A meta-analysis is a statistical method that systematically combines the results of multiple studies to obtain a more precise estimate of an effect.
In a medical context, meta-analysis is often part of a systematic review and is used in clinical guidelines, benefit assessments, and health technology assessments.
Distinction: systematic review vs. meta-analysis
A systematic review describes the structured literature search, study selection based on predefined criteria, quality assessment, and presentation of results. Meta-analysis is the quantitative component that statistically combines individual effect measures.
It is possible to conduct a systematic review without a meta-analysis (e.g., when studies are highly heterogeneous), just as a meta-analysis without a rigorous systematic approach would be methodologically problematic.
Typical effect measures and models
Depending on the endpoint, different effect measures are combined, such as risk ratio or odds ratio for binary endpoints, mean difference or standardized mean difference for continuous endpoints, or hazard ratios for time-to-event data.
Fixed-effect models or random-effects models are commonly used for statistical pooling. Random-effects models account for additional variability between studies and are common in many health-science research questions.
Heterogeneity and its assessment
A key step is assessing heterogeneity: if studies differ in terms of population, intervention, comparator, endpoints, or methodology, pooling may be misleading.
Statistically, heterogeneity is often described using the I^2 approach or the Q test; from a content perspective, subgroup analyses and sensitivity analyses are important to assess robustness.
Risk of bias and quality assurance
Meta-analyses can be distorted by publication bias if studies with negative or neutral results are less likely to be published. Selective outcome reporting within studies can also influence the overall estimate.
Funnel plots, registered protocols, transparent documentation of the search strategy, and sensitivity analyses help reduce the risk of bias, but they do not replace a critical appraisal of the quality of the evidence.
Practical relevance in HTA, benefit assessment, and reimbursement
In HTA dossiers, meta-analyses are used to summarise the evidence on benefits, harms, and patient-relevant endpoints. Depending on the research question, indirect comparisons or network meta-analyses may be relevant.
For German benefit assessments, transparency and traceability are particularly important, because documentation of the study search, selection, and statistical synthesis is crucial.
FAQ
When should a meta-analysis not be conducted?
If the studies are too different (e.g., markedly different populations, interventions, or endpoints) or if data quality is very low, a quantitative synthesis may be more misleading than a narrative summary.
What is the difference between fixed-effect and random-effects?
A fixed-effect model assumes that all studies estimate the same “true” effect. A random-effects model allows the true effect to vary between studies and often yields wider confidence intervals.
How should publication bias be addressed?
A comprehensive search (including registries), use of registered protocols, transparency regarding inclusion and exclusion decisions, and sensitivity analyses can reduce the risk; however, it can rarely be eliminated completely.
Regulatory references: ICH E6(R3) Good Clinical Practice; EU Regulation (EU) No 536/2014 (Clinical Trials Regulation).
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.
From a sponsor and CRO perspective, it is important that definitions and analysis rules are specified in advance in the protocol and statistical analysis plan. Consistent data collection, plausibility checks, and transparent reporting reduce room for interpretation vis-à-vis ethics committees, authorities, and HTA bodies.