Criteria for considering studies for this review

Constructs of cognitive function, cognitive deficits and AD

Cognitive function is referred to as the mental action or process of acquiring knowledge and understanding through thoughts, experiences and senses.30 It encompasses processes, such as knowledge, attention, memory and working memory, judgement and evaluation, reasoning and ‘computation’, problem solving and decision making, comprehension and production of language, etc.30 Any measures of cognition (or its domains) (eg, the presence or absence of cognitive deficits determined using a single question, a case definition of any measure of cognitive domain by a standardised clinical tool, or cognitive domain scores reported as a continuous variable) will be considered.

Development of a cognitive deficit in patients with CNS trauma will be assigned based on the Diagnostic and Statistical Manual of Mental Disorders Fifth edition (DSM-V) criteria for MCI, namely, (1) evidence of modest cognitive decline from a previous level of performance in one or more cognitive domains (complex attention, executive function, learning and memory, language, perceptual motor, or social cognition) based on: (i) concern of the individual, a knowledgeable informant, or the clinician that there has been a mild decline in cognitive function; and (ii) modest impairment in cognitive performance, preferably documented by standardised neuropsychological testing or, in its absence, another quantified clinical assessment; and (2) the cognitive deficits do not interfere with the capacity for independence in everyday activities (ie, complex instrumental activities of daily living such as paying bills or managing medications are preserved, but greater effort, compensatory strategies or accommodation may be required.)31

Exclusion criteria

Search methods for the identification of studies

In collaboration with clinical experts and a medical information specialist, a comprehensive search strategy for studying cognitive function in CNS trauma was developed. All English language, peer-reviewed studies with prospective or retrospective data collection and a longitudinal design, found through Medline, Central, Embase, Scopus, PsycINFO and supplemental PubMed, were eligible. Searches in individual databases covered publications’ time frame from inception until early December 2016. The complete search strategy can be found in online supplementary file 1. We refer the reader to the Cochrane handbook and other published sources for justification of the selected databases.

Searching other resources

Reference lists of included studies will be reviewed to identify any additional studies.

Selection of studies

All searches will be saved in EndNote with duplicates removed. For the first level of screening, two reviewers (TM and NP or TM and AD) will read the titles and abstracts of all the citations from the electronic database searches and remove all citations not related to the primary research objectives. For the second level of screening, each reviewer will independently assess the full article. If the title or abstract suggests that the study might meet the inclusion criteria, each reviewer individually will assess the full text; any conflicting views will be resolved by discussion between reviewers and systematic review team members. If needed, clinical and research experts on the field will be contacted. Studies failing to meet the inclusion criteria will be excluded and the reason for exclusion will be reported.

Data extraction and management

For studies fulfilling the inclusion criteria, two review authors (AS or NP for descriptive data; TM and AS or NP for outcome data) will independently extract data into data collection forms grouped according to study design. The data from observational longitudinal studies will be used to address all research objectives. Randomised control trial (RCT) studies will be treated as cohort studies by abstracting data from the control (eg, untreated group) to address the first research objective (eg, to determine the course of cognitive deficits) in patients with CNS trauma.

The abstracted data will include the following: (1) study characteristics (author names, publication year, country of study, study setting, study design, sample size, method of measuring cognition and cognitive domains covered, number of participants assessed for at each time point, time between assessments and time since injury to each follow-up); (2) participant characteristics (mean age, sex, education, definition of CNS (ie, TBI, SCI), localisation/level of injury and injury severity); (3) medication regimen, if reported; and (4) results (reported frequencies of cognitive impairments, and all reported predictive associations between cognition and other variables (ie, age, sex, education, TBI severity, comorbidity)).

Data synthesis

Results of each study will be divided into two main categories: the course of cognitive deficits after CNS trauma and prognostic factors of cognitive function across assessment times.

To determine the course of cognitive deficits, matching assessment times (ie, time since diagnosis when cognitive function was measured) will be grouped by their corresponding frequencies and a sample size-weighted mean frequency value will be calculated for time points with more than one contributing frequency value (ie, more than one study reporting cognitive deficit at that time point). Results will also be grouped, if sufficient data exists, taking into account measurements used to assess cognitive function.

Prognostic factors associated with cognition and/or its domain(s) will be extracted from all cohorts (and untreated/without an effect RCTs). All factors influencing the course of cognitive values/status, as reported by the author, will be considered as prognostic factors. A prognostic association will be considered as significant if the reported p-value is ≤0.05, authors reported association as significant, or the 95% confidence intervals around a rate ratio or similar statistic did not exceed one, when adjusted for at least age and sex (ie, minimum set of confounders) in a multivariable model. Where a prognostic factor was assessed with respect to the outcome at several time points in the same cohort, data will be extracted and reported for each follow-up time.

Methodological quality and risk of bias assessment

Study quality will be assessed independently by two experienced reviewers using the Quality in Prognosis Studies (QUIPS) guidelines.37 38 The appraisal will consist of two steps: (1) assessment of six potential sources of bias (study participation and attrition, prognostic factors and outcome measurements, confounding measurement and account, and analyses) and (2) grading the presence of potential biases as ‘Yes’, ‘Partly’, ‘No’ or ‘Unsure’. To summarise the level of evidence, the Scottish Intercollegiate Guidelines Network (SIGN) methodology38 will be utilised: (1) ‘++’ when all or most of the quality criteria proposed by QUIPS are fulfilled (allowing one ‘Partly’ while appraising all potential sources of bias); (2) ‘+’ when some of the criteria are fulfilled; and (3) ‘-’ when few or none of the criteria are fulfilled (at least one ‘Yes’). As proposed by SIGN, studies with retrospective design will not receive a ‘++’ rating. Group (1) will be referred to as ‘high-quality studies’ and group (2) as ‘moderate-quality studies’. Results will be reported in a table format.

Finally, the consistency of the level of evidence will be summarised. Evidence will be designated ‘strong’ when consistent findings are found in multiple ‘moderate-quality’ or one or more ‘high-quality studies’ and the total sample size of combined eligible studies is ≥100; ‘moderate’ when consistent findings in multiple ‘moderate-quality’ or one ‘high-quality’ quality study with a total sample ≥50, or at least one ‘moderate-quality’ or ‘high-quality’ quality study with a total sample of 50–99; ‘limited’ when findings are found in at least one ‘moderate-quality,’ or ‘high-quality’ quality study with total sample size between 25 and 49; and ‘unknown’ when findings are of indeterminate rating, in studies with poor methodological quality or with a sample of <25.38 To ensure the explicit basis for bias assessment, aspects of the trial methods on which the judgement for ‘high risk of bias’ will be based as well as the judgement itself—including the trial method on which the decision of exclusion was based—will be reported.

Measurements: description and properties evaluation

Using a standardised form developed for a previous systematic review on measurements’ properties,39 details of included measurements will be extracted from original studies and manuals—where available—and studies that evaluated their psychometric properties. The following descriptors will be extracted and reported: general characteristics, purpose and content, method of administration, respondent burden, language (and translations), psychometric properties, particularly test–retest reliability and construct validity, and strengths and cautions for application in patients with CNS trauma.39 Measurements will then be categorised into the following groups: (1) global measures of cognition; (2) domain-specific measures of cognition; and (3) multi-domain measures, which include items (subscales) of cognition among other functions.40 Based on the above descriptors and using the Holmbeck et al’s evidence-based assessment criteria,41 a rating will be assigned—‘well-established assessment’, ‘approaching well-established assessment’ or ‘promising assessment’ in patients with CNS trauma—to each included measurement.

Given the diversity in measurements of domains of cognitive function, severity and localisation of CNS trauma, as well as the high likelihood of the diversity of statistical methodology used to express associations, a best-evidence synthesis approach42 will be applied, synthesising findings from studies with sufficient quality through tabulation and qualitative description. Some features falling under meta-analysis constructs (ie, sample size-weighted mean frequencies) will be considered only for the first research question (ie, the course of cognitive deficits). As indicated above, for studies focusing on the same domain of cognitive deficits prevalence (ie, executive function, memory or other), sample size-weighted mean frequencies will be collected and matching assessment times (ie, time post-injury that each domain of the cognitive function was measured) will be grouped with their corresponding frequencies. A sample size-weighted mean frequency value will be calculated for time points with more than one contributing frequency value (ie, more than one study reporting values at that time point post-injury).

Dealing with missing data

In cases of missing data, primary authors will be contacted. The proportion of missing data will be reported along with reasons, where indicated. In the case of duplicate publications and companion papers of a primary study, we will attempt to yield maximum scientific information by abstraction of all available data. Nonetheless, original publication (usually the oldest version) will take priority in data analysis.

Ethics and dissemination

This systematic review will provide increased scientific certainty about the prognostic factors and natural history of cognitive status in patients with CNS trauma, informing clinicians, policymakers and subsequent studies funded by the Alzheimer’s Association. The strength of this systematic review protocol and research programme is in its methodology, making it possible to identify associations longitudinally, thus improving the quality of inductive inferences about the natural progression of cognitive status in patients with CNS trauma. The multilevel risk of bias assessment will allow researchers to detect main flaws in the designs of individual studies and to inform future research on the topic of cognition in CNS trauma. During the development phase of the project, concepts and hypotheses were formulated based on a synthesis of relevant discoveries across various disciplines. It is evident that the content of measurements and their psychometric properties can influence estimates of cognitive impairment across time points (ie, natural history research aim) as well as the predictors of cognition in patients with CNS trauma. Likewise, earlier research has highlighted that different domains of cognition are affected differently in TBI and SCI based on type, localisation and severity of injury and that the perception of a person with CNS trauma can also be influenced by the duration of cognitive impairment (ie, individuals could have adjusted to their long-term impairment, making them less likely to recognise their degree of impairment). To mitigate these issues, a variety of data related to the domains of each measure will be collected and reported on. This approach will help researchers think about elements that constitute cognition in patients with CNS trauma, increasing the comprehensiveness of understanding and the appreciation/relevance of the measurement theory in the study of cognition. Finally, multi-level knowledge translation activities (publications, presentations, research-knowledge user collaborations) throughout this research activity will be performed. This will ensure that the programme results reach their intended knowledge users with the goal of informing innovations in AD and CNS trauma research, health policy and practice.

Limitations

Certain features of this review are open to debate. Those features include the following: (1) The assumption of expected heterogeneity in the primary studies with respect to sample characteristics (ie, age, injury/localisation of injury, time since injury) and the applied measurements of cognition properties and content; (2) potentially unequal sex distribution in primary studies, given that historically both TBI and SCI have been considered an injury-specific to males; (3) age, a potential predictor of cognitive decline, may not be always reported as a continuous variable or in similar ranges; therefore, assessment of the age factor would be limited to those studies that reported it; (4) potential confounders, such as medication effects/illicit substance use and comorbidity load, may not be adequately explored given the lack of consistent reports and effect consideration in TBI research43; nonetheless, where possible, such effects will be explored; (5) baseline assessments of cognition in patients with CNS trauma are expected to vary between studies: while the issue was predicted and attempts will be made to mitigate the zero-time effect by reporting results with baseline assessments up to 1-month post-injury and after 1 month; separately, pre-morbid assessments of cognition are unlikely to be understood, and therefore, the generalisability of the results may remain unclear due to inadequate control of the confounding effects of premorbid cognitive functioning; (6) additional limitations relate to the exclusion of grey literature, non-English language articles and unpublished manuscripts and their potentially relevant results; this decision was based on the extensive number of studies identified within databases we have searched, as well as limited empiric evidence about the potential impact of selective searching and inclusion of these works on the results of systematic reviews.44

Despite the outlined limitations, this is the first review of its kind to consciously and comprehensively synthesise evidence on the cognitive status in patients with CNS trauma, aiming to advance scientific knowledge in the field and enrich the care provided to patients with cognitive impairments stemming from TBI and traumatic SCI.

Implications

An ageing population will increase the burden of CNS trauma as the number of people surviving after the injury progressively increases. While the neurological consequences of CNS trauma are well described, evidence is emerging on sex-dependent/age-dependent associations between a prior injury (most frequently TBI) and the development of senile Alzheimer’s-type dementia many years later. Diagnostic criteria, first established by the National Institute of Neurological and Communicative Disorders and Stroke, were revised in light of the discovery of new markers—amyloid-B and tau proteins—accumulation of which brings cognitive dysfunction and neuronal death. These markers, shown to be non-specific in AD, appeared to be similar in persons with TBI and fatal familial insomnia. To fill the gap that is left by current evidence-based practice, the proposed study will lay the ground for further research aimed at determining the underlying basis behind any observed patterns. The significant economic and human costs of cognitive dysfunction years after CNS trauma merit the call for systematic efforts to understand the factors that contribute to its development. This research looks to motivate future investigations of AD and CNS trauma in numerous directions.