Systematic review statistics and synthesis services

The included studies are raw material; the synthesis is the contribution. A review that pools incorrectly, ignores heterogeneity, or reports a single number where one is not justified will be sent back by a competent reviewer no matter how thorough the search was. The statistics are judged on whether the model fits the data, whether the heterogeneity was explored rather than buried, and whether the certainty of the conclusions matches the strength of the evidence. We treat the analysis as the decisive stage it is, choosing between a fixed-effect and a random-effects model on a stated rationale rather than habit.

Heterogeneity is the question that most often decides whether pooling is even defensible. We quantify it, then explore its sources with subgroup analysis and meta-regression, and where it is too high to support a single estimate we say so plainly. Our guide on heterogeneity in meta-analysis sets out how we read it, and where your data are already extracted, the analysis follows straight from your extraction dataset.

1. 1

   Effect measure, model, and pooling

   We choose the effect measure and the fixed-effect or random-effects model, then compute the pooled estimate with its confidence interval and interpretation.

2. 2

   Heterogeneity and meta-regression

   We quantify heterogeneity with Q, I-squared, and tau-squared, then explore it through subgroup analysis and meta-regression where there are enough studies.

3. 3

   Network meta-analysis

   Where you compare three or more interventions, we combine direct and indirect evidence, check transitivity and consistency, and rank treatments.

4. 4

   Bias, robustness, and certainty

   We test small-study effects with funnel plots and Egger's test, run sensitivity analyses, and rate the certainty of each outcome with GRADE.

You receive the pooled results, publication-quality forest plots and funnel plots, the heterogeneity and publication bias diagnostics, a GRADE certainty table, and a methods-and-results write-up you can drop into the manuscript, all with the script that produced it. If you want to sanity-check the numbers first, our free meta-analysis calculator and heterogeneity calculator run the core computations on your own data, and the publication bias calculator checks for small-study effects. For the appraisal that feeds GRADE, our risk of bias and quality assessment service completes the picture.

• A pooled analysis with the model choice justified in writing
• Heterogeneity quantified and explored, not just reported
• Subgroup, meta-regression, and sensitivity analyses where the data allow
• A network meta-analysis with transitivity and consistency checks where relevant
• A GRADE certainty table and a drafted statistical methods and results section
• The annotated analysis script, so a reviewer can rerun every figure

This service is broader than a single pooled estimate. It is a statistician embedded in the quantitative stage of your review, which matters most in the situations below.

We quote a fixed fee per project. The drivers are the number of outcomes to analyse, whether the synthesis is a standard pairwise pooling or a network meta-analysis, how much heterogeneity exploration the data justify, and whether you need the full statistical write-up or only the analysis. Send your extracted data and your outcomes and we will return a fixed quote for the synthesis.