Derive standard participant characteristics

Diversity of characteristics and their coding

Time and resources required to standardise them

Loss of detail in characteristics through categorisation, for example, physical activity coded as any activity versus predominantly inactive

Analysis of subgroup effects and treatment–covariate interactions

Poor recruitment of participants with a given characteristic in primary trials

Limits to clinical usefulness of analysis where covariates are coded in a crude way

Derive standard outcome definitions

Complexity or lack of consensus concerning outcome definition

Time and resources required to standardise them

Unaddressed between-study heterogeneity

Delay in the development of final dataset

Incorporation of unreported outcomes

Format of variables in shared datasets may prevent harmonisation

For outcomes, data collected as mean change, lack of baseline and final values in dataset make them unusable in IPD meta-analysis

Acquisition of IPD from all eligible trials

Uncooperative trial investigators

Lost or corrupted trial data

Logistically unfeasible acquisition of IPD (trials identified late in the project life)

Trials in the IPD meta-analysis might be substantially different in comparison with those for which IPD is not available

The IPD meta-analysis does not reflect the totality of evidence on the subject and is at risk of being out-of-date

Check the integrity of trial data

Matching participants reported in trial publication and with those in shared datasets

Delay in development of final dataset

Dataset credibility

Imputation of missing data

Variation in reasons for data missingness across trials

Effect sizes using imputed data may be harder to defend

Incorporation of unpublished trials

Data integrity could not be verified against peer-reviewed publication

Potential undetectable data errors can have serious impact on the findings of the analysis and its reliability

Analyse all important outcomes

Lack of consensus among trialists and clinicians (core outcome set) means that important outcomes are not recorded in each study

Evidence pooled for research-driven outcomes rather than for those critical and important to patient care