Study Participants
We used a nationally representative sample of cross-sectional data from the National Health and Nutrition Examination Survey (NHANES) 2011-2012 cycle. NHANES is a survey conducted by the Center for Disease Control and Prevention (CDC) every two years, including demographic, interviews, laboratory, and examination data. We selected the 2011-2012 survey cycle because this survey cycle exclusively, after 1999-2000, contains tuberculosis data. Out of 13,431 people selected for NHANES from 30 different study locations in 2011-2012, 756 completed the interview, and 9,338 were examined. Considering the low prevalence of Tuberculosis in the US, no inclusion and exclusion criteria were considered to account for the sample size. The analytic sample consisted of all 9,756 people who responded to the survey.
Study Variables
We included demographics, questionnaires, and examination data for the study. Total cholesterol level was considered the primary exposure of interest. The total cholesterol was defined as the sum of HDLs, LDLs, and 20% of their triglyceride level. The total cholesterol level was categorized into three categories: low cholesterol level (0 mg/dL – 120 mg/dL), average cholesterol level (120 mg/dL – 240 mg/dL), and high cholesterol level (240 mg/dL and more). The TB prevalence included people who had (1) positive reactions to the skin test, (2) induration greater than 15mm, (3) positive blood and tine test, and (4) people who were prescribed TB medicine. The covariates included age, gender, race, education, birth country, household size, and smoking status. The birth country was categorized into people born in the US and outside the US. We included the birth country in the analysis as there is a high prevalence of TB in many other countries compared to the US. Smoking status was included, given its pulmonary and immunity effects among people who smoke.
Statistical Analysis
The overall sample population was described using frequencies for the categorical variables and means and medians for the continuous variables. The sample population was also assessed based on TB prevalence. We used the student’s t-test and chi-square tests to compare the covariates among those with and without TB. Fisher’s exact test was used for the demography with low TB prevalence. Eventually, since TB seems to be a rare outcome, the odds ratio was calculated to understand the odds of developing TB based on the cholesterol level. The estimates and crude odds of tuberculosis based on the cholesterol level were assessed using bivariate logistic regression. Multivariate logistic regression was performed using sex, age, race, household size, and country of birth as covariates. The sample weights could not be used as TB prevalence data included self-reported and examination data.
Results
Of 9,756 sample participants, 986 (10.11%) contributed to TB prevalence. The median age of the sample population was 26, and it did not significantly differ based on TB prevalence. There were approximately four people in a household, regardless of TB prevalence, on average, in this study. The mean cholesterol level was 190.09 mg/dL among people with TB compared to 182.32 mg/dL among people without TB. Participants with TB were primarily female, non-Hispanic white had less than a high school of education, US-born, and non-smokers, as presented in Table 1 in the Appendix.
The bivariate logistic regression yielded an odds ratio of 1.26 (1.004, 1.59) for people with high cholesterol levels compared to people with average cholesterol levels suggesting a 26% increase in odds of tuberculosis among people with high cholesterol compared to average cholesterol levels. In addition, the bivariate logistic regression also yielded an odds ratio of 0.52 (0.32, 0.84) for people with low cholesterol levels compared to average cholesterol levels suggesting a 48% decrease in the odds of TB among people with low cholesterol levels compared to people with average cholesterol levels. The bivariate analysis suggested high cholesterol level as the risk factor and low cholesterol level as the protective factor per our hypothesis, as presented in Table 2 in the Appendix.
On the contrary, after the model was adjusted for gender, race, education, birth country, smoking, and household size, the multivariate logistic regression yielded an odds ratio of 1.40 (1.05, 1.88) for people with high cholesterol levels compared to people with average cholesterol level suggesting 40% increase in odds of TB among people with high cholesterol level. The odds ratio of the low cholesterol level compared to the average cholesterol level was 0.65 (0.36, 1.20), suggesting an insignificant relationship between the total cholesterol level and TB prevalence. For the covariates, Mexican Americans and non-Hispanic Black, compared to non-Hispanic white, seemed to have lower odds of TB. The fully adjusted multivariate model is demonstrated in Table 3 in the Appendix.