This study found significantly larger reductions in HbA1c in overweight and obese patients with type 2 diabetes who were prescribed a low-GL diet, versus an isoenergetic low-fat diet, as part of a lifestyle modification program. The advantage for diabetes control was observed despite lack of significant differences in body weight, fasting glucose, and dietary fiber between the two groups. This finding suggests that the quality and quantity of carbohydrate consumed play an important role in broad glycemic exposure among patients with type 2 diabetes, even in a state of negative energy balance necessary for concomitant weight loss.
The effect of reducing GI or GL on glycemic control in persons with diabetes has been summarized in several meta-analyses [21
]. However, the studies included in those analyses differed notably from the present trial. In the Cochrane review by Thomas and Elliot, for instance, only 5 of the 11 included studies tested diets for at least 12 weeks [28
]. Of those, only two were of participants with type 2 diabetes; they were both cross-over studies with a total of only 22 subjects between them. The meta-analysis found that low-GI or low-GL diets were associated with reductions in HbA1c
that were approximately 0.3% to 0.5% greater than with higher-GI or GL alternatives [28
]. In the present study, we found that the advantage for the low-GL diet was 0.4% at the midpoint of treatment (i.e., week 20) and 0.7% at the conclusion. The widening gap between groups was more attributable to an increase in HbA1c
(from week 20 to 40) among those in the low-fat group, than to a continued reduction in the low-GL group.
A recent 12-month randomized controlled comparison of low-GI, high-GI, and low-carbohydrate diets in patients with type 2 diabetes found no differences among groups in HbA1c
]. Furthermore, HbA1c
in all groups at study's end than at baseline. Similarly, another trial found no significant change in HbA1c
at 1 year among patients with type 2 diabetes who were assigned to follow a low-carbohydrate (and by extension, low-GL) or low-fat diet [30
]. Both of those studies, however, offered far less frequent and intense clinical contacts than the level of treatment provided in the present study.
Studies of lifestyle modification in participants with type 2 diabetes typically show that changes in HbA1c
track with changes in weight in this population [31
]. Thus, diminished improvements in HbA1c
are expected when patients with type 2 diabetes reach a weight loss plateau or begin to regain weight. The low-calorie diet that is typically prescribed, however, is a low-fat diet. Without explicit instruction to choose low-GI carbohydrates, participants who follow a low-fat (i.e., high-carbohydrate) diet may be inadvertently following an eating plan that puts them at risk for increasing glycemic exposure when they cease to create an energy deficit. That participants in our low-GL group maintained their HbA1c
reduction in full, in the absence of additional weight loss in the second half of the intervention, suggests that they continued to adhere to the principles of low-GL eating, even as treatment became less frequent.
In contrast to the significant HbA1c
advantage for the low-GL diet, the between-groups differences in weight loss (~ 1 and 2 kg at weeks 20 and 40, respectively) were not statistically significant and were consistent with those reported previously. A meta-analysis of six randomized controlled trials found that low-GI/GL diets induced a mean weight loss that was 1 kg greater than that achieved with higher-GI/GL alternatives in healthy participants [20
]. We had hypothesized a significant difference in weight loss, in part, due to findings that the reduction in metabolic rate after a 10% weight loss was significantly but modestly smaller with a low-GL diet, compared with a low-fat diet [32
]. Additionally, feeding laboratory-based findings that lower-GL meals are associated with greater satiety and less energy intake at subsequent meals suggested that consumption of low-GL foods would facilitate adherence to a low-calorie diet [4
]. Whether those mechanisms were engaged in the present study is unknown. However, the study is underpowered to detect changes in body weight that might result from relatively subtle effects on metabolic rate, hunger, or satiety, especially when calorie prescriptions were identical in both groups.
At least two studies, published after the launch of the present one, found that insulin secretion rates – as assessed by insulin concentration 30 minutes after oral glucose consumption – moderated the effect of a low-GL diet on weight loss. Pittas et al. compared weight loss achieved with isocaloric low- and high-GL provided diets in healthy overweight participants [33
]. Although there was no effect of diet on weight loss for those with low insulin secretion, the low-GL diet produced a significantly greater reduction in weight for those with a high insulin response. Similarly, Ebbeling et al. found no main effect of diet (low-GL vs. low-fat) on weight loss among obese young adults [34
]. However, a significant effect was apparent among those with high baseline insulin secretion; the low-GL and low-fat diets produced weight losses of 5.8 kg and 1.2 kg, respectively, at 18 months. In both of the studies described above, participants were overweight or obese, but with no history of type 2 diabetes. It is unclear whether a similar moderating effect would have been found in the present study had we conducted an oral glucose tolerance test at baseline.
Several previous studies that examined the effects of low-GL diets provided diet-consistent food to participants. We instead opted to instruct participants in how to reduce the GL of diets they selected autonomously. This approach has the advantage of external validity, as long-term food provision is not practical for the long-term management of obesity and type 2 diabetes. The disadvantage of providing instruction, rather than food, is that adherence to the diet is likely less complete. We note, however, that between-groups differences in dietary intake (as measured by analysis of 3-day food records) were consistent with expectations based on the dietary instruction that each group received. That is, participants in the low-fat group reduced fat intake to a greater extent than did those in the low-GL group, who in turn, reduced carbohydrate intake, GI, and GL to a greater degree than their low-fat counterparts.
Aside from the significant advantage of the low-GL diet for reducing HbA1c, there were no other significant differences between groups in other metabolic outcomes, including fasting glucose. This finding highlights the contribution of postprandial glycemic exposure to overall diabetes control. Both systolic and diastolic blood pressure improved to a marginally greater degree in the low-fat group at weeks 20 and 40, compared to the low-GL group. Examination of group means revealed a modest increase in blood pressure among participants in the low-GL group. This finding is difficult to interpret. We did not control for changes in antihypertensive medications, sodium intake, or other factors that may explain the observed differences in blood pressure.
A weakness of this study is the attrition rate of 36.7%. We suspect that dissatisfaction with treatment (including with initial weight loss) contributed to attrition, particularly among those who dropped out before week 20 (24.1%) or were lost to follow-up (16.5%). In addition, the lack of financial compensation provided little incentive for participants who achieved unsatisfactory results to return for assessments. Our greatest protection against the effects of attrition was the use of mixed models for analyses. This approach is considered superior to alternative methods of handing missing data such as including only participants with complete data or assigning a pre-specified score to the missing data. A major strength of the trial includes the long duration and large sample size relative to most previous comparably designed studies. An additional strength was the incorporation of each diet into a standard-format lifestyle modification program with identical energy intake goals, exercise prescriptions, and behavioral interventions in both conditions.
The current position of the American Diabetes Association [35
] is that “the use of the glycemic index and glycemic load may provide a modest additional benefit for glycemic control over that observed when total carbohydrate is considered alone” (Evidence grade B: supportive evidence from well-conducted cohort studies.) The present study was not designed to address this issue directly, as the comparison diet was a higher-GL low-fat diet. Future investigations should compare low- and high-GI/GL diets that control for total carbohydrate intake.
This study demonstrated that prescribing a calorie-restricted low-GL diet to overweight and obese adults with type 2 diabetes resulted in greater glycemic control than was achieved with an isoenergetic low-fat prescription. The advantage of the low-GL diet for improving HbA1c was apparently not attributable to weight change or calorie-restriction, as these were equivalent between groups. These results add to a growing literature on the benefits of following a low-GL diet for diabetes control.