DEVELOPMENT OF AN OPTIMAL POLICY FOR STATIN INITIATION AMONG TYPE 2 DIABETES PATIENTS

Development of an Optimal Policy for Statin Initiation Among Type 2 Diabetes Patients

Purpose: Statins can be used to treat lipid abnormalities in the primary prevention of CHD and stroke events for patients with Type 2 diabetes, but may have some negative-side effects. We investigate the problem of optimal timing of statin initiation using a sequential decision model that maximizes the patients quality-adjusted life years (QALYs) prior to their first CHD or stroke events.

Methods: We develop a discrete-time Markov decision process (MDP) model and describe the state of the process by the patient's lipid-ratio (LR) levels. We also consider other risk factors including the patient's age at diagnosis, gender, ethnicity and smoking status, systolic blood pressure (SBP) and HbA1c levels, and assume the latter two evolve deterministically. We model the effect of using statins by a treatment-effect factor that denotes the relative reduction in LR levels by treatment, and reflect the negative side-effects of statins by decrementing the patient's quality of life by a quality-adjustment factor. We calibrate our MDP model using longitudinal data from Mayo Clinic and use the UKPDS risk model to predict the patient's CHD and stroke probabilities.

Results: We observe that a patient's optimal policy follows a threshold structure where the optimal decision is to initiate treatment if and only if the patient's LR level falls into an interval which is above a certain threshold, and wait for a year, otherwise. As illustrated in Figure 1, the patients' optimal LR-range thresholds decrease with age and the optimal policy is more aggressive for male patients than for female patients: male patients do not have higher LR-range thresholds than the females at any age. We also compute the patients' expected gains in QALYs from following the recommended policies rather than initiating treatment immediately at the time of diagnosis (Table 2). We observe that female patients have higher gains than males in all LR-ranges and the patients gain less as their LR increase.

Conclusions: Patient's gender and LR-progression rates play an important role in determining when to initiate statin treatment. Given the factors that describe the patient's CHD and stroke risk profiles, our MDP model can be used as a guide for deciding whether to initiate or delay statin treatment.