In this large prospective cohort of women, we found that after adjustment for confounding variables, especially BMI, a higher low-carbohydrate-diet score was not associated with risk of type 2 diabetes. This dietary score was associated with a modest decreased risk of type 2 diabetes when vegetable sources rather than animal sources of fat and protein were chosen.
Although little research has been conducted on the association between low-carbohydrate-diet score and risk of type 2 diabetes, several investigations have examined the effects of a low-carbohydrate diet on risk factors for type 2 diabetes for 6 mo or longer (24
). After 6 mo, Samaha et al (24
) found that insulin sensitivity improved more in subjects who consumed a low-carbohydrate diet than in those who consumed a low-fat diet. Among diabetic subjects, the mean fasting glucose concentration decreased more in the low-carbohydrate-diet group. In a 1-y follow-up study, Stern et al (25
) found that, in diabetic subjects, hemoglobin A1c
concentrations improved more in persons who consumed a low-carbohydrate diet than in those who consumed a low-fat diet. Foster et al (26
), in a 1-y trial, compared a low-carbohydrate diet with a conventional low-fat diet and found no differences in insulin sensitivity between the 2 groups. Finally, Brehm et al (27
) compared a low-carbohydrate diet with a low-fat diet for 6 mo and reported no differences in fasting glucose and fasting insulin between the 2 groups. These studies are difficult to interpret because of the various degrees of weight loss between the diet groups.
In shorter investigations, Boden et al (28
) found a significantly lower hemoglobin A1c
concentration, fasting plasma glucose concentration, mean 24-h insulin concentration, and improved insulin sensitivity in obese persons with type 2 diabetes after 2 wk of a low-carbohydrate diet. McAuley et al (29
) observed no difference in fasting insulin or fasting glucose concentrations after 16 wk of either a low-fat or a low-carbohydrate diet. In a 5-wk randomized crossover trial, Gannon and Nuttall (30
) found that, compared with a low-fat diet, a low-carbohydrate diet significantly reduced fasting glucose and glycated hemoglobin concentrations. In contrast, Swinburn et al (31
) reported a significant improvement in oral glucose tolerance and a significant reduction in fasting plasma glucose in subjects who consumed a high-carbohydrate diet relative to those who consumed a low-carbohydrate diet. Once again, these studies are difficult to interpret because of various degrees of weight loss between the diet groups and because oral-glucose-tolerance tests are greatly affected by carbohydrate consumption over the few days before the test is conducted (32
When compared with low-fat, high-carbohydrate diets, low-carbohydrate diets tend to have a relatively higher percentage of fat and protein and a lower percentage of carbohydrate. When evaluating the association between low-carbohydrate-diet score and risk of type 2 diabetes, each of the macronutrients must be taken into consideration.
Distinct types of fats have various effects on risk factors for type 2 diabetes. Substituting unsaturated fats for saturated fats increases insulin sensitivity in diabetic (33
), overweight (34
), and healthy (35
) subjects. In epidemiologic studies, polyunsaturated fat has been shown to be associated with a reduced risk of type 2 diabetes (36
). Generally, no association has been found between saturated fat (37
) or monounsaturated fat (38
) and risk of type 2 diabetes. Results from cohort studies on the association between trans
fat and risk of type 2 diabetes have not been consistent (36
). Dietary interventions in humans have shown no consistent adverse effects of high-fat diets on insulin sensitivity (31
) and in epidemiologic studies, total fat has not been shown to increase risk of type 2 diabetes (36
). Therefore, the increase in total fat common in low-carbohydrate diets would not be expected to increase the risk of type 2 diabetes.
In low-carbohydrate diets, dietary protein is substituted for some of the carbohydrate. Dietary protein has not been shown to raise peripheral glucose concentrations after ingestion in healthy subjects or in persons with type 2 diabetes (46
). However, protein does tend to stimulate insulin secretion in healthy persons (52
) and even more so in persons with type 2 diabetes (54
). In metabolic studies, Gannon et al (55
) found that 24-h integrated glucose area response and glycated hemoglobin decreased significantly more after 5 wk of a high-protein diet than after 5 wk of a low-fat control diet. In a similar investigation, Sargrad et al (56
) found no beneficial effects of a high-protein diet relative to a low-fat diet. Epidemiologic studies of protein intake and risk of type 2 diabetes are limited. In a 6-y follow-up of subjects in the Nurses' Health Study, Colditz et al (38
) found no association between protein consumption and risk of type 2 diabetes. In these updated analyses, we found no association between total protein, animal protein, or vegetable protein and risk of type 2 diabetes.
The lack of an association between a low-carbohydrate-diet score and risk of type 2 diabetes when adjusted for confounders may also be explained by the amount and quality of carbohydrate present in the diet. High-carbohydrate diets generally result in high postprandial glucose and insulin responses. The total percentage of energy from carbohydrate has generally not been found to increase the risk of type 2 diabetes (39
). However, in the present analysis, we found a modest but positive association between carbohydrate consumption and risk of type 2 diabetes.
A carbohydrate-restricted diet tends to have a lower glycemic index and lower glycemic load than does a high-carbohydrate diet. The glycemic index of a carbohydrate is a measure of how much that food raises blood glucose compared with a standard carbohydrate (usually glucose or white bread) (60
). The glycemic load takes into account the amount of carbohydrate in addition to its glycemic index (61
). Compared with higher glycemic diets, low glycemic diets have been shown in epidemiologic studies to decrease glucose and insulin responses (62
) and glycated hemoglobin (65
) and to increase insulin sensitivity (64
). In addition, several prospective studies have shown an association between dietary glycemic index or glycemic load and risk of type 2 diabetes (39
), whereas 2 prospective studies have not reported this association (57
). In the present investigation we found a statistically significant positive association for both dietary glycemic load and total carbohydrate and risk of type 2 diabetes. The positive association between dietary glycemic load and type 2 diabetes was much stronger than that for total carbohydrate because glycemic load captures both the quality and quantity of carbohydrate.
Adequate power for this investigation was provided by the large sample size and 20-y follow-up with updated dietary data. The prospective design and high follow-up rate served to minimize bias. Because diet was assessed with a self-reported questionnaire, some degree of misclassification of intakes of fat, protein, and carbohydrate will have occurred. Measurement error in assessing long-term diet was reduced in this analysis by using the average of all available measurements of diet up to the start of each 2-y follow-up interval.
In this investigation we measured and adjusted for a variety of potential confounding variables. However, we cannot rule out the possibility of residual confounding. A concern in this analysis was whether to consider body mass index as a mediator of the relation between low-carbohydrate-diet score and risk of type 2 diabetes or as a potential confounder of the relation. In our cohort, total calories were similar across deciles of low-carbohydrate-diet score (). Furthermore, in most weight-loss trials, a low-carbohydrate-diet has not been associated with a significant increase in body weight (24
). Therefore, we considered body mass index as a potential confounder and included it in the multivariate analyses.
The Nurses' Health Study consists of mostly white women with some college education. Although this homogeneity increases the internal validity of the study by reducing confounding by factors that are difficult to measure, the association between low-carbohydrate-diet score and risk of type 2 diabetes among women of other educational and racial backgrounds should also be investigated.
In conclusion, a diet lower in carbohydrate and higher in protein and fat did not increase the risk of type 2 diabetes in this cohort of women. In fact, when vegetable sources of fat and protein were chosen, these diets were associated with a modest reduction in the risk of type 2 diabetes. These data support a potential benefit in reducing the glycemic load of the diet and for substituting low-glycemic fruit, vegetables, whole grains, and healthful sources of fat and protein for high-glycemic refined carbohydrates.