Capabilities
Meta-analysis
When the studies measure the same thing in comparable ways, Julie pools them. Whether they do is your call.
EXAMPLE · A GUIDELINE UPDATEQuestion seven asks whether a maintenance inhaler improves lung function. Three trials report the outcome on the same scale; a fourth reports it only as a percentage change. The methodologist signs the analysis plan before any number is pooled, decides that the fourth trial is described in the text rather than pooled, and records why. What comes back is a forest plot, a certainty rating and a Results paragraph that already states that decision.
Meta-analysis is where a review turns several results into one, and where a careless step does the most damage: a pooled estimate looks more certain than any single trial, whether it deserves to or not. Julie prepares the analysis, computes it, carries it into the certainty assessment and the manuscript, and stops wherever the judgement belongs to you. You can start from a Julie systematic review, or from a set of studies you already have.
-
Agrees the analysis plan with you before the data
You set out the analysis before any pooled result exists: the outcomes, the effect measures, the model and the sensitivity analyses. Julie checks that the plan is complete and consistent, and you sign it. If you change it later, the change is recorded; a change made after you have seen the data is reported as one.
What you get: a plan a peer reviewer can hold you to, and a methods section that states every revision instead of smoothing it over.
-
Pools only what you agree belongs together
Every study that could contribute is listed with the values both readers extracted. Where the two readers disagree, the analysis waits for you. Where a study cannot be pooled, you decide whether it is described in the text, and your reason goes with it into the Methods and the Discussion.
What you get: a pooled estimate in which every contributing number has a recorded decision behind it, and no study left out without a word.
-
Leaves the arithmetic to statistical software
The pooled numbers are not produced by an AI model. They are calculated by established meta-analysis software from the extracted values, so the same inputs give the same results, and a statistician can recompute them.
What you get: forest plots, pooled estimates with confidence and prediction intervals, and heterogeneity statistics (I², τ², Q) that can be reproduced.
-
Covers the analyses a systematic review usually needs
Pairwise comparisons of two-arm studies: risk ratios, odds ratios and risk differences for events; mean and standardised mean differences for continuous outcomes; and reported effect estimates such as hazard ratios, with their intervals. Random-effects and common-effect models, methods suited to rare events and zero cells, and sensitivity analyses that change the method or leave out one study at a time. Where a trial reports only a median or a confidence interval, any conversion is made openly and stated in the methods.
What you get: the standard toolkit, applied as you declared it, with every model choice and conversion written into the methods.
-
Carries the pooled result into GRADE
On the GRADE route, inconsistency and imprecision are judged on the pooled estimate itself. Absolute effects per 1,000 people, and where relevant the number needed to treat, are calculated from a baseline risk you source. Randomised and non-randomised evidence are analysed separately, unless you decide on the record to combine them as GRADE guidance 44 describes.
What you get: a Summary of Findings table that rests on the same numbers as the forest plot.
-
Writes the analysis into the manuscript
The Methods describe the plan and every revision, the model, each conversion and each decision to describe a study rather than pool it, with the name of the person who made it. The Results and the Discussion report the pooled estimates together with their certainty, and the draft names the questions still waiting for the authors.
What you get: a methods section a methods editor can check line by line, and a draft that does not pretend to be finished.
If you are running a systematic review, sitting on a guideline panel, or working in an evidence synthesis unit, tell us what you are working on.