Psoriasis is a chronic inflammatory disease, and patients typically cycle through multiple treatment options. We aimed to assess and compare the performance of traditional logistic regression (LR) and machine learning (ML) models in predicting treatment outcome in psoriasis. A systematic search was performed on Medline, Embase, and Web of Science databases for studies published between January 2000 to December 2021, that developed or validated at least one statistical or ML model that could predict treatment outcome in psoriasis. We included studies on psoriasis treatment and models developed using clinical variables, excluding genetic and blood biomarkers. Two reviewers independently screened abstracts and full texts for eligibility, assessed methodological quality and risk of bias using the Prediction model Risk Of Bias Assessment Tool. Out of 20285 records identified, 8 studies met inclusion criteria. The sample size used to develop the models ranged from 23 to 3388. The number of predictor variables ranged from 7 to 20; with age, weight, baseline PASI and history of previous biological therapy most consistently included in the final prognostic models. The LR algorithm was applied most frequently, followed by state-of-the-art ML methods (generalised linear model, elastic net, gradient boosted tree and artificial neural network). ML models exhibited superior performance (estimated area under the curve, AUC 0.80-0.95) compared to LR models (0.59-0.82). Considerable methodological heterogeneity was observed, due to lack of external validation, variation in treatment outcome endpoints and follow-up times. Most studies had high risk of bias in their analysis. Although ML methods offered limited improvement over traditional LR in predicting psoriasis treatment outcome, their clinical feasibility and reliability are still unclear. Improvements in psoriasis prediction models' methodology and reporting are needed before they can be considered for clinical use.

Psoriasis is a chronic inflammatory disease, and patients typically cycle through multiple treatment options. We aimed to assess and compare the performance of traditional logistic regression (LR) and machine learning (ML) models in predicting treatment outcome in psoriasis. A systematic search was performed on Medline, Embase, and Web of Science databases for studies published between January 2000 to December 2021, that developed or validated at least one statistical or ML model that could predict treatment outcome in psoriasis. We included studies on psoriasis treatment and models developed using clinical variables, excluding genetic and blood biomarkers. Two reviewers independently screened abstracts and full texts for eligibility, assessed methodological quality and risk of bias using the Prediction model Risk Of Bias Assessment Tool. Out of 20285 records identified, 8 studies met inclusion criteria. The sample size used to develop the models ranged from 23 to 3388. The number of predictor variables ranged from 7 to 20; with age, weight, baseline PASI and history of previous biological therapy most consistently included in the final prognostic models. The LR algorithm was applied most frequently, followed by state-of-the-art ML methods (generalised linear model, elastic net, gradient boosted tree and artificial neural network). ML models showed modest superior performance (estimated area under the curve, AUC 0.80-0.95) compared to LR models (0.59-0.82). Considerable methodological heterogeneity was observed, due to lack of external validation, variation in treatment outcome endpoints and follow-up times. Most studies had high risk of bias in their analysis. Although ML methods offered limited improvement over traditional LR in predicting psoriasis treatment outcome, their clinical feasibility and reliability are still unclear. Improvements in psoriasis prediction models' methodology and reporting are needed before they can be considered for clinical use.

ESDR116 - Comparison of machine learning and logistic regression models in predicting psoriasis treatment outcome: A scoping review

The scientific meeting of the European Society for Dermatological Research (ESDR) is held yearly, usually in September. With an international attendance of more than 1000 skin scientists from around the globe, the four-day event includes short oral presentations from peer-reviewed abstracts, keynote lectures from international experts, sponsored symposia, satellite programs, educational events as well as an important focus on posters, networking and building collaborations amongst scientists. For more information about this year’s meeting visit our website.