Dose-Response Meta-Analysis Services
Some questions are not “does it work?” but “how does the effect change as the exposure changes?” Dose-response meta-analysis pools studies across multiple exposure levels to estimate the shape of that relationship—linear or curved—so you can see thresholds, plateaus, and where more is not better.
Dose-response meta-analysis pools studies that report an outcome across two or more exposure or dose levels to estimate the shape of the relationship between exposure and effect. It can model linear and non-linear trends (for example with restricted cubic splines), revealing thresholds, plateaus, and turning points that a single pooled estimate would miss.
What dose-response meta-analysis does
A standard meta-analysis compares two conditions—treated versus control, exposed versus unexposed—and pools a single effect. But many exposures are not on/off: they come in levels. A drug has a dose, a programme has an intensity or duration, a behaviour has a frequency, a risk factor has an amount. For these, the interesting question is the shape of the relationship: does the effect keep growing with the dose, level off, or even reverse?
Dose-response meta-analysis answers this by pooling studies that each report an outcome across multiple exposure levels, using the within-study trend across those levels as well as the differences between studies. The result is an estimated dose-response curve—with a confidence band—rather than a single number. Crucially, it can model non-linear relationships (commonly with restricted cubic splines), so it can reveal a threshold below which nothing happens, a plateau beyond which more adds little, or a J- or U-shaped curve where the effect turns.
When to use it
Use dose-response meta-analysis when your exposure is quantitative and reported at several levels across studies, and when the shape of the relationship—not just whether an effect exists—matters for the decision. It is the right tool for questions like: is there a threshold dose? Are the returns diminishing? Is there an optimum beyond which the effect worsens?
It has real data requirements: studies must report results for at least two exposure levels (ideally more), with the number exposed and the outcome at each level, and usually a defined reference category. Where studies only compare a single treated group to a control, ordinary meta-analysis or meta-regression on a dose moderator is the better fit. Dose-response methods earn their place when the within-study exposure gradient is actually reported.
Standard vs dose-response meta-analysis
| Standard meta-analysis | Dose-response meta-analysis | |
|---|---|---|
| Exposure | Two conditions (on/off) | Multiple levels (a gradient) |
| Answers | Does it work? | How does the effect change with dose? |
| Output | One pooled effect | A dose-response curve with a band |
| Shape | Not modelled | Linear or non-linear (splines) |
| Reveals | Average effect | Thresholds, plateaus, turning points |
One-stage, two-stage, and the shape question
Dose-response meta-analysis is fitted in one of two ways. The two-stage approach estimates a dose-response trend within each study first, then pools those trends. The one-stage approach models all the data together in a single mixed model; it uses the data more efficiently and is often preferred, particularly when studies report few dose levels. We choose the approach that fits the data and justify it.
The central modelling decision is the shape. Assuming a straight line when the truth curves—or forcing a complex curve on sparse data—both mislead. We test whether a non-linear model (typically restricted cubic splines) is warranted, compare it against a linear fit, and let the evidence decide, rather than imposing a shape. Correct handling of the reference level and the correlation among the estimates within each study is essential, and getting these details right is much of what separates a sound dose-response analysis from a fragile one.
Don’t assume the shape—test for it. The value of a dose-response meta-analysis is telling you whether the relationship is linear, has a threshold, or plateaus. Forcing a straight line on a curved relationship hides exactly the finding that matters.
Interpretation & software
Extrapolating beyond the observed dose range is unreliable—the curve is only supported where studies actually reported exposures—so we interpret within range and say so. We fit these models in established, reproducible tools (R’s dosresmeta and related packages) and deliver versioned code.
How we deliver a dose-response meta-analysis
Dose-response meta-analysis sits within our wider meta-analysis and evidence-synthesis service, run on a full systematic-review workflow—so the curve is built on a sound, reproducible review.
We begin with a registered protocol, a comprehensive search, and careful extraction of the exposure levels, the number exposed, and the outcome at each level, with the reference category defined. We then fit the dose-response model (one- or two-stage), test linear against non-linear shapes, and assess heterogeneity and bias.
Reporting follows PRISMA standards, with the model, shape testing, and dose range reported transparently.
You receive the estimated dose-response curve with its confidence band, effect estimates at meaningful dose levels, the test of non-linearity, a heterogeneity and sensitivity assessment, and reproducible code and data—interpreted within the observed dose range. The result shows not just whether the exposure matters, but how.
Dose-response meta-analysis across Management & Allied Studies
Dose-response meta-analysis originated in—and is most established in—health and epidemiological research. Within Management and Allied Studies it applies wherever studies report an outcome across several ordered exposure levels; where they only report a single continuous moderator, meta-regression is usually the more appropriate route. We are explicit about which of the two your evidence base supports.
Health & Behavioural Science
The classic and most data-rich setting—dose, duration, or frequency of exposure and its curved relationship to risk or outcome, with graded exposure levels routinely reported.
Marketing & Consumer Research
How response varies with advertising exposure or frequency—an established “dose” setting, including saturation and wear-out—where studies report several exposure levels.
Education & Learning Sciences
How learning outcomes change with instructional dosage—time or intensity—a recognised concept in intervention research, where graded-dosage studies exist.
Economics & Public Policy
Graded treatment intensity or transfer size where studies report ordered exposure levels; otherwise handled as a meta-regression on a continuous moderator.
Management & Organisational Studies
Applicable where an intervention’s intensity or duration is reported at several levels (e.g. training dosage); more often, a meta-regression on the continuous moderator is the honest fit.
When it does not fit
If studies compare only a single treated group to a control, or report just one continuous predictor, dose-response methods do not apply—we use standard meta-analysis or meta-regression instead, and say so.
Dose-response meta-analysis: common questions
Need the shape, not just the average?
If your exposure comes in levels and the threshold, plateau, or optimum matters, a dose-response meta-analysis will show you the curve. We design and deliver it—shape tested, not assumed.