1. Continuous-time Markov models to assess the effects of obesity, smoking and cholesterol- lowering drugs on heart disease and expected future lifetime Howard Waters Joint work with: Tushar Chatterjee & Angus Macdonald Heriot-Watt University, Edinburgh Seminar: University of Barcelona, 5 May 2008

2. Model Motivation: Actuaries need to be able to predict future lifetimes. Future lifetimes will depend on: Lifestyle: smoking, obesity, … Medical advances: cholesterol-lowering drugs, … ….

3. Model Objective: Construct and parameterise a continuous-time Markov model for a person’s lifetime from age 20 to death. Individual states represent combinations of risk factors for heart disease and also incidents of heart disease and stroke.

4. Model Data –Framingham Heart Study (1948 – 1998) –Health Survey for England (2003) Risk factorRisk FactorLevels AgeBody Mass Index (BMI)5 SexDiabetes2 SmokingHypertension4 Hypercholesterolaemia4

5. Model Technicalities: Model specified by transition intensities between states Transition intensities estimated using Occurrence/Exposure rates Significant factors determined using Generalised Linear Models Probabilities calculated by solving the Kolmogorov equations Model has 1761 states 2 (diab) x 4 (h’tens) x 4 (h’chol) x 5 (BMI) x 11 (IHD+stroke) + 1 (dead)

6. Application - Smoking Direct risk factor for: IHD Stroke Mortality Models for relative risk: Depend on smoking profile: current smoker, never smoked or duration since stopping Do not depend on sex or age Do not depend on number of cigarettes smoked Do not depend on years as a smoker

7. Application - Smoking

8. Expected future lifetime from age 20 Expected future `Event free’ lifetime from age 20 Smoking profileMaleFemaleMaleFemale Never smokes58.662.453.258.2 Gives up at age 3057.261.251.856.9 Gives up at age 5055.860.250.055.7 Gives up at age 7052.757.846.853.0 Always smokes51.556.145.851.4

9. Application - Statins Hypercholesterolaemia: direct risk factor for IHD and mortality, but not stroke. Statins: drugs which lower cholesterol (LDL) reduce the relative risk of IHD and stroke whatever the initial level of LDL were first licensed in 1987 have been developing since then (Rosuvastatin 2003) have been widely prescribed since the mid-1990s

10. Application - Statins Standard dose: reduction in relative risk Data: Wald and Law (BMJ 2003) Stroke: 17% IHD (MI): Year of treatmentReduction in relative risk % 119 239 3 - 551 6 +55

11. Application - Statins Treatment protocol: Males aged 50+, Females aged 60+ (Prof R Boyle, National Director for Heart Disease and Stroke, July 2007) Expected future lifetime from age 20* Expected future `Event free’ lifetime from age 20* MalesFemalesMalesFemales Treated59.262.754.358.8 Untreated58.662.453.258.2 * Never smokes

12. Application - Obesity Body Mass Index = (Weight kgs)/(Height mtrs) ² CategoryBMI range Underweight BMI ≤ 18.5 Lightweight18.5 < BMI ≤ 25 Overweight25 < BMI ≤ 30 Moderately obese30 < BMI ≤ 40 Morbidly obese40 < BMI

13. Application - Obesity Prevalence of obesity (BMI > 30 kg/m 2 ) in England for males

14. Application - Obesity BMI is a direct risk factor for: Diabetes Hypertension Mortality BMI is an indirect risk factor for: IHD (through diabetes and hypertension) Stroke (through hypertension) Mortality (through diabetes, hypertension, IHD and stroke)

15. Application - Obesity

17. Three models for changes in an individual’s BMI All matching prevalence in England in 2003 I No calendar time trend II Matches prevalence in 1994 and changes to 2003, with changes continuing for a further 20 years III Matches prevalence in 1994 and changes to 2003, with changes continuing indefinitely

18. Application - Obesity Predicted prevalence (%) of obesity for females starting from age 20: Age 20406080 England 200313.122.227.926.3 Model I13.123.329.930.7 Model II13.160.482.380.0 Model III13.160.493.597.5

19. Application - Obesity Expected future lifetime and prevalence of diabetes, hypertension and IHD/stroke at age 70: males, non-smokers, starting from age 20 BMI modelEFL years Diabetes % H’tension % IHD/stroke % I58.612.656.019.2 II58.217.464.120.1 III57.918.565.420.2

20. Papers Chatterjee T, Macdonald A S & Waters H R A model for ischaemic heart disease and stroke I: The model II: Modelling obesity III: Applications www.ma.hw.ac.uk/~andrewc/actuarial/#papers