Data Interpretation and Confounding Exercise:

Table below describes data from a prospective cohort study examining the association between type of work schedule and the risk of prostate cancer in men.

Table: Incidence of prostate cancer in 14,052 male workers according to type of work schedule
Work schedule
Daytime Fixed night Rotating shift

No. of subjects with no diagnosis of prostate cancer at the time of enrollment in the study 11,269 982 1,801
No. of person-years of follow-up 89,179 8,272 14,523
No. of new prostate cancer cases 21 3 7

Express your answers as per 10,000 person-years of follow-up where appropriate.

A. What is the incidence density of prostate cancer in all participants in the study?

B. What is the incidence density of prostate cancer in daytime workers?

C. What is the incidence density of prostate cancer in fixed night workers?

D. What is the incidence density of prostate cancer in rotating shift workers?

E. What is the ratio of the person-time incidence (or incidence density) of prostate cancer in fixed night shift to the person-time incidence of prostate cancer in Daytime workers?

F. What is the ratio of the person-time incidence of prostate cancer in rotating shift to the person-time incidence of prostate cancer in Daytime workers?

G. How do rates of prostate cancer among fixed night and rotating shift compare to the rate of prostate cancer in daytime shift? What would you conclude about the relationship between work schedule and risk or rate of prostate cancer? Why?

H. Assuming that all participants were followed up for a total of 8 years with complete follow-up, what would be the cumulative incidence (risk) of prostate cancer in daytime workers for the 8-year period?

(1) Below are excerpts from a study of Urinary Tract Infections (UTI) in hospitals by Reintjes et al. (Epidemiology 2000;11:81-83).

Background - Severijnen et al. conducted a prospective multicenter study in eight hospitals to determine the feasibility of standardized surveillance of nosocomial infections in The Netherlands. After this study had been completed, the authors used the dataset from gynecologic patients to measure the influence of possible risk factors on the development of urinary tract infections (UTI). UTI are known to be associated with a variety of risk factors (host and intervention related). These factors are not independent, and confounding is an obvious potential problem. Antibiotic prophylaxis has been shown to be effective in randomized clinical trials and is sometimes used to prevent UTI. In non-experimental studies antibiotic prophylaxis has been shown to be associated positively with hospital acquired infections. The authors studied the association between UTI and antibiotic prophylaxis using univariate (crude) and stratified analyses.

Results: Table above shows the results from the univariate (crude) analysis and Table second (next page) shows the results from a stratified analysis.

a) Compute a relative risk of UTI for antibiotic prophylaxis for the univariate (crude) results presented in Table 1 (2 points).

b) What does the RR suggest about the effect of antibiotic prophylaxis on the risk of UTI? (2 points)

c) Compute relative risks for the 2 strata of hospitals presented in Table 2 (4 points).

d) How do the results from the crude and stratified analysis compare? Is prophylaxis treatment protective? (5 points)

e) Which findings are consistent with what you would expect? Which findings do you believe and why? (5 points)

(2) The Epidoria Birth Weight Study

Background - In the small island nation of Epidoria, a team of reproductive epidemiologists has been studying the relationship between very low birth weight and risk of cognitive, motor, and behavioral problems. Five years ago these investigators initiated a study. Using birth certificate files and delivery room entry logs, these investigators attempted to identify all full-term births in Epidoria over a 6-month period. The investigators enrolled all low birth weight babies and a representative sample of normal birth weight babies into their study. The investigators then examined the children every year until age 3 years. During the last examination, the investigators administered a standardized developmental screening test to assess personal-social, language, and motor-adaptive skills. Based on this test, the investigators classified the children into two groups: normal development and delayed development.

The results from the study were:
Development
Delayed Normal Total
Birth Weight Low 140 220 360
Normal 77 283 360
Total 217 503 720

a) What kind of a study is this (1 point)? What is the exposure under study? What is the outcome under study?

b) Calculate the crude cumulative incidence ratio (relative risk).

Background (cont.) - To account for the possibility that environmental lead exposure might confound the relationship between birth weight and developmental status, blood lead levels were determined from blood samples collected at the age 3-year visit. Elevated lead levels (> 10 micrograms/ dL) were found in 173 of the low birth weight children (88 of whom had delayed development according to their screening test). Elevated lead levels were also found in 72 of the normal birth weight children (24 of whom had delayed development).

c) Carry out a stratified analysis of birth weight and developmental delay, controlling for blood lead level. Create 2x2 tables for each stratum of lead level and estimate the relative risk for each stratum.

d) Comparing the findings from the crude and stratified analysis, is there a suggestion of confounding by lead exposure in this example?

e) Is there evidence of an association between blood lead level and the primary exposure variable in this study (hint create a 2x2 table for blood lead level and exposure)?

f) Is there evidence of an association between blood lead level and the outcome in this study?

g) Based on all of this information would you judge blood lead level to be a confounder of the association between birth weight and delayed development? Explain your answer.

h) What are two changes in the study design would have avoided the potential confounding effects of blood lead level? What are the advantages and disadvantages of these alternatives?