The Women's Health Initiative (WHI) has been a major contributor to diet and chronic disease research among postmenopausal US women over its 30+ year history (1993 to present). The WHI program ...included full-scale randomized trials of a low-fat dietary pattern high in fruits, vegetables, and grains, and of calcium and vitamin D supplementation, each with designated primary and secondary chronic disease outcomes. The history of these trials will be briefly reviewed here, along with principal findings that included evidence for breast cancer-related benefits for each of the 2 interventions. In recent years, WHI investigators have developed an active research program in nutritional biomarker development and in the application of these biomarkers in WHI cohorts, among various other nutritional epidemiology uses of WHI observational study resources. The intake biomarker work, which primarily relies on blood and urine metabolomics profiles, lends support to the low-fat dietary pattern trial results, and supports chronic disease benefits of higher carbohydrate diets more generally, especially through the fiber component of carbohydrate.
Dietary intake biomarkers that can be written as actual intake, plus 'error' that is independent of actual intake and confounding factors can substitute for actual intake in disease association ...analyses. Also, such biomarkers can be used to develop calibration equations using self-reported diet and participant measures, and biomarker-calibrated intakes can be calculated in larger cohorts for use in disease association analyses. Criteria for biomarkers, and for biomarker-calibrated intakes, arise by working back from properties needed for valid disease association analyses. Accordingly, arguments for a potential biomarker are strengthened if error components are small relative to actual intakes, and important sources of reduced sensitivity or specificity are not apparent. Feeding study biomarker development can then involve regression of actual intake on putative biomarkers, with regression R2 values playing a role in biomarker evaluation. In comparison, 'predictive' biomarker status, as argued in this issue by Freedman and colleagues for 24-hour urinary sucrose plus fructose as biomarker for total sugars, involves regression of potential biomarker on actual intake and other variables, with parameter stability across populations and limited within-person variability as criteria. The choice of criteria for biomarkers and for biomarker-calibrated intakes, is discussed here, in the context of total sugars intake. See related article by Freedman et al., p. 1227.
Semiparametric, multiplicative-form regression models are specified for marginal single and double failure hazard rates for the regression analysis of multivariate failure time data. Cox-type ...estimating functions are specified for single and double failure hazard ratio parameter estimation, and corresponding Aalen-Breslow estimators are specified for baseline hazard rates. Generalization to allow classification of failure times into a smaller set of failure types, with failures of the same type having common baseline hazard functions, is also included. Asymptotic distribution theory arises by generalization of the marginal single failure hazard rate estimation results of Lin et al. The Péano series representation for the bivariate survival function in terms of corresponding marginal single and double failure hazard rates leads to novel estimators for pairwise bivariate survival functions and pairwise dependency functions, at specified covariate history. Related asymptotic distribution theory follows from that for the marginal single and double failure hazard rates and the continuity, compact differentiability of the Péano series transformation and bootstrap applicability. Simulation evaluation of the proposed estimation procedures is presented, and an application to multiple clinical outcomes in the Women's Health Initiative Dietary Modification Trial is provided. Higher dimensional marginal hazard rate regression modeling is briefly mentioned.
Supplementary materials
for this article are available online.
We pooled data from 5 large validation studies of dietary self-report instruments that used recovery biomarkers as references to clarify the measurement properties of food frequency questionnaires ...(FFQs) and 24-hour recalls. The studies were conducted in widely differing US adult populations from 1999 to 2009. We report on total energy, protein, and protein density intakes. Results were similar across sexes, but there was heterogeneity across studies. Using a FFQ, the average correlation coefficients for reported versus true intakes for energy, protein, and protein density were 0.21, 0.29, and 0.41, respectively. Using a single 24-hour recall, the coefficients were 0.26, 0.40, and 0.36, respectively, for the same nutrients and rose to 0.31, 0.49, and 0.46 when three 24-hour recalls were averaged. The average rate of under-reporting of energy intake was 28% with a FFQ and 15% with a single 24-hour recall, but the percentages were lower for protein. Personal characteristics related to under-reporting were body mass index, educational level, and age. Calibration equations for true intake that included personal characteristics provided improved prediction. This project establishes that FFQs have stronger correlations with truth for protein density than for absolute protein intake, that the use of multiple 24-hour recalls substantially increases the correlations when compared with a single 24-hour recall, and that body mass index strongly predicts under-reporting of energy and protein intakes.
Metabolomics profiles from blood, urine, or other body fluids have the potential to assess intakes of foods and nutrients objectively, thereby strengthening nutritional epidemiology research. ...Metabolomics platforms may include targeted components that estimate the relative concentrations for individual metabolites in a predetermined set, or global components, typically involving mass spectrometry, that estimate relative concentrations more broadly. While a specific metabolite concentration usually correlates with the intake of a single food or food group, multiple metabolites may be correlated with the intake of certain foods or with specific nutrient intakes, each of which may be expressed in absolute terms or relative to total energy intake. Here, I briefly review the progress over the past 20 years on the development and application intake biomarkers for foods/food groups, nutrients, and dietary patterns, primarily by drawing from several recent reviews. In doing so, I emphasize the criteria and study designs for candidate biomarker identification, biomarker validation, and intake biomarker application. The use of intake biomarkers for diet and chronic disease association studies is still infrequent in nutritional epidemiology research. My comments here will derive primarily from our research group's recent contributions to the Women's Health Initiative cohorts. I will complete the contribution by describing some opportunities to build on the collective 20 years of effort, including opportunities related to the metabolomics profiling of blood and urine specimens from human feeding studies that approximate habitual diets.
Dietary assessment poses a major challenge in nutritional epidemiology research. In an observational cohort setting, even if individual dietary intake history could be measured precisely over the ...life span, substantial efforts would be required to quantify the role of specific nutrients, food groups, and dietary patterns in chronic disease risk. In particular, related analyses would need to allow for a complex mixture of correlated and interactive intake variables, as well as the usual confounding issues. Based on comparisons with the few established dietary intake biomarkers, that available self-report dietary assessment approaches have a major deficiency. Observational study designs will continue to provide a major approach to diet and chronic disease epidemiology. The use of intake biomarkers, either to correct self-report intake estimates for measurement error or as direct intake assessments, could provide a fresh and reliable look at a broad range of diet and chronic disease associations.
IMPORTANCE: The influence of menopausal hormone therapy on breast cancer remains unsettled with discordant findings from observational studies and randomized clinical trials. OBJECTIVE: To assess the ...association of prior randomized use of estrogen plus progestin or prior randomized use of estrogen alone with breast cancer incidence and mortality in the Women’s Health Initiative clinical trials. DESIGN, SETTING, AND PARTICIPANTS: Long-term follow-up of 2 placebo-controlled randomized clinical trials that involved 27 347 postmenopausal women aged 50 through 79 years with no prior breast cancer and negative baseline screening mammogram. Women were enrolled at 40 US centers from 1993 to 1998 with follow-up through December 31, 2017. INTERVENTIONS: In the trial involving 16 608 women with a uterus, 8506 were randomized to receive 0.625 mg/d of conjugated equine estrogen (CEE) plus 2.5 mg/d of medroxyprogesterone acetate (MPA) and 8102, placebo. In the trial involving 10 739 women with prior hysterectomy, 5310 were randomized to receive 0.625 mg/d of CEE alone and 5429, placebo. The CEE-plus-MPA trial was stopped in 2002 after 5.6 years’ median intervention duration, and the CEE-only trial was stopped in 2004 after 7.2 years’ median intervention duration. MAIN OUTCOMES AND MEASURES: The primary outcome was breast cancer incidence (protocol prespecified primary monitoring outcome for harm) and secondary outcomes were deaths from breast cancer and deaths after breast cancer. RESULTS: Among 27 347 postmenopausal women who were randomized in both trials (baseline mean SD age, 63.4 years 7.2 years), after more than 20 years of median cumulative follow-up, mortality information was available for more than 98%. CEE alone compared with placebo among 10 739 women with a prior hysterectomy was associated with statistically significantly lower breast cancer incidence with 238 cases (annualized rate, 0.30%) vs 296 cases (annualized rate, 0.37%; hazard ratio HR, 0.78; 95% CI, 0.65-0.93; P = .005) and was associated with statistically significantly lower breast cancer mortality with 30 deaths (annualized mortality rate, 0.031%) vs 46 deaths (annualized mortality rate, 0.046%; HR, 0.60; 95% CI, 0.37-0.97; P = .04). In contrast, CEE plus MPA compared with placebo among 16 608 women with a uterus was associated with statistically significantly higher breast cancer incidence with 584 cases (annualized rate, 0.45%) vs 447 cases (annualized rate, 0.36%; HR, 1.28; 95% CI, 1.13-1.45; P < .001) and no significant difference in breast cancer mortality with 71 deaths (annualized mortality rate, 0.045%) vs 53 deaths (annualized mortality rate, 0.035%; HR, 1.35; 95% CI, 0.94-1.95; P= .11). CONCLUSIONS AND RELEVANCE: In this long-term follow-up study of 2 randomized trials, prior randomized use of CEE alone, compared with placebo, among women who had a previous hysterectomy, was significantly associated with lower breast cancer incidence and lower breast cancer mortality, whereas prior randomized use of CEE plus MPA, compared with placebo, among women who had an intact uterus, was significantly associated with a higher breast cancer incidence but no significant difference in breast cancer mortality.
There are several different topics that can be addressed with multivariate failure time regression data. Data analysis methods are needed that are suited to each such topic. Specifically, marginal ...hazard rate models are well suited to the analysis of exposures or treatments in relation to individual failure time outcomes, when failure time dependencies are themselves of little or no interest. On the other hand semiparametric copula models are well suited to analyses where interest focuses primarily on the magnitude of dependencies between failure times. These models overlap with frailty models, that seem best suited to exploring the details of failure time clustering. Recently proposed multivariate marginal hazard methods, on the other hand, are well suited to the exploration of exposures or treatments in relation to single, pairwise, and higher dimensional hazard rates. Here these methods will be briefly described, and the final method will be illustrated using the Women’s Health Initiative hormone therapy trial data.
IMPORTANCE: Health outcomes from the Women’s Health Initiative Estrogen Plus Progestin and Estrogen-Alone Trials have been reported, but previous publications have generally not focused on all-cause ...and cause-specific mortality. OBJECTIVE: To examine total and cause-specific cumulative mortality, including during the intervention and extended postintervention follow-up, of the 2 Women’s Health Initiative hormone therapy trials. DESIGN, SETTING, AND PARTICIPANTS: Observational follow-up of US multiethnic postmenopausal women aged 50 to 79 years enrolled in 2 randomized clinical trials between 1993 and 1998 and followed up through December 31, 2014. INTERVENTIONS: Conjugated equine estrogens (CEE, 0.625 mg/d) plus medroxyprogesterone acetate (MPA, 2.5 mg/d) (n = 8506) vs placebo (n = 8102) for 5.6 years (median) or CEE alone (n = 5310) vs placebo (n = 5429) for 7.2 years (median). MAIN OUTCOMES AND MEASURES: All-cause mortality (primary outcome) and cause-specific mortality (cardiovascular disease mortality, cancer mortality, and other major causes of mortality) in the 2 trials pooled and in each trial individually, with prespecified analyses by 10-year age group based on age at time of randomization. RESULTS: Among 27 347 women who were randomized (baseline mean SD age, 63.4 7.2 years; 80.6% white), mortality follow-up was available for more than 98%. During the cumulative 18-year follow-up, 7489 deaths occurred (1088 deaths during the intervention phase and 6401 deaths during postintervention follow-up). All-cause mortality was 27.1% in the hormone therapy group vs 27.6% in the placebo group (hazard ratio HR, 0.99 95% CI, 0.94-1.03) in the overall pooled cohort; with CEE plus MPA, the HR was 1.02 (95% CI, 0.96-1.08); and with CEE alone, the HR was 0.94 (95% CI, 0.88-1.01). In the pooled cohort for cardiovascular mortality, the HR was 1.00 (95% CI, 0.92-1.08 8.9 % with hormone therapy vs 9.0% with placebo); for total cancer mortality, the HR was 1.03 (95% CI, 0.95-1.12 8.2 % with hormone therapy vs 8.0% with placebo); and for other causes, the HR was 0.95 (95% CI, 0.88-1.02 10.0% with hormone therapy vs 10.7% with placebo), and results did not differ significantly between trials. When examined by 10-year age groups comparing younger women (aged 50-59 years) to older women (aged 70-79 years) in the pooled cohort, the ratio of nominal HRs for all-cause mortality was 0.61 (95% CI, 0.43-0.87) during the intervention phase and the ratio was 0.87 (95% CI, 0.76-1.00) during cumulative 18-year follow-up, without significant heterogeneity between trials. CONCLUSIONS AND RELEVANCE: Among postmenopausal women, hormone therapy with CEE plus MPA for a median of 5.6 years or with CEE alone for a median of 7.2 years was not associated with risk of all-cause, cardiovascular, or cancer mortality during a cumulative follow-up of 18 years. TRIAL REGISTRATION: clinicaltrials.gov Identifier: NCT00000611