Med Biostat&Epidemiology
Homework #3
Due November 4, 2020 Name ________________________
- Past studies of chronic obstructive pulmonary disease (COPD) have found that overweight and obese individuals with COPD have lower mortality rates compared to normal weight individuals with COPD. Those studies included only current or former smokers, so researchers conducted a study that included never smokers with COPD (Am J Epidemiol 2019; 188:1977).
A total of 1723 adults with COPD were enrolled and followed for mortality for a mean of 9.5 years. Participants were classified as overweight or normal weight based in body mass index (BMI) computed at the beginning of the study.
(a) What study design is this?
(b) What is one advantage of using this study design to examine BMI and mortality in people with COPD?
(c) What is one disadvantage of using this study design to examine BMI and mortality in people with COPD?
(d) Cigarette smoking is a known cause of mortality, and smokers tend to have a lower BMI than nonsmokers.
(d-i) In addition to being a cause of the outcome and associated with the exposure, what else must be true for smoking to be a confounder in the relationship between BMI and mortality?
(d-ii) Do you think the criterion in (d-i) is true (i.e., do you think smoking could be a confounder)? Why or why not?
(e) The association between BMI and mortality among people with COPD was examined using proportional hazards regression, based on person-years of follow-up for each participant. Hazard ratios were obtained from the proportional hazards model, with the hazard ratios (HRs) adjusted for age, sex, ethnicity, and education.
(e-i) What is a hazard ratio?
(e-ii) What is the null hypothesis for a hazard ratio?
Investigators stratified the analysis on smoking status, examining the hazard ratio for overweight vs normal weight individuals separately for ever smokers and never smokers.
(f) In ever smokers with COPD, the hazard ratio for mortality for overweight individuals compared to normal weight individuals was 0.59, with a 95% confidence interval of 0.45-0.90.
(f-i) Interpret the hazard ratio.
(f-ii) Interpret the 95% confidence interval.
(f-iii) Is this hazard ratio statistically significant? Why or why not?
(g) In never smokers with COPD, the hazard ratio for mortality for overweight individuals compared to normal weight individuals was 1.37, with a 95% confidence interval of 0.60-3.18.
(g-i) Interpret this hazard ratio.
(g-ii) Is this hazard ratio statistically significant? Why or why not?
(h) The test of homogeneity (i.e., test of heterogeneity, test of interaction) comparing the hazard ratios of 0.59 for ever smokers and 1.37 for never smokers gave a p value of 0.01. Based on this p value, the researchers concluded that in this study “of individuals with COPD, ever/never smoking modified the association between body mass index and all-cause mortality.” What does this statement mean?
