 key set points -3 clinical visits during 1 pregnancy and a postnatal visit. The clinical visits will be held at 1) between 19-21 2 weeks; 2) between 27-29 weeks; 3) between 37-39 weeks. At first, second, third and 3 postnatal visits, additional information regarding maternal history, details on 4 pregnancy complications, maternal biophysical data (weight, height, skinfolds) and 5 pregnancy adverse outcomes will be collected following a specific Standard Operating 6 Procedure (SOP) specially developed for MAES-I study. Additionally, the Perceived 7 Stress Scale [42] and Resilience Scale [43] will be applied during 27-29w visit. Figure 1   8 shows the set points of MAES-I study. 9 Eligible women will be invited to use a small wrist device similar to a regular watch 10 (GENEActiv Original -Activinsights®). The device contains sensors that estimate PA and 11 sleep-awake patterns through a proper software algorithm. At the first set point of 12 MAES-I study (between 19+0 -21+0 weeks of gestation), eligible women who agreed 13 to participate will start using the wrist bracelet device on the non-dominant arm 14 during day and night (24h/day) uninterruptedly until childbirth (including bathing or 15 aquatic activities). The acquisition of actigraphy data can be performed in different 16 frequencies (from 10Hz to 100Hz). Since the frequency of data acquisition impacts on 17 the battery life of the device (inverse relationship), the measurement frequency will be 18 set up according to the participant´s gestational age (Table 3). This information will be 19 registered in the database accordingly. The data accumulated will be downloaded 20 during participant´s antenatal care visits, according to the maximum return periods 21 showed in Table 3. The maximum return periods were calculated taking into 22 consideration the expected battery life. At each antenatal care visit, the used device 23 will be returned to the research team and a new charged device will be provided to the 24 participant. 25 A leaflet with detailed information and FAQ (Frequently asked questions) on the device 26 will also be provided to the women. They will also have a cell phone number to call 27 whether doubts arise regarding the procedures for using the device, or if any technical 28 or medical concern arises. 29 All actigraphy data collected will be entered into proper software to interpret data and 30 generate an output file. Then, the actigraphy data will be uploaded to an online 31 also be registered. The actigraphy software uses several algorithms to estimate 2 physical activity and sleep patterns. The database in centralized, secure, internet-3 based and allows several procedures for prospective and retrospective monitoring, 4 hierarchical access (local user, general manager, etc.). The database will be translated 5 into Portuguese and English, facilitating data collection for Portuguese-speaking team 6 and international monitoring. A correspondent paper form will be available for data 7 collection if necessary (e.g. internet connection failure for instance). 8 Decision to start monitoring PA and sleep patterns between 19-21weeks 9 There are various underlying mechanisms involved in the development of the maternal 10 and perinatal adverse outcomes that will be assessed, as preterm birth, preeclampsia, 11 gestational diabetes, fetal growth restriction and small for gestational age. The pre- 12 clinical phase, stage where there are no clinical signs or symptoms, might be different 13 for each disease and dependent on environmental and individual aspects. The study of 14 adverse maternal and perinatal predictors has been focused in early pregnancy so far 15 (first trimester, <14 weeks of gestation), aiming to maximize the window of 16 opportunity for preventative interventions. However, we hypothesized that the 17 modification of PA and/or sleep pattern due to maternal underlying changes of 18 biological function might not be evident at a very early stage in pregnancy before the 19 beginning of the pre-clinical phase. Our hypothesis is that it possibly occurs shortly 20 before symptoms. 21 Additionally, we took into account that the occurrence of the main maternal 22 complications, as preeclampsia, fetal growth restriction, and preterm birth, are more 23 common in late pregnancy to establish the period between 19-21 weeks as 24 appropriate to start the assessment of PA and sleep patterns. A recent cross-sectional 25 study conducted in 20 referral centres in Brazil, including the five participating centres 26 of this proposal, showed that the occurrence of preterm birth before 28 weeks 27 comprised less than 1% of all births and less than 8% of all preterm births [44]. In 28 addition, the early onset of preeclampsia (before 34 weeks of gestation) complicates 29 less than 0.4% of all pregnancies, according to a large retrospective cohort of more 30 than 450,000 deliveries in USA [45]. Figure 2 outlines the predicted prevalence of 31 preterm birth and preeclampsia in the second trimester, highlighting its clinical 1 presentation, when classic symptoms and signs of a certain disease/complication are 2 presented, through pregnancy in red. Our hypothesis is that PA and sleep patterns 3 might be altered closely to the clinical presentation, still in preclinical phase when 4 there is no symptoms or signs. 5 In brief, as an exploratory study, we indeed needed to make an arbitrary decision 6 regarding interval of monitoring PA and sleep patterns. For that, we had taken into 7 consideration: 1) the main maternal/perinatal complications of interest occur in the 8 second half of pregnancy, more precisely in late pregnancy ( Figure 2); 2) we 9 hypothesize that any potential change on PA or sleep patterns might occur days or 10 weeks before the onset of maternal or perinatal complication. Then, we focused 11 monitoring women during second half of pregnancy. 12 Thus, the start of assessment between 19-21 weeks seems to be very reasonable, 13 providing a wide interval to monitor and predict the main maternal and perinatal 14 adverse outcomes. 15 Actigraph device 16 The actigraph device that will be used to monitor PA and sleep-wake patterns is 17 GENEActiv Original (GENEActiv, Activinsights Ltd, Huntingdon, UK). The device has 18 multiple sensors as microelectromechanical (MEMS) accelerometer, temperature 19 (linear active thermistor) and light (silicon photodiode), providing crude raw data for a 20 variety of applications. 21 Wrist vs waist wear: advantages and performance 22 Wrist wear of actigraph devices provides more comfortable use during wake and sleep 23 periods and highest wear time compared to waist monitors [33,46]. A non-systematic 24 review published in 2011 showed that actigraphy is a useful and reliable tool to assess 25 sleep patterns and circadian rhythm disorders, although there are some limitations on 26 diagnosing sleep diseases or measuring sleep stages [25]. Actigraphy showed excellent 27 concordance with polysomnography in assessing sleep parameters in healthy subjects 28 (sensitivity >90% in estimating total sleep time, for instance). A recent study evaluated 29 the concordance of physical activity estimation of wrist device in free-living settings in  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   12 forty overweight or obese women [34]. They used both wrist and hip devices, and a 1 small camera that captured participant behaviour for 7 days, enabling the monitoring 2 of physical activity behaviour (gold-standard comparison). The hip and wrist machine 3 learning (ML) classifiers used are different due to the different methods/algorithms to 4 estimate physical activity [34]. The sensitivity and specificity of hip and wrist 5 estimations according to Ellis et al are showed in Table 4 [34]. 6 Two years ago, the same author had published a similar evaluation using 40 adults 7 (women and men), showing that the hip and wrist accelerometers obtained an average 8 accuracy of 92.3% and 87.5%, respectively, in predicting PA types [47]. 9 Staudenmayer et al developed an investigation with 20 participants also using two 10 devices (wrist-and hip-worn), and showed that wrist actigraphy can estimate energy 11 expenditure accurately and relatively precisely [48]. Another study evaluating PA 12 patterns in a free-living environment with wrist devices showed that women in the top 13 40% or bottom 40% of the distribution of daily PA, hip and wrist accelerometers 14 agreed on the classification for about 75% of the women [49]. Additionally, the total 15 activity (counts per day) was moderately correlated (Spearman's r = 0.73) between the 16 wrist and hip worn devices. 17 At the best of our knowledge, there are no systematic reviews or other high-quality 18 evidence-based recommendation supporting a particular method. Although wrist wear 19 of actigraphy is not the more traditional method, it might be the best choice for 20 assessing long periods of PA or sleep patterns, even more considering the similar 21 performance of the waist wear. The current proposal does not intend to diagnose 22 pathologic behaviours or diseases, but to identify different patterns along pregnancy 23 and in different subgroups of women. Therefore, supported by the evidence that wrist 24 wear of actigraphy devices can accurately and more comfortably estimate PA and 25 sleep patterns, mainly for long periods and in the free-living environment, the MAES-I 26 study group addopted wrist wear devices. 27 Main variables 28 The independent variables assessed as potential predictors of maternal complications 29 will be related to sleep-wake cycle and mobility as:  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58 6 sleep-ratio of total sleep time to time in bed. 7 The actigraph device collects many pieces of information related to body position and 8 body movements to estimate the described sleep variables. Then, actigraphy software 9 will be used to analyse the data and generates the output variables. 10 "Physical activity" variables 11 Actigraphy technology estimates physical activity through various parameters 12 collected by the actigraph device. Briefly, according to Freedson et al, the triaxial 13 sensors stressed by acceleration forces can estimate the intensity of movements. The 14 acceleration signal is converted to digital signal and summed over a user specified 15 time interval (epoch). At the end of each epoch the activity count is stored. Then, 16 according to Count per minute (CPM) cut points, the PA intensity can be categorized 17 [50]. The information is translated by the software using proper algorithms into 18 quantitative variables as following: 19 -Sedentary time (hours/day): the number of hours per day of count per minute 20 between 0-99. 21 -Light activity (hours/day): the number of hours per day having count per minute 22 between 100 -1951. 23 -Moderate activity (minutes/day): the number of hours per day having count per 24 minute between 1952 -5724. 25 -Vigorous activity (minutes/day): the number of hours per day having count per 26 minute between 5725 -9498. 27 -Very vigorous activity (minutes/day): the number of hours per day having count 28 per minute between 9499 -∞. the intensity of physical activities. One MET is the energy cost of resting quietly, 2 often defined in terms of oxygen uptake as 3.5 mL·kg -1 ·min -1. MET rate 3 expresses a person's working metabolic rate relative to their resting metabolic 4 rate. Briefly, the triaxial piezoelectric sensors stressed by acceleration forces can 5 estimate the intensity of movements, converted to the oxygen consumption 6 required to perform such movement. 7 -Step counts/day: estimated steps count per day (estimated by proper 8 algorithms using accelerometer data.) 9 Outcomes 10 The primary outcomes are late pregnancy complications as: 11 -Preeclampsia: Hypertension after 20 weeks of gestation, systolic BP ≥ 140mmHg 12 and/or diastolic BP ≥ 90mmHg (Korotkoff V) on at least 2 occasions 4h apart with:  -Provider-Initiated Preterm Birth: defined as childbirth occurring at less than 37 4 weeks, medically indicated due to maternal/fetal compromise or both; 5 -Maternal Hemorrhage: Classified in 1) Antepartum haemorrhage defined as 6 bleeding from the genital tract after 24 weeks of gestation; 2) Primary Postpartum 7 haemorrhage as the loss of 500 ml blood or more from the genital tract within 24 8 hours of the childbirth. 9 Secondary outcomes include childbirth variables and neonatal adverse outcomes as 10 fetal death, caesarean section, small for gestational age (defined as birth weight below 11 percentile 10 for gestational age), Apgar score < 7 at 5 minutes, neonatal severe 12 morbidity (Table 5) and neonatal mortality before discharge. 13 Plans for analyses 14 Sample size estimation 15 This is an exploratory and innovative study focused on a specific population (pregnant 16 women) and therefore there are no previously published parameters available for 17 sample size estimation. Considering a relatively wide range of frequency of 18 complications arising in pregnancy from 3 to 20% (including preeclampsia, fetal growth 19 restriction, gestational diabetes, hemorrhage, preterm birth, etc.), a theoretical large 20 population size (above 1 million pregnant women), an acceptable margin of error of 21 4%, the involvement of 5 clusters (participating centers) and a 95% level of confidence, 22 384 women would be necessary. Therefore, we are rounding up this estimation for at 23 least 400 initially low-risk pregnant women to be enrolled in the study. We estimated 24 the incidence of some main maternal complications considering the following studies: 25 -Pre-eclampsia: An international prospective cohort study with nulliparous women 26 called SCOPE used similar criteria for the low-risk profile, having 5% of incidence of 27 pre-eclampsia [53].  [44]. 3 -Gestational Diabetes: SCOPE international cohort, previously mentioned, had a 4 prevalence of 8.9% of gestational diabetes in low-risk screened nulliparous women, 5 according to mainly to the NICE guidelines [54]. 6 -Fetal growth restriction/small for gestational age: SCOPE international cohort, 7 previously mentioned, had a prevalence of 10.7% of newborns small for gestational 8 age, according to the customized centiles of birthweight (<10%) [55]. 9 Analyzes and statistics details 10 According to these studies above, the predicted incidence of these complications 11 seems reasonable and reproducible in our cohort. Then, sample size estimation might 12 assure enough cases of maternal and perinatal complications for the current proposal. 13 Firstly, we will identify PA and sleep-wake patterns of women who did not develop 14 adverse maternal or perinatal outcomes. It will permit the recognition of a normal PA 15 and sleep-wake patterns in low-risk population without complication during 16 pregnancy. Using the same population, we will analyze changes in PA and sleep-wake 17 patterns through pregnancy, allowing for gestational age periods. 18 Then, we will compare the PA and sleep-wake patterns of women who developed 19 specific adverse maternal or perinatal outcomes with those who did not. The 20 differences between groups might be identified to be used as potential markers for 21 specific pregnancy complications. 22 After that, we will analyze changes in PA and sleep-wake patterns of women who 23 developed adverse maternal or perinatal outcomes through pregnancy, comparing the 24 patterns and trying to discover which changes and when before the onset it would be 25 related to pregnancy complications. If possible, we will conduct subgroup analysis 26 including subpopulation with potential higher risk for maternal complications 27 (confounder variabels), inclusing obesity, smoking, etc. Finally, we will develop a predictive model for screening pregnant women for risk of 1 specific adverse maternal and perinatal outcomes using PA and sleep-wake data 2 estimated with actigraphy technology. 3 The analysis will be performed using the actigraph software that translates the 4 collected information into PA and sleep-wake parameters. Additionally, Friedman and 5 Wilcoxon for paired samples, t-test, and ANOVA for repeated measures will be applied 6 to achieve statistical analyses. Also, we will address sensitivity, specificity and 7 likelihood ratio for altered PA and sleep patterns or for their changes throughout 8 pregnancy. 9 Discontinuation of participants 10 The criteria for discontinuation include: 11 -Withdrawal of consent; 12 -Not regularly using the actigraph device for long periods, less than 50% of all 13 planned time. The information that they are not using the device properly will be 14 recorded if women notice the MAES-I team. Otherwise, the low use of the device 15 will be noticed after data discharge during antenatal care visits. 16 -The loss to follow-up, not allowing the download of actigraphy data. 17 18 All entered data will be prospectively and retrospectively monitored by local research 19 assistants and a global monitor. Internal consistency of variables will be constantly 20 performed by the database and error messages are automatically flagged. A local 21 research assistant will be responsible for checking all forms and actigraphy data before 22 locking forms, assuring good quality of data (double-checking entered data and 23 checking for inconsistencies between variables, for instance). Then, the local principal 24 investigator (PI) will be responsible for signing the case, enabling its incorporation to 25 the final database. The University of Campinas will coordinate, implement and monitor 26 the study in the five participating centres. A general manager and a global monitor are 27 also part of the team of the coordinating centre. The local team of each participating 28 centre is comprised of a Local PI and research assistants. 29  considered of low risk for maternal and perinatal complications, they are not free of 2 suffering complications. Furthermore, the first and second delays, defined as a delay in 3 deciding to seek care and delay in reaching a health care facility [56], are not 4 uncommon, establishing a barrier between earlier recognition of symptoms and timely 5 interventions capable to successfully treat potentially life-threatening conditions. We 6 believe that women will feel encouraged, empowered and willing to participate in the 7 study that aims to develop a potentially useful prenatal care tool to identify the risk for 8 maternal and perinatal morbidity and mortality. Following national ethical regulations, 9 the participants will not receive any financial compensation. 10 Women who agree to participate in the study will not have any disadvantages or injury 11 of their prenatal care. On the contrary, they will receive a telephone number to 12 contact the clinical researchers at any time (24/7 service), which enables a closer 13 contact with researchers and providers of care, since the MAES-I team are committed 14 to contacting providers of care if any potential complication is noticed by participants. 15 The participating women will not be responsible if loss, theft or damage to the wrist 16 device occurs. However, they will be asked to return the device after they finished the 17 participation in the study, in order to use it for new women entering the study. No self- 18 damage is expected in those who use the device. 19 Participating women will not be able to identify any PA or sleep parameters at any 20 stage of the study. The download of the data is only possible through the own licensed 21 software of the device. Actigraphy devices provided for participating women have a 22 unique code which will be recorded in the database together with the interval of use 23 for each women. Actigraphy data will be labelled using participant ID, device number, 24 gestational age when starting using each device and return date of each device. The 25 use of such codes, ID`s and numbers will ensure confidential identify for all 26 participating women. The identity of all women will be kept confidential. 27 MAES-I study has been reviewed and approved by the National Committee for Ethics in 28 Research of Brazil (CONEP) and by the Institutional Review Board (IRB) of the 29 coordinating centre (Letter of approval 1.834.116 issued on 24 th November 2016) and of all other Brazilian participating centres. All women who will be enrolled in the 1 MAES-I cohort will sign an informed consent form. 2 The ethical principles stated in the Brazilian National Heath Council (Resolution CNS 3 466/12) will be respected in every stage of this study. The anonymity of the source of 4 information will be guaranteed and the care for the women will be provided 5 independently of her agreement to participate in the study. The study also complies 6 with the Declaration of Helsinki amended in Hong Kong in 1989.The methodological 7 and ethical aspects of MAES-I study protocol were developed following STROBE 8 guidelines [57]. 9 Patient and Public Involvement 10 Patients and public were not involved in this study for the development of the 11 research question and outcome measures. However, the choice for a wrist device was 12 based on the preference of users as reported. Participants of the study will have access 13 to the results by its webpage that will be open access. 14 Detailed information of the study is provided in the Brazilian Cohort website 15 (www.medscinet.com/samba) and findings will be publicized in scientific literature and 16 Institutional webpages. We intend to disseminate our findings in scientific peer- 17 reviewed journal, general free access website, specialists' conferences, and to our 18 funding agencies. 19 20

21
The actigraphy is an innovative, non-invasive, non-operator dependent, wearable 22 technology, which enables the estimative under real life conditions of diverse variables 23 related to mobility, physical activity, sleep-wake, and circadian cycle patterns. 24 Actigraph devices show high sensitivity in sleep-wake parameters detection and are 25 currently highly recommended by the American Sleep Disorder Association for 26 diagnosis and therapy response of circadian rhythm disorders [27,28,58]. Although 27 some studies show that 7 to 14 days using the actigraph device provides reliable 28 estimates of physical activity behavior in older adult, it is not absolutely clear how 29 many days is needed to estimate habitual PA by using the wrist/waist device during 30 pregnancy. In general, it seems to depend mainly on the type of actigraph device, 1 position of wear and target population [30,33]. Nevertheless, MAES-I study will 2 provide sufficient data to assess different patterns along pregnancy. 3 The use of wearable physical activity monitors has grown enormously due to the 4 interest about the relationship between the pathophysiology of diseases and physical 5 activity and sleep patterns. A recent study on the use of physical activity monitors in 6 human physiology research unravels the current and potential uses of actigraph device 7 as in strategies to promote healthier behaviour or to predict outcomes [59]. The 8 authors conclude that physical activity monitors, as others new 21 st century 9 technologies, have already transformed physiology research, revolutionizing the way 10 we assess patients and opening new areas of interest. In addition, the use of objectives 11 measures to evaluate habitual sleep duration and outcomes in pregnancy is critical, 12 taking into account recent investigations reporting little agreement between objective 13 and subjective assessments of sleep time [60]. 14 Alterations in sleep patterns, as less deep sleep and more nocturnal awakenings, can 15 be observed in pregnancy as early as 10-12 weeks gestation [61]. Sleep disturbances 16 during pregnancy have been associated with preterm delivery, gestational 17 hypertensive disorders, glucose intolerance and increased risk of caesarean delivery 18 [19]. Shorter night time sleep was also associated with hyperglycemia [62]. Persistent 19 sleep deficiency is correlated with depressive symptoms and stress perception by 20 pregnant women [61]. These studies lay correlation between PA patterns and sleep 21 disturbances determining complications, in a well-established relationship of cause and 22 consequence, although sometimes it could not be adequately determined due to the 23 study design [17]. 24 In a distinct way, our analysis intends to figure out if the maternal complication could 25 be identified by physical activity and/or sleep patterns modifications, even during its 26 pre-clinical period, previous the appearance of clinical signs. Considering the existing 27 evidence, we speculate that the PA and/or sleep patterns change days or weeks before 28 the clinical presentation of the complication. In general, the signs and symptoms of 29 some maternal outcomes are part of the gold-standard criteria for diagnosis (high 30 blood pressure, proteinuria and/or edema in the case of preeclampsia; premature 31 contractions and cervical ripening/dilation in preterm birth; abnormal placental blood 1 flow and insufficient fetal growth in Intrauterine Growth Restriction). We acknowledge 2 the fact that there are potential confounders and limitations in predicting maternal and 3 perinatal complications using PA and sleep patterns estimated by actigraph devices. 4 We expect that our studied population will have different subgroups of women with 5 different risks and associated factors playing a role on maternal complication. It 6 includes obesity, smoking, extremes of age, for instance. None of those factors was 7 considered exclusion criteria and, if possible, we intend to assess subgroup analysis for 8 those maternal subgroups at they might present different PA and sleep patterns. 9 Nonetheless, we decided to perform a pragmatic approach, not excluding such 10 common factor from our sample. 11 The use of actigraph device during prenatal visits has a potential to become a new tool 12 to monitor pregnant women, improving maternal health care, identifying altered PA 13 and/or sleep patterns, measured objectively through actigraphy, before the occurrence 14 of those signs and symptoms. Therefore, the focus would be offering new technology 15 to monitor the development of a potential maternal complication. Other positive 16 points of our study are the period of data collection (from 19 weeks till delivery) and 17 the low-risk profile of the cohort. Through which, it would be possible to describe a PA 18 and sleep patterns in a low-risk pregnant population and better interpret actigraphy 19 data among pregnant women. The current clinical and biological predictors for the 20 main maternal complications as preeclampsia, preterm birth, maternal haemorrhage, 21 and gestational diabetes still lack for effective sensitivity and specificity. 22 If this is confirmed to be true, an important step will be achieved for a possible 23 introduction of screening non-invasive procedures during prenatal care with the 24 purpose of identifying women at higher risk of developing those conditions. Therefore, 25 they could receive specific orientation on prevention and earlier detection of the onset 26 of condition for taking immediate action to look for professional health care and 27 receiving appropriate interventions, avoiding delays that are the most striking factor 28 for the low quality of care the women usually receive in low and middle-income 29 settings, contributing to the still high burden of maternal morbidity and mortality. If 30 we were successful in identifying such "specific patterns of physical activity and sleep"  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   22 as predictors for pregnancy complications, further validation studies will necessarily be 1 recommended for assessing its effectiveness for the whole management of such 2 conditions. Additionally, MAES-I will enable further specific studies among high risk 3 population and also will help to identify the best gestational age for monitoring, giving 4 the means to target a specific gestational age interval.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58

Competing interests
The authors declare that they have no competing interests.

Strengths and limitations of this study
2  This multicentre cohort will collect comprehensive data on the main maternal 3 and perinatal complications as pre-eclampsia, small for gestational age/fetal 4 growth restriction, preterm birth and gestational diabetes mellitus. 5  Physical activity and sleep patterns will be estimated through an innovative 6 wearable device used in the natural environment of the study subject. 7  Physical activity and sleep patterns will be estimated from the beginning of the 8 second half of pregnancy until delivery, covering a wide interval during 9 pregnancy and enabling the study of PA and sleep patterns changes throughout 10 pregnancy. 11  One possible limitation refers to the uncovered first half of pregnancy regarding 12 this information. 13 14 Reducing the global maternal mortality ratio to less than 70 per 100,000 live births by 3 2030 is one of the targets of the new United Nations Sustainable Development Goals 4 [1]. Multiple challenges need to be tackled to achieve this target, but the 2016-2030 5 health and development agenda goes well beyond mortality reduction. The Global 6 Strategy for Women's, Children's and Adolescent's Health aims to ensure that every 7 newborn, women and child not only survive but thrive. This will only be possible if a 8 transformative agenda, with innovation at the central stage, is put into action [2]. 9 One of the major challenges that need to be addressed is optimizing the recognition of 10 earlier predictors and identifiers of maternal and perinatal complications. Delays in 11 diagnosing and managing maternal complications have been associated with poor 12 outcomes [3]. The reduced self-perception of clinical signs related to maternal 13 complications, difficulties in accessing the health system and poor quality of care may 14 contribute to the late identification of complications and worsened prognosis. The 15 development of a non-invasive Antenatal Care (ANC) tool capable of identifying 16 maternal sub-clinical signs during pregnancy may provide the window of opportunity for 17 the earlier identification of abnormal patterns of physiologic parameters related to 18 pregnancy complications. We consider earlier identification when the recognition could 19 be made before clinical presentation, when standard criteria based on clinical signs, 20 symptoms, and supplementary tests are presented. Shortening the time between the 21 onset of a complication and the initiation of the appropriate management allow for 22 secondary prevention and reduction of maternal morbidity and mortality [3][4][5][6][7]. 23 Pervasive computing (i.e. the trend towards embedding microprocessors in everyday- 24 life objects so they can generate data) and wearable technology (i.e. clothing and 25 accessories incorporating computer and advanced electronic technologies such as wrist 26 and/or waistband sensors) are ubiquitous and able to generate a new dataset that needs 27 to be correlated with pregnancy outcomes. Preterm birth and preeclampsia, for 28 instance, are two important pregnancy complications that have a relatively long 29 subclinical phase before the appearance of signs or symptoms [8,9]. It is plausible that 30 during this subclinical phase of certain conditions the pattern of physical activity (PA) or 31 sleep-wake rhythm is affected in some way and this change could be captured through maternal complications are very scarce. 5 The human circadian rhythm is ruled by endogenous physiologic mechanisms and 6 environmental stimuli [12]. There is solid evidence showing that modification of 7 circadian rhythm or sleep and PA patterns are an underlying condition related to 8 inflammatory, degenerative and/or metabolic chronic diseases as diabetes, 9 hypertension, and cancer [13]. Circadian misalignment is defined as having 10 inappropriate timed sleep and wake, misplaced feeding periods and modification of 11 activity behavior. 12 The determination of cause or consequence effect between these modifications and the 13 development of pathological conditions is a complex task. It seems that changes in 14 appetite-stimulating hormones, glucose metabolism, inflammatory markers, and mood 15 are some of the related pathways [13][14][15]. Leproult et al. evaluated the effect of 16 circadian misalignment on metabolic and inflammation markers in cardiovascular 17 disease [15]. The insulin action and release, and also the levels of some inflammatory 18 markers that are predictors for cardiovascular diseases, were abnormal in individuals 19 with circadian misalignment. The mechanisms involved in the association between 20 changes of PA pattern and pathologic conditions seem to be of multiple etiologies. Sani 21 et al. assessed the circadian rithm of more than 2,300 African descendant adults. More 22 than evaluating physical activity itself, the authors aimed to identify chronobiologic 23 patterns of adults from different socioeconomic settings. The study identified that 24 chronobiologic behavior can vary depending on individual BMI, socioeconomic 25 background, work type and time of sunlight exposure. Possibly, many other factors are 26 involved in modifications of chronobiologic behavior, such as pathologic conditions. 27 Some metabolic, cognitive, cardiovascular and other chronic degenerative diseases have 28 been associated with particular patterns of PA and sleep [10,11,[16][17][18]. A previous 29 observational study assessed various sleep parameters during pregnancy, e.g. sleep 30 onset latency (SOL), wake after sleep onset (WASO) and total nocturnal sleep time (TST). 31 Difficulty in initiating sleep in early pregnancy was associated with higher body mass outcomes, discussing common mechanisms of stress system activation [19]. Low-quality 4 evidence suggests an association between sleep loss and prenatal depression, 5 gestational diabetes, preeclampsia, abnormal length of labour, caesarean delivery, 6 abnormal fetal growth, and preterm birth. Those results corroborate with other findings 7 regarding pregnancy and sleep disorders [20][21][22][23]. 8 The assessment of PA and the sleep patterns can be performed using small wrist (or 9 waist) devices similar to a regular watch (actigraphy technology). The type of sensor, 10 batteries, materials and output data have been substantially developed in recent years, 11 enabling low cost, comfort, discretion and performance [24]. Nowadays there are 12 devices that are portable, lightweight and with a large capacity to storage information, 13 including a software with automatic scoring algorithms packages for the detection of 14 wakefulness, sleep periods and PA [24,25]. The actigraphy estimation of PA and sleep 15 patterns is validated as a proxy for chronobiologic behavior [26][27][28][29] and 7 to 14 days 16 using the actigraph device provides reliable estimates of PA behavior in older adults [30- 17 32]. Both hip and wrist devices show reliable and acceptable performance in estimating 18 PA and sleep-wake patterns [33][34][35][36]. 19 The main advantages of using wearable devices for actigraphy is the non-invasiveness, 20 24/7 monitoring of the circadian pattern and PA, and informing sleep habits and 21 parameters in the user natural environment [24,25,28]. We propose an innovative and 22 strategic approach to monitor PA and sleep-wake patterns during pregnancy, 23 establishing a large database comprised of clinical, epidemiological, PA and sleep-wake 24 variables potentially capable of composing a prediction model for maternal 25 complications during pregnancy. The main goal of this study is to identify earlier 26 predictors of pregnancy complications by correlating data generated on PA and sleep 27 patterns through wearable devices (wristband sensors) with maternal and perinatal 28 complications and outcomes. 29  We will conduct a cohort study of 400 pregnant women using wrist sensor bands able 2 to capture information on daily physical activity and sleep patterns (exposure). This 3 cohort study will be implemented in 5 ANC clinics linked to obstetric units in 3 different 4 regions of Brazil that are already part of the Brazilian Network for Studies on 5 Reproductive and Perinatal Health [37], as shown in Table 1. During a period of eight 6 months, the ANC clinics will identify eligible cases for using the wristband sensors. 7 Wearable technology data will be correlated with the occurrence of pregnancy and 8 childbirth complications and outcomes, such as hypertensive disorders, gestational 9 diabetes mellitus, fetal growth restriction and prematurity. 10 Eligible women will be identified up to 21 weeks of gestation and invited to participate. 11 A proper consent form will be applied and the women who agree to participate will 12 receive a wristband sensor to be used starting at 19-21 weeks until childbirth, 13 uninterruptedly.

Study design
14 Study setting and population 15 Brazil is a multi-ethnic mixed-race population of diverse resourced settings [38]. Despite 16 the considered high global overall human development index (HDI 0.727) in 2010, the 17 HDI of Brazilian municipalities ranged from 0,862 to 0,418 [39]. The possibility of 18 considering such mixed population is suitable to explore information regarding maternal 19 patterns of mobility and sleep, maximizing external validity and comparisons to other 20 populations. The following reasons support the study population being focused in low-21 risk nulliparous women: 1) Previous obstetric history can refer to known risk factors for 22 many maternal complications such as preterm birth, preeclampsia, and diabetes [13, 23 40]. Therefore nulliparous women enable unbiased sampling regarding obstetric history. 24 2) Women with previous morbidity such as hypertension, diabetes, nephropathy or 25 others chronic/degenerative diseases are more likely to present abnormalities of sleep- 26 wake rhythm or physical activity patterns during pregnancy. 27 Sampling 28 The five participating centers are regional referral obstetric units responsible for 29 antenatal care assistance mainly for high-risk pregnant women. Participating centers are 30 listed in Table 1. Nevertheless, there are primary health care units strategically linked with these participating centers, enabling the identification and enrollment of women 2 with non-pathological pregnancies. The recruitment strategies include approaching all 3 eligible women in these participating centers and their linked facilities. An informed 4 consent form will be applied for women who agree to participate. 5 Eligible women: Low-risk pregnant subjects 6 There is no international consensus on the criteria for low-risk pregnancies, although 7 there are several known factors associated with maternal and perinatal adverse 8 outcomes. A recent study evaluating complications of "low-risk" pregnancies of US 9 Americans (10 million births from 2011 through 2013) showed that 29% of low-risk 10 women had an unexpected complication requiring no routine obstetric/neonatal care 11 [41]. This shows the difficulty in establishing a "low-risk profile" for maternal/perinatal 12 complications. In order to better identify eligible low-risk pregnant women, we excluded 13 potential known confounders of pre-pregnancy conditions that could be related to 14 adverse maternal or perinatal outcomes as shown in Table 2, so we could assess PA and 15 sleep patterns of a mostly "normal" population. Nonetheless, lifestyle habits and body 16 composition (Body mass index, height, etc.) characteristics, and some non-severe 17 chronic diseases as non-severe anaemia and/or asthma are not among the exclusion 18 criteria in this study but may be part of subgroup analyses (composition of any previous 19 disorder, e.g.). Intra and inter-individual analyses of PA and sleep patterns enable the 20 identification of potential confounders affecting primary outcomes, avoiding potential 21 biases. It means that comparison of PA and sleep pattern parameters collected in 22 different stages of pregnancy from the same participant (intra-individual analysis) and 23 collected at the same stage of pregnancy from different participants (inter-individual 24 analysis) will be carried out. 25 Eligible women will be enrolled between 04-21 weeks. The inclusion and exclusion 26 criteria are presented in Table 2. 27 Data collection methods 28 Essentially, MAES I study is comprised of 4 key set points -3 clinical visits during 29 pregnancy and a postnatal visit. The clinical visits will be held at 1) between 19-21 weeks; 30 2) between 27-29 weeks; 3) between 37-39 weeks. At first, second, third and postnatal and Resilience Scale [43] will be applied during 27-29w visit. Figure 1 shows the set 6 points of MAES-I study. 7 Eligible women will be invited to use a 43mmx40mmx13mm water-resistant wrist device 8 similar to a regular watch (GENEActiv Original -Activinsights®). The device contains 9 accelerometer to estimate PA and sensors to estimate sleep-wake patterns through light 10 and temperature measurements, using a proper software algorithm. 11 At the first set point of MAES-I study (between 19+0 -21+0 weeks of gestation), eligible 12 women who agreed to participate will be instructed that the wrist bracelet device should 13 be worn on the non-dominant arm during day and night (24h/day) uninterruptedly until 14 childbirth (including bathing or aquatic activities). The participant will not have to press 15 any button or take any special care regarding the functioning of the device, which will 16 be configured to register physical activity and sleep-wake data automatically from the 17 moment it is delivered at the antenatal care visit. Also the battery charge will be held by 18 the research assistant before delivering the device to the participant woman. 19 The acquisition of actigraphy data can be performed in different frequencies (from 10Hz 20 to 100Hz). Since the frequency of data acquisition impacts on the battery life of the 21 device (inverse relationship), the measurement frequency will be set up according to 22 the participant´s gestational age (Table 3). This information will be registered in the 23 database accordingly. The data accumulated will be downloaded during participant´s 24 antenatal care visits, according to the maximum return periods showed in Table 3. The 25 maximum return periods were calculated taking into consideration the expected battery 26 life. At each antenatal care visit, the used device will be returned to the research team 27 and a new charged device will be provided to the participant. 28 A leaflet with detailed information and FAQ (Frequently asked questions) on the device 29 will also be provided to the women. They will also have a cell phone number to call 30 whether doubts arise regarding the procedures for using the device, or if any technical 31 or medical concern arises. During each antenatal care visit, the wrist device will be connected to a charge base 2 which can be connected to a computer through an USB connection. All actigraphy data 3 will be extracted to the computer as a raw data ".bin" file. An open source proper 4 software (Geneactiv Software®) will allow to convert this file into ".csv" compressed 5 epoch files for each 30 minutes of registered data, which can be read in Excel® program. 6 Then, the actigraphy data will be uploaded to an online database platform developed by 7 MedSciNet®, where all clinical data of the study will also be registered. 8 The actigraphy software uses several algorithms to translate numerical information 9 obtained though the epoch files into physical activity and sleep-wake patterns, which 10 will compose the independent variables of this study. The database in centralized, 11 secure, internet-based and allows several procedures for prospective and retrospective 12 monitoring, hierarchical access (local user, general manager, etc.). The database will be 13 translated into Portuguese and English, facilitating data collection for Portuguese- 14 speaking team and international monitoring. A correspondent paper form will be 15 available for data collection if necessary (e.g. internet connection failure for instance). 16 Decision to start monitoring PA and sleep patterns between 19-21weeks 17 There are various underlying mechanisms involved in the development of the maternal 18 and perinatal adverse outcomes that will be assessed, as preterm birth, preeclampsia, 19 gestational diabetes, fetal growth restriction and small for gestational age. The pre- 20 clinical phase, stage where there are no clinical signs or symptoms, might be different 21 for each disease and dependent on environmental and individual aspects. The study of 22 adverse maternal and perinatal predictors has been focused in early pregnancy so far 23 (first trimester, <14 weeks of gestation), aiming to maximize the window of opportunity 24 for preventative interventions. However, we hypothesized that the modification of PA 25 and/or sleep pattern due to maternal underlying changes of biological function might 26 not be evident at a very early stage in pregnancy before the beginning of the pre-clinical 27 phase. Our hypothesis is that it possibly occurs shortly before symptoms. 28 Additionally, we took into account that the occurrence of the main maternal 29 complications, as preeclampsia, fetal growth restriction, and preterm birth, are more 30 common in late pregnancy to establish the period between 19-21 weeks as appropriate 31 to start the assessment of PA and sleep patterns. A recent cross-sectional study conducted in 20 referral centres in Brazil, including the five participating centres of this 2 proposal, showed that the occurrence of preterm birth before 28 weeks comprised less 3 than 1% of all births and less than 8% of all preterm births [44]. In addition, the early 4 onset of preeclampsia (before 34 weeks of gestation) complicates less than 0.4% of all 5 pregnancies, according to a large retrospective cohort of more than 450,000 deliveries 6 in USA [45]. Figure 2 outlines the predicted prevalence of preterm birth and 7 preeclampsia in the second trimester, highlighting its clinical presentation, when classic 8 symptoms and signs of a certain disease/complication are presented, through 9 pregnancy in red. Our hypothesis is that PA and sleep patterns might be altered closely 10 to the clinical presentation, still in preclinical phase when there is no symptoms or signs. 11 In brief, as an exploratory study, we indeed needed to make an arbitrary decision 12 regarding interval of monitoring PA and sleep patterns. For that, we had taken into 13 consideration: 1) the main maternal/perinatal complications of interest occur in the 14 second half of pregnancy, more precisely in late pregnancy ( Figure 2); 2) we hypothesize 15 that any potential change on PA or sleep patterns might occur days or weeks before the 16 onset of maternal or perinatal complication. Then, we focused monitoring women 17 during second half of pregnancy. 18 Thus, the start of assessment between 19-21 weeks seems to be very reasonable, 19 providing a wide interval to monitor and predict the main maternal and perinatal 20 adverse outcomes. 21 Actigraph device 22 The actigraph device that will be used to monitor PA and sleep-wake patterns is 23 GENEActiv Original (GENEActiv, Activinsights Ltd, Huntingdon, UK). The device has 24 multiple sensors as microelectromechanical (MEMS) accelerometer, temperature 25 (linear active thermistor) and light (silicon photodiode), providing crude raw data for a 26 variety of applications. 27 Wrist vs waist wear: advantages and performance 28 Wrist wear of actigraph devices provides more comfortable use during wake and sleep 29 periods and highest wear time compared to waist monitors [33,46]. A non-systematic 30 review published in 2011 showed that actigraphy is a useful and reliable tool to assess  [25]. Actigraphy showed excellent 3 concordance with polysomnography in assessing sleep parameters in healthy subjects 4 (sensitivity >90% in estimating total sleep time, for instance). A recent study evaluated 5 the concordance of physical activity estimation of wrist device in free-living settings in 6 forty overweight or obese women [34]. They used both wrist and hip devices, and a 7 small camera that captured participant behaviour for 7 days, enabling the monitoring of 8 physical activity behaviour (gold-standard comparison). The hip and wrist machine 9 learning (ML) classifiers used are different due to the different methods/algorithms to 10 estimate physical activity [34]. The sensitivity and specificity of hip and wrist estimations 11 according to Ellis et al are showed in Table 4 [34]. 12 Two years ago, the same author had published a similar evaluation using 40 adults 13 (women and men), showing that the hip and wrist accelerometers obtained an average 14 accuracy of 92.3% and 87.5%, respectively, in predicting PA types [47]. 15 Staudenmayer et al developed an investigation with 20 participants also using two 16 devices (wrist-and hip-worn), and showed that wrist actigraphy can estimate energy 17 expenditure accurately and relatively precisely [48]. Another study evaluating PA 18 patterns in a free-living environment with wrist devices showed that women in the top 19 40% or bottom 40% of the distribution of daily PA, hip and wrist accelerometers agreed 20 on the classification for about 75% of the women [49]. Additionally, the total activity 21 (counts per day) was moderately correlated (Spearman's r = 0.73) between the wrist 22 and hip worn devices. 23 At the best of our knowledge, there are no systematic reviews or other high-quality 24 evidence-based recommendation supporting a particular method. Although wrist wear 25 of actigraphy is not the more traditional method, it might be the best choice for 26 assessing long periods of PA or sleep patterns, even more considering the similar 27 performance of the waist wear. The current proposal does not intend to diagnose 28 pathologic behaviours or diseases, but to identify different patterns along pregnancy 29 and in different subgroups of women. Therefore, supported by the evidence that wrist 30 wear of actigraphy devices can accurately and more comfortably estimate PA and sleep patterns, mainly for long periods and in the free-living environment, the MAES-I study 2 group addopted wrist wear devices. 3 Main variables 4 The independent variables assessed as potential predictors of maternal complications 5 will be related to sleep-wake cycle and mobility as:  13 The actigraph device collects many pieces of information related to body position and 14 body movements to estimate the described sleep variables. Then, actigraphy software 15 will be used to analyse the data and generate the output variables. 16 "Physical activity" variables 17 Actigraphy technology estimates physical activity through various parameters  per minute between 9499 -∞. 7 -MET rates: Metabolic Equivalents (METs) are commonly used to also express the 8 intensity of physical activities. One MET is the energy cost of resting quietly, often 9 defined in terms of oxygen uptake as 3.5 mL·kg -1 ·min -1. MET rate expresses a 10 person's working metabolic rate relative to their resting metabolic rate. Briefly, 11 the triaxial piezoelectric sensors stressed by acceleration forces can estimate the 12 intensity of movements, converted to the oxygen consumption required to 13 perform such movement. 14 - Step counts/day: estimated steps count per day (estimated by proper algorithms 15 using accelerometer data.) 16 Outcomes 17 The primary outcomes are late pregnancy complications as: 18 -Preeclampsia: Hypertension after 20 weeks of gestation, systolic BP ≥ 140mmHg 19 and/or diastolic BP ≥ 90mmHg (Korotkoff V) on at least 2 occasions 4h apart with: Pulmonary oedema confirmed by chest x-ray [51]. -Provider-Initiated Preterm Birth: defined as childbirth occurring at less than 37 10 weeks, medically indicated due to maternal/fetal compromise or both; 11 -Maternal Hemorrhage: Classified in 1) Antepartum haemorrhage defined as 12 bleeding from the genital tract after 24 weeks of gestation; 2) Primary Postpartum 13 haemorrhage as the loss of 500 ml blood or more from the genital tract within 24 14 hours of the childbirth. 15 Secondary outcomes include childbirth variables and neonatal adverse outcomes as 16 fetal death, caesarean section, small for gestational age (defined as birth weight below 17 percentile 10 for gestational age), Apgar score < 7 at 5 minutes, neonatal severe 18 morbidity (Table 5) and neonatal mortality before discharge. 19 Plans for analyses 20 Sample size estimation 21 This is an exploratory and innovative study focused on a specific population (pregnant 22 women) and therefore there are no previously published parameters available for 23 sample size estimation. Considering a relatively wide range of frequency of 24 complications arising in pregnancy from 3 to 20% (including preeclampsia, fetal growth 25 restriction, gestational diabetes, hemorrhage, preterm birth, etc.), a theoretical large 26 population size (above 1 million pregnant women), an acceptable margin of error of 4%, 27 the involvement of 5 clusters (participating centers) and a 95% level of confidence, 384 28 women would be necessary. Therefore, we are rounding up this estimation for at least 29 400 initially low-risk pregnant women to be enrolled in the study. We estimated the 30 incidence of some main maternal complications considering the following studies: according to mainly to the NICE guidelines [54]. 10 -Fetal growth restriction/small for gestational age: SCOPE international cohort, 11 previously mentioned, had a prevalence of 10.7% of newborns small for gestational 12 age, according to the customized centiles of birthweight (<10%) [55]. 13 Statistical Analysis details 14 According to these studies above, the predicted incidence of these complications seems 15 reasonable and reproducible in our cohort. Then, sample size estimation might assure 16 enough cases of maternal and perinatal complications for the current proposal. 17 The epoch files obtained from Geneactv Software by reading data of sleep variables and 18 physical activity parameters will be translated into numerical results and then averaged 19 in periods of 7 days. By doing this, there will be one value to be used in statistical analysis 20 for each variable per week of use of the wrist device. 21 Firstly, we will identify PA and sleep-wake patterns of women who did not develop 22 adverse maternal or perinatal outcomes. It will permit the recognition of a normal PA 23 and sleep-wake patterns in low-risk population without complication during pregnancy. 24 Using the same population, we will analyze changes in PA and sleep-wake patterns 25 through pregnancy, allowing for gestational age periods. 26 Then, we will compare the PA and sleep-wake patterns of women who developed 27 specific adverse maternal or perinatal outcomes with those who did not. The differences 28 between groups might be identified to be used as potential markers for specific 29 pregnancy complications. After that, we will analyze changes in PA and sleep-wake patterns of women who 2 developed adverse maternal or perinatal outcomes through pregnancy, comparing the 3 patterns and trying to discover which changes and when before the onset it would be 4 related to pregnancy complications. If possible, we will conduct subgroup analysis 5 including subpopulation with potential higher risk for maternal complications 6 (confounder variables), including obesity, smoking, etc. 7 Finally, we will develop a predictive model for screening pregnant women for risk of 8 specific adverse maternal and perinatal outcomes using PA and sleep-wake data 9 estimated with actigraphy technology. 10 The analysis will be performed using the actigraph software that translates the collected 11 information into PA and sleep-wake parameters. Additionally, data regarding time of 12 sleep onset latency, wake after sleep onset and total sleep time as well as sleep 13 efficiency will be compared between participants along the whole pregnancy time using 14 Friedman and Wilcoxon for paired samples. ANOVA and t-test will be used to compare 15 the sleep parameters between the participants for each week of gestational age for 16 repeated measures. The same tests will be applied to analyze quantitative data 17 regarding the median of number of hours per day having different types of physical 18 activity (sedentary, light, moderate, vigorous and very vigorous), MET rates and 19 estimative os steps/day through the entire gestational period examined, and the 20 comparison between the participants for each week of gestational age. Also, we will 21 address sensitivity, specificity and likelihood ratio for altered PA and sleep patterns or 22 for their changes throughout pregnancy. 23 Discontinuation of participants 24 The criteria for discontinuation include: 25 -Withdrawal of consent; 26 -Not regularly using the actigraph device for long periods, less than 50% of all planned 27 time. The information that they are not using the device properly will be recorded if 28 women notice the MAES-I team. Otherwise, the low use of the device will be noticed 29 after data discharge during antenatal care visits. 30 -The loss to follow-up, not allowing the download of actigraphy data. Those women who decide to leave follow-up will be asked by telephone call to return 2 the wrist device and a last visit will be set in order to regain the wrist monitor and direct 3 the woman to a proper antenatal care service to continue their consultations. 4 Data and Sample Quality 5 All entered data will be prospectively and retrospectively monitored by local research 6 assistants and a global monitor. Internal consistency of variables will be constantly 7 performed by the database and error messages are automatically flagged. A local 8 research assistant will be responsible for checking all forms and actigraphy data before 9 locking forms, assuring good quality of data (double-checking entered data and checking 10 for inconsistencies between variables, for instance). Then, the local principal 11 investigator (PI) will be responsible for signing the case, enabling its incorporation to the 12 final database. The University of Campinas will coordinate, implement and monitor the 13 study in the five participating centres. A general manager and a global monitor are also 14 part of the team of the coordinating centre. The local team of each participating centre 15 is comprised of a Local PI and research assistants. 16 Ethics and Dissemination 17 MAES-I study focuses on low-risk nulliparous Brazilian pregnant women. Despite being 18 considered of low risk for maternal and perinatal complications, they are not free of 19 suffering complications. Furthermore, the first and second delays, defined as a delay in 20 deciding to seek care and delay in reaching a health care facility [56], are not uncommon, 21 establishing a barrier between earlier recognition of symptoms and timely interventions 22 capable to successfully treat potentially life-threatening conditions. We believe that 23 women will feel encouraged, empowered and willing to participate in the study that 24 aims to develop a potentially useful prenatal care tool to identify the risk for maternal 25 and perinatal morbidity and mortality. Following national ethical regulations, the 26 participants will not receive any financial compensation. 27 Women who agree to participate in the study will not have any disadvantages or 28 compromise of their prenatal care. On the contrary, they will receive a telephone 29 number to contact the clinical researchers at any time (24/7 service), which enables a 30 closer contact with researchers and providers of care, since the MAES-I team are  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   19   1 committed to contacting providers of care if any potential complication is noticed by 2 participants. 3 The participating women will not be responsible if loss, theft or damage to the wrist 4 device occurs. They will be asked only to use the device just as a regular wrist-watch 5 would be worn and no self-damage is expected in those who use it. 6 Participating women will not be able to identify any PA or sleep parameters at any stage 7 of the study. The download of the data is only possible through the own licensed 8 software of the device. Actigraphy devices provided for participating women have a 9 unique code which will be recorded in the database together with the interval of use for 10 each women. Actigraphy data will be labelled using participant ID, device number, 11 gestational age when starting using each device and return date of each device. The use 12 of such codes, ID`s and numbers will ensure confidential identify for all participating 13 women. The identity of all women will be kept confidential.
14 MAES-I study has been reviewed and approved by the National Committee for Ethics in 15 Research of Brazil (CONEP) and by the Institutional Review Board (IRB) of the 16 coordinating centre (Letter of approval 1.834.116 issued on 24 th November 2016) and 17 of all other Brazilian participating centres. All women who will be enrolled in the MAES-I 18 cohort will sign an informed consent form. 19 The ethical principles stated in the Brazilian National Heath Council (Resolution CNS 20 466/12) will be respected in every stage of this study. The anonymity of the source of 21 information will be guaranteed and the care for the women will be provided 22 independently of her agreement to participate in the study. The study also complies with 23 the Declaration of Helsinki amended in Hong Kong in 1989.The methodological and 24 ethical aspects of MAES-I study protocol were developed following STROBE guidelines 25 [57]. 26 Patient and Public Involvement 27 Patients and public were not involved in this study for the development of the research 28 question and outcome measures. However, the choice for a wrist device was based on 29 the preference of users as reported. Participants of the study will have access to the 30 results by its webpage that will be open access.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Detailed information of the study is provided in the Brazilian Cohort website 2 (www.medscinet.com/samba) and findings will be publicized in scientific literature and 3 Institutional webpages. We intend to disseminate our findings in scientific peer-4 reviewed journal, general free access website, specialists' conferences, and to our 5 funding agencies. 6 7

8
The actigraphy is an innovative, non-invasive, non-operator dependent, wearable 9 technology, which enables the estimative under real life conditions of diverse variables 10 related to mobility, physical activity, sleep-wake, and circadian cycle patterns. Actigraph 11 devices show high sensitivity in sleep-wake parameters detection and are currently 12 highly recommended by the American Sleep Disorder Association for diagnosis and 13 therapy response of circadian rhythm disorders [27,28,58]. Although some studies 14 show that 7 to 14 days using the actigraph device provides reliable estimates of physical 15 activity behavior in older adult, it is not absolutely clear how many days is needed to 16 estimate habitual PA by using the wrist/waist device during pregnancy. In general, it 17 seems to depend mainly on the type of actigraph device, position of wear and target 18 population [30,33]. Nevertheless, MAES-I study will provide sufficient data to assess 19 different patterns along pregnancy. 20 The use of wearable physical activity monitors has grown enormously due to the interest 21 about the relationship between the pathophysiology of diseases and physical activity 22 and sleep patterns. A recent study on the use of physical activity monitors in human 23 physiology research unravels the current and potential uses of actigraph device as in 24 strategies to promote healthier behaviour or to predict outcomes [59]. The authors 25 conclude that physical activity monitors, as others new 21 st century technologies, have 26 already transformed physiology research, revolutionizing the way we assess patients and 27 opening new areas of interest. In addition, the use of objectives measures to evaluate 28 habitual sleep duration and outcomes in pregnancy is critical, taking into account recent 29 investigations reporting little agreement between objective and subjective assessments 30 of sleep time [60].  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   21   1 Alterations in sleep patterns, as less deep sleep and more nocturnal awakenings, can be 2 observed in pregnancy as early as 10-12 weeks gestation [61]. Sleep disturbances during 3 pregnancy have been associated with preterm delivery, gestational hypertensive 4 disorders, glucose intolerance and increased risk of caesarean delivery [19]. Shorter 5 night time sleep was also associated with hyperglycemia [62]. Persistent sleep deficiency 6 is correlated with depressive symptoms and stress perception by pregnant women [61]. 7 These studies lay correlation between PA patterns and sleep disturbances determining 8 complications, in a well-established relationship of cause and consequence, although 9 sometimes it could not be adequately determined due to the study design [17]. 10 In a distinct way, our analysis intends to figure out if the maternal complication could be 11 identified by physical activity and/or sleep patterns modifications, even during its pre- 12 clinical period, previous the appearance of clinical signs. Considering the existing 13 evidence, we speculate that the PA and/or sleep patterns change days or weeks before 14 the clinical presentation of the complication. In general, the signs and symptoms of some 15 maternal outcomes are part of the gold-standard criteria for diagnosis (high blood 16 pressure, proteinuria and/or edema in the case of preeclampsia; premature contractions 17 and cervical ripening/dilation in preterm birth; abnormal placental blood flow and 18 insufficient fetal growth in Intrauterine Growth Restriction). We acknowledge the fact 19 that there are potential confounders and limitations in predicting maternal and perinatal 20 complications using PA and sleep patterns estimated by actigraph devices. We expect 21 that our studied population will have different subgroups of women with different risks 22 and associated factors playing a role on maternal complication. It includes obesity, 23 smoking, extremes of age, for instance. None of those factors was considered exclusion 24 criteria and, if possible, we intend to assess subgroup analysis for those maternal 25 subgroups at they might present different PA and sleep patterns. Nonetheless, we 26 decided to perform a pragmatic approach, not excluding such common factor from our 27 sample. 28 The use of actigraph device during prenatal visits has a potential to become a new tool 29 to monitor pregnant women, improving maternal health care, identifying altered PA 30 and/or sleep patterns, measured objectively through actigraphy, before the occurrence 31 of those signs and symptoms. Therefore, the focus would be offering new technology to  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   22   1 monitor the development of a potential maternal complication. Other positive points of 2 our study are the period of data collection (from 19 weeks till delivery) and the low-risk 3 profile of the cohort. Through which, it would be possible to describe a PA and sleep 4 patterns in a low-risk pregnant population and better interpret actigraphy data among 5 pregnant women. The current clinical and biological predictors for the main maternal 6 complications as preeclampsia, preterm birth, maternal haemorrhage, and gestational 7 diabetes still lack for effective sensitivity and specificity. 8 If this is confirmed to be true, an important step will be achieved for a possible 9 introduction of screening non-invasive procedures during prenatal care with the 10 purpose of identifying women at higher risk of developing those conditions. Therefore, 11 they could receive specific orientation on prevention and earlier detection of the onset 12 of condition for taking immediate action to look for professional health care and 13 receiving appropriate interventions, avoiding delays that are the most striking factor for 14 the low quality of care the women usually receive in low and middle-income settings, 15 contributing to the still high burden of maternal morbidity and mortality. If we were 16 successful in identifying such "specific patterns of physical activity and sleep" as 17 predictors for pregnancy complications, further validation studies will necessarily be 18 recommended for assessing its effectiveness for the whole management of such 19 conditions. Additionally, MAES-I will enable further specific studies among high risk 20 population and also will help to identify the best gestational age for monitoring, giving 21 the means to target a specific gestational age interval.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Competing interests
The authors declare that they have no competing interests.

Strengths and limitations of this study
2  This multicentre cohort will collect comprehensive data on major maternal and 3 perinatal complications such as pre-eclampsia, small for gestational age/fetal 4 growth restriction, preterm birth and gestational diabetes mellitus. 5  Physical activity and sleep patterns will be estimated by an innovative wearable 6 device used in the natural environment of the study subject. 7  Physical activity and sleep patterns will be estimated from the beginning of the 8 second half of pregnancy until delivery, covering a wide interval during 9 pregnancy, allowing for the study of changes in PA and sleep patterns 10 throughout pregnancy. 11  One possible limitation is the first half of pregnancy at a time when this 12 information was not covered. 13 14 Reducing the global maternal mortality ratio to less than 70 per 100,000 live births by 3 2030 is one of the targets of the new United Nations Sustainable Development Goals 4 [1]. Multiple challenges need to be tackled to achieve this target, but the 2016-2030 5 health and development agenda goes well beyond mortality reduction. The aim of the 6 Global Strategy for Women's, Children's and Adolescent's Health is to ensure that every 7 newborn, woman, and child not only survives but thrives. This will only be possible if a 8 transformative agenda centered on innovation is put into action [2]. 9 One of the major challenges lies in optimizing earlier predictors and identifiers of 10 maternal and perinatal complications. Delays in diagnosing and managing maternal 11 complications have been associated with poor outcomes [3]. Decreased self-perception 12 of clinical signs related to maternal complications, difficulties in accessing the health 13 system and poor quality of care may contribute to late identification of complications 14 and a worse prognosis. The development of a non-invasive Antenatal Care (ANC) tool 15 for identifying maternal sub-clinical signs during pregnancy may provide a window of 16 opportunity for an earlier identification of abnormal patterns of physiological 17 parameters related to pregnancy complications. Earlier identification occurs when 18 recognition is made before clinical presentation by standard criteria based on clinical 19 signs, symptoms, and supplementary tests. Shortening the time between the onset of a 20 complication and the initiation of appropriate management enables secondary 21 prevention and reduction of maternal morbidity and mortality [3][4][5][6][7]. 22 Pervasive computing (i.e. the trend towards embedding microprocessors in everyday- 23 life objects so they can generate data) and wearable technology (i.e. clothing and 24 accessories incorporating computer and advanced electronic technologies such as 25 sensor wristbands and/or waistbands) are ubiquitous and can generate a new dataset 26 that requires correlation with pregnancy outcomes. Preterm birth and preeclampsia are 27 two important pregnancy complications that have a relatively long subclinical phase 28 before the appearance of signs or symptoms [8,9]. It is plausible that during subclinical 29 phases of certain conditions the pattern of physical activity (PA) or sleep-wake rhythm 30 is affected in some way and wearable devices could capture these changes. Although 31 some studies have shown that PA patterns (actigraphy parameters) may be related to systemic inflammation and diseases in the general population [10,11], there is a paucity 2 of published literature that correlates wearable technology data with maternal 3 complications. 4 The human circadian rhythm is regulated by endogenous physiological mechanisms and 5 environmental stimuli [12]. Solid evidence indicates that modification in circadian 6 rhythm or sleep and PA patterns are underlying conditions related to inflammatory, 7 degenerative and/or metabolic chronic diseases such as diabetes, hypertension, and 8 cancer [13]. Circadian misalignment is defined as inappropriately timed sleep and wake, 9 misplaced feeding periods and modification in physical activity behaviour. 10 Determining a cause or effect relationship between these modifications and the 11 development of pathological conditions is a complex task. It seems that changes in 12 appetite-stimulating hormones, glucose metabolism, inflammatory markers, and mood 13 are some of the related pathways [13][14][15]. Leproult et al. evaluated the effect of 14 circadian misalignment on metabolic and inflammation markers in cardiovascular 15 disease [15]. Insulin action and release, and also levels of some inflammatory markers 16 that are predictors of cardiovascular disease, were abnormal in individuals with 17 circadian misalignment. The mechanisms involved in the association between changes 18 in PA pattern and pathologic conditions seem to have multiple etiologies. Sani et al. 19 assessed circadian rhythms of more than 2,300 African adult descendants. In addition 20 to the evaluation of physical activity itself, the aim of those authors was to identify 21 chronobiological patterns of adults from different socioeconomic settings. The study 22 described that chronobiological behaviour can vary depending on individual BMI, 23 socioeconomic background, work type and time of sunlight exposure. Many other 24 factors, such as pathologic conditions, may be potentially involved in a modification in 25 chronobiological behaviour. Some metabolic, cognitive, cardiovascular and other 26 chronic degenerative diseases have been associated with particular patterns of PA and 27 sleep [10,11,[16][17][18]. A previous observational study assessed various sleep parameters 28 during pregnancy, e.g. sleep onset latency (SOL), wake after sleep onset (WASO) and 29 total nocturnal sleep time (TST). Difficulty in initiating sleep in early pregnancy was 30 associated with higher body mass index, greater weight gain and higher blood pressure 31 during pregnancy [17]. Palagini  sleep loss and adverse pregnancy outcomes, discussing common mechanisms of stress 2 system activation [19]. Low-quality evidence suggests an association between sleep loss 3 and prenatal depression, gestational diabetes, preeclampsia, abnormal length of labour, 4 caesarean delivery, abnormal fetal growth, and preterm birth. Those results corroborate 5 with other findings regarding pregnancy and sleep disorders [20][21][22][23]. 6 Assessment of PA and sleep patterns can be performed by wearing small wrist (or waist) 7 devices similar to a regular watch (actigraphy technology). More recently, substantial 8 advance has been made in types of sensors, batteries, materials and output data, 9 leading to lower cost, comfort, discretion and performance of the devices [24]. 10 Nowadays, portable, lightweight devices have a large capacity to store data, including 11 software with automatic scoring algorithm packages for the detection of wakefulness, 12 sleep periods and PA [24,25]. Actigraphy estimation of PA and sleep patterns is 13 validated as a proxy for chronobiological behaviour [26][27][28][29] and the use of an actigraphy 14 device for 7 to 14 days provides reliable estimates of PA behaviour in older adults [30-15 32]. The performance of both hip and wrist devices has been shown to be reliable and 16 acceptable for estimating PA and sleep-wake patterns [33][34][35][36]. 17 The main advantages of using wearable devices for actigraphy are non-invasiveness, 18 24/7 monitoring of PA and circadian patterns, and information about sleep habits and 19 parameters in the natural environment of the subject [24,25,28]. We propose an 20 innovative and strategic approach to monitor PA and sleep-wake patterns during 21 pregnancy, establishing a large database comprised of clinical, epidemiological, PA and 22 sleep-wake variables that are potentially capable of composing a prediction model for 23 maternal complications during pregnancy. The main goal of this study is to identify 24 earlier predictors of pregnancy complications by establishing a correlation between data 25 on PA and sleep patterns using wearable devices (sensor wristbands) and maternal and 26 perinatal complications and outcomes.   3 We will conduct a cohort study of 400 pregnant women using sensor wristbands 4 capable of capturing information on daily physical activity and sleep patterns 5 (exposure). This cohort study will be implemented in 5 ANC clinics linked to obstetric 6 units in 3 different regions of Brazil that are already part of the Brazilian Network for 7 Studies on Reproductive and Perinatal Health [37], as shown in Table 1. During an 8-8 month period, the ANC clinics will identify cases that are eligible to use the sensor 9 wristband. Wearable technology data will be correlated with the occurrence of 10 pregnancy and childbirth complications and outcomes, such as hypertensive 11 disorders, gestational diabetes mellitus, fetal growth restriction and prematurity. 12 Eligible women will be identified up to 21 weeks of gestation and invited to participate 13 in the study. A proper consent form will be applied and women who agree to participate 14 will receive a sensor wristband to wear continuously from 19-21 weeks until childbirth. 15 16 Brazil is a multi-ethnic mixed-race population of diverse resourced settings [38]. Despite 17 the high global overall human development index (HDI 0.727) in 2010, the HDI of 18 Brazilian municipalities ranged from 0,862 to 0,418 [39]. A mixed population is suitable 19 for exploring information on patterns of maternal mobility and sleep, maximizing 20 external validity and comparisons to other populations. The following reasons support 21 a study population of low-risk nulliparous women: 1) Previous obstetric history can refer 22 to known risk factors for many maternal complications such as preterm birth, 23 preeclampsia, and diabetes [13,40]. Therefore, nulliparous women permit unbiased 24 sampling regarding obstetric history. 2) Women with previous morbidities such as 25 hypertension, diabetes, nephropathy or other chronic/degenerative diseases are more 26 likely to present abnormalities in sleep-wake rhythms or physical activity patterns during 27 pregnancy. 28 29 The five participating centres are regional referral obstetric units responsible for 3 antenatal care of mainly high-risk pregnant women. Participating centres are listed in 4 Table 1. Nevertheless, there are primary health care units strategically linked to these 5 participating centres, enabling the identification and enrolment of women with non- 6 pathological pregnancies. Recruitment strategies include approaching all eligible 7 women in these participating centres and their linked facilities. An informed consent 8 form will be applied for women who agree to participate. 9 Eligible women: Low-risk pregnant subjects 10 There is a lack of international consensus on criteria for low-risk pregnancies, although 11 several factors are known to be associated with maternal and perinatal adverse 12 outcomes. A recent study evaluating complications of "low-risk" pregnancies of US 13 Americans (10 million births from 2011 to 2013) indicated that 29% of low-risk women 14 experienced an unexpected complication that required no routine obstetric/neonatal 15 care [41]. This illustrates the difficulty in establishing a "low-risk profile" for 16 maternal/perinatal complications. To make a better identification of eligible low-risk 17 pregnant women, we excluded known potential confounders of pre-pregnancy 18 conditions that could be related to adverse maternal or perinatal outcomes as shown in 19 Table 2, so we could assess PA and sleep patterns of a mostly "normal" population. 20 Nonetheless, features such as lifestyle habits and body composition (body mass index, 21 height), and some non-severe chronic diseases including non-severe anaemia and/or 22 asthma are not exclusion criteria in this study. However, these features and conditions 23 may be a part of subgroup analyses (composition of any previous disorder, e.g.). Intra 24 and inter-individual analyses of PA and sleep patterns can avoid possible bias by 25 identifying potential confounders that may affect primary outcomes. A comparative 26 analysis will be conducted, in which parameters of PA and sleep patterns will be 27 collected in different stages of pregnancy from the same participant (intra-individual 28 analysis) and compared to data collected at the same stage of pregnancy from different 29 participants (inter-individual analysis). 30 Eligible women are to be enrolled at 04-21 weeks of gestation. Inclusion and exclusion 31 criteria are shown in Table 2.  4 Essentially, MAES I study is comprised of 4 key set points -3 clinical visits during 5 pregnancy and a postnatal visit. Clinical visits will be held at 1) 19-21 weeks; 2) 27-29 6 weeks; and 3) 37-39 weeks. On the first, second, third and postnatal visits, additional 7 information on maternal history, details of pregnancy complications, maternal 8 biophysical data (weight, height, skinfolds) and adverse pregnancy outcomes will be 9 collected following a specific Standard Operating Procedure (SOP) specially developed 10 for MAES-I study. Furthermore, the Perceived Stress Scale [42] and Resilience Scale [43] 11 will be applied during the 27-29 weeks visit. Figure 1 shows the set points of MAES-I 12 study. 13 Eligible women will be invited to use a 43mmx40mmx13mm water-resistant wrist device 14 similar to a regular watch (GENEActiv Original -Activinsights®). The device contains an 15 accelerometer for PA calculation and sensors for estimation of sleep-wake patterns by 16 light and temperature measurements, using a proper software algorithm. 17 At the first set point of MAES-I study (between 19+0 -21+0 weeks of gestation), eligible 18 women who agreed to participate will be instructed to wear the wrist bracelet device 19 on the non-dominant wrist night and day (24h/day), uninterruptedly until childbirth 20 (including bathing or recreational water activities). Participants will not need to press 21 any buttons and functioning of the device requires no special care. The device will be 22 configured to register physical activity and sleep-wake data automatically from the 23 moment it is delivered to the participant during antenatal care visit. In addition, the 24 battery charge will be held by the research assistant before delivering the device to the 25 study participant. 26 The acquisition of actigraphy data can be performed in different frequencies (from 10Hz 27 to 100Hz). Since the frequency of data acquisition has an impact on battery life of the 28 device (inverse relationship), measurement frequency will be set according to 29 gestational age of the participant (Table 3). This information will be registered in the 30 database accordingly. Cumulative data will be downloaded during antenatal care visits,  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 Table 3. Calculation of maximum return 2 periods will be based on expected battery life. At each antenatal care visit, the used 3 device will be returned to the research team and a new charged device will be provided 4 to the participant. 5 A leaflet with detailed information and FAQ (Frequently asked questions) about the 6 device will also be provided. Women will also have a cell phone number to call in case 7 of any doubts regarding use of the device, or if any technical or medical concern arises. 8 During each antenatal care visit, the wrist device will be connected to a charge base 9 which can be connected to a computer through an USB connection. All actigraphy data 10 will be extracted to the computer as raw data ".bin" file. A proper open source software 11 (Geneactiv Software®) will allow the conversion of this file into ".csv" compressed epoch 12 files for each 30 minutes of registered data, which can be read in Excel® program. The 13 actigraphy data will then be uploaded to an online database platform developed by 14 MedSciNet®, where all clinical study data will also be registered. 15 The actigraphy software uses several algorithms to translate numerical information 16 obtained from epoch files into physical activity and sleep-wake patterns, which will 17 compose the independent variables of this study. This is a centralized, secure, internet- 18 based database that allows several procedures for prospective and retrospective 19 monitoring, hierarchical access (local user, general manager, etc.). The database will be 20 translated into Portuguese and English, facilitating data collection for Portuguese- 21 speaking teams and international monitoring. A correspondent paper form will be 22 available for data collection if necessary (e.g. internet connection failure for instance). 23 Decision to start monitoring PA and sleep patterns between 19-21weeks 24 There are various underlying mechanisms involved in the development of maternal and 25 perinatal adverse outcomes that will be assessed, such as preterm birth, preeclampsia, 26 gestational diabetes, fetal growth restriction and small for gestational age. Each disease 27 may have a different pre-clinical phase, depending on environmental and individual 28 aspects. In this phase, there are no clinical signs or symptoms. So far, the study of 29 adverse maternal and perinatal predictors has been focused on early pregnancy (first 30 trimester, <14 weeks of gestation) to maximize the window of opportunity for the  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   11   1 performance of preventive interventions. However, we hypothesized that modifications 2 in PA and/or sleep pattern due to underlying changes in maternal biological function 3 might not be evident at a very early stage in pregnancy before the beginning of the pre- 4 clinical phase. Our hypothesis is that changes might occur shortly before the 5 manifestation of symptoms. 6 Furthermore, we took into account that major maternal complications, including 7 preeclampsia, fetal growth restriction, and preterm birth, occur more commonly in late 8 pregnancy and established the period between 19-21 weeks as an appropriate time to 9 start assessment of PA and sleep patterns. A recent cross-sectional study conducted in 10 20 referral centres in Brazil, including the five participating centres of this proposal, 11 showed that the occurrence of preterm birth before 28 weeks comprised less than 1% 12 of all births and less than 8% of all preterm births [44]. In addition, the early onset of 13 preeclampsia (before 34 weeks of gestation) complicates less than 0.4% of all 14 pregnancies, according to a large retrospective cohort of more than 450,000 deliveries 15 in the USA [45]. Figure 2 outlines the predicted prevalence of preterm birth and 16 preeclampsia in the second trimester. Clinical presentation, when classic symptoms and 17 signs of a certain disease/complication occur, is highlighted by pregnancy in red. Our 18 hypothesis is that alterations in PA and sleep patterns may occur closer to clinical 19 presentation, still in the preclinical phase when there are no symptoms or signs. 20 Briefly, an exploratory study required an arbitrary decision about the interval for 21 monitoring PA and sleep patterns. To that end, we considered that: 1) the main 22 maternal/perinatal complications of interest occur in the second half of pregnancy, 23 more precisely in late pregnancy ( Figure 2); 2) any potential change in PA or sleep 24 patterns occurred hypothetically days or weeks before the onset of maternal or 25 perinatal complications. Then, we focused on monitoring women during the second half 26 of pregnancy. 27 Thus, starting assessment at 19-21 weeks seems to be quite reasonable, providing a 28 wide interval to monitor and predict major maternal and perinatal adverse outcomes. 29 30 Actigraphy device  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  The actigraphy device that will be used for monitoring PA and sleep-wake patterns is the 2 GENEActiv Original (GENEActiv, Activinsights Ltd, Huntingdon, UK). The device has 3 multiple sensors including a microelectromechanical (MEMS) accelerometer, 4 temperature (linear active thermistor) and light (silicon photodiode) sensors, providing 5 crude raw data for a variety of applications. 6 Wrist vs waist wear: advantages and performance 7 Wrist-worn actigraphy devices are more comfortable to use during wake and sleep 8 periods and provide the highest wear time compared to waist-worn monitors [33,46]. 9 A non-systematic review published in 2011 showed that actigraphy is a useful and 10 reliable tool to assess sleep patterns and circadian rhythm disorders, although there are 11 some limitations in the diagnosis of sleep disorders or measurement of sleep stages [25]. 12 Actigraphy had a very good concordance with polysomnography for assessment of sleep 13 parameters in healthy subjects (i.e., sensitivity >90% in estimating total sleep time). A 14 recent study evaluated the concordance of physical activity estimation by wrist device 15 in free-living settings in forty overweight or obese women [34]. Those women used both 16 wrist and hip devices, and a small camera that captured participant behaviour for 7 days, 17 monitoring physical activity behaviour (gold-standard comparison). There was a 18 difference in hip and wrist machine learning (ML) classifiers, resulting from different 19 methods/algorithms used to measure physical activity [34]. The sensitivity and 20 specificity of hip and wrist estimations according to Ellis et al are shown in Table 4 [34]. 21 Two years previously, the same author published a similar evaluation of 40 adult women 22 and men, showing that hip and wrist accelerometers predicted types of PA with an 23 average accuracy of 92.3% and 87.5% respectively [47]. 24 Staudenmayer et al investigated 20 participants who also wore two devices (wrist and 25 hip), and concluded that wrist actigraphy can estimate energy expenditure in an 26 accurate and relatively precise manner [48]. Another study evaluated PA patterns in 27 women at the top 40% or bottom 40% of the distribution of daily PA who wore wrist 28 devices in a free-living environment. There was agreement in classification between hip 29 and wrist accelerometers in about 75% of those women [49]. Additionally, total activity 30 (counts per day) was moderately correlated (Spearman's r = 0.73) with wrist-worn and 31 hip-worn devices.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   13   1 To the best of our knowledge, there are no systematic reviews or other high-quality 2 evidence-based recommendations that support a particular method. Although a wrist-3 worn actigraphy device is not the most traditional method, it might be the best choice 4 for assessment of prolonged periods of PA or sleep patterns, considering that it 5 performs similarly to a waist-worn device. The current proposal has no intention of 6 diagnosing pathological behaviours or diseases, but it plans to identify different patterns 7 throughout pregnancy and in different subgroups of women. Evidence suggests that 8 wrist-worn actigraphy devices can accurately and more comfortably estimate PA and 9 sleep patterns, mainly during prolonged periods and in free-living environments. 10 Therefore, the MAES-I study group adopted a wrist-worn device. 11 Main variables 12 Independent variables assessed as potential predictors of maternal complications will 13 be related to the sleep-wake cycle and mobility as:  18 -Wake after sleep onset (WASO): defined as total amount of time awake after sleep. 19 -Sleep Efficiency (SE): the ratio between total sleep time and time in bed. 20 The actigraphy device collects many pieces of information related to body position and 21 body movements to estimate the described sleep variables. The actigraphy software will 22 then be used to analyse data and generate output variables. 23 "Physical activity" variables 24 Actigraphy technology estimates physical activity through various parameters 25 collected by the actigraphy device. Briefly, according to Freedson et al, the triaxial 26 sensors stressed by acceleration forces can estimate movement intensity. The 27 acceleration signal is converted to a digital signal and summed over a user-specified 28 time interval (epoch). At the end of each epoch the activity count is stored. Then, 29 according to count per minute (CPM) cut points, PA intensity can be categorized [50].  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   14   1 The software translates information into quantitative variables using appropriate 2 algorithms as follows: 3 -Sedentary (hours/day): the number of hours per day when the count per minute 4 ranges from 0-99. 5 -Light activity (hours/day): the number of hours per day when the count per 6 minute ranges from 100 -1951. 7 -Moderate activity (minutes/day): the number of hours per day when the count 8 per minute ranges from 1952 -5724. 9 -Vigorous activity (minutes/day): the number of hours per day when the count per 10 minute ranges from 5725 -9498. 11 -Very vigorous activity (minutes/day): the number of hours per day when the 12 count per minute is 9499 -∞. 13 -MET rates: Metabolic Equivalents (METs) are also commonly used to express the 14 intensity of physical activity. One MET is the energy cost of resting quietly, often 15 defined by oxygen uptake as 3.5 mL·kg -1 ·min -1. MET rate expresses the working 16 metabolic rate of subjects in comparison to their resting metabolic rate. Briefly, 17 the triaxial piezoelectric sensors stressed by acceleration forces can estimate 18 movement intensity, converted to oxygen consumption required to perform such 19 a movement. 20 -Step counts/day: estimated step counts per day (estimated by proper algorithms 21 using accelerometer data.) 22 Outcomes 23 Primary outcomes are late pregnancy complications such as: 24 -Preeclampsia: Hypertension after 20 weeks of gestation, systolic BP ≥ 140mmHg 25 and/or diastolic BP ≥ 90mmHg (Korotkoff V) on at least 2 occasions 4h apart with: 26 1) Proteinuria 300 mg/24h or spot urine protein: creatinine ratio 30 mg/mmol 27 creatinine or urine dipstick protein ≥ (+) OR, in the absence of proteinuria, 28 hypertension and 2) any multi-system complication that are: Haematological 29 abnormalities; thrombocytopenia (platelets < 100 x 10 9 /L); Disseminated  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59 13 -Spontaneous Preterm Birth: spontaneous onset of preterm labour or premature 14 rupture of membranes leading to preterm birth, childbirth before 37 weeks of 15 gestation. 16 -Provider-Initiated Preterm Birth: defined as childbirth occurring at less than 37 17 weeks, medically indicated due to maternal/fetal compromise or both; 18 -Maternal Hemorrhage: Classified as 1) Antepartum haemorrhage defined as 19 bleeding from the genital tract after 24 weeks of gestation; 2) Primary Postpartum 20 haemorrhage defined as the loss of at least 500 ml blood from the genital tract 21 within 24 hours of childbirth. 22 Secondary outcomes include childbirth variables and neonatal adverse outcomes such 23 as fetal death, caesarean section, small for gestational age (defined as birth weight 24 below percentile 10 for gestational age), Apgar score < 7 at 5 minutes, neonatal severe 25 morbidity (Table 5) and neonatal mortality before discharge. 26 Plans for analyses 27 Sample size estimation 28 This is an exploratory and innovative study focused on a specific population (pregnant 29 women) and therefore there are no previously published parameters available for 30 sample size estimation. Considering that the rate of pregnancy-related complications is 31 3 to 20% (including preeclampsia, fetal growth restriction, gestational diabetes,  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   16   1 hemorrhage, preterm birth, etc.), assuming a large population (above 1 million pregnant 2 women), an acceptable margin of error of 4%, involvement of 5 clusters (participating 3 centres) and a 95% level of confidence, the study would require 384 women. Therefore, 4 we rounded up this estimation to 400 initially low-risk pregnant women for enrolment 5 in the study. We estimated the incidence of some main maternal complications 6 considering the following studies: 7 -Pre-eclampsia: An international prospective cohort study with nulliparous women 8 termed SCOPE used similar criteria for low-risk profile, with a 5% of incidence of pre-9 eclampsia [53]. 10 -Preterm Birth: A recent cross-sectional study conducted in 20 referral obstetric 11 centres in Brazil, including the five participating centres, showed that preterm birth 12 was prevalent in 12.3% of all births [44]. 13 -Gestational Diabetes: In the previously mentioned SCOPE international cohort, the 14 prevalence of gestational diabetes was 8.9% in screened low-risk nulliparous 15 women, according to the NICE guidelines [54]. 16 -Fetal growth restriction/small for gestational age: the previously mentioned SCOPE 17 international cohort had a prevalence of 10.7% of small for gestational age 18 newborns, according to customized centiles of birthweight (<10%) [55]. 19 20 According to the studies above, the predicted incidence of complications seems 21 reasonable and reproducible in our cohort. Therefore, sample size estimation may 22 ensure a sufficient number of cases of maternal and perinatal complications for the 23 current proposal. 24 The epoch files obtained from Geneactv Software by reading data on sleep variables and 25 physical activity parameters will be translated into numerical results and then averaged 26 in 7-day periods. Therefore, only one value will be employed in statistical analysis for 27 each variable per week of use of the wrist-worn device. 28 First, we will identify PA and sleep-wake patterns of women who did not develop 29 adverse maternal or perinatal outcomes. This will permit the recognition of normal PA 30 and sleep-wake patterns in a low-risk population without complications during  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   17   1 pregnancy. We will use the same population to analyse changes in PA and sleep-wake 2 patterns throughout pregnancy, allowing for gestational age periods. 3 Subsequently, we will compare PA and sleep-wake patterns of women who developed 4 specific adverse maternal or perinatal outcomes with those who did not have any 5 complications. Differences between groups may be identified and used as potential 6 markers for specific pregnancy complications. 7 Afterwards, we will analyse changes in PA and sleep-wake patterns of women who 8 developed adverse maternal or perinatal outcomes throughout pregnancy, comparing 9 patterns in an attempt to discover which changes occurred before the onset of 10 symptoms that could be related to pregnancy complications. If possible, we will conduct 11 a subgroup analysis including a subpopulation with a potentially higher risk for maternal 12 complications (confounder variables), including obesity, smoking, etc. 13 Finally, we will develop a predictive model for screening pregnant women at risk of 14 specific adverse maternal and perinatal outcomes using PA and sleep-wake data 15 estimated by actigraphy technology. 16 Analysis will be performed using the actigraphy software that translates collected 17 information into PA and sleep-wake parameters. In addition, sleep onset latency, wake 18 after sleep onset and total sleep time as well as sleep efficiency will be compared 19 between participants throughout pregnancy using the Friedman and Wilcoxon tests for 20 paired samples. The ANOVA and t-test will be used to compare sleep parameters 21 between participants per week of gestational age for repeated measures. The same 22 tests will be applied to analyse quantitative data on the median number of hours per 23 day that different types of physical activity (sedentary, light, moderate, vigorous and 24 very vigorous) are performed, MET rates and estimate of steps/day through the entire 25 gestational period examined, and the comparison between participants per week of 26 gestational age. Also, we will address the sensitivity, specificity and likelihood ratio for 27 altered PA and sleep patterns or for their changes throughout pregnancy. 28 29 Criteria for discontinuation include: 30 -Withdrawal of consent;  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   18   1 -Irregular use of the actigraphy device for prolonged periods, less than 50% of the 2 whole planned time. Information of improper use of the device will be recorded if 3 women notify the MAES-I team. Otherwise, the low level of use of the device will be 4 observed after data discharge during antenatal care visits. 5 -Loss to follow-up, preventing us from downloading actigraphy data. 6 Women who decide to withdraw from follow-up care will be called by telephone and 7 asked to return the wrist device. The last visit will be scheduled to regain the wrist 8 monitor and direct the woman to a proper antenatal care service to continue medical 9 consultations. 10 11 All entered data will be prospectively and retrospectively monitored by local research 12 assistants and a global monitor. Internal consistency of variables will be constantly 13 performed by database and error messages are automatically flagged. A local research 14 assistant will be responsible for checking all forms and actigraphy data before locking 15 forms, assuring the good quality of data (i.e. double-checking entered data and checking 16 for inconsistencies between variables). The local principal investigator (PI) will be in 17 charge of signing the case, which will then be incorporated into the final database. The 18 University of Campinas will coordinate, implement and monitor the study in the five 19 participating centres. A general manager and a global monitor are also part of the 20 coordinating team. The local team of each participating centre is comprised of a Local PI 21 and research assistants. 22 Ethics and Dissemination 23 MAES-I study focuses on low-risk nulliparous Brazilian pregnant women. Although 24 classified as low risk for maternal and perinatal complications, these women are not free 25 from suffering complications. Furthermore, first and second delays, defined as a delay 26 in deciding to seek care and delay in reaching a health care facility [56], are not 27 uncommon. A barrier is created between earlier recognition of symptoms and timely 28 intervention for the successful treatment of potentially life-threatening conditions. We 29 believe that women will feel encouraged, empowered and willing to participate in a 30 study aimed at developing a potentially useful prenatal care tool to identify the risk for  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   19   1 maternal and perinatal morbidity and mortality. Following national ethical regulations, 2 the participants will not receive any financial compensation. 3 Women who agree to participate in the study will not have any disadvantage or 4 difficulties in prenatal care. On the contrary, they will receive a contact number to find 5 clinical researchers at any time (24/7 service), maintaining a closer contact with 6 researchers and care providers. The MAES-I team is committed to contact health care 7 providers if any potential complication arises. 8 Participating women will not be held accountable for any loss, theft or damage to the 9 wrist device. These women will only be required to wear the device as a regular wrist- 10 watch and no self-damage is expected. 11 Participating women will not be able to identify any PA or sleep parameters at any stage 12 of the study. Data can only be downloaded through proper licensed software of the 13 device. The actigraphy devices provided to participating women have a unique code 14 which will be recorded in the database along with the interval of use per woman. 15 Actigraphy data will be labelled using participant ID, device number, gestational age 16 when the device was initially used and the return date of each device. Codes, ID number 17 and numbers will ensure confidentiality of all participating women. The identity of all 18 women will be kept confidential. 19 MAES-I study has been reviewed and approved by the National Committee for Ethics in 20 Research of Brazil (CONEP) and by the Institutional Review Board (IRB) of the 21 coordinating centre (Letter of approval 1.834.116 issued on 24 th November 2016) and 22 of all other Brazilian participating centres. All women enroled in the MAES-I cohort will 23 sign an informed consent form. 24 Ethical principles of the Brazilian National Heath Council (Resolution CNS 466/12) will be 25 upheld at every stage of this study. Anonymity of the source of information will be 26 guaranteed and the woman will receive care irrespective of her agreement to 27 participate in the study. The study also complies with the Declaration of Helsinki 28 amended in Hong Kong in 1989. Methodological and ethical aspects of MAES-I study 29 protocol were developed following STROBE guidelines [57].  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  Patients and the public were not involved in this study for the development of the 3 research question and outcome measures. However, the choice of a wrist device was 4 based on user preference as reported. Participants of the study will have access to 5 information available at the open-access website. 6 Detailed information about the study is provided in the Brazilian Cohort website 7 (www.medscinet.com/samba). Publications of the results of the study can be found in 8 the scientific literature and Institutional webpages. We intend to disseminate our 9 findings to a scientific peer-reviewed journal, general free access website, specialist 10 conferences, and our funding agencies. 11

13
Actigraphy is an innovative, non-invasive, non-operator dependent, wearable 14 technology, that is capable of measuring diverse variables related to mobility, physical 15 activity, sleep-wake, and circadian cycle patterns under real-life conditions. Actigraphy 16 devices have a high sensitivity in detecting sleep-wake parameters and are currently 17 highly recommended by the American Sleep Disorder Association for diagnosis and 18 therapy response of circadian rhythm disorders [27,28,58]. Although some studies 19 show that using the actigraphy device for 7 to 14 days provides reliable estimates of 20 physical activity behaviour in older adults, it is not absolutely clear how many days are 21 needed to estimate habitual PA by using the wrist/waist device during pregnancy. In 22 general, it seems to depend mainly on the type of actigraphy device, wear location and 23 target population [30,33]. Nevertheless, MAES-I study will provide sufficient data to 24 assess different patterns throughout pregnancy. 25 The use of wearable physical activity monitors has increased considerably, owing to 26 interest in the relationship between the pathophysiology of diseases and patterns of 27 physical activity and sleep. A recent study on the use of physical activity monitors in 28 human physiology research unravels current and potential use of the actigraphy device. 29 The device can be applied in strategies that promote a healthier behaviour or predict 30 outcomes [59]. The authors conclude that physical activity monitors, as well as other  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   21   1   new 21 st century technologies, have already transformed physiology research,   2 revolutionizing how we assess patients and opening new areas of interest. In addition, 3 the use of objective measures to evaluate habitual sleep duration and outcomes in 4 pregnancy is critical, considering recent reports of little agreement between objective 5 and subjective assessments of sleep time [60]. 6 Alterations in sleep patterns, including less deep sleep and more nocturnal awakenings 7 can be observed in pregnancy as early as in 10-12 weeks gestation [61]. Sleep 8 disturbances during pregnancy have been associated with preterm delivery, gestational 9 hypertensive disorders, glucose intolerance and increased risk of caesarean delivery 10 [19]. Shortened nocturnal sleep time was also associated with hyperglycemia [62]. 11 Persistent sleep deprivation has been correlated with depressive symptoms and stress 12 perception by pregnant women [61]. These studies explored a correlation between PA 13 patterns and sleep disturbances that determine complications through a well- 14 established relationship between cause and effect. However, this correlation could not 15 always be adequately determined due to study design [17]. 16 In a distinct manner, the intent of our analysis is to discover whether a maternal 17 complication can be identified before the manifestation of its clinical signs, by evaluating 18 physical activity and/or sleep patterns modifications of pregnant women. Considering 19 existing evidence, we speculate that patterns of PA and/or sleep change days or weeks 20 before clinical presentation of the complication. In general, the signs and symptoms of 21 some maternal outcomes are part of the gold-standard criteria for diagnosis (high blood 22 pressure, proteinuria and/or edema in preeclampsia; premature contractions and 23 cervical ripening/dilation in preterm birth; abnormal placental blood flow and 24 insufficient fetal growth in Intrauterine Growth Restriction). We acknowledge that there 25 are potential confounders and limitations in predicting maternal and perinatal 26 complications using PA and sleep patterns estimated by actigraphy devices. The 27 population in our research is expected to have different subgroups of women with 28 different risks and associated factors contributing to maternal complications, such as 29 obesity, smoking habit, and with age under twenty or over forty years old, for instance. 30 None of those factors was considered an exclusion criterion. If possible, we intend to 31 conduct a subgroup analysis of the maternal subgroups, since they may have different  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   22   1 PA and sleep patterns. Nonetheless, we decided to adopt a pragmatic approach and not 2 exclude such a common factor from our sample. 3 The use of actigraphy device during prenatal visits has the potential to become a new 4 tool for monitoring pregnant women. It may improve maternal health care and identify 5 altered PA and/or sleep patterns. Changes can be objectively measured by actigraphy 6 before the occurrence of signs and symptoms. The focus is on providing new technology 7 to monitor the development of potential maternal complications. Other positive points 8 in our study are the data collection period (from 19 weeks until delivery) and the low- 9 risk profile of the cohort, enabling us to describe PA and sleep patterns in a low-risk 10 pregnant population and make a better interpretation of actigraphy data among 11 pregnant women. Current clinical and biological predictors of major maternal 12 complications such as preeclampsia, preterm birth, maternal haemorrhage, and 13 gestational diabetes still lack effective sensitivity and specificity. 14 If our hypothesis is confirmed, this will be an important step for introducing non-invasive 15 screening procedures into prenatal care to identify women at higher risk for those 16 conditions. Women could receive specific advice on the prevention and earlier detection 17 of the condition, take immediate action and seek professional health care to receive 18 appropriate treatment. This would avoid delays, the most significant factors 19 contributing to low-quality health care in underprivileged women, which increase the 20 still substantial burden of maternal morbidity and mortality. If we succeed in identifying 21 "specific patterns of physical activity and sleep" that are predictors of pregnancy 22 complications, further validation studies are recommended to assess the effectiveness 23 of screening procedures in management of these conditions. In addition, MAES-I will 24 permit further specific studies among a high-risk population and also help to identify the 25 best gestational age for monitoring, targeting a specific gestational age interval.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59

Competing interests
The authors declare that they have no competing interests.