A machine learning (ML)-based approach accurately predicts the risk for discontinuation of biologic therapy for psoriasis within 5 years of treatment, according to a research letter published in the Journal of the American Academy of Dermatology.
Researchers compared the accuracy of ML with a traditional statistical risk factor-based model for predicting the 5-year probability of biologic discontinuation in patients with psoriasis. Data were obtained from a Danish registry with 3388 patients.
The biologic agents analyzed were adalimumab, etanercept, guselkumab, infliximab, ixekizumab, secukinumab, and ustekinumab. Variables included age, sex, body mass index, age at diagnosis, age at first biologic prescription, previous biologic exposure, concurrent psoriatic arthritis, concurrent methotrexate therapy, comorbidities, baseline Psoriasis Area Severity Index (PASI) score, and baseline Dermatology Life Quality Index (DLQI) score. Hazard ratios were computed for the predictive factors using Cox regression analysis, with biologic discontinuation as the outcome and adalimumab as the reference value.
The machine learning models included Generalized Linear Model, Naive Bayes, Deep Learning, Decision Tree, Random Forest, and Gradient Boosted Trees. Model performance was evaluated with use of the area under the receiver operating characteristic (AUROC) curve.
Ustekinumab and ixekizumab were associated with the lowest risk for biologic discontinuation compared with adalimumab, and etanercept had the highest risk for discontinuation. Previous exposure to biologic therapy and sex were also significant variables. Men with no previous biologic exposure had the longest drug survival. Weight and baseline PASI score were statistically significant predictors, although the effect was negligible compared with the other predictors.
Using adalimumab as a reference, hazard ratios (HR) for variables included in the nomogram for predicting probability of biologic agent survival in Cox regression analysis were as follows:
- Etanercept (HR, 1.59; 95% CI, 1.34-1.88, P <.001);
- Guselkumab (HR, 0.45; 95% CI, 0.20-1.01; P =.05);
- Infliximab (HR, 1.32; 95% CI, 1.07-1.64; P =.01);
- Ixekizumab (HR, 0.63; 95% CI, 0.39-0.97; P =.04);
- Secukinumab (HR, 0.91; 95% CI, 0.74-1.12; P =.37);
- Ustekinumab (HR, 0.61; 95% CI, 0.52-0.72; P <.001)
- Biologic-naivete (HR, 0.72; 95% CI, 0.64-0.82; P <.001);
- Sex (HR, 0.77; 95% CI, 0.68-0.87; P <.001); and
- Weight (HR, 1.00; 95% CI, 1.00-1.00; P =.02).
The AUROC curve was 0.61, indicating a low discriminatory value.
All ML algorithms predicted the likelihood of biologic discontinuation within 5 years with high accuracy, ranging from 65.3% for Naive Bayes to 77.5% for Gradient Boosted Trees. The most efficient ML algorithm predicted treatment outcome with less than 23% classification error, using only basic patient information generally available to clinicians.
The most important predictive parameters were biologic drug (weight, 0.247), patient sex (weight, 0.076), and body weight (weight, 0.069). The AUROC curve for the Gradient Boosted Trees was 0.85, which indicated excellent performance.
“Ultimately, a machine learning-based approach, more so than a traditional statistical model, accurately predicted the risk of discontinuation of biologic therapy within 5 years of treatment based on simple patient variables available to dermatologists in clinical practice,” the researchers conclude. “The results of our study could fulfill an unmet need to predict the long-term effectiveness of biologics in psoriasis patients.”
Disclosure: Some of the study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures.
Du AX, Ali Z, Ajgeiy KK, et al. Machine learning model for predicting outcomes of biologic therapy in psoriasis. J Am Acad Dermatol. Published online January 30, 2023. doi:10.1016/j.jaad.2022.12.046