Study design

Multicentre observational cohort study.

Study population

The IndEcho study started in November 2016 and will continue until 2019. Participants will be recruited from two Indian birth cohorts: the New Delhi Birth Cohort (NDBC)29 45 and Vellore Birth Cohort (VBC).46 Both were established in 1969–1973, originally to study maternal health and pregnancy outcomes. The participants were measured at various time points through infancy, childhood, adolescence and early adulthood to assess growth and were subsequently followed up at approximately 30 and 40 years of age to measure a range of CVD risk markers. Flow charts of the different stages of follow-up and various measurements are provided in figure 3. Currently, the participants are aged 43–50 years and new cases of glucose intolerance, T2DM and hypertension are emerging alongside rapid transitions in lifestyle and socioeconomic status (SES). For the IndEcho study, we aim to recruit approximately 3000 individuals from both cohorts combined.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Flow chart of various stages of follow-up of the New Delhi and Vellore Birth cohorts and measurements recorded at each stage. BMI, body mass index; DXA, dual energy X-ray absorptiometry; HC, hip circumference; OGTT, oral glucose tolerance test; PA, physical activity; SES, socioeconomic status.

The NDBC

In 1969–1972, 20 755 married women of reproductive age living in a 12 km² area of South Delhi were recruited (figure 3). The cohort included 8181 singleton live births from 9169 pregnancies among these women. Gestational age was derived from last menstrual period (LMP) dates. The birth weight, length and head circumference of the babies were recorded within 72 hours of birth (n=7119) and thereafter 3 monthly up to the age of 12 months (n=4104) and 6 monthly until the age of 21 years (n=2892). An average 23 sets of measurements were recorded for each individual from birth until 21 years. The first adult study took place in 1998–2002; 2584 (32%) of the original cohort were retraced, of whom 1526 men and women (then aged 28–31 years) participated in a study of CVD risk factors.29 Data on their SES, attained education, family history of disease, tobacco and alcohol consumption, diet and physical activity were obtained using standardised questionnaires. Clinical measurements included anthropometry, blood pressure, an ECG and analysis of plasma glucose (during an oral glucose tolerance test), insulin, lipids and proinflammatory markers. In 2008 (at age 34–39 years), the same measurements were repeated in 1100 cohort members and additionally, cIMT and brachial artery endothelial function were measured47 and body composition was assessed using dual-energy X-ray absorptiometry (DXA) in a subset. Excluding 18 known deaths (from a total of 1526 who were followed up during 1998–2002), we aim to recruit as many as possible of the 1508 alive NDBC participants.

The VBC

From 1969–1973, 20 626 women of reproductive age were recruited within defined areas of Vellore town and adjoining rural villages in Tamil Nadu, in South India48 (figure 3). The areas of Vellore town were selected to represent different socioeconomic groups. The cohort comprised 10 691 singleton live babies born to the 20 626 women recruited. Weight and length were recorded at birth (n=10 676) and subsequently in infancy (in the first 3 months, n=5753), childhood (6–8 years, n=5541) and adolescence (10–15 years, n=2672). Gestational age was determined from the mother’s LMP dates. Depending on available funding, VBC members had up to 3 measurements in the first 3 months, up to 2 measurements between 6 and 8 years and up to 5 measurements between 10 and 15 years. The first adult follow-up took place in 1998–2002. Cohort members for whom all birth measurements were available (n=4092) were retraced to participate in an adult follow-up during 1998–2002. Of these 2572 were contactable, and 2218 men and women, then aged 26–32 years, consented to participate in the study. CVD risk factors were assessed using a similar protocol to that for NDBC.49 Subsequently in 2013–2014, 1080 participants (50.1% urban), aged 41–45 years, took part in a body composition study, in which the same anthropometric and biochemical parameters were recorded along with detailed body composition using DXA. Excluding 62 known deaths (from a total of 2218 who were followed up during 1998–2002), we aim to recruit as many as possible of the 2156 alive VBC members.

The total number of participants in phases 1 and 2 of adult follow-up in the NDBC and VBC and reasons for losses to follow-up are summarised in online supplementary table 1. We will minimise loss to follow-up through the following algorithm until all non-responders are accounted for: telephone calls/postal letters, meetings with neighbours by field workers and contact with local municipality/postal services to track addresses of migrated individuals.

Measurements in the IndEcho study

Measurements will include anthropometry (height, weight, waist and hip circumferences, skinfold thickness), body composition by DXA (Vellore only) and bioelectrical impedance, hand grip strength, blood pressure, biochemical measurements (an oral glucose tolerance test, fasting plasma insulin and lipids and urinary albumin/creatinine ratio, as a measure of microalbuminuria). Details of the methods used for these measurements are given in table 1. Lifestyle factors (tobacco and alcohol consumption, diet, physical activity, occupation and SES) will be reassessed. Standard questionnaires for diet (food frequency questionnaire),50 SES (Standard of Living Index),51–53 Global Physical Activity Questionnaire,54 55 smoking and alcohol consumption56 will be used. Details on the key covariates relating to life style are summarised in online supplementary table 2.

Impaired glucose tolerance, impaired fasting glucose and T2DM will be diagnosed based on fasting glucose concentration and glucose concentration 120 min after a 75 g oral glucose load, using WHO criteria.57 Hypertension will be defined as systolic pressure >140 mm Hg or diastolic pressure >90 mm Hg or on treatment for hypertension.58 We will use International Diabetes Federation criteria for the following outcomes59: overweight and obesity: BMI >25 and >30 kg/m^2, respectively; central obesity: a waist circumference >90 cm in men and > 80 cm in women; hypertriglyceridaemia: plasma triglyceride concentration >1.7 mmol/L and low HDL cholesterol:<1.03 mmol/L in men and <1.29 mmol/L in women.

Echocardiography

Cardiac chamber dimensions and systolic and diastolic function will be assessed using transthoracic echocardiography. All measurements will be performed according to the American and European Societies for Echocardiography guidelines for chamber quantification60 61 and left ventricular diastolic function.62 The same machine (Philips CX50 Compact Xtreme, Bothell, USA) and transducer (C5-1 Purewave curved array transducer) will be used in both centres and images will be analysed using the Freeland digitiser and software packages (Alpharetta, USA). M-mode and two-dimensional (2D) echocardiography in the parasternal long axis, midpapillary short axis and apical two-chamber and four-chamber views will be used to measure relative wall thickness (RWT; an index of wall thickness relative to internal dimensions). LVH will be defined based on any one of the following criteria: (1) LV mass >150 g in women and >200 g in men; (2) posterior wall thickness >1.1 cm or (3) LV mass indexed to body surface area (BSA) >95 g/m² in women and 115 g/m² in men.61 Concentric LVH, which is an independent CVD risk factor, particularly in hypertensive patients, will be defined as increased RWT >0.42 in the presence of normal LV mass. LV systolic dysfunction will be assessed using measurements of fractional shortening, ejection fraction and global longitudinal strain (GLS). Ejection fraction will be measured using Simpson’s biplane method in apical 2 and 4 chamber views. GLS will be calculated offline using speckle-tracking technology from the acquired 2D; two-chamber, three-chamber and four-chamber views using the in-built QLAB software in the CX50 echo machine. LV diastolic dysfunction will be assessed using mitral valve inflow velocities, mitral annular tissue Doppler velocities, left atrial volume index and tricuspid regurgitation jet velocity as per the guidelines.62

Scans will be performed by experienced echo technicians, two per centre, after a joint 2-day protocol-specific training session. Following this, the technical quality of scans will be checked by the senior cardiologist locally in each centre, and a set of 10 scans will be exchanged between centres for an independent quality assessment. Scans will be read by experienced cardiologists in each centre (AR in New Delhi and VTS and AGA in Vellore). A random 10% of all scans recorded in the study will be exchanged and read by the other centre to assess interobserver variability.

cIMT measurement

Measurements of cIMT will be made bilaterally using high-resolution B-mode ultrasound with the same machine (Philips CX50 Compact Xtreme) and a 10-MHz linear array probe (Philips) and quantified using the CX50-IMT Philips Quantification application according to American Society for Echocardiography guidelines.63 Each common carotid artery will be imaged in three different projections (anterior, lateral and posterior) proximal to the bifurcation. Measurements of IMT will be taken at a plaque-free zone in the far wall of the common carotid artery, captured at the end of diastole. The mean of three readings will be taken. The presence or absence of plaque, defined as a focal structure that encroaches on the arterial lumen by at least 0.5 mm or is more than 50% of the surrounding IMT or has a thickness greater than or equal to 1.5 mm will also be recorded. Between measurer reproducibility in the LV parameters and cIMT will be assessed by the exchange of 10% of scans from each site.

ECG

All participants will undergo a 12-lead ECG recording. ECGs will be reviewed by a cardiologist, and if there are changes suggestive of ischaemia (ST-depression and/or T wave inversion), detailed Minnesota Coding will be performed. We will use Minnesota codes 1–1 ​and ​​1–2, 4–1 and 4​–2, 5–1 and 5–2 and 7–1 to indicate ischaemic heart disease.64

Body composition assessments

Body composition (total lean mass, fat mass and fat per cent) will be assessed using bioelectrical impedance in both centres (Tanita BC-418 in NDBC and Bodystat-2500 in VBC). In VBC, a DXA examination will be additionally performed (Hologic Discovery) to obtain total and depot specific lean mass, fat mass and total fat percentage. DXA body composition analysis was assessed among NDBC participants in 200965 and will not be re-examined in IndEcho.

A copy of the full protocol can be obtained from the corresponding author.

Sample size and analysis

We will target the surviving cohort members (1508 from NDBC and 2156 from VBC) who participated in follow-ups between 1998 and 2002. From initial tracing in early 2016, a total of 3500 individuals were recontacted, among whom we expect approximately 1250 in Delhi and 1750 in Vellore to participate (total 3000) in the current protocol.

Exposures in early life will include weight and length at birth: weight, height and BMI and independent conditional weight and height estimates of growth during infancy, childhood and adolescence, as previously described.48 Because early growth measurements were measured at different time points and at slightly different ages in both cohorts, and owing to the skewness of BMI, the data will be transformed into SD scores using the lambda-mu-sigma (LMS) method.66 The main analysis will use all available data at each time point; it will also be repeated using the subset of participants having data for all time points (birth, infant, childhood and adolescent measurements). Exposures in young adult life will include lifestyle characteristics (diet, physical activity, smoking and alcohol consumption, SES, urban/rural residence and history of rural–urban migration (Vellore only)), anthropometry, lean and fat mass measured by DXA and bioimpedance, metabolic risk markers (plasma glucose, insulin and lipids) and categorical abnormalities as defined above.

We will check the distributions of all variables and perform appropriate transformations. We will explore relationships between LV wall thickness and mass (as continuous variables) with current body size and make appropriate adjustments, including the conventional adjustment for body surface area and height.60 We will create categorical variables to represent LVH and ventricular dysfunction based on standard cut-offs for LV mass as described above.61 We will examine associations of early life and young adult exposures with outcomes, using linear and logistic regression as appropriate, with and without adjustment for potential confounders and covariates (including age, sex, history of migration, current body size and composition). We will also use the conditional growth modelling approach to relate growth in BMI and height during discrete age periods with adult outcomes.67 68 We will explore interactions between early life and young adult exposures and between early life exposures and current measures, using product terms. The role of young adult lifestyle and CVD risk markers as mediators of associations between early-life exposures and outcomes will also be examined. To test the representativeness of the IndEcho participants and assess the risk of bias, we will compare their early growth variables (birth length, birth weight and height, weight and BMI in infancy, childhood and early adulthood) with members of the original cohorts who did not participate as we have done previously (table 2). Since all analyses in the IndEcho study will be ‘internal’ (within the sample of participants) bias will occur only if the associations between early growth and adult outcomes differ between those who participate in IndEcho and those who were not studied. We cannot think of a mechanism by which such differences could occur. We will examine the data from men and women, from Delhi and Vellore and (in Vellore) among rural and urban participants separately and assess heterogeneity of associations and scope for pooling.

Using tests at 5% significance and a total sample size of 3000, we will have 80% power to detect an association of 0.05 SDs of a continuous outcome (eg, LV mass) per SD change in a continuous predictor (eg, birth weight or conditional weight gain in infancy). For the Vellore cohort alone (based on n=1750), this figure is 0.07 SDs and for the Delhi cohort alone (based on n=1250), it is 0.08 SDs. These figures compare favourably with the association shown between infant weight and adult LV mass in the Hertfordshire cohort (0.17 SD per SD change in infant weight).26 For a binary outcome such as LV diastolic dysfunction (Grades 1 and 2), assuming 10% prevalence and a test at 5% significance level, we will have 80% power to detect an association of 0.17 log odds per SD change in a continuous predictor, equivalent to an OR of 1.19 (Vellore 1.25, Delhi 1.30).