Meta-Analysis of Low–Glycemic Index Diets in the Management of Diabetes

Response to Franz

In the editorial by Franz (1) on the value of low–glycemic index diets, the results of our meta-analysis, which were published in the same issue of Diabetes Care (2), have been misinterpreted and therefore misrepresented. Specifically, the author contrasts the overall decrease in HbA_1c of ∼1–2% seen with various nutrition interventions with a reduction of 0.43% from low–glycemic index diets. The 0.43% reduction, however, is not the overall effect of a low–glycemic index diet, but the incremental effect of a low–glycemic index diet over and above that seen with an equally intensive nutrition intervention. Thus if a nutrition intervention improves HbA_1c by 1%, then the meta-analysis predicts that a low–glycemic index version of that intervention will result in an overall reduction of 1.43%.

In the reviews and studies cited by Franz, the change in HbA_1c levels of a nutrition intervention is compared with either a control group given a “basic” or “usual” level of care or with the baseline HbA_1c. For example, in the Dose Adjustment For Normal Eating (DAFNE) study (3), the control group consisted of patients who simply continued to receive usual care for 6 months versus a group who received training in flexible intensive insulin treatment combining dietary freedom and insulin adjustment. Similarly, in discussing the U.K. Prospective Diabetes Study (4), Franz compares the glycemic control at baseline with that after 3 months of intensive dietary intervention. In both of these instances, intensive nutritional interventions are compared with basic care. In the meta-analysis, however, we included only studies in which two equally intensive nutrition interventions were compared.

The author goes on to state that the 0.43% reduction in HbA_1c is equivalent to a 7.4% reduction. This is incorrect. A 7.4% reduction in glycated proteins, as shown in the combined analysis with both fructosamine and HbA_1c data, is larger and equivalent to >0.6% HbA_1c in an individual with an HbA_1c >8%. This is addressed in our discussion (2).

Lastly, the author states that the meta-analysis found that “in subjects with type 1 diabetes, HbA_1c was reduced by ∼0.4% units and in type 2 diabetes by ∼0.2% units.” This is also incorrect; we did not perform a subgroup analysis of HbA_1c in type 1 versus type 2 diabetic subjects because there were insufficient data. However, we reported that the incremental reduction in glycated proteins was ∼10% (equivalent to 0.8% HbA_1c) in type 1 diabetic subjects and ∼6% (equivalent to ∼0.5% HbA_1c) in type 2 diabetic patients.

The editorial’s title, “The Glycemic Index: Not the Most Effective Nutrition Therapy Intervention,” thus represents a misinterpretation of the results of the meta-analysis. It may, unfortunately, lead readers to dismiss the study’s findings and potential value.