Question
These data are from a case-control study on biomass cooking fuel exposure and tuberculosis [Perez-Padilla et al. Int J Tuberc Lung Dis 2001;5:1-7.]. Cases were 288 patients with an active smear or culture positive tuberculosis, and controls were 545 patients with ear, nose and throat ailments from the same hospital, seen at the same time. Past or present exposure to biomass smoke was obtained by interview.
The table gives the data for the association between current exposure to biomass smoke and TB:
TB
| ||
Biomass Smoke Exposure Status
Exposed
Unexposed
|
Cases Controls
50 21
238 524
|
Total
71
762
|
Total
|
288 545
|
833
|
The authors found a significant association between indoor smoke and TB (OR = 5.2). After adjusting for several confounders, a significant association remained (adjusted OR = 2.2).
A. In this study, exposure assessment was made by interviews. If tuberculosis patients provide more accurate histories about past exposure to biomass smoke compared to controls, what bias can occur? Would it be differential or non-differential?
B. If tuberculosis patients changed their cooking fuel to cleaner fuels like gas after developing the disease, how could it affect the odds ratio for current use of a biomass stove?
C. In this study, patients with ear, nose and throat (ENT) conditions were chosen as controls. If some ENT conditions are associated with biomass smoke (like allergic rhinitis), could this affect the odds ratio? What would this type of bias represent?
D. If cases and controls had similar problems in remembering the cooking fuels they used in the past, what type of misclassification could have occurred? How will that affect the odds ratio?
E. If the interviewers who did the exposure assessment had known the disease status of the cases and controls, how could this have affected the odds ratio? What type of bias would this be?
F. Biomass smoke is known to contain many toxic chemicals. If exposure to biomass smoke is one cause of tuberculosis, and if it leads to rapid progression of disease and death among these patients, how would it affect the odds ratio in a case-control study examining biomass smoke as a risk factor for TB?
G. The authors could have improved the measurement of exposures in this study by actually quantifying indoor air pollution levels due to biomass smoke using instruments. If they had done that, how would it have affected the precision of their OR estimates?
H. Suppose instead of interviews, biomass smoke exposure was assessed using air monitors worn by individuals in the study. You classify exposure to biomass smoke as high, medium and low using the monitor results. After you finish your study, the monitor company calls to tell you their monitors have been acting strangely. A random set of monitors in your study malfunctioned such that some over-estimated smoke exposure while others under-estimated smoke exposure (malfunctions affected cases and controls equally). What, if anything, can you say about the odds ratios you calculated in your study?
Source: 250B Problem Set
