Audit Effect Direction Before Drawing a Forest Plot

Define what left and right mean

A forest plot places effect estimates and intervals on a common axis. Its visual direction is meaningful only when outcome coding and comparator direction are consistent. Write the effect measure, numerator group or subtraction order, and whether higher outcomes are beneficial or harmful before adding favors labels.

Cochrane’s meta-analysis guidance explains synthesis methods and effect measures. A correctly calculated pooled estimate can still be communicated incorrectly if the displayed axis or event definition reverses its interpretation.

Audit every study row

Check the original outcome definition and contrast. Some studies may report recovery while others report failure, or intervention versus control in the opposite order. A ratio can be inverted when reversing the same contrast, but switching from an event to its complement is not generally accomplished by taking the reciprocal of a risk ratio.

For continuous scales, establish whether larger scores mean improvement or deterioration. Any sign reversal should be documented and applied consistently to the point estimate and interval endpoints. Preserve original values in the extraction record.

Reconcile a hypothetical reversed contrast

Suppose a study reports a risk ratio of 0.8 with interval [0.6, 1.1] for treatment versus control. Reversing the same comparison gives 1/0.8 = 1.25 with interval [1/1.1, 1/0.6], approximately [0.9091, 1.6667]. The endpoints reverse order during reciprocation.

If the event is harmful, a ratio below one favors treatment. If the event is beneficial, the same numerical direction has a different interpretation. Therefore a label such as favors treatment cannot be assigned from the number alone.

Check the finished display

Verify the null line: zero for differences and one for ratios. Confirm that ratio estimates are plotted on the intended logarithmic axis and that displayed labels remain on the ratio scale. Inspect truncated intervals and make their continuation explicit rather than clipping them invisibly.

Compare plotted points and printed numbers against the extraction table for several rows, including reversed contrasts and extreme intervals. Check pooled diamonds and subgroup labels separately. The final forest plot should communicate a harmonized scientific comparison, with direction rules that can be traced back to each study rather than inferred from the position of its marker.

Source verification: 6 November 2026. Examples are hypothetical.