How to make a forest plot from a spreadsheet

Standard in meta-analysis and increasingly expected for any paper reporting several effect estimates side by side. Most tutorials assume R. This one does not.

Laying out your data

One row per study or subgroup, with columns for the label, the effect estimate, and the lower and upper confidence limits. If you have counts rather than computed effects, the measure can be calculated for you.

Choosing the effect measure

Odds ratio (OR) for case-control designs and logistic models. Risk ratio (RR) for cohort studies and trials with a clear baseline risk. Easier to interpret than an OR and preferred where both are available. Hazard ratio (HR) for time-to-event data. Mean difference (MD) when every study used the same scale. Standardised mean difference (SMD) when they used different scales.

Do not mix measures on one plot.

Reading it

The vertical line is the null: 1 for ratio measures, 0 for differences. Each horizontal line is one study's confidence interval, its square the point estimate sized by weight. The diamond at the foot is the pooled estimate. Any interval crossing the null is non-significant at that level.

The mistake to avoid

Plotting ratio measures on a linear axis. Ratios are multiplicative, so an OR of 0.5 and one of 2.0 are equal and opposite. On a linear axis they look wildly different. Use a log scale for OR, RR and HR.

In Plotara

Forest plots support OR, RR, HR, MD and SMD, with a log axis where appropriate, and the Python exports with the figure.