Slide 1
It is the aim of this presentation to examine the cost-effectiveness of primary prevention with the focus on statins.
In case of questions or remarks please contact:
Guido Wendland, Statistician
Universität zu Köln (University of Cologne)
Lindenthalgürtel 22
50935 Köln (Cologne)
Tel.: +49-221/478-6467
Fax.: +49-221/478-6675
e-mail: guido.wendland@medizin.uni-koeln.de
Slide 2
The Status-Quo model reflects the current status of primary prevention with statins.
Model 2 can be regarded as total prevention of CHD and preludes Model 3. Modes 3 investigates the effect of prevention on the health care expenditures and life years.
Slide 3
We present a flexible and comprehensive approach, which incorporates some new methodological aspects and data sources.
It is assumed in this short presentation that the treatment is continued until death and it incurs benefits and costs over the whole time span.
Slide 4
The Status-Quo model follows a cohort of persons in the "no CHD" state over time. In this Markov Model circles are health states (with different costs) and the branches reflect transitions with a certain transition probabitity. The CHD-state comprises of MI, coronary insufficiency and stable angina. There is evidence that the underlying type of state does only moderately affect future mortality.
The estimates of transition probabilities and costs for different health states form the foundation of the model. The Markov model can be represented in various forms. Here the emphasis is on a cohort simulation.
Slide 5
Mortality is based on the Gompertz Function R(x) = R(0) exp(bx), where x is the age, and R(0) and b are the two parameters of the Gompertz function.
Dots and circles represent the figures of the Life Table. The two parameters of the Gompertz-Function R(0) and b were found by non-linear regression.This figure illustrates the excellent agreement between 30-90 years of age with the German Life Table from 1998.
In the model, the mortality of a "CHD-cohort" is found by a different parameter set. It is assumed that parameters sets which reflect different prognoses of CHD-patients can all be found along the straigth line ln(R(0)) =a*b + c, where "a" is the slope parameter and "c" is a constant. This assumptions around the Gompertz-function form the mathematical framework of survival in the model.
Slide 6
This slide shows how incidence of CHD was estimated. The individual CHD-risks were based on the German Health Survey Data of 1998. The Framingham formula of Anderson (1991) was used to derive this risk. Several publications show a good transferability of Framingham results to a German population.
Here dots and circles are mean values for each age group and the lines were fitted on individual data of 7124 representative persons.
The percentage of incident cases who die was also estimated using Framingham data. This portion increases with age.
Slide 7
In the elimination model, CHD is completely excluded as cause of death. Those previously dying from CHD now survive. However, the model allows for a higher mortality of these survivors, reflecting a higher level of risk factors in this group. It is plausible that particular risk factors for CHD such as diabetes and smoking also increase mortality from other causes. This can be incorporated in a higher transition probability for the "prognosis of eliminated CHD"-branch as compared to the "mortality"-branch.
Slide 8
The prevention model is somewhat a mixture between the Status-Quo model and the Elimination model (Model 1 & 2). The numbered branches show new transitions which have to be estimated.
1) Reduction of CHD-events by prevention: Estimates from WOSCOPS and AFCAPS/TexCAPS.
2 & 3) CHD-Risk of prevented cases: On the one hand, prevented cases can probably regarded more susceptible to CHD-events. On the other hand, these patients are also susceptible to prevention. The effect of the chosen assumption can be investiged in sensitivity analyses.
4) Other mortality of prevented cases: Like in the Elimination model, the other mortality is assumed to be increased, due to an overlap of risk factors such as smoking and diabetes.
Slide 9
This figures displays data of the statutory health insurance. The "CHD"-line and "no CHD" line reflect real data on the sum of hospital, medication and sickness benefits expenditures, which make up roughly 50% of all expenditures.
The dashed line sketches the expenditures of prevented cases. In concordance with their increased mortality it is also assumed that this group incurs higher health care expenditures as the "no CHD" group, e.g. because diabetes is more prevalent in this group. It was assumed that the increase was 30% of the increase of the CHD group.
For the model calculations the displayed data were extrapolated to include all health care expenditures of the statutory health insurance. It was assumed that expenditures for dentists were the same in all groups, and the same percentage increase in "CHD group" was assumed for physicians and "other health care expenditures".
Slide 10
This slide shows the results of elimination model with respect to life-years gained. The age marks the starting age of the observed cohort.
The calibration model shown here is an alternative model in which the CHD-incidence was estimated such that the percentage of persons dying from CHD in the model agrees with the current figures from the Statistical Yearbook. Except for the incidence, all other parameters were held constant. It indicates here that especially for younger persons the CHD-risk estimates might be too high for the whole population but too low for high age groups. However, the data from the Statistical Yearbook might be subject to bias. Apart from that, the future trend can be different from current status. Another possible calibration parameter is the prevalence of CHD, but there are no firm estimates of CHD-prevalence.
Slide 11
If CHD is eliminated, the corresponding health care expenditures would be saved as well. However, people would also live longer (previous slide) and incur health care costs attributable to other diseases.
If younger persons are observed over lifetime CHD incurs additional costs at 3% discount rate. If persons older than 50 years are followed over time, the presence of CHD prevents health care costs for the statutory health insurance.
Slide 12
To present Cost-Effectiveness (CE) results for subgroups we devided the whole population in decentiles of risk (baseline model) and fitted polynoms to the corresponding data which show the age-dependency of the risk.
Slide 13
The CE-Ratio depends strongly on baseline risk. A population approach would increase life expectancy by 0.33 years at an additional cost of 5300$ per person which yields a CE-Ratio of 16,000$. The ten-year risks for the decentiles are (1: 2.7%, 2: 4.2%, 3: 5.5%, mean: 6.6%, 7: 9.8%, 8: 12.8%, 9: 15.6%, 10: 23.3%).
Slide 14
The horizontal axis shows the variation of the sensitivity parameter, the vertical axis the outcome arameter, here the CE-ratio.
All parameters of the model have been subjected to sensitivity analyses. Sensitive parameters are also the costs of statins (300$ in baseline scenario) and the discount rate. An increase of discount rates also increases the CE-Ratio. Further sensitive parameters are baseline risk and age (-> next slide).
Slide 15
This slide shows the interaction of Basline risk and age on the CE-Ratio, which is particularly strong in the basline model, but still present in the calibration model.
Females reach the same average risk 10-15 years later as compared to men in the baseline model and the calibration model. The time gap between men and women increases with age. At the same risk and age CE-Ratios for men and women are comparable.
