Epidemiology Example

Epidemiology Example

Sample Size for confidence interval for odds ratio

What sample size is required for an experiment to estimate an odds ratio with a confidence interval of specified width? The study plan is to evaluate an educational program for expectant mothers. The educational program attempts to reduce the risk of preterm birth among expectant mothers who are seen in county prenatal clinics.

Initial plans call for randomly assigning 300 pregnant women to the control group and 300 to the educational intervention group. Will the 95% two-sided confidence interval for the odds ratio be likely to be narrow enough to demonstrate the effectiveness of the educational intervention? nQuery Advisor® can provide sample size and interval width calculations for the odds ratio in a two-sample design.

• Select File … New, and in the Study Goal and Design Box.
• Select Proportions, Two Groups, and Confidence Interval and select Confidence interval for log odds ratio.
• In the table, fill in the desired confidence level.
• Previous studies found preterm birth rates of 7-9%. So we enter an expected proportion of pre-term births of 8% for the control group.
• You can specify either the expected proportion for the intervention group, or the odds ratio, or the ln odds ratio. The intervention would be of interest if it reduced the rate of preterm births enough to produce an odds ratio of 0.5.
• You can specify the planned sample size for each group and solve for the interval width or specify the desired interval width and solve for the required sample size per group. Here we enter a sample size of 300 per group and request the sample size statement to see what the expected confidence limits on the odds ratio will be.

For this example, a sample of 300 women per group would be expected to result in a confidence interval which includes 1.0.

In the second column we request the sample size which will shorten the distance to the limit for the ln odds ratio to 0.55; the answer 491 suggests that we study 500 women per group which would result in expected 95% limits of 0.29 and 0.86 on an observed odds ratio of 0.5.

In additional columns you can assess the expected confidence interval width for other possible preterm rates for the control group.

nQuery Advisor® provides tables, plots, and standardized sample size justification statements which can be printed directly or copied to clipboard and pasted into your manuscript or proposal.