Deterioration of Health-Related Quality of Life After Withdrawal of Risankizumab Treatment in Patients with Moderate-to-Severe Plaque Psoriasis: A Machine Learning Predictive Model
Abstract
Introduction: Risankizumab has demonstrated efficacy in treating moderate-to-severe psoriasis. The phase-3 IMMhance trial (NCT02672852) examined the effect of continuing versus withdrawing from risankizumab treatment on psoriasis severity, including the Psoriasis Area and Severity Index (PASI) and static Physician Global Assessment (sPGA). However, the effect of withdrawal on health-related quality of life (HRQL) was not assessed. Therefore, this study was conducted to evaluate the impact of risankizumab withdrawal on HRQL measured by the Dermatology Life Quality Index (DLQI). Because DLQI was not measured beyond week 16 in IMMhance, a machine learning predictive model for DLQI was developed. Methods: A machine learning model for DLQI was fitted using repeated measures data from three phase-3 trials (NCT02684370, NCT02684357, NCT02694523) (pooled N = 1602). An elastic-net algorithm performed automated variable selection among candidate predictors including concurrent PASI and sPGA, demographics, and interaction terms. The machine learning model was used to predict DLQI at weeks 28–104 of IMMhance among patients re-randomized to continue (N = 111) or withdraw from (N = 225) risankizumab after achieving response (sPGA = 0/1) at week 28. Results: The machine learning predictive model demonstrated good statistical fit during tenfold cross-validation and external validation against observed DLQI at weeks 0–16 of IMMhance (N = 507). Predicted improvements in DLQI from baseline were lower in the withdrawal versus the continuation cohort (mean DLQI change at week 104, −5.9 versus −11.5, difference [95% CI] = 5.6 [4.1, 7.3]). Predicted DLQI deteriorated more extensively than PASI (49.7% versus 36.4%) after treatment withdrawal. Conclusions: The predicted DLQI score deteriorated more rapidly after risankizumab withdrawal than the PASI score, an objective measure of disease. These findings suggest that the deterioration in HRQL reflects more substantial impacts after risankizumab discontinuation than those measured by PASI only.
My Take
An interesting question in psoriasis treatment is what happens when patients stop therapy after achieving good control of their disease. For biologics like risankizumab, some patients maintain their response for extended periods after discontinuation while others experience rapid relapse. Understanding who falls into which category has practical implications for treatment planning and patient counseling. This study used machine learning to predict which patients would experience meaningful deterioration in quality of life after risankizumab withdrawal.
The IMMhance trial provided a unique opportunity to study this question, as it included a randomized withdrawal period where responders were assigned to continue treatment or switch to placebo. We applied gradient boosting methods to build a predictive model, which offers advantages over traditional regression when dealing with complex interactions between predictors. The model identified several factors associated with quality of life deterioration, including baseline disease severity and duration of response prior to withdrawal. These insights can help clinicians have more informed conversations with patients about the likely course after stopping treatment.
What I find valuable about this work is that it bridges the gap between clinical trial data and real-world decision-making. Rather than just reporting average outcomes, we’re asking a more granular question: for this particular patient, what’s the likely trajectory? Machine learning approaches are well-suited to this kind of individualized prediction, and applying them to quality of life outcomes—rather than just clinical endpoints—keeps the focus on what matters most to patients living with chronic skin disease.