With use of machine learning, researchers developed a predictive tool that may identify patients with ulcerative colitis (UC) who would not achieve steroid-free clinical remission (SFCR) at week 22 during vedolizumab (VDZ) therapy, according to a study in Scientific Reports.
The investigators retrospectively obtained clinical data at baseline (week 0) and assessed the clinical activity of UC at week 22 in 34 patients who began VDZ for the induction of remission at a hospital in Japan between September 2019 and April 2020. Patients underwent blood testing at week 0, and were examined at week 22 (training cohort, cohort 1).
In addition, 35 patients with UC who initiated VDZ for the induction of remission at a medical center from January 2019 to June 2020, underwent blood testing at week 0, and were examined at week 22 (cohort 2).
Cohort 1 included 25 men and 9 women with a median age of 37 years (range, 17-92 years), and their median disease duration was 5.0 years (range, 0.1-31.0 years). Cohort 2 included 22 men and 13 women with a median age of 42 years (range, 17-90 years), and their median disease duration was 7.2 years (range, 0.6-38.0 years).
Data regarding 49 clinical features for cohort 1 were collected and included sex, age, height, body weight, body mass index (BMI), disease duration, disease type (inflammation distribution), treatment history for UC, clinical activity, endoscopic activity, and 25 blood test items.
The 8 leading clinical features — partial Mayo score, mean corpuscular hemoglobin, BMI, blood urea nitrogen, concomitant use of azathioprine, lymphocyte fraction, height, and C-reactive protein — at week 0, according to random forest findings, were selected for logistic regression to predict achievement or no achievement of SFCR at week 22.
At week 22, 18 patients in cohort 1 (52.9%) and 13 patients in cohort 2 (37.1%) achieved SFCR. No patients in either cohort terminated VDZ treatment due to adverse events.
When the top 8 clinical features were assessed using random forest findings, the predictive accuracy was 100% in cohort 1 and 68.6% in cohort 2. The analysis showed that the positive predictive value was 54.5%, and the negative predictive value was 92.3% in cohort 2.
The investigators noted that their study was limited by its small sample size and that larger training and test cohorts are needed to improve the prediction model and its accuracy.
“The concept and findings in this study will promote personalized medicine in UC, and they could possibly be extrapolated to other medications and diseases,” the researchers commented.
Disclosure: Some of the study authors declared affiliations with pharmaceutical companies. Please see the original reference for a full list of authors’ disclosures.
Miyoshi J, Maeda T, Matsuoka K, et al. Machine learning using clinical data at baseline predicts the efficacy of vedolizumab at week 22 in patients with ulcerative colitis. Sci Rep. 2021;11(1):16440. doi: 10.1038/s41598-021-96019-x
Source: Medical Bag https://www.medicalbag.com/home/news/machine-learning-tool-may-successfully-predict-sfcr-promote-personalized-medicine-in-uc/