Evaluating the Impact of Different Perspectives on the Optimal Start Time for Statins Jennifer E. Mason Edward P. Fitts Department of Industrial and Systems Engineering, NC State University, Raleigh, NC Nilay D. Shah, Brian T. Denton, Murat Kurt, Steven Smith
Diabetes 20.8 million people have diabetes in the US 7% of the US population 90% of diabetics have type 2 diabetes Two out of three people with diabetes will die from either stroke or coronary heart disease (CHD). HMG Co-A reductase inhibitors (statins) are an important part of treatment for preventing stroke and CHD events for patients with Type 2 diabetes.
Statins Statins can reduce Total Cholesterol by up to 24% and increase HDL by up to 8% 1, greatly reducing the risk of cardiovascular events. Patients who have already suffered a heart attack or stroke can reduce the risk of future events by starting statins after an event has occurred. 2 1 Maron DJ, et al., Current Perspectives on Statins. Circulation. 2000;101:207-213. 2 http://www.bhf.org.uk/news_and_campaigning/press_office/latest_news views/tonight_statins_show.aspx
Optimization Models Objective Function Decision Variables Constraints An optimization model seeks to find values of the decision variables that optimize (maximize or minimize) the objective function subject to the given constraints.
Stages: Markov Decision Process Model Decision Horizon: Ages 40-80 Post-decision Horizon: Ages 80-100 40 41 42 43 79 80 81 82 100 Decision Horizon Post-decision Horizon States: Metabolic: Total cholesterol and HDL (each can be L, M, H, V) Actions: Event: Either a stroke or CHD event has occurred Before an event has occurred, the decision to initiate or delay statin treatment must be made.
State Transition Diagram r(e,i) Non-Fatal Events (On Statins) On Statins r(s,i) Metabolic States before an event has occurred. Action to initiate or delay statins is taken here. r(l,w) L r(m,w) M H r(h,w) V r(v,w) r(d,d) Death
Transition Probabilities Among metabolic states Diabetes Electronic Management System (DEMS) For an event UKPDS cardiovascular risk models For death from other causes CDC Mortality tables
Total Cholesterol Level Metabolic States High-density Lipoprotein Level L / L L / M L / H L / V M / L... V / V TC and HDL have four possible levels each, so there are 16 states in all.
Total Cholesterol Values (mg/dl) L M H V < 160 160 200 200 240 > 240 High-density Lipoprotein Values (mg/dl) L M H V < 40 40 50 50 60 > 60
Perspectives Society Maximize a weighted combination of patient rewards for life years minus costs of treatment and health services Patient Maximize quality adjusted lifetime until an event Third-party Payer Minimize costs of treatment and health services
Society This objective function includes a reward for the QALYs gained, where quality adjustment factors are used to reflect the impact of nonfatal cardiovascular events and the side effects of medication. Objective: Maximize rewards costs. One-time Costs Weighted Reward Follow-up Costs Statins Cost
Weighted Reward for the Society Perspective Stroke Decrement Factor Statins Decrement Factor Reward in Dollars CHD Decrement Factor
Patient No costs are considered in this model -- only the patient s QALYs are considered. Before you have an event, a year of life is worth 1 if not on statins and 0.97 if on statins. Objective: Maximize QALYs before an event.
Third-party Payer The patient s quality of life is not considered here at all. The third-party payer s costs are all that are considered. Objective: Minimize Costs
Value functions The objective is to determine the action a t that optimizes the value function at each state s t : The following optimality equation is solved at each decision epoch using backwards induction:
Age Optimal Start Times - Males (statins decrement = 0.03) 80 75 70 65 60 55 50 Patient Society Third-party Payer 45 40 35 30 V/L H/L M/L L/L V/M H/M M/M L/M V/H H/H M/H L/H V/V H/V M/V L/V TC/HDL
tatins decrement = 0.03) NC STATE UNIVERSITY Age Optimal Start Times - Males (statins dec 80 75 70 65 60 Optimal Start Times for Males with very high total cholesterol and low HDL. Statins utility decrement = 0.03. 55 50 45 Patient Society Third-party Payer 40 35 30 V/L H/L M/L L/L V/M H/M M/M L/M V/H H/H M/H L/H
Age Optimal Start Times - Males (statins decrement = 0.03) 80 75 70 65 60 55 50 Patient Society Third-party Payer 45 40 35 30 V/L H/L M/L L/L V/M H/M M/M L/M V/H H/H M/H L/H V/V H/V M/V L/V TC/HDL
Age Optimal Start Times - Males (statins decrement = 0.02) 80 75 70 65 60 55 50 Patient Society Third-party Payer 45 40 35 30 V/L H/L M/L L/L V/M H/M M/M L/M V/H H/H M/H L/H V/V H/V M/V L/V TC/HDL
Age Optimal Start Times - Females (statins decrement = 0.03) 80 75 70 65 60 55 50 Patient Society Third-party Payer 45 40 35 30 V/L H/L M/L L/L V/M H/M M/M L/M V/H H/H M/H L/H V/V H/V M/V L/V TC/HDL
Age Optimal Start Times - Females (statins decrement = 0.02) 80 75 70 65 60 55 50 Patient Society Third-party Payer 45 40 35 30 V/L H/L M/L L/L V/M H/M M/M L/M V/H H/H M/H L/H V/V H/V M/V L/V TC/HDL
Conclusions The optimal ages to start statin treatment vary depending on which perspective is being considered. For each perspective and statins decrement factor, the earliest optimal age for women is much later than for men. Under some perspectives it is optimal for low risk patients to start statins relatively late in life, as age is a strong predictor for cardiovascular events. The society and patient perspective optimal start times are very sensitive to the disutility factor for statins.
Related Projects Modeling the optimal start times for statins using different adherence levels and the associated distributional medication effects to account for imperfect adherence. Modeling the optimal start times for blood pressure medications with stochastic blood pressure states. Modeling the optimal start times for statins and ace inhibitors with stochastic TC, HDL, and blood pressure states. We are in the process of adding the event of kidney failure to the blood pressure models.
Thank You! A special thanks to those that helped make my internship and this research possible: Dr. Doug Wood, Nilay Shah, Brian Denton, Dr. Steve Smith, Hari Balasubramanian, Todd Huschka, Jason Egginton, and Murat Kurt.