Split-Face Study

The split-face design is a within-subject comparison applied to a symmetrical organ. Side allocation is randomised per participant, product and vehicle go on opposite halves under identical instructions, and outcomes are compared as paired differences. Its appeal is statistical: between-subject variance, enormous in dermatology because skin type, sun history, age and baseline severity differ wildly, drops out entirely, so twenty to thirty participants can do work a parallel-group design would need many times more for.

The design works only when its assumptions hold. The condition must be reasonably symmetrical, which is why it suits melasma, acne and photoageing and is useless for anything unilateral. There must be no systemic carryover, so an oral or injected agent cannot be studied this way. Product must not migrate across the midline, which requires separated application and instructions about sleeping position. Assessors, ideally participants too, must be blinded to side.

What a well-run split-face trial buys is a clean internal comparison; what it does not buy is external validity. Effects are measured under supervised application in people whose other side is being treated with vehicle, and subjects who notice a difference may change behaviour on both sides. It also cannot answer questions about systemic exposure or about how the product performs used alone.

The analytical error to look for is treating the two halves as independent observations. Reporting a split-face study of twenty-five people as fifty independent units and running an unpaired test doubles the apparent sample size and inflates significance, when the correct analysis is paired. Two related tells are unrandomised allocation, the active always on the same side, and grading by the investigator who applied it.