Meta-analytic Dose-Response Analysis
Also known as: dose-response meta-analysis, DRMA, pooled dose-response modeling, trend meta-analysis
Meta-analytic dose-response analysis pools summary statistics from multiple epidemiological studies to characterize how disease risk changes across ordered levels of an exposure. Rather than comparing a single high-exposure group against a reference, it reconstructs a continuous or categorical exposure-risk curve across the full range of doses, providing far richer evidence about the shape and magnitude of an association than any single study can supply.
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 meta-analytic dose-response analysis when multiple studies report risk estimates across ordered exposure categories and the research question concerns the shape — not merely the presence — of the exposure-risk relationship (linear vs. J-shaped vs. threshold). It is the method of choice for nutrition, environmental, and occupational epidemiology where continuous exposures are categorized in primary studies. Do NOT use it when fewer than three exposure categories are available per study, when studies use incompatible exposure metrics that cannot be harmonized, or when individual-level data are available — in that case a one-stage individual patient data meta-analysis is preferable and more powerful.
Strengths & limitations
- Reconstructs a continuous exposure-risk curve from aggregated published data, without requiring access to individual patient records.
- Detects non-linear exposure-risk shapes (J-curves, thresholds, saturation) that simple two-group meta-analyses cannot reveal.
- Increases statistical power by borrowing information across all exposure categories and studies simultaneously.
- Restricted cubic splines allow flexible, data-driven curve fitting while controlling overfitting through choice of knot number.
- Produces visually interpretable risk-vs-dose plots that communicate evidence to clinicians and policymakers clearly.
- Requires that each included study reports at least three ordered exposure categories with cell counts or variance-enabling statistics; many papers report only one contrast.
- Relies on the Greenland-Longnecker covariance approximation, which is exact only when the reference category is unexposed and cell counts are available; misapplication yields incorrect standard errors.
- Cannot correct for within-study confounding; any residual confounding in primary studies propagates into the pooled curve.
- Between-study heterogeneity in exposure assessment, follow-up length, and confounding control can make a pooled curve biologically misleading even when statistically estimable.
Frequently asked
What software implements meta-analytic dose-response analysis?
The dosresmeta package (available for both Stata and R) by Orsini and colleagues is the standard tool. It implements the Greenland-Longnecker covariance reconstruction, GLS fitting, and random-effects pooling. The R package also supports restricted cubic splines. A newer R package, dodr, offers additional functionality for one-stage approaches.
How many studies do I need for a reliable dose-response curve?
There is no formal minimum, but fewer than five studies typically yield very wide confidence bands and make spline fitting unstable. Practical guidance suggests at least five studies with at least three exposure categories each. With fewer studies, a simpler linear trend test is more defensible than a flexible spline curve.
Can I include both cohort and case-control studies in the same analysis?
Yes, but with caution. Cohort studies yield RR or HR; case-control studies yield OR. Under rare-disease assumptions ORs approximate RRs, but this assumption breaks down for common outcomes. If designs are mixed, run design-stratified sensitivity analyses and note that pooling assumes the log-risk scale is comparable across study types.
How do restricted cubic splines differ from a simple quadratic model?
Both allow a non-linear curve, but RCS (natural splines) are piecewise cubic polynomials constrained to be linear beyond the boundary knots. This prevents the erratic behaviour at extreme exposure values that polynomial models often show. The number and placement of interior knots determine flexibility; three to five knots are typical in dose-response meta-analysis.
What if some studies do not report cell counts needed for covariance reconstruction?
If a study reports only the log-RR and its standard error per category but not cell counts, it cannot contribute to the full GLS analysis. Options include: excluding those studies and noting it as a limitation, imputing counts from the reported variance using an iterative algorithm (implemented in dosresmeta), or running a sensitivity analysis restricted to studies with complete data.
Sources
- Greenland, S., & Longnecker, M. P. (1992). Methods for trend estimation from summarized dose-response data, with applications to meta-analysis. American Journal of Epidemiology, 135(11), 1301–1309. DOI: 10.1093/oxfordjournals.aje.a116237 ↗
- Orsini, N., Li, R., Wolk, A., Khudyakov, P., & Spiegelman, D. (2012). Meta-analysis for linear and nonlinear dose-response relations: Examples, an evaluation of approximations, and software. American Journal of Epidemiology, 175(1), 66–73. DOI: 10.1093/aje/kwr265 ↗
How to cite this page
ScholarGate. (2026, June 3). Meta-analytic Dose-Response Analysis. ScholarGate. https://scholargate.app/en/epidemiology/meta-analytic-dose-response-analysis
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.
- Generalized Least SquaresStatistics↔ compare
- Network Meta-AnalysisEvidence Synthesis↔ compare