The Influence of Metabolic Syndrome in Predicting Mortality Risk Among US Adults

Importance of Metabolic Syndrome Even in Adults With Normal Weight

Ting Huai Shi, BA; Binhuan Wang, PhD; Sundar Natarajan, MD, MSc


Prev Chronic Dis. 2020;17(5):E36 

In This Article


Study Design and Sample Population

We used data from 1999–2010 NHANES. NHANES is a national publicly available database that has de-identified health and nutritional data on the US population. The data are compiled through surveys using interviews, physical examinations, and laboratory results. Participants are selected according to a complex sample design that clusters and stratifies the US population for the corresponding year. Some underrepresented groups are oversampled to provide more precise and reliable estimates. The sample was weighted to be representative of the US population for the given years using NHANES analytic guidelines for combining data across years.[14] NHANES surveys are conducted continuously in 2-year intervals; from 1999 through 2010 (6 cycles), our study period, 62,160 people participated. Participants are interviewed about demographic, lifestyle, and health-related information. Medical and physiologic measurements are taken during a physical examination.[15] We linked NHANES data with data from the National Death Index from 1999 to 2011; this database provides follow-up mortality data for up to 150 months for NHANES participants aged 18 or older.[16] A minimum of 10 years is suggested for observing the effects of MetS on mortality.[17]

We preliminarily excluded NHANES participants if they were younger than 20 years (n = 29,696). We then excluded participants if they had BMI less than 18.5 (n = 487); were missing data on education (n = 90), poverty-income ratio (n = 2,939), mortality follow-up (n = 49), MetS criteria (n = 18,183), or BMI (n = 2,472); or had a nonpositive survey weighting value (n = 1,934). In addition, we excluded participants with follow-up periods shorter than 12 months after the time of the survey to account for any frailty due to serious preexisting conditions (n = 648). The final analytic sample of 12,047 participants aged 20 to 85 had data for all variables examined in our study, eligible follow-up mortality data, and no preexisting frailty. NHANES collects data for people older than 85 but does not report these extreme values to protect privacy.

Random subsampling accounted for most missing data points. Subsamples of participants were randomly selected to participate in various survey topics or laboratory testing. For example, less than one-third of all participants were tested for fasting glucose or triglycerides. Each subsample was further weighted so that each represents the US population for the given year. Less than 10% of missing data were due to nonresponse or refusals. Similarly, less than 10% of mortality data was lost at follow-up.


We categorized the study sample into 3 weight groups based on BMI according to standard definitions: normal weight (18.5 to <25.0 kg/m2), overweight (25.0 to <30.0 kg/m2), and obese (≥30.0 kg/m2). We further divided each weight group into 2 groups according to whether the participant met criteria for MetS. We defined MetS according to criteria provided by the American Diabetes Association, in which MetS is indicated by the presence of 3 or more of the following 5 criteria: central or abdominal obesity (men, >40 in waist circumference; women, >35 in waist circumference), triglycerides ≥150 mg/dL; high-density lipoprotein cholesterol (men <40 mg/dL; women <50 mg/dL), blood pressure ≥130/85 mm Hg, and fasting glucose ≥100 mg/dL.[18]

We included the following covariates in the multivariate adjusted analyses: age (20–85), sex (male, female), race/ethnicity (non-Hispanic white, non-Hispanic black, Mexican American, other Hispanic, other/multiracial), education (<9th grade, 9th–11th grade, high school graduate, some college, college graduate), poverty-income ratio (0 to ≥5), smoking status (never smoked, former smoker, current smoker), and physical activity level (active, insufficiently active, and inactive). Physical activity level was defined by using categories proposed by Zhao et al,[19] which were created according to physical activity guidelines published by the US Department of Health and Human Services[20] as follows: 1) physically active if they reported ≥150 minutes per week of moderate-intensity activity or ≥75 minutes per week of vigorous-intensity activity or ≥150 minutes per week of an equivalent combination (≥150 min/week); 2) insufficiently active if they reported some physical activity but not enough to meet the active definition (>0 to <150 min/week), or 3) inactive if they reported no (0 min/week) moderate-intensity or vigorous-intensity physical activity. Because of differences in questionnaires between NHANES cycles, we included only leisure-time physical activity in our analysis.

Statistical Analysis

Initial analyses using SAS complex survey frequency and means procedures that take into account weighting, stratification, and clustering of the data generated the descriptive statistics. We used the LIFETEST procedure to generate the unadjusted mortality data for each MetS–BMI category and the log rank test to determine significant differences between categories.

We evaluated the independent effect of obesity and MetS categories on mortality by using Cox proportional hazards models that accounted for the complex sampling design (weighting, stratification, and clustering), adjusted for age, sex, race/ethnicity, education, poverty-income ratio, smoking status, and physical activity to account for covariates commonly associated with mortality. Other important risk factors such as blood pressure, cholesterol, and blood glucose were already included in the definition of MetS. We excluded covariates, such as alcohol consumption, that did not significantly improve the statistical model. We used the 6-level BMI–MetS variable to find the hazard ratio of each group compared with the normal-weight–no-MetS group for all-cause mortality, cardiovascular mortality, and cancer mortality. The 6 groups were normal-weight–MetS, normal-weight–no-MetS, overweight–MetS, overweight–no-MetS, obese–MetS, and obese–no-MetS. We chose the normal-weight–no-MetS group as the referent because we hypothesized that it would be the healthiest. We then tested the moderating effect of BMI on MetS and mortality by testing the interaction between weight groups and MetS. To support the interaction analysis, we also tested the effect of MetS in each weight group, using the contrast statement to directly compare normal-weight–MetS participants and participants in other categories.

In a further analysis, while accounting for the complex sampling design and controlling for the same covariates, to determine the incremental influence of MetS on mortality, we compared each MetS group with their no-MetS counterparts in each BMI group. We performed all statistical analyses in 2017 using SAS version 9.4 (SAS Institute Inc), and a 2-sided P value <.05 was considered significant.