Subgroup Specific Medicine: Finding Patients With Poor Therapy Response

Abstract

Personalized medicine aims at a better medical treatment by tailoring the treatment for the patient’s individual characteristics. One step towards personalized medicine is subgroup specific medicine that adjusts the treatment for groups of patients. This work aims at detecting patient subgroups, which react worse to a specific medication compared to the rest of the population. Therefore, we propose a framework that finds characteristics of patients that correlate with weak therapy results. We used the health claims database of Arvato Health Analytics that contains diagnosis codes and prescriptions of 3 million German insurants for the years 2008-2015. First, we selected all patients suffering from rheumatoid arthritis (ICD-10 GM: M05, M06.0) and their respective medication intervals. We implemented a quality of life (QoL) metric for each interval based on the number of emergencies, admissions, side effects and outpatient/inpatient visits. Then, we applied subgroup mining to identify patient groups with significantly worse QoL result. Finally, we explored the characteristics that lead to this worse QoL outcome. Our analysis included n=36,756 RA patients. For female RA patients, the biologicals Etanercept (μQoL=1.05, p=0.03), Infliximab (IFX) (μQoL=0.93, p=0.003) and Adalimumab (μQoL=0.97, p=0.002) showed a poor QoL outcome. In contrast, Golimumab had no significant worse impact on the QoL of the female patients (μQoL=0.9, p=0.83). According to the product information of IFX, it is advisable to use IFX concomitant with Methotrexate (MTX). Although this has not been verified in clinical studies, our algorithm identified that RA patients using MTX react better to IFX than patients without MTX (μQoL=1.12, p=0.009). Furthermore, we determined poor QoL scores for patients >50 years taking Etanercept (μQoL=1.05, p=0.008) and Adalimumab (μQoL=1.02, p=0.0009) that has not been reported in literature yet. By this approach, we can guide the development of new drugs for the identified patient subgroups that react poor to approved medication.