Design: Phase 2, randomised, double-blind, vehicle-controlled mechanism-of-action trial
Population: 60 adults with vitiligo
Primary outcome: Change in interferon-related biomarkers and VASI
Effect: Facial VASI −49.1% at week 24 (p<0.001); CXCL10 fall correlated with T-VASI improvement

This phase 2, double-blind trial randomised 60 adults with vitiligo 2:1 to ruxolitinib 1.5% cream or vehicle twice daily for 24 weeks, then all used the active cream to week 52. Its aim was mechanism rather than efficacy.

Ruxolitinib cream reduced interferon-gamma-related proteins in serum, including CXCL10, at every time point, and CXCL10 fell in lesional skin from weeks 12 to 40. Facial VASI fell by 32.9% at week 12 and 49.1% at week 24, and total VASI by 21.2% at week 24. Falls in CXCL10 correlated with total VASI improvement, and rising melanocyte counts correlated with facial improvement.

The practical value is in confirming why the face responds better than the body and why treatment takes months. CXCL10 is not a test to order in clinic, and the small vehicle group limits comparisons. Topical ruxolitinib availability and cost in India are not established, so standard topical steroids, calcineurin inhibitors and phototherapy remain the routine choices.

Set expectations that facial vitiligo responds faster and more fully than body or acral patches to topical JAK inhibition.
Advise at least 24 weeks of consistent twice-daily use before judging response to a topical JAK inhibitor.
Continue to use topical steroids, calcineurin inhibitors and narrowband UVB as established options.
Do not order CXCL10 testing in routine practice; it remains a research marker.

Why it matters

It links a measurable immune pathway to visible repigmentation, which supports targeting interferon signalling in vitiligo.

The statistics, in plain English

The VASI reductions are within-group changes in the treated arm, not differences from vehicle, so they overstate the drug's effect a little. Correlation between CXCL10 and repigmentation supports the mechanism but does not show CXCL10 can predict who will respond.