Methods: Using data from the Mayo Clinic electronic medical record we construct a discrete time Markov process that defines the progression of patients through health states defined by cardiovascular events, and metabolic factors including: total cholesterol, high density lipoproteins (HDL), systolic blood pressure, and HbA1C. Our model assumes yearly decision epochs at which statin treatment may be initiated between ages 40 and 80 years. The model seeks to maximize a patients quality adjusted time until a cardiovascular event. We consider negative effects of statin treatment based on conservative assumptions about decrements to quality adjusted life years. To reflect the existence of multiple risk models, the Markov process assumes a probability range for CHD and stroke probabilities based on the upper and lower bounds defined by the three risk models mentioned above. The decision of when to intiate statin therapy over the course of a patient's lifetime is made to minimize the worst case across each of the cardiovascular risk models. Therefore the optimal timing of therapeuthic intervention based on our model is robust against uncertainties in predicted cardiovascular risk.

Results: We find that the mean start times of statin therapy for male and female patients, across all metabolic states, are 47 and 48 years respectively . The earliest start time, for the highest risk metabolic states, is age 40 for males and females. The latest start time, for the lowest risk metabolic states, is age 67 for males. For female patients that remain in the lowest risk state throughout their lifetime it is never optimal to initiate statin therapy.

Conclusions: Patient gender and metabolic state play an important role in the decision to initiate treatment to manage cardiovascular risk. Under conservative assumptions of our model, which considers the three most popular cardiovascular risk models for diabetes patients, there is significant variation in the optimal start time of statin therapy which depends on gender and metabolic factors.