Identification of Limitations in Epidemiologic Studies

Identification of Limitations in Epidemiologic Studies (30 points, 2 points each)

Indicate the one major type of limitation that is possible in each of the situations described below. Choose from the following: selection bias, information bias, confounding, random error, or limited generalizability. Choose only one per example.

1. A survey of coronary artery disease prevalence is conducted based on the question, “Has a doctor ever told you that you have coronary artery disease?”
2. A clinical trial is conducted in which middle aged people at high risk for coronary heart disease are randomized to treatment (physical activity coaching and classes provided) and control (lectures on their health) groups.  The randomization process was able to prevent any significant age differences between the treatment and control groups.  What sort of bias was prevented?
3. An epidemiologist reported the following conclusion from a large study during a national meeting: “Daily use of 500 milligrams of vitamin C for a period of one year was associated with reduced frequency of upper respiratory infections in children under 10 years of age.” (Relative Risk = 0.56, 95% Confidence Interval = 0.11- 4.21).
4. A population survey of adults residing in Chicago is conducted, where a probability sample is drawn from voter registration rolls.
5. In order to prevent confounding by gender and smoking status, a study of the relationship between exercise and heart attacks is conducted with a sample of men who do not smoke.  Although confounding by smoking status was prevented, another limitation was thereby introduced into the study.  What was this limitation?
6. An ecological study of ambient air pollution and infant mortality rates among 50 world cities is conducted.  An association is found.  However, the cities with the highest levels of air pollution are also the poorest.
7. A case-control study was designed to test whether persons exposed to certain types of pesticides during early childhood were more likely to develop neurological problems in later adulthood. The results were statistically significant. The cases consisted of those with severe but treatable neurological disorders, and the controls consisted of those without any diagnosed neurological disorders. Both the cases and controls were asked detailed questions about their pesticide exposures during early childhood.
8. In a study of weight loss, a scale was used that added a few pounds more or subtracted a few pounds less each time the scale was used.  This resulted in an imprecise measurement of weight, the direction of which, was completely unpredictable.  Since the researcher did not realize this, no efforts were made to adjust for the problem by either getting another scale or by taking repeated measurements that would average out over time. 
9. In a study to determine the incidence of a chronic disease, 150 people were examined at the end of a three-year period. Twelve cases were found, giving a cumulative risk of 8%. Fifty other members of the initial cohort could not be examined; 20 of these 50 could not be examined because they died.
10. A study of the relationship between contact lens use and the risk of eye ulcers is conducted.  The crude relative risk was 3.0.  The investigators had failed to adjust for age, which, had they done so, would have yielded a relative risk of 1.5.
11. A study investigating the risk factors associated with clear-cell carcinoma in young women collects exposure data from live controls and from mothers of the deceased cases.
12. A survey is taken at the end of an MPH course in which students are asked to evaluate the course and their professor.  Out of 20 students in the class, only 5 respond to the online survey.  These 5 include the two who were failing the class, one who got into an argument with the professor over his exam grade, one who loved the class so much she decided to become an epidemiologist, and one “average” student.
13. A case-control study is conducted testing the hypothesis that fatal motor vehicle accidents are more likely if the driver is a smoker.  A positive association is found.  Later heavy drinking is also investigated.  When heavy drinking was controlled for in the final analysis, there was no association between smoking and fatal motor vehicle accidents.  Which limitation was influencing the crude analysis?
14. A study was conducted of 3,000 military troops deployed at an atomic test site in Nevada to detect the occurrence of leukemia.  A total of 1,870 persons were successfully traced by the investigators, and an additional 443 contacted the investigators on their own as a result of the publicity about the study.  Four cases of leukemia occurred among the 1,870 individuals traced by the investigators, and an additional four occurred among those individuals who contacted the investigators on their own.
15. In a study of AIDS and esophageal candidiasis, exposed participants with AIDS are compared to a group of unexposed persons without a diagnosis of AIDS on their respective frequencies of esophageal candidiasis.  The presence of esophageal candidiasis is assessed by questionnaire, and no association is found (OR = 1.1).  However, in a different study, esophageal candidiasis was ascertained by biopsy, yielding an OR of 8.4. Which limitation existed in the first study?