We Can Change the Natural History of Type 2 Diabetes

Lawrence S. Phillips; Robert E. Ratner; John B. Buse; Steven E. Kahn


Diabetes Care. 2014;37(10):2668-2676. 

In This Article

Abstract and Introduction


As diabetes develops, we currently waste the first ~10 years of the natural history. If we found prediabetes and early diabetes when they first presented and treated them more effectively, we could prevent or delay the progression of hyperglycemia and the development of complications. Evidence for this comes from trials where lifestyle change and/or glucose-lowering medications decreased progression from prediabetes to diabetes. After withdrawal of these interventions, there was no "catch-up"—cumulative development of diabetes in the previously treated groups remained less than in control subjects. Moreover, achieving normal glucose levels even transiently during the trials was associated with a substantial reduction in subsequent development of diabetes. These findings indicate that we can change the natural history through routine screening to find prediabetes and early diabetes, combined with management aimed to keep glucose levels as close to normal as possible, without hypoglycemia. We should also test the hypothesis with a randomized controlled trial.


Diabetes is the major cause of kidney failure, blindness, and nontraumatic leg amputations in U.S. adults and a leading cause of stroke and heart disease.[1] Diabetes cost the U.S. $176 billion in direct costs in 2012, amounting to 11% of U.S. health care dollars,[2] and is a critical public health problem in other countries as well.[3] Moreover, National Health and Nutrition Examination Surveys (NHANES) from 2003–2006 and 2007–2010 show that while there were slight improvements in the percentage of U.S. adults with diagnosed diabetes who had A1C levels >9% (75 mmol/mol) (decreasing from 13.0 to 12.6%) and those who had A1C levels <8% (64 mmol/mol) (increasing from 78.0 to 79.1%), there was no improvement in those with A1C <7.0% (53 mmol/mol) (decreasing from 56.8 to 52.2%).[4] We believe it does not have to be like this.

The ongoing "epidemic" of diabetes, which currently affects 11% of U.S. adults and 27% of those over 65,[1] is mostly type 2 diabetes and largely reflects the "success of society." We are living longer, we eat too much, and we are inactive. Being older, overweight, and sedentary makes us resistant to insulin, and if our bodies cannot make enough insulin to compensate, glucose levels rise.[5]

The earliest stage of type 2 diabetes is prediabetes (impaired glucose tolerance [IGT] and/or impaired fasting glucose [IFG]), where glucose levels are higher than normal but not in the diabetes range.[1] Prediabetes tends to progress to diabetes, and over time, persistent hyperglycemia leads to the complications that are the major source of morbidity, mortality, and cost (Fig. 1A). This natural history reflects underlying loss of β-cell function,[6,7] due in part to factors such as elevated glucose and lipid levels, inflammation, amyloid, and oxidative and endoplasmic reticulum stress.

Figure 1.

Diagram illustrating the natural history of diabetes (progression from prediabetes to diabetes and development of diabetes complications over time) without interventions (A); with interventions such as lifestyle change or a glucose-lowering medication that are successful in decreasing progression from prediabetes to diabetes, but then are stopped (B); and with interventions that are titrated to keep glucose and A1C levels in the normal range and are not stopped (C).

Unfortunately, we waste the first ~10 years[8] of the natural history—when the disorder is easiest to treat. There are 79 million Americans with prediabetes and 7 million with early type 2 diabetes[1] who are largely unrecognized, because we do not screen to find these states of dysglycemia. Moreover, when we do make the diagnosis, we do not treat in a way that lowers glucose levels to normal. American Diabetes Association (ADA) guidelines call for adjusting therapy when A1C reaches 7.0% (53 mmol/mol),[9] which corresponds to glucose levels that average 154 mg/dL.[10] And as adjustments in therapy are frequently delayed,[11,12] many patients have glucose levels that are well above normal. For example, Nichols et al.[13] reported in 2007 that over half of Kaiser Permanente Northwest (KPNW) patients who initiated metformin-sulfonylurea combination therapy attained but failed to maintain A1C levels below 8.0% (64 mmol/mol), and continued combination therapy for an average of almost 3 years before insulin was added, with glucose exposure equivalent to 32 months of A1C levels of 9% (75 mmol/mol). Similarly, Khunti et al.[14] reported in 2013 that in the U.K. Clinical Practice Research Datalink (CPRD) database—representative of the U.K. general population—patients who were using two oral glucose-lowering drugs and had A1C levels of at least 8.0% (64 mmol/mol) had a median time of over 6.9 years before further intensification of therapy, with a mean A1C level of 9.1% (76 mmol/mol) when therapy was finally intensified.

As a consequence, the disease tends to progress, and patients need more and more medications. Many patients eventually come to need mealtime insulin and other drugs that can cause hypoglycemia—itself a potential cause of acute cardiovascular events. Johnston et al.[15] found ICD-9-CM coding for outpatient hypoglycemic events in patients with type 2 diabetes to be independently associated with acute cardiovascular events in a retrospective analysis of a large health care claims database. Glucose levels of 41 to 70 mg/dL (2.3 to 3.9 mmol/L) were associated with increased mortality in the prospective Normoglycemia in Intensive Care Evaluation–Survival Using Glucose Algorithm Regulation (NICE-SUGAR) study.[16] As NICE-SUGAR involved intensive care unit inpatients, the findings may not be generalizable to outpatients. However, outpatients with type 2 diabetes who are treated with insulin and/or sulfonylureas exhibit an increased frequency of asymptomatic mild and severe hypoglycemia, and severe hypoglycemia is associated with increased ventricular ectopy.[17,18] Severe hypoglycemia is also independently associated with QTc interval prolongation in both type 2 and type 1 diabetes,[17,19] and there are plausible biological mechanisms through which hypoglycemia could trigger acute cardiovascular events.[20,21]

But this natural history is not inevitable. If we found prediabetes and early diabetes when they first presented and treated them more effectively, we could change the natural history—preventing or delaying the need for the use of mealtime medications and other drugs that can cause hypoglycemia, as well as the development of complications. What is the evidence that we could do this?

Patients with diabetes who are given glucose-lowering medications earlier in their natural histories—when glucose levels are lower and/or lower glucose nadirs can be achieved—can go longer periods of time before another medication is needed. In another KPNW study,[22] half of the patients initiated on metformin monotherapy who achieved an A1C nadir of 7–8% (53–64 mmol/mol) were given another drug after 3 additional years, but half of those reaching a nadir of 6–7% (42–53 mmol/mol) were given another drug after 6.5 years, and half of those achieving an A1C nadir of less than 6% (42 mmol/mol) were given another drug after 7.5 years. In the Treatment Options for type 2 Diabetes in Adolescents and Youth (TODAY) study,[23,24] diabetic youth were given metformin and the end point was preserving A1C levels below 8% (64 mmol/mol). Those who failed or succeeded with treatment were comparable in age, sex, and BMI. However, those who failed were begun on treatment when their A1C averaged 6.5% (48 mmol/mol), while those who succeeded were begun when their A1C averaged 5.7% (39 mmol/mol) and they had correspondingly higher measures of β-cell function. Such retrospective analyses might be subject to length time bias, but suggest the potential benefit of starting treatment earlier.

Stronger evidence for early intervention comes from randomized controlled trials: the Da Qing IGT and Diabetes Study in China,[25] the Finnish Diabetes Prevention Study,[26] the U.S. Diabetes Prevention Program (DPP),[27] and the multinational Diabetes Reduction Assessment With Ramipril and Rosiglitazone Medication (DREAM) study.[28] These studies all showed that treatment with lifestyle change or glucose-lowering medications could reduce progression from prediabetes to diabetes. On follow-up after the studies ended and the interventions were stopped or reduced, additional study participants developed diabetes. However, there was little "catch-up"—the cumulative development of diabetes in the previously treated groups remained less than that in the untreated control groups.[29–32] The follow-up findings of the DREAM study are shown in Fig. 2, and the pattern is particularly clear with the DPP subjects who had been randomized to the troglitazone arm (which was stopped after a little less than a year), as shown in Fig. 3.[33] During 0.9 years of treatment with troglitazone, the diabetes incidence rate was 3.0 cases/100 person-years, considerably lower than the rate of 12.0 cases/100 person-years with placebo. However, during the 3 years after troglitazone was stopped, the diabetes incidence was virtually identical to that of the placebo group—without evidence of "catch-up" (Figs. 3 and 4).

Figure 2.

Cumulative diabetes incidence in the DREAM study, where subjects with prediabetes were given rosiglitazone or placebo, including time points before and after the primary study was stopped (vertical dashed line). Adapted with permission from Gerstein et al. (32).

Figure 3.

Cumulative diabetes incidence in the U.S. DPP study, showing subjects with prediabetes who were given troglitazone or placebo, including time points before and after the primary study was stopped (vertical dashed line). Adapted with permission from Knowler et al. (33).

Figure 4.

Cumulative diabetes incidence in the U.S. DPP study, showing subjects with prediabetes who were given troglitazone or placebo, including only time points after the primary study was stopped on 4 June 1988. Adapted with permission from Knowler et al. (33).

These patterns are consistent with a change in the natural histories of the subjects in the treated groups (Fig. 1B). Moreover, in the Da Qing study, follow-up 14 years after the trial ended showed that the lifestyle change group had a significant decrease in incident severe diabetic retinopathy compared with the control group (Fig. 5);[34] after an additional 3 years of follow-up, the lifestyle change group had decreases in cardiovascular and all-cause mortality.[35] These are clinical benefits accompanying the glucose level benefit.

Figure 5.

Cumulative incidence of severe diabetic retinopathy in the Da Qing study, showing subjects with prediabetes who were randomized to receive instruction in diet + exercise or to be control subjects, including time points before and after the primary study was stopped (vertical dashed line). Adapted with permission from Gong et al. (34).

In the DPP subject population, achieving normal glucose levels appeared to be particularly beneficial. Among subjects who had not developed diabetes at the end of the primary study, those who achieved normal fasting and 2-h oral glucose tolerance test (OGTT) glucose levels at least once during the average 3.2 years of the primary study had a 56% decrease in development of diabetes during follow-up after the primary study ended.[36] Moreover, it did not matter how normal glucose was achieved, as the reduction in the tendency to develop diabetes was comparable among subjects in the lifestyle change, metformin, and control groups who achieved normal glucose levels at least once. In addition, the risk of subsequent development of diabetes was decreased progressively according to the number of times normal glucose levels were achieved; the risk reduction was 47, 61, and 67% when normal glucose levels were achieved once, twice, or three times, respectively. Attaining normal glucose levels was predicted somewhat more strongly by better β-cell function than by better insulin sensitivity, although both were statistically significant.