Epidemiology EPIB-695 Final Exam Monday, 10 April 2006

Epidemiology EPIB-695Final examMonday, 10 April 20061.This is an open-book exam. Textbooks, notes, and dictionaries are allowed.However, these items cannot be shared with other students.2.You must answer all questions. The total number of points for this exam is 100.3.The exam lasts two hours. No extra time will be granted.4.Desk calculators are allowed (but cannot be shared). Unless indicated otherwise,however, you do not need to carry out any computation. For example, if in a 2x2table a 2, b 8, c 24, and d 74, it is sufficient in an answer to write that the oddsratio is (2x74)/(8x24).The use of other electronic devices such as computers or BlackBerries is strictlyforbidden.5.When the answer is a measure with units, always indicate those units.6.The vocabulary used in this exam is always taken in its strict specific sense, asused in this course.7.You may answer in English or in French.S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc1

Question 1 ( 35 points)The abstract and the table below present results of a study of mammography and breastcancer mortality conducted by Cooper et al.S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc2

a)b)Rothman (course textbook, page 203) describes biases that may affect theassessment of screening programs, including lead-time bias, and the effect oflength-biased sampling. Define lead-time bias and length-biased sampling. (3-5 sentences each) For each bias, indicate whether it may have been present in the studyconducted by Cooper et al. Justify your answer. (3-5 sentences each)Rothman also identifies prognostic selection bias as a possible bias affectingscreening studies.List three methods that may be used to handle this type of bias in the design or theanalysis of studies of screening efficacy.(1 sentence for each method)c)Define ecological fallacy. (3-5 sentences)Is it possible that the Cooper et al. study was affected by the ecological fallacy?Justify your answer. (3-5 sentences)d)The authors included 105,525 women in their analyses. In this type of study,which source of error is affected by sample size? (1 sentence)e)Table 1 shows mammography rates for 29 states. Do these rates representaggregate measures or intrinsically population level measures?Explain. (1-3 sentences)S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc3

Question 2 (30 points)The abstract and the table below present results of a study of infant homicide.S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc4

a)b)Consider the data on marital status (table 2: not married versus married). Estimate the risk difference ( attributable risk to the exposed). Estimate the attributable risk to the population.Relative risks for No prenatal visit (relative risk 10.4) and for Second orsubsequent child, maternal age 17 yr (relative risk 10.9) were very similar.Does this observation (i.e., very similar relative risks) imply that risk differencesfor these two factors should also be similar? Justify your answer. (1-3 sentences)c)Table 2 indicates that 3.1% of the births in the United States in 1983-1991occurred among Asian women.What does this observation suggest about the fertility of Asian women in theUnited States during that time period?d)A medical student has calculated that all risk factors listed in table 2 accountedtogether for 60% of the cases of infant homicide, and he concluded that 40% ofthe cases remained unaccounted for. Criticize this statement. (3-5 sentences)e)Death certificates were used to identify cases of homicide. It is known that datafrom such certificates are sometimes inaccurate. Under what circumstances wouldthe presence of errors in birth certificate data not be associated with a bias inrelative risk estimates shown in table 2?S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc5

Question 3 (35 points)The abstract and the table below present results of a study of oral contraceptives andmyocardial infarction (NEJM 2001).S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc6

a)The body-mass index was calculated on the basis of measures of height andweight obtained at index date (date of diagnosis of myocardial infarction, orcorresponding date in controls).Does this feature of the study design (i.e., measures obtained at index date)invalidate conclusions of this study regarding obesity, contraceptives, andmyocardial infarction?Explain. (3-5 sentences)b)Twenty-eight cases of myocardial infarction were excluded from the studybecause these women died rapidly after diagnosis.Under what circumstances would the exclusion of these 28 cases lead to selectionbias (as distinct from confounding)?Explain. (3–5 sentences)c)On the basis of table 4 data, assess the presence of additive interaction betweensmoking and use of oral contraceptives (point estimate analysis only; no statisticaltest).Assess the presence of multiplicative interaction.Compare the conclusions of these two analyses and discuss briefly.(3-5 sentences)d)The authors of this study observed that the average age at diagnosis of myocardialinfarction in white women was lower than that of non-white women (data notshown).What may be the significance of this observation with respect to the epidemiologyof myocardial infarction in women? (2-3 sentences)S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc7

e)This case-control study determined the risk of myocardial infarction associatedwith current exposure to oral contraceptives (i.e., exposure during the monthbefore the index date).Could this scientific question have been addressed with the use of a casecrossover design?Justify your answer. (5-8 sentences)S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc8

S:\BOIVIN\695\Winter 2006\final exam 11 Apr 06.doc 1 Epidemiology EPIB-695 Final exam Monday, 10 April 2006 1. This is an open-book exam. Textbooks, notes, and dictionaries are allowed. However, these items cannot be shared with other students. 2. You must answer all quest