Meta-analytic case-crossover design — pooled synthesis of case-crossover studies
Meta-Analysis of Case-Crossover Studies · Also known as: pooled case-crossover analysis, case-crossover meta-analysis, MACCO, systematic pooling of case-crossover studies
The meta-analytic case-crossover design combines the within-person control structure of the case-crossover study with formal meta-analytic pooling across multiple studies. Each contributing study uses cases as their own controls by comparing exposure windows immediately preceding an acute event to matched reference windows in the same individual. The pooled approach synthesizes conditional odds ratios across studies, maximizing statistical power and generalizability — commonly applied to short-term environmental exposures such as air pollution, temperature extremes, and drug triggers of acute events.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use when multiple case-crossover studies have examined the same short-term, transient exposure and acute health event and individual study estimates are imprecise or geographically restricted. The design is particularly well suited to environmental epidemiology (air pollution, heat, cold spells), pharmacoepidemiology (drug-triggered arrhythmia or bleeding), and occupational triggers. Do not use when contributing studies used incompatible control-window strategies without harmonization, as pooling estimates from bidirectional and time-stratified designs introduces overlap bias. Avoid when the research question concerns a chronic or sustained exposure — the case-crossover logic requires a transient exposure with rapid onset and brief hazard window. Also avoid when individual-participant data are unavailable and aggregate-level heterogeneity is high, since aggregate meta-analysis cannot fully untangle exposure-lag-response complexity.
Strengths & limitations
- Each contributing study is internally self-matched, eliminating stable individual confounders such as socioeconomic status, genetics, and chronic health behaviors — this advantage is preserved in the pooled estimate.
- Pooling substantially increases statistical power to detect modest transient effects that are difficult to estimate in any single study.
- Geographic and seasonal diversity of contributing studies improves external validity of the summary estimate.
- The meta-analytic layer allows explicit quantification and exploration of between-study heterogeneity, informing public health guidance across different settings.
- Requires that all contributing studies use compatible, non-overlapping control windows; mixing bidirectional and time-stratified designs without harmonization produces biased pooled estimates.
- The self-matching design handles only stable confounders; time-varying confounders (e.g., concurrent illness, behavioral changes around the event) can still bias individual study estimates and propagate into the pool.
- Aggregate-level meta-analysis cannot reconstruct individual-level exposure-lag-response curves as precisely as individual-participant data meta-analysis.
- Publication bias may be more severe than in cohort meta-analyses because small negative studies of transient triggers are less likely to be published.
Frequently asked
Why is the case-crossover design preferred over cohort studies for transient triggers?
The case-crossover design uses each case as their own control, comparing the hazard period just before the acute event to reference periods from the same individual. This self-matching removes all stable between-person confounders (genetics, chronic comorbidities, socioeconomic status) without having to measure them. A cohort study would need to measure and adjust for all these factors, which is rarely fully achievable. For short-term, transient triggers the case-crossover is therefore more efficient and less confounded.
What is the most important methodological requirement before pooling?
Harmonization of control-window strategies. Studies using bidirectional control windows risk overlap bias when the exposure or outcome has a time trend. Studies using time-stratified windows avoid this. Before pooling, the analyst must verify that contributing studies used non-overlapping, trend-adjusted control windows, or restrict pooling to studies within the same design category.
Can I use individual-participant data rather than aggregate estimates?
Yes, and individual-participant data (IPD) meta-analysis is strongly preferred when feasible. IPD allows a one-stage pooled regression that simultaneously estimates exposure-lag-response curves with correct standard errors, models effect modifiers at the individual level, and avoids ecologic bias inherent in aggregate pooling. However, IPD requires data-sharing agreements across contributing study teams, which is logistically demanding.
How should I handle different exposure metrics across studies?
Standardize to a common unit before pooling — for example, express all air pollution effects per 10 micrograms per cubic meter increase in PM2.5. When studies used different exposure metrics (e.g., interquartile range, percentile change), convert to a common unit using the reported means and standard deviations before meta-regression. Studies that cannot be converted to a compatible unit should be reported separately in sensitivity analyses.
Is a random-effects or fixed-effects model appropriate?
Random-effects models are almost always more appropriate for case-crossover meta-analyses because effect sizes genuinely vary across populations, climates, pollutant compositions, and healthcare systems. Fixed-effects pooling assumes all studies estimate exactly the same parameter, which is implausible for multi-geographic environmental or pharmacological triggers. Use DerSimonian-Laird or REML random-effects estimators and report the between-study variance (tau-squared) alongside I-squared.
Sources
- Maclure, M. (1991). The case-crossover design: a method for studying transient effects on the risk of acute events. American Journal of Epidemiology, 133(2), 144–153. DOI: 10.1093/oxfordjournals.aje.a115853 ↗
- Bateson, T. F., & Schwartz, J. (2001). Selection bias and confounding in case-crossover analyses of environmental time-series data. Epidemiology, 12(6), 654–661. DOI: 10.1097/00001648-200111000-00013 ↗
How to cite this page
ScholarGate. (2026, June 3). Meta-Analysis of Case-Crossover Studies. ScholarGate. https://scholargate.app/en/epidemiology/meta-analytic-case-crossover-design
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Case-crossover designEpidemiology↔ compare
- Matched case-control studyEpidemiology↔ compare
- Nested case-controlEpidemiology↔ compare