Risk and potential risk reduction in diabetes type 2 patients in Germany

Abstract

Avoiding serious complications such as stroke, myocardial infarction, and amputations in diabetes patients is the main interest of long-term treatment. Given the considerable prevalence of diabetes type 2 in industrialized countries this is a major public health concern as well as a burden to health care systems. The present study estimated the current risk of major complications occurring in the German diabetes type 2 population and explored the potential for further risk reduction. Risk reduction can be achieved when physiological and behavioral parameters (HbA1c, blood pressure, cholesterol level, body mass index, smoking) are set to target values recommended in guidelines. To estimate individual risk and potential risk reduction the multifactor disease model Mellibase was employed. Data were obtained from the German Health Survey of 1998, which includes a sample of 7,124 individuals representative of the German population. The survey shows a prevalence rate of 6.3% for diabetes type 2 in persons older than 35 years. The analyses reveal that the overall potential for risk reduction is moderate (e.g., the average reduction potential of the 10-year risk of stroke is 5.7%). A majority of parameter ranges found in the patient population are either already close to the recommended values (HbA1c), are not alarmingly higher than in the general population (blood pressure) or have little impact on risk reduction. In addition nonmodifiable risk factors such as duration of the illness and advanced age constrain possible improvements. However, there is a wide variation in the actual risk between individuals (e.g., the 10-year risk of stroke varies between 2.2% and 79.8%), and thus a wide variation in potential risk reduction (the risk reduction potential for stroke varies between 0% and 53.4%). Intensified treatment should therefore (a) focus on relevant subgroups of patients taking their risk reduction potential into account and (b) aim at improvement in the overall metabolic profile rather than concentrating on single risk factors.

Appendix: Mellibase disease model description

The Institute for Medical Informatics and Biostatistics (IMIB, Basle) began developing the type-1 and type-2 diabetes models in 1996 to perform risk stratification in the field of health insurances [5]. Adapted versions of this diabetes model were used for the assessment of oral antidiabetic medications in several European countries [3, 14]. The model used in the present study reflects published medical evidence as identified by a regular systematic review. The cohort analysis version of the IMIB model was used in referenced publications. The latest version of the model is called Mellibase. This is based on the IMIB model and allows automated individual risk analysis.

Mellibase is a Markov-based calculation engine with transition matrices and underlying variables, some of which are time-dependent. The diabetes model reflects the course of diabetes mellitus and its associated long-term micro- and macrovascular complications in a range of relevant submodels including the complications of nephropathy, retinopathy, acute myocardial infarction, stroke, and amputation. Simulations are performed for a certain number of years or until death. In each yearly cycle the number of events (including death) and costs associated for each patient are estimated. Surviving patients enter the model in the next cycle, and this process is repeated until the end of the analysis or elimination of the cohort. The results are then combined to estimate the cumulated events and total treatment costs as well as incremental cost-effectiveness for each option considered. The probabilities of developing events are considered to be time, age, gender, and state specific. The events are defined on the basis of key medical prognostic factors such as control of blood glucose levels, hypertension, and hyperlipidemia. The applied transition matrices are based on logical and chronological decision trees, which in a submodel comprise the complete history of a disease-related complication, starting with initial risk profiles, including intermediate outcomes, and leading to the endpoints of each submodel. This approach offers transparency since the model is available in software form with a graphic user interface. The model was published in detail in peer reviewed journals [5, 6, 15, 16].

The diabetes model is based on a summary of data representing epidemiological and clinical evidence from studies such as the United Kingdom Prospective Diabetes Study and many other prospective clinical trials, together with diabetes registery data and meta-analyses appraising diabetes treatments as well as associations between HbA1C levels and a range of micro- and macrovascular events. It is also based on data from the Framingham studies, which assess the relationship between lipid profile and coronary heart disease.