Epidemiology module of practical skills for mbbs student

Practical exercise

Using sample size / power calculations in protocol development

Case studies and some recent epidemiologic studies have suggested a possible link between some occupational and environmental exposures and idiopathic pulmonary fibrosis (IPF).

1. To address the question, “Is there an association between ‘idiopathic’ pulmonary fibrosis (IPF) and occupational exposure to solvents or metals?” describe several possible study designs that could be considered.

2. One of your colleagues is the medical director for a large metal trades employer and you have access to health records for all employees in the company for the past 35 years. Therefore, you have decided on a retrospective cohort study. All employees with 1 or more years of employment since January 1, 1965 will be enrolled in the cohort. They will be divided into exposure subgroups according to job records and company exposure information. You will compare rates of IPF among the exposure sub-groups.

What additional information do you need in order to calculate the whether or not the available workforce is large enough to do this kind of study?

3. Using Epi-Info, calculate the sample size needed for a cohort study given the following assumptions and decisions:

Null hypothesis: There is no difference in IPF rates between workers exposed to solvents and metals combined compared to workers not exposed to either.
alpha=0.05 ; power = 0.80
Effect size, based on:

expected ‘baseline’ IPF rate = 6 / 1,000,000 person years

expected risk ratio = 2 (ie. risk of IPF expected to be double in the exposed group)

Sample size required: _______________

4. Do the same calculation several more times, varying one or more of: alpha, power, and ‘expected risk ratio’, to complete this table:

Sample size / power estimates for a cohort study of occupational exposure to solvents and metals and IPF

assuming a baseline IPF rate (among non-exposed) of 6/1,000,000 persons per year.

	Risk ratio = 2	Risk ratio=2.5	Risk ratio=3	Risk ratio=4
Power = 80%
Alpha=0.01
Alpha=0.05
Power=90%
Alpha=0.01
Alpha=0.05

5. If the baseline rate of IPF is actually lower than 6 / 1,000,000 per year, would you expect the required sample size to be lower or higher? Check your answer by doing another Epi-Info calculation.

6. After investigating the company records, it is evident that there are only 1000 employees who have been exposed to solvents and metals. Therefore, based on the sample size calculations, the research team decides that a cohort study is not feasible and decides to explore the case-control design instead. All pulmonologists and pathologists in the region are polled and they agree to submit all their IPF cases over the 2 years to the study. Based on estimates of the number of IPF cases seen in the previous 2 years, you anticipate about 200 cases for the study.
Using Epi-Info, calculate what odds ratio a study of this size should be able to detect (with alpha=0.05 and power of 80%). What additional information about the expected study population is needed to do this calculation?

7. Compare the minimum detectable odds ratios for this kind of study, for expected baseline exposure prevalences of 2%, 20%, 50%, and 90%. What do you conclude about study power in relation to the prevalence of exposure in the population?

Practical exercise