This is one of the first studies to report on the associations of GL with CVD risk factors within race/ethnicity and BMI strata. While this study corroborated the results of previous studies confirming an association of GL with certain CVD risk factors (lipoproteins/lipids) and apparent lack of an association with others (glucose metabolism factors), previous studies either did not include race/ethnicity- and BMI-diverse samples or did not present their results by race/ethnicity or BMI strata.
In this sample of WHI Observational Study participants as a whole, we observed that GL was inversely associated with HDL cholesterol. This corroborates results from several observational studies, including the Nurses' Health Study [10
], the Women's Health Study [22
], the Whitehall II study [23
], and the Third National Health and Nutrition Examination Survey (1988-1994) [8
]. GL also was inversely associated with HDL cholesterol in studies of Japanese women [24
], healthy adults [25
], and healthy young males and females [9
], but in men only in the Insulin Resistance Atherosclerosis Study (IRAS) [26
]. We believe ours is the first study to extend these analyses by reporting the association between GL and HDL cholesterol by race/ethnicity and BMI.
We observed that GL was positively associated with triglycerides, again corroborating results from many previous observational studies. Dietary GL was positively associated with triglycerides in the Women's Health Study [22
], the Nurses' Health Study [10
], and IRAS [26
]. In a study of Japanese women, the lowest concentration of triglycerides was observed in the lowest tertiles of both GL and GI [24
]. However, dietary GI, but not GL, was positively associated with triglycerides in the Whitehall II study [23
], and there was no association between GL and triglycerides in one study of healthy adults [25
]. Again, our analyses go beyond those published previously by demonstrating that there were possible race/ethnicity and BMI differences in the association between GL and triglycerides, although the test for interaction was not statistically significant. It is interesting to note that high-GI carbohydrates compared to low-GI carbohydrates have been shown to create a detrimental postprandial pattern in triglyceride-rich lipoproteins derived from both hepatic and intestinal sources in studies of obese, insulin-resistant persons [27
]. However, the positive association between GL and triglycerides was not observed in obese participants in our study.
It may be tempting to attribute associations of GL with HDL cholesterol and triglycerides to effects of GL on circulating insulin concentrations and IR since hyperinsulinemia plays a critical role in the development of dyslipidemia [2
]. However, in our study GL was not associated with glucose or insulin concentrations, or with IR, either overall or by specific race/ethnicity or BMI groups. Previous studies have produced equivocal results on these associations. GL was not associated with fasting glucose or insulin concentrations in a study of Japanese women [24
]. However, lower fasting glucose concentrations were observed with increasing dietary GL, while fasting insulin was inversely related to GL in women in the Whitehall II study [23
]. In female participants of the Health, Aging and Body Composition study, GL was not associated with either fasting glucose or insulin concentrations [28
], while GL was positively associated with HOMA-IR in the Framingham Offspring Cohort [29
]. GL was not associated with fasting glucose, fasting insulin, insulin sensitivity (assessed by frequently sampled intravenous glucose tolerance test [FSIVGTT]), or acute insulin response in IRAS [30
], nor was GL associated with HOMA-IR in cross-sectional analyses within the Inter99 study [32
The inverse association between GL and SBP in Whites in this study was somewhat unexpected. Very few observational studies have investigated the association between GL (or GI) and BP. In the British National Diet and Nutritional Survey, GI was not associated with SBP or DBP in participants over 65 years of age [33
]. Intervention studies have produced equivocal results, with some showing no effect of reduced-GL/GI diets on BP [34
] and others showing trends toward reductions in both SBP and DBP on such diets [35
]. There was no effect of dietary GI on SBP in the Canadian Trial of Carbohydrates in Diabetes [37
], a clinical trial utilizing low-GI carbohydrates.
The strengths of this study include the race/ethnicity- and BMI-diverse sample, the comprehensive data collection which took place within the WHI, and the availability of a comprehensive GI/GL database which was added to the existing WHI FFQ database. There were also potential limitations. In addition to the well-documented limitations associated with assessing diet with FFQs in general, there were further issues specific to this study which could have attenuated the results. While the addition of GI/GL values to the FFQ nutrient database was done in a systematic and well-documented manner [14
], the structure of the FFQ itself was not optimal for assessment of GI and GL, and the FFQ has not been validated for this purpose. Further, energy underreporting has been noted among Black and Hispanic postmenopausal women compared to Whites, and among women with higher compared to lower BMI [38
], which potentially dampens the ability to test for diet-metabolic effects among minority or heavier participants. WHI participants were postmenopausal women who were 50-79 years of age; thus, there is a potential limitation in generalizing our study results to other populations. The WHI Observational Study population may be healthier than US postmenopausal women in general [12
], thus limiting the metabolic range of the study. Finally, the investigation was cross-sectional, leaving the potential for spurious correlations despite adjusting for relevant covariates.
Finally, we recognize that given the distribution of race/ethnicity within the subsample, power to detect linear trends was greatest for White participants. Nevertheless, the overall sample provided the opportunity to examine differences by race/ethnicity in the associations of GL with the outcomes. Though formal tests of interaction were conducted to examine whether the associations between GL and the risk factors varied among the categories of race/ethnicity, the statistical power of these tests were low due to the small sample sizes within the strata. However, the stratified analyses were presented as a set of descriptive secondary analyses. These results should be interpreted with caution as exploratory and hypothesis generating. Further research is warranted to examine whether the differences in the degree of associations are observable in larger sample sizes for Blacks and Hispanics and, if so, to investigate factors contributing to these differences.