Managing People at High Risk for Diabetes

IN RESPONSE:

We are pleased to address the issues raised by Dr. Herman and his colleagues. Concerning glycemic progression, they are correct that we used UKPDS data to help build our model. We see that as a strength, not a weakness. The resulting model accurately simulates the progression of diabetes across the entire time horizon of our analysis—from impaired fasting glucose and impaired glucose tolerance (it independently predicted the rates in the Diabetes Prevention Program) through to clinical diabetes and for at least 15 years thereafter.

Dr. Herman and associates argue for a faster rate of glycemic progression. They assume that everyone progresses from “diabetes onset” to “clinical diabetes” in exactly 10 years (1), which implies a rate that is greater than than twice that seen in the UKPDS. We see several problems with this assumption, beyond the fact that it contradicts the UKPDS. They based their assumption on 2 papers (2, 3) that estimated that retinopathy first begins to appear in populations about 4 to 7 years before clinical diagnosis. These papers, in turn, cited a study that followed 30 people after onset of diabetes and noted that the first case of retinopathy appeared 5 years after onset (4). The 10-year assumption comes from adding 5 years (from onset to retinopathy) and 4 to 7 years (from retinopathy to clinical diagnosis). First, 5 years was the shortest time from diabetes onset to retinopathy. The study actually showed a wide variation, and retinopathy developed in only 8 of the 30 (27%) patients after a minimum of 10 years of follow-up. Second, 4 to 7 years is the time of the first case in a sample. The authors appear to have mistakenly interpreted these values to mean that retinopathy appears 4 to 7 years before clinical diabetes in everyone. In fact, the time of occurrence of retinopathy varied widely and was 10 to 20 years after clinical diagnosis on average. These papers actually contradict a fixed 10-year transit time.

Dr. Herman and colleagues point to the frequency of early retinopathy in people in the Diabetes Prevention Program after onset of diabetes (13% within 3 years) and suggest that the complications we calculated are too low. The rates we reported were for advanced retinopathy in people who did not have diabetes at the start of the calculation. We calculated that 8% of people with prediabetes developed early retinopathy within 3 years, which fits with the estimate of 13% in people who have diabetes. We also note that our model matches quite closely the progression of retinopathy seen in the UKPDS and other trials (5).

Concerning myocardial infarctions, the 12% and 50% values are measuring entirely different things and are not comparable. To evaluate the accuracy of our model for calculating cardiovascular events, we recommend looking at the rate actually seen in the Diabetes Prevention Program after 3 years, 7.3 events per 1000 person-years (6). We calculated 8.4 events per 1000 person-years. About 40% of people with diabetes die of ischemic heart disease (7). For the population from the Diabetes Prevention Program, we calculated 28%, which is right on target given the differences in populations and time horizons. Additional validations of myocardial infarction rates in 12 clinical trials, including UKPDS and a prospective prediction of CARDS (Collaborative Atorvastatin Diabetes Study) are described elsewhere (1, 5).

We can correct other misunderstandings. First, we did not assume that glycemic progression is a “linear process”; in our model, progression varies from person to person and is not linear even for any particular person. Second, we did not “constrain” diabetes progression nor did we fix HbA_1c levels at 6.6%. The physicians in our simulation followed American Diabetes Association guidelines and treated people to achieve a target of 7% (making the average less than 7%). From that point on, we assumed the degree of control would gradually deteriorate as seen in the intensive care group of the UKPDS. Third, we used the same model that had been validated against 18 clinical trials (5); those validations do apply to this analysis. Fourth, we have validated the model for analyzing prediabetes; we did a blinded validation against the Diabetes Prevention Program itself (1, 5). Fifth, we have searched the literature again and still cannot find any validations of the model that Dr. Herman and his colleagues used. Specific references might help.

In addition to the issues that the authors raise in their letter, other aspects of their model may help explain why they achieved different results. Some examples from their paper and technical report (2) are as follows: HbA_1c levels increase at a constant annual rate of 0.2% during clinical diabetes (technical report, page 22; note that the rate was 0.11% in the UKPDS); at the beginning of clinical diabetes, no one yet has any signs of retinopathy (technical report, page 88); the annual probability that a person with prediabetes progresses to diabetes does not depend on how long they have had impaired fasting glucose or impaired glucose tolerance (paper, page 324); lifestyle reduces the probability of progressing from prediabetes to diabetes by a fixed 58% (paper, page 324; note that for patients in the Diabetes Prevention Program, the effect declined steadily over time and was about 44% after 4 years of follow-up); costs are multiplicative (paper, page 326); retinopathy and neuropathy do not affect costs (paper, page 326); blood pressure and levels of total serum cholesterol and high-density lipoprotein can only be categorized as “normal” or “above normal” (technical report, pages 39 and 43); risk for coronary heart disease in people with early (preclinical) diabetes is calculated from the UKPDS risk engine (paper, page 325), which was designed to be used after a patient develops clinical diabetes; and the annual risk for end-stage kidney disease does not depend on how long someone has had diabetes, clinical nephropathy, or blood pressure levels (technical report, page 12). These and other assumptions, some of which are summarized in the appendix to our paper, may deserve reconsideration.