NHANES I Epidemiologic Follow-up Study Research Articles

Tuberculosis and Risk of Emphysema among US Adults in the NHANES I Epidemiologic Follow-Up Study Cohort, 1971-1992.

(1) Background: History of TB is a known risk factor for long-term respiratory impairment affecting lung functions in both restrictive and obstructive lung disease. (2) Methods: We analyzed data from the NHANES I Epidemiologic Follow-up Study (NHEFS), a longitudinal study conducted on a noninstitutionalized adult US population aged 25-74 years. Approximately 93 percent of the original NHANES I cohort was successfully traced by the end of the survey period and was available for analysis. The final adjusted model included age groups, gender, family income, lifetime smoking, body mass index (BMI), and frequency of alcohol consumption as potential confounders. (3) Results: The estimated hazards ratio of developing emphysema during follow-up for individuals with a past diagnosis of TB was 54% lower (95% CI = 0.35, 0.61) that that in individuals with no past TB, after controlling for potential confounders and using proportional hazards regression appropriate to the complex sample design. The association, however, was not statistically significant (HR = 0.86, p-value = 0.38) when only a self-reported history of TB was considered as the exposure in an unadjusted model. (4) Conclusions: Tuberculosis (self-reported or LTBI) was strongly (but inversely) associated with emphysema incidence. The association was not statistically significant with only a self-reported history of TB as exposure.

Weight Fluctuation and Postmenopausal Breast Cancer in the National Health and Nutrition Examination Survey I Epidemiologic Follow-Up Study.

Objective. The aim of this study is to investigate if weight fluctuation is an independent risk factor for postmenopausal breast cancer (PBC) among women who gained weight in adult years. Methods. NHANES I Epidemiologic Follow-Up Study (NHEFS) database was used in the study. Women that were cancers-free at enrollment and diagnosed for the first time with breast cancer at age 50 or greater were considered cases. Controls were chosen from the subset of cancers-free women and matched to cases by years of follow-up and status of body mass index (BMI) at 25 years of age. Weight fluctuation was measured by the root-mean-square-error (RMSE) from a simple linear regression model for each woman with their body mass index (BMI) regressed on age (started at 25 years) while women with the positive slope from this regression were defined as weight gainers. Data were analyzed using conditional logistic regression models. Results. A total of 158 women were included into the study. The conditional logistic regression adjusted for weight gain demonstrated positive association between weight fluctuation in adult years and postmenopausal breast cancers (odds ratio/OR = 1.67; 95% confidence interval/CI: 1.06–2.66). Conclusions. The data suggested that long-term weight fluctuation was significant risk factor for PBC among women who gained weight in adult years. This finding underscores the importance of maintaining lost weight and avoiding weight fluctuation.

Abstract 4603: Male pattern baldness in relation to prostate cancer-specific mortality: A prospective analysis in the NHANES I Epidemiologic Followup Study (NHEFS)

Abstract Background: Androgenic action underlies prostate gland development and prostate cancer progression. However, the role of such in prostate carcinogenesis remains unclear. Results from studies that have quantitated pre-diagnostic, circulating androgens at a single time-point in relation to prostate cancer are inconsistent, possibly due to the failure to capture cumulative or relevant age-specific hormone exposure. Therefore, we used male pattern baldness as a proxy of long-term androgen exposure, and investigated the association between dermatologist-assessed male pattern baldness and prostate cancer-specific mortality in the NHANES-I Epidemiologic Followup Study (NHEFS). Methods: We included 4,316 men from NHEFS, who were 25-74 years, received dermatologic exams, and had no prior cancer diagnosis at baseline. Hazard ratios (HRs) and 95% confidence intervals (95%CIs) were estimated using Cox proportional hazards regression with age as the time-metric and baseline hazard stratified by age at baseline. A hybrid model was used to account for stratification and clustering of the survey design, while adjusting for variables used to calculate sample weights. Results: During follow-up (median = 21 years), 3,284 deaths occurred, 107 of which had the underlying cause of prostate cancer. Any degree of baldness was associated with a 56% increased risk of prostate cancer-specific mortality (HR = 1.56; 95%CI = 1.02, 2.37) and, specifically, moderate balding was associated with an 83% increased risk of the outcome (HR = 1.83; 95%CI = 1.15, 2.92), each compared with no balding. Conversely, male pattern baldness was not statistically significantly associated with all-cause mortality. Conclusion: Our analysis suggests that male pattern baldness is associated with an increased risk of fatal prostate cancer, and supports the hypothesis of overlapping pathophysiological mechanisms. Impact: If the association between male pattern baldness and fatal prostate cancer is substantiated in future studies, male pattern baldness may contribute to predictive algorithms of prostate cancer risk, helping guide individuals as to whether they should opt to undergo cancer screening. Citation Format: Cindy Ke Zhou, Paul H. Levine, Sean D. Cleary, Heather J. Hoffman, Barry I. Graubard, Michael B. Cook. Male pattern baldness in relation to prostate cancer-specific mortality: A prospective analysis in the NHANES I Epidemiologic Followup Study (NHEFS). [abstract]. In: Proceedings of the 106th Annual Meeting of the American Association for Cancer Research; 2015 Apr 18-22; Philadelphia, PA. Philadelphia (PA): AACR; Cancer Res 2015;75(15 Suppl):Abstract nr 4603. doi:10.1158/1538-7445.AM2015-4603

Is Stress Worth it? Stress‐Related Health and Wealth Trade‐Offs for Entrepreneurs

Occupational stress is associated with numerous health problems that cost organisations considerable resources. We explore whether the detrimental effects of stress on individual health are accompanied by productive effects on individual performance for self‐employed people, thereby making stress somewhat “worth it” for this occupational group. Given that positive affect can serve as a stress‐buffering resource, we also examine the potential for positive affect (PA) to moderate these relationships. Our hypotheses are tested using data from the NHANES I Epidemiologic Follow‐up Study (NHEFS) that incorporated extensive demographic, medical history, nutritional, clinical, and laboratory data representative of the non‐institutionalised civilian US population. From this dataset we created a longitudinal matched sample of 688 self‐employed individuals and 688 employees, incorporating self‐reported and physiological measures of stress and health. Our findings indicate that (controlling for past income and prior health) self‐employed people experience greater stress than employees, and they experience a positive impact of stress on income despite a negative impact on physical health. These relationships are moderated by positive affectivity, where PA accentuates the positive effect of stress on personal income and mitigates the negative effect of stress on physical health.

Nutritional risk factors for tuberculosis among adults in the United States, 1971-1992.

The risk of developing tuberculosis (TB) may be related to nutritional status. To determine the impact of nutritional status on TB incidence, the authors analyzed data from the First National Health and Nutrition Examination Survey (NHANES I) Epidemiologic Follow-up Study (NHEFS). NHANES I collected information on a probability sample of the US population in 1971-1975. Adults were followed up in 1982-1992. Incident TB cases were ascertained through interviews, medical records, and death certificates. TB incidences were compared across different levels of nutritional status after controlling for potential confounding using proportional hazards regression appropriate to the complex sample design. TB incidence among adults with normal body mass index was 24.7 per 100,000 person-years (95% confidence interval (CI): 13.0, 36.3). In contrast, among persons who were underweight, overweight, and obese, estimated TB incidence rates were 260.2 (95% CI: 98.6, 421.8), 8.9 (95% CI: 2.2, 15.6), and 5.1 (95% CI: 0.0, 10.5) per 100,000 person-years, respectively. Adjusted hazard ratios were 12.43 (95% CI: 5.75, 26.95), 0.28 (95% CI: 0.13, 0.63), and 0.20 (95% CI: 0.07, 0.62), respectively, after controlling for demographic, socioeconomic, and medical characteristics. A low serum albumin level also increased the risk of TB, but low vitamin A, thiamine, riboflavin, and iron status did not. A population's nutritional profile is an important determinant of its TB incidence.

Comparison of a single global item and an index of a multi-item health status measure among persons with and without diabetes in the US

This study examined the hypothesis that a single global item can be substituted for an index of a multi-item assessment and lead to equivalent interpretative outcomes. Substitutability would be demonstrated if: (1) the two measures were strongly correlated, and regression analysis showed that the same variables accounted for variation in each measure, and (2) difference scores between multi-item and global scores were close to zero and remained so as socio-demographic and co-morbid conditions varied. A multi-item assessment was constructed by mapping items from the NHANES I Epidemiologic Follow-up Study (NHEFS), using available data for persons with and without diabetes, onto the health-status classification system of the Health Utilities Index Mark 1 (HUI), creating the NHEFS-HUI. NHEFS-HUI data, when correlated to the self-assessed health status (SAHS) item, revealed a coefficient of 0.55. Regression analyses identified 9 of 14 variables contributed to the variability of each health status index, but differences existed in which variables were significant for which measure. Five of the possible 14 difference scores for persons with diabetes and non-diabetics approached zero. Persons with diabetes had lower NHEFS-HUI scores than non-diabetics. These data were considered insufficient for demonstrating substitutability. Suggestions were made on how optimal substitutability could be achieved.

Counts of Neutrophils, Lymphocytes, and Monocytes, Cause-specific Mortality and Coronary Heart Disease: The NHANES-I Epidemiologic Follow-up Study

To examine the association of elevated counts of white blood cell types with increased risk of coronary heart disease (CHD) and death. Data were examined from the NHANES-I Epidemiologic Follow-up Study. Relative risks for death at ages 25 to 74 comparing the upper and lower tertiles of neutrophil count were: all causes 1.29 (95% CL, 1.14, 1.47), and cardiovascular causes 1.39 (95% CL, 1.15, 1.67) after adjusting for baseline risk factors. The increased risk of CHD and death from all causes and cardiovascular diseases appeared to be only partially due to effects of smoking. No association was seen for lymphocytes or monocytes.

Bone mineral density and the subsequent risk of cancer in the NHANES I follow-up cohort

BackgroudBone mineral density (BMD) is a marker of long-term estrogen exposure. BMD measurement has been used in this context to investigate the association of estrogen with breast cancer risk in three cohorts. In order to assess further BMD as a predictor of estrogen related cancer risk, the association of BMD with colorectal and corpus uteri cancer was investigated in the NHANES I Epidemiologic Followup Study (NHEFS) cohort along with breast cancer and prostate cancer.MethodsParticipants were members of the NHEFS cohort who had BMD measurement in 1974–1975. Age, race, and BMI adjusted rate ratios and 95% confidence intervals were calculated for incidence of cancers of the corpus uterus, breast, colorectum, prostate, and of osteoporosis and hip fracture related to baseline BMD.ResultsData were available for 6046 individuals. One hundred cases of breast cancer, 94 prostate cancers, 115 colorectal cancers, 29 uterine cancers, 110 cases of hip fracture and 103 cases of osteoporosis were reported between 1974 and 1993. Hip fracture and osteoporosis were both significantly inversely associated with BMD. Uterine cancer was positively associated (p = 0.005, test for linear trend) and colorectal cancer negatively associated (p = 0.03) with BMD. No association was found between elevated BMD and incidence of breast cancer (p = 0.74) or prostate cancer (p = 0.37) in the overall cohort, although a weak association was seen between BMD and subsequent breast cancer incidence when BMD was measured in post-menopausal women (p = 0.04).ConclusionThe findings related to cancers of the uterus and colorectum as well as the weak association of BMD with breast cancer strengthen the use of BMD as a marker of estrogen exposure and cancer risk.

Tea consumption and the reduced risk of colon cancer -- results from a national prospective cohort study.

This study examines the relationship between tea consumption and colon cancer risk in the US population. Data from the NHANES I Epidemiologic Follow-up study (NHEFS) were used to examine the hypothesis. Cox proportional hazard models were used to examine the hypothesis of a protective effect of frequent tea consumption on colon cancer occurrence. Due to differences in the precision of the exposure data, we analysed two cohort periods based on the NHEFS. Cohort I was based on the survey conducted at the NHEFS baseline and Cohort II began at the first follow-up. After excluding non-incidence cases and cases lost to follow-ups, there were 2359 tea users and 6498 non-tea users at baseline and 7656 tea users and 4514 non-tea users at the first follow-up. : After adjusting for confounders, the relative risks of colon cancer are 0.57 (95% confidence interval (CI) 0.42, 0.78) and 0.59 (95% 1.00) for subjects who consumed <or=1.5 cups and >1.5 cups per day, respectively, compared with non-tea users in Cohort II. Although more women consumed tea and the mean intake was higher, the preventive effect of tea consumption on colon cancer was found predominantly in men. The relative risks of colon cancer are 0.41 (95% 0.66) for men who consumed <or=1.5 cups day-1 and 0.30 (95% 0.98) for >1.5 cups day-1 of tea consumption (P-value for trend <0.01). No significant results were found in Cohort I. This study suggests an inverse association between colon cancer risk and habitual tea consumption.

Nutritional Status of Folate and Colon Cancer Risk: Evidence from NHANES I Epidemiologic Follow-up Study

PURPOSE: This manuscript utilized the NHANES I Epidemiologic Follow-up Study (NHEFS), a national probability sample of the U.S. non-institutionalized population, to examine whether the intake of folate at baseline is associated with colon cancer risk. METHODS: The NHEFS consists of 14,407 subjects with 20 years of follow-up. Sociodemographic status, dietary information, family history of colon cancer, alcohol and aspirin use, smoking status, and body mass index (BMI) are included in the Cox proportional hazard model to examine confounding effects. RESULTS: After adjusting for confounders, a marginally significant association was observed between folate intake and reduced colon cancer risk. Gender and alcohol consumption appears to have an interactive effect with this association. The stratified results suggest that dietary folate is significantly inversely associated with colon cancer in men (relative risk (RR) = 0.40, 95% confidence interval (CI) = 0.18, 0.88) who consumed more than 249 μg/day of folate and that there is a significant dose-response relationship ( p = 0.03). The association did not reach statistical significance in women. Using a composite dietary profile, we found that there is a significantly increased risk for men who consumed low-folate, low-methionine, and high alcohol diets when compared to male non-drinkers who consumed high-folate and high methionine diets (RR = 2.67, 95% CI = 1.16, 6.16). CONCLUSIONS: This study found significant association between folate intake and reduced colon cancer risk among men and non-drinkers, but not women or drinkers. The study supports a synergistic interaction between intakes of folate, methionine and alcohol and colon cancer risk.

Preventable smoking and exercise-related hospital admissions: A model based on the NHEFS

Background: The NHANES I Epidemiologic Follow-Up Study (NHEFS), a longitudinal study of a representative sample of U.S. adults, makes it possible for the first time to develop a simulation model relating hospital admissions to baseline clinical risk factors for the general adult population. The model is presented here and used to project the impact on hospital admissions of changes in smoking behavior and sedentary lifestyle. Methods: Three kinds of projections were calculated for the cohort of adults aged 45 to 74 at baseline: projections of hospital admissions in the absence of the risk factor; projections that reflect a 10-year lag between behavior change and full health benefit; and projections that reflect both lag and incomplete adherence to behavior change. For incomplete adherence we assumed that only 10% of the at-risk population changed their behavior. Results: Tests of the simulation model showed that it agreed with a Cox analysis of the hospital data and accurately projected observed hospital admissions over the study period. The projections showed that eliminating smoking would reduce annual rates of all-cause hospitalization among older adults by 8.9% 20 years after baseline. Eliminating inactivity would reduce them 4.6%. Introducing a lag of 10 years between behavior modification and full benefit delayed the impact on hospitalization rates but the effect at 20 years was the same. When only 10% of the population at risk stopped smoking or became physically active, a percentage that reflects the effectiveness of current interventions, annual hospitalization rates at 20 years fell by 0.9% and 0.5%, respectively. Conclusions: Substantial reductions in hospital admissions can be achieved by interventions to prevent smoking, help smokers quit, or encourage inactive persons to become active. Improving adherence can markedly improve the impact on hospitalizations. The costs of these efforts would be rewarded not only by better health, but by lower expenditures for hospitalization.

Incidence of hypertension and educational attainment: the NHANES I epidemiologic followup study. First National Health and Nutrition Examination Survey.

Previous research has demonstrated the association between cardiovascular disease and education. However, few studies have described the incidence of hypertension, a risk factor for cardiovascular disease, by education or other socioeconomic status indicators. To examine the association between hypertension incidence and education, the authors analyzed data from the First National Health and Nutrition Examination Survey (NHANES I) Epidemiologic Followup Study (NHEFS) (1971-1984). The relative risk of hypertension incidence (blood pressure > or =160/95 and/or using antihypertensive medication) by education was calculated for non-Hispanic Whites (aged 25-64 years) and non-Hispanic Blacks (aged 25-44 years) normotensive at baseline using Cox proportional hazards models. The age-adjusted relative risk of hypertension incidence among persons with less than 12 years of education compared with those with more than 12 years was significant among non-Hispanic Whites aged 25-44 years (men: relative risk (RR) = 2.14, 95% confidence interval (CI): 1.29, 3.54; women: RR = 2.06, 95% CI: 1.39, 3.05) but not among non-Hispanic Blacks (RR = 1.16, 95% CI: 0.63, 2.14). Relative risks for non-Hispanic White men remained stable after adjusting for age, systolic blood pressure, body mass index, and region of residence; relative risks for non-Hispanic White women were reduced but remained significant. Non-Hispanic White men and women aged 45-64 years with less than 12 years of education were not at higher risk of developing hypertension compared with their more educated counterparts. These results demonstrate a significant interaction between age and education with an independent association between education and hypertension incidence among younger but not older non-Hispanic White men and women.

Dietary sodium intake and incidence of congestive heart failure in overweight U.S. men and women: NHANES I epidemiologic follow-up study (NHEFS)

Dietary sodium intake and incidence of congestive heart failure in overweight U.S. men and women: NHANES I epidemiologic follow-up study (NHEFS)

Serum uric acid and cardiovascular mortality the NHANES I epidemiologic follow-up study, 1971-1992. National Health and Nutrition Examination Survey.

Although many epidemiological studies have suggested that increased serum uric acid levels are a risk factor for cardiovascular mortality, this relationship remains uncertain. To determine the association of serum uric acid levels with cardiovascular mortality. Cross-sectional population-based study of epidemiological follow-up data from the First National Health and Nutrition Examination Survey (NHANES I) from 1971-1975 (baseline) and data from NHANES I Epidemiologic Follow-up Study (NHEFS). A total of 5926 subjects who were aged 25 to 74 years and had serum uric acid level measurements at baseline. Ischemic heart disease mortality, total cardiovascular mortality, and all-cause mortality, compared by quartiles of serum uric acid level. In an average of 16.4 years of follow-up, 1593 deaths occurred, of which 731 (45.9%) were ascribed to cardiovascular disease. Increased serum uric acid levels had a positive relationship to cardiovascular mortality in men and women and in black and white persons. Deaths due to ischemic heart disease in both men and women increased when serum uric acid levels were in the highest quartile compared with the lowest quartile (men, >416 vs <321 micromol/L; risk ratio, 1.77 [95% confidence interval [CI], 1.08-3.98]; women, >333 vs <238 micromol/l; risk ratio, 3.00 [95% CI, 1.45-6.28]). Cox regression analysis showed that for each 59.48-micromol/L increase in uric acid level, cardiovascular mortality and ischemic heart disease mortality increased. Hazard ratios for men were 1.09 (95% CI, 1.02-1.18) and 1.17 (95% CI, 1.06-1.28), and for women were 1.26 (95% CI, 1.16-1.36) and 1.30 (95% CI, 1.17-1.45), respectively, after adjustment for age, race, body mass index, smoking status, alcohol consumption, cholesterol level, history of hypertension and diabetes, and diuretic use. Further analysis, stratifying by cardiovascular risk status, diuretic use, and menopausal status, confirmed a significant association of uric acid and cardiovascular mortality in all subgroups except among men using diuretics (n=79) and men with 1 or more cardiovascular risk factors (n=1140). Our data suggest that increased serum uric acid levels are independently and significantly associated with risk of cardiovascular mortality.

Discussion

The long-term impact of smoking cessation on mortality is assessed among two U.S. populations: a large cohort of U.S. veterans aged 55-64 at entry and followed from 1954 through 1979 and the NHANES I Epidemiologic Followup Study (NHEFS) cohort of a national sample of U.S. adults aged 55-74 at entry and followed from 1971 through 1992. Direct and indirect survey data indicate that 50-70% of those who were current cigarette smokers at entry had quit smoking during the 19- to 26-year follow-up periods. The impact of smoking cessation on mortality among the cigarette smokers as a whole has been assessed by determining the time trend of the relative risk (RR) of death and 95% confidence interval (CI) for the cigarette smokers compared with never-smokers over the entire follow-up period in both cohorts. The total death rates for the 1954/57 U.S. veteran smokers as a whole (63,159 males) have converged only slightly toward those of never-smokers, from RR = 1.65 (1.58-1.72) during 1954-1959 to RR = 1.61 (1.58-1.63) during 1954-1979. The lung cancer death rates for 1954/57 smokers as a whole have not converged toward those of never-smokers, with RR = 10.89 (7.70-15.41) during 1954-1959 and RR = 11.10 (9.78-12.61) during 1954-1979. The total death rates for the 1971-1975 NHEFS smokers as a whole (694 males and 1116 females) have not converged toward those of never-smokers. For males, RR = 1.92 (1.46-2.52) during 1971-1982 and RR = 1.96 (1.63-2.36) during 1971-1992; for females, RR = 1.79 (1.31-2.46) during 1971-1982 and RR = 1.79 (1.47-2.17) during 1971-1992. The lung cancer death rates have diverged, based on small numbers of deaths. For males, RR = 15.76 (2.06-120.61) during 1971-1982 and RR = 22.20 (5.31-92.92) during 1971-1992; for females, RR = 2.92 (0.57-15.06) during 1971-1982 and RR = 4.74 (1.94-11.59) during 1971-1992. These trends are contrary to the substantial convergence predicted by the death rate trends among U.S. veterans who were former smokers at the beginning of follow-up. While these results confirm that those former smokers who survive for at least 5 years experience death rates that converge toward those of never-smokers, they also indicate that a cohort of cigarette smokers that undergoes substantial cessation experiences a death rate that does not converge toward the death rate of never-smokers. The fact that there has been no convergence for lung cancer is quite surprising, as this is the disease most strongly linked to smoking and smoking cessation and less likely to be influenced by other lifestyle factors. Further investigation is needed for a complete understanding of the impact of smoking cessation.

Prospective Assessment of Coronary Heart Disease Risk Factors: The NHANES I Epidemiologic Follow-up Study (NHEFS) 16-Year Follow-up

Objective: Our specific aim in the 16 year prospective NHANES I epidemiologic follow-up study (NHEFS) was to assess the important roles of modifiable dietary and behavioral factors in causation and prevention of deaths and hospitalizations for coronary heart disease (CHD).Methods: Using NHEFS 16 year follow-up data (1971 to 1987), we studied 5811 subjects, 1958 with and 3853 without CHD events, using logistic regression.Results: In age groups 40 to 49, 50 to 59, 60 to 69, and 70 to 74 years (at study entry in 1971–4), the numbers of men and women were respectively 597 and 1019, 570 and 619, 932 and 1042, and 486 and 546. The following factors were independently positively associated (p<.05) with CHD: age, serum cholesterol, body mass index, cigarette use, and region (Midwest, Northeast). The following factors were independently negatively associated (p<.05) with CHD: gender (female), race (black), fish intake, alcohol, high school education, moderate exercise, and moderate and heavy habitual physical activity. Subjects with serum cholesterol >249 mg/dl benefitted less (p=.04) from fish intake than those with 209 to 249 or <209, and benefitted less (p=.03) from alcohol intake (CHD incidence [%]):Cholesterol<209209–249>249<209209–249>249FishNone39%40%45%AlcoholNone38%39%41%<1/week33%32%41%<1/week27%28%38%≥1/week26%30%38%1–6/week23%26%37%≥1/day30%29%42%Conclusions: These associations emphasize the important role of modifiable dietary and behavioral factors in the causation and prevention of CHD.

An Introduction to the Analysis of Paired Hazard Rates in Studies of the Family

Hazard rate models in studies of the family are becoming more common. Ironically family studies scholars have generally used these techniques based on individuals rather than explicitly on couples or families. We describe and apply selected techniques used to analyze paired hazard rates when event times are right censored. We illustrate the techniques by looking at mortality patterns in husbands and wives. Recently developed measures and models are introduced and applied to a sample of 1031 married couples from the [U.S.] National Health and Nutrition Examination Survey (NHANES I) and the NHANES I Epidemiologic Follow-Up Study (NHEFS). The advantages and disadvantages of the measures are discussed. (EXCERPT)

Establishing the Nadir of the Body Mass Index-Mortality Relationship: A Case Study

Many studies have demonstrated a nonmonotonic relationship between mortality and body mass index (BMI), with excess mortality occurring at both low and high levels. Although much discussion and many different analyses have appeared, to our knowledge no attempt has been made to quantitatively establish the BMI at which minimum mortality (BMImin) occurs or to establish confidence intervals for this BMI, accounting for the asymmetry of the relationship. We model the nonmonotonic relationship between BMI and mortality in 13,242 black and white participants of the NHANES-I Epidemiologic Follow-up Study to estimate the BMI at which minimum mortality occurs. In our analyses we consider the joint relationship of age, smoking status, and BMI to mortality. We present two methodologies for estimating BMImin: a logistic regression model with a transformation of BMI to accommodate asymmetry and a changepoint model as suggested by Goetghebeur and Pocock. We establish confidence intervals for BMImin using the delta method and bootstrap sampling for the logistic and the profile likelihood and bootstrap sampling for the changepoint model. We also present formal tests for the heterogeneity of BMImin by smoking status, sex, race, and age. Only the interaction between race and BMI is significant; the BMImin is somewhat higher for blacks than whites. Finally, we discuss the problem of goodness-of-fit statistics when the relationship between the characteristic and the outcome is nonmonotonic.

Establishing the Nadir of the Body Mass Index-Mortality Relationship: A Case Study

Abstract Many studies have demonstrated a nonmonotonic relationship between mortality and body mass index (BMI), with excess mortality occurring at both low and high levels. Although much discussion and many different analyses have appeared, to our knowledge no attempt has been made to quantitatively establish the BMI at which minimum mortality (BMImin) occurs or to establish confidence intervals for this BMI, accounting for the asymmetry of the relationship. We model the nonmonotonic relationship between BMI and mortality in 13,242 black and white participants of the NHANES-I Epidemiologic Follow-up Study to estimate the BMI at which minimum mortality occurs. In our analyses we consider the joint relationship of age, smoking status, and BMI to mortality. We present two methodologies for estimating BMImin: a logistic regression model with a transformation of BMI to accommodate asymmetry and a changepoint model as suggested by Goetghebeur and Pocock. We establish confidence intervals for BMImin using the delta method and bootstrap sampling for the logistic and the profile likelihood and bootstrap sampling for the changepoint model. We also present formal tests for the heterogeneity of BMImin by smoking status, sex, race, and age. Only the interaction between race and BMI is significant; the BMImin is somewhat higher for blacks than whites. Finally, we discuss the problem of goodness-of-fit statistics when the relationship between the characteristic and the outcome is nonmonotonic.

Relative weight and mortality in U.S. blacks and whites: Findings from representative national population samples

PURPOSE: To examine the impact of relative weight on mortality in black and white men and women. METHODS: Two representative national populations samples were used: the NHANES-I Epidemiologic Follow-up Study (NHEFS), and the National Health Interview Survey (NHIS). The principal analysis focused on 13,242 participants in the NHEFS and 114,954 in the NHIS. Minimum mortality was estimated from both categorical analysis and a logistic model. RESULTS: Minimum mortality ranged from a body mass index (BMI) of 25 to 32 kg/m 2. The modelestimated BMI of minimum mortality for NHEFS was 27.1 (24.8–29.4, 95% CI), 26.8 (24.7–28.9, 95% CI), 24.8 (23.8–25.9, 95% CI) and 24.3 (23.2–25.4, 95% CI); for black men, black women, white men and white women, respectively, whereas for NHIS the corresponding values were 30.2 (24.8–35.6, 95% CI) 26.4 (24.2–28.7, 95% CI), 27.1 (25.5–28.7, 95% CI), and 25.6 (24.2–27.0, 95% CI). In all groups the shape of the relative risk curve was virtually identical and a broad range of BMI values in the middle of the distribution was associated with low relative mortality risk. Averaging the results from both surveys, the observed BMI of minimum risk was 3.1 kg/m 2 higher in black men and 1.5 kg/m 2 higher in black women than in their white counterparts; when adjusted for covariates these differences were only of borderline statistical significance, however. CONCLUSIONS: Because of the wide range of BMI values associated with low risk, and the consistency of the point of the up-turn in risk, group specific definitions of optimal values do not appear to be warranted.