Quantifying the advantages of conducting a prospective meta-analysis (PMA): a case study of early childhood obesity prevention

Abstract

Background: For prospective meta-analyses (PMA), eligible studies are identified and the PMA hypotheses, selection criteria and analysis methods are pre-specified before results of any of the studies are known. This reduces publication bias and selective outcome reporting and provides a unique opportunity for outcome standardisation/harmonisation. We conducted a world-first PMA of four trials investigating interventions to prevent early childhood obesity. The aims of this study were to quantitatively analyse the effects of prospective planning on variations across trials, outcome harmonisation, and the power to detect intervention effects, and to derive recommendations for future PMA.Methods: We examined intervention design, participant characteristics and outcomes collected across the four trials included in the EPOCH PMA using their registration records, protocol publications and variable lists. The outcomes trials that planned to collect prior to inclusion in the PMA were compared to the outcomes trials collected after PMA inclusion. We analysed the proportion of matching outcome definitions across trials, the number of outcomes per trial, and how collaboration increased the statistical power to detect intervention effects.Results: The included trials varied in intervention design and participants, which improved external validity and the ability to perform subgroup analyses for the meta-analysis. While individual trials had limited power to detect the main intervention effect (BMI z-score), synthesising data substantially increased statistical power. Prospective planning led to an increase in number of collected outcome categories (e.g. weight, child’s diet, sleep), and greater outcome harmonisation. Prior to PMA inclusion, only 18% of outcome categories were included in all trials. After PMA inclusion this increased to 91% of outcome categories. However, whilst trials mostly collected the same outcome categories after PMA inclusion, some inconsistencies in how the outcomes were measured remained (such as measuring physical activity by hours of outside play versus using an activity monitor).Conclusion: Prospective planning led to greater outcome harmonisation and greater power to detect intervention effects, while maintaining acceptable variation in trial designs and populations, which improved external validity. Recommendations for future PMA include more detailed harmonisation of outcome measures, and careful pre-specification of analyses to avoid research waste by unnecessary over-collection of data.

A Review of Registered Randomized Controlled Trials for the Prevention of Obesity in Infancy

“…The four included trials had minimal power on their own (all less than 0.35) to detect the observed intervention effect of 0.12 on BMI z-score at p < 0.05. However, their combined power was 0.83 [ 62 ].…”

Transforming Obesity Prevention for CHILDren (TOPCHILD) Collaboration: protocol for a systematic review with individual participant data meta-analysis of behavioural interventions for the prevention of early childhood obesity

“…This strengthens the chance of detecting intervention effect differences, and enables us to determine the size of such effects with greater certainty, 49 while also allowing variation in study designs and population which heightens generalisability and allows a greater diversity to study effect modification for different subgroups of individuals or trial characteristics. 50 Moreover, this collaborative approach maximises the use of existing data, thereby reducing research waste.…”

Prospective meta-analyses and Cochrane's role in embracing next-generation methodologies

“…[8] In a recent PMA on childhood obesity prevention, the decision to collaborate increased the number of core outcomes collected by all trials (and thus the number of outcomes that could be combined in a meta-analysis) from 18% to 91%. [9] Second, by specifying eligibility criteria and outcomes before results are known, PMA can prevent bias introduced by prior knowledge of study context and findings. [5] Third, PMA can facilitate access to individual participant data (IPD) and allow more complete interrogation of primary datasets.…”