Personalised Therapy Dermatology
Use Case Description
Psoriasis is a chronic autoimmune disorder characterized by persistent skin inflammation, resulting in red, scaly patches. Recognized by the World Health Organization (WHO) as a chronic and disabling non-communicable disease, psoriasis significantly impacts not only physical health but also psychological and social well-being.
Addressing psoriasis effectively requires more than just clinical intervention; it demands a holistic, personalized approach that integrates both physical and psychosocial care. However, achieving this comprehensive care presents several challenges:
Undertreatment and Mismanagement: Psoriasis is often underrecognized in its severity, leading to suboptimal treatment. Patients may receive treatments that do not adequately address the full spectrum of the disease, including its psychological and social ramifications.
Psychological and Social Impact: The visible nature of psoriasis can lead to significant psychological distress, including depression and anxiety. Social interactions and relationships may be negatively impacted, leading to a diminished quality of life. In severe cases, psoriasis can be life-ruining, contributing to a vicious cycle where stress exacerbates the condition.
Comorbidities: Psoriasis is associated with a range of comorbid conditions such as depression, anxiety, cardiovascular diseases, obesity, diabetes, and Crohn's disease. These comorbidities complicate treatment and necessitate a comprehensive, personalized approach to care.
To effectively manage psoriasis, a shift in the approach to treatment is essential. Current practices often fall short of addressing this full spectrum of the disease, leading to gaps in care that can significantly affect patients' quality of life. The complexity of psoriasis calls for a more tailored and holistic approach, an approach that should not only focus on alleviating the physical symptoms but also on the broader psychosocial dimensions of the disease. To achieve this, the following goals are proposed:
Optimization of Treatment Pathways:
Develop and implement treatment strategies that are tailored to individual patient needs, taking into account both clinical and psychosocial factors.
Shift towards personalized dosing therapy to optimize treatment efficacy and minimize side effects.
Ensure that treatment pathways are aligned with a treat-to-target approach, where the goal is to achieve and maintain specific clinical outcomes.
Holistic Management of Psoriasis:
Move beyond focusing solely on clinical symptoms to incorporate psychosocial aspects of psoriasis into treatment plans.
Address the psychological impact of psoriasis, including the management of associated conditions such as depression and anxiety.
Value Optimization:
Balance the cost of treatment with the outcomes achieved, ensuring that patients receive the most effective care without unnecessary expenditure.
Address the issue of over- and undermanagement by optimizing treatment plans to meet individual patient needs without under- or overtreating.
The expected outcome is to enhance the quality of life for psoriasis patients through comprehensive, personalized care that effectively manages the disease, reduces psychological burden, and optimizes healthcare resource use.
Background
Therapeutic drug monitoring in dermatology: the way towards dose optimization of secukinumab in chronic plaque psoriasis
Secukinumab is a biologic therapy with a favorable efficacy profile for treating psoriasis, yet clinicians often observe varying clinical responses among patients. These variations may be linked to issues of under- or overdosing, which can significantly impact treatment outcomes. Given the high cost of biologics like secukinumab, ensuring their rational and effective use is crucial. Therapeutic drug monitoring (TDM) offers a potential solution, guiding clinicians in making informed decisions about dose adjustments to optimize patient outcomes. This multicenter, prospective study was designed to address the need for a more tailored approach to secukinumab dosing. The study also aimed to identify a therapeutic window that could serve as a target for optimal dosing in psoriasis patients. This shows the potential benefits of TDM in managing secukinumab treatment and the obtained findings situate the use case in the broader context of personalized medicine, where the goal is to tailor treatment strategies to the individual characteristics of each patient, thereby enhancing the overall effectiveness and efficiency of biologic therapies in psoriasis management.
Publication:
[1] Soenen, Rani, et al. "Therapeutic drug monitoring in dermatology: the way towards dose optimization of secukinumab in chronic plaque psoriasis." Clinical and Experimental Dermatology 47.7 (2022): 1324-1336.
Results, Highlights & Research Outputs
Establishing Baselines for Next-Visit PASI Prediction Using the BIOLOPTIM Dataset
We investigated the prediction of next-visit PASI in patients with psoriasis as a first step toward developing baseline models for treatment response forecasting. The analysis was based on the BIOLOPTIM dataset, which follows patients receiving guselkumab through regular clinical visits. At each visit, PASI scores were recorded alongside additional clinical measures and patient-reported outcomes, including psychosocial assessments such as the DLQI.
Two complementary modeling strategies were explored. In the first, PASI was treated as a continuous outcome, with regression models developed to predict the next numerical value. In the second, PASI was categorized into clinically relevant severity levels (mild, moderate, severe), allowing the task to be framed as a classification problem. Models were trained using different sets of input features. Some relied only on baseline visit characteristics, while others incorporated longitudinal information from previous visits to capture temporal response patterns under treatment. This comparison enabled the establishment of a set of reference models for next-visit PASI prediction.
These results form a foundation for more advanced predictive modeling, with the aim of improving individualized treatment strategies and decision support in psoriasis management.
Publication:
[2] Chevaillier, Ayko; Avereyn, Warre; Steenwinckel, Bram; Soenen, Rani; Lambert, Jo; Van Hoecke, Sofie; Ongenae, Femke "Baseline and Longitudinal Modeling for Next-Visit PASI Prediction " (in preparation for submission in 2026).
Predicting therapy response in psoriasis: a Machine Learning approach based on demographic, clinical, psychosocial, treatment data from the IRIS trial
This work investigated the application of machine learning models to support personalized management of plaque psoriasis. The primary goal was to predict patient response to therapy, distinguishing responders from non-responders based on clinical outcomes such as PASI and DLQI scores. A secondary objective was to explore the potential influence of psychosocial factors, captured through patient-reported questionnaires, on treatment response.
Both linear regression and tree-based models were developed and evaluated. The modeling considered two complementary approaches: classification, to predict responder status, and regression, to estimate continuous disease measures. Models were trained using clinical, treatment, and psychosocial features, and evaluated across multiple time points to capture temporal dynamics in patient response.
Feature analyses were used to assess which variables, including psychosocial measures, contributed most to predictions. These efforts provide a first framework for integrating diverse patient data into predictive models and offer initial insights into the role of psychosocial factors in treatment outcomes, supporting more individualized decision-making in psoriasis care.
Publication:
[3] van den Bulck, Marie; Chevaillier, Ayko; Avereyn, Warre; Steenwinckel, Bram; Vyvey Emma; Soenen, Rani; Lambert, Jo; Van Hoecke, Sofie; Ongenae, Femke "Predicting therapy response in psoriasis: a Machine Learning approach based on demographic, clinical, psychosocial, treatment data from the IRIS trial" (in preparation for submission in 2026).
Applicability of Contextualized Models in PASI Prediction: Insights from the BIOLOPTIM Trial
This work investigated the potential of contextualized machine learning models to enhance predictive performance in psoriasis treatment outcomes. The primary objective was to evaluate whether contextualized Support Vector Machine (SVM) regression models could improve the PASI score predictions and enable early identification of patients likely to achieve a 75% reduction in PASI (PASI 75).
Models were developed for multiple biologics, including GUS, IXE, and SEC, and evaluated using regression metrics (R², MAE). Feature analyses highlighted distinct contributions: for GUS and IXE, performance was optimized by combining baseline PASI with demographic, temporal, and behavioral features, while for SEC, patient-reported outcomes from EQ5D and DLQI questionnaires proved most informative. Contextualized models for GUS were further applied to construct PASI response curves, predicting the number of treatment days required to achieve PASI 75.
These results demonstrate that contextualization provides added value in PASI prediction tasks, integrating diverse clinical, behavioral, and psychosocial features into robust predictive frameworks. The findings support the use of contextualized SVM models as a step toward more individualized treatment planning in psoriasis care, as suggested by prior work on contextual boosting for explainable SVM classification [4].
Publication:
[4] Loor, Marcelo; Tapia Rosero, Ana; De Tré, Guy "Contextual boosting to explainable SVM classification," in Fuzzy Logic and Technology, and Aggregation Operators, EUSFLAT, 14069 (2023), 480–491.
Outputs & Resources
Publications
Soenen, Rani, et al. "Therapeutic drug monitoring in dermatology: the way towards dose optimization of secukinumab in chronic plaque psoriasis." Clinical and Experimental Dermatology 47.7 (2022): 1324-1336.
Chevaillier, Ayko; Avereyn, Warre; Steenwinckel, Bram; Soenen, Rani; Lambert, Jo; Van Hoecke, Sofie; Ongenae, Femke "Baseline and Longitudinal Modeling for Next-Visit PASI Prediction" (in preparation for submission in 2026).
van den Bulck, Marie; Chevaillier, Ayko; Avereyn, Warre; Steenwinckel, Bram; Vyvey Emma; Soenen, Rani; Lambert, Jo; Van Hoecke, Sofie; Ongenae, Femke "Predicting therapy response in psoriasis: a Machine Learning approach based on demographic, clinical, psychosocial, treatment data from the IRIS trial" (in preparation for submission in 2026).
Loor, Marcelo; Tapia Rosero, Ana; De Tré, Guy "Contextual boosting to explainable SVM classification," in Fuzzy Logic and Technology, and Aggregation Operators, EUSFLAT, 14069 (2023), 480–491.