ACNE vulgaris (AV) remains one of the most prevalent dermatological conditions, particularly affecting adolescents and young adults. While topical treatments are commonly prescribed for mild-to-moderate AV, adherence to these regimens is often poor due to a range of factors, including perceived ineffectiveness and inconvenience. As a result, interest has grown in personalised approaches to acne care, particularly those driven by technological innovations. A recent study explored whether machine learning could enhance outcomes by recommending customised treatments tailored to individuals. A key finding showed that machine learning-led recommendations resulted in significantly better clinical outcomes and improvements in quality of life compared to self-selected skincare.
This randomised, evaluator-blinded, parallel-group trial enrolled 68 participants (mean age 27 years) with mild-to-moderate acne from an online database. Participants were allocated to one of two groups: Group A received tailored product recommendations generated by a Bayesian machine learning model based on self-assessments and objective data, while Group B chose their own treatments. After an 8-week treatment period, patients submitted self-reported disease scores and facial photographs for evaluation. Primary outcomes included change in acne severity, assessed via the Investigator Global Assessment (IGA) by dermatologists. Secondary outcomes measured changes in quality of life using the Dermatology Life Quality Index (DLQI).
Group A showed statistically significant improvement in IGA scores (mean difference: 0.32; p=0.04), while Group B’s improvement was not significant (mean difference: 0.09; p=0.54). DLQI scores in the machine learning group significantly decreased from 7.75 to 3.5 (p<0.001), indicating an improved quality of life. In contrast, the self-selected treatment group showed a non-significant reduction from 7.53 to 5.3 (p>0.05). Additionally, a significant correlation was found between IGA and DLQI scores in Group A but not in Group B. No adverse reactions were reported in Group A, whereas three occurred in Group B.
These results suggest that personalised skincare recommendations powered by machine learning could be a valuable approach in managing mild-to-moderate acne. The study highlights the potential clinical utility of AI-driven decision support tools in dermatology. However, limitations include a relatively small sample size and short follow-up period, which may impact the generalisability of results. Further research in more diverse and larger populations is needed, but the findings offer promising implications for improving treatment adherence and outcomes in clinical practice.
Reference
Ghazanfar NM et al. Effectiveness of a Machine Learning-Enabled Skincare Recommendation for Mild-to-Moderate Acne Vulgaris: 8-Week Evaluator-Blinded Randomized Controlled Trial. JMIR Dermatol. 2025;DOI: 10.2196/60883.
