A randomized trial by Wilt et al. (2012) compared surgery versus observation for localized prostate cancer. If 21 of 364 who had surgery died and 31 of 367 in the observation group died, which test would compare these proportions?

A randomized trial by Wilt et al. (2012) compared surgery versus observation for localized prostate cancer. If 21 of 364 who had surgery died and 31 of 367 in the observation group died, which test would compare these proportions?

• A. Fisher's exact test
• B. Paired t-test
• C. 2-proportion z-test
• D. ANOVA

Answer: (C)

When you want to compare death rates between two independent groups in a randomized trial, focus on the proportions in each group and whether they differ. Here, the two groups have deaths out of totals: 21/364 in the surgery group and 31/367 in the observation group. With such large sample sizes, the difference between these two proportions has a roughly normal distribution, so you can test whether the proportions are different using a two-proportion z-test. This test specifically assesses if p1 equals p2 or if there’s a statistically meaningful difference between the two independent proportions. Paired t-test isn’t appropriate because the data are binary outcomes, not continuous measurements from paired observations. ANOVA is for comparing means of a continuous variable across three or more groups, not for proportions. Fisher’s exact test could be used for a 2x2 table, especially with small counts, but with these sample sizes the two-proportion z-test is the standard and more straightforward choice.

When you want to compare death rates between two independent groups in a randomized trial, focus on the proportions in each group and whether they differ. Here, the two groups have deaths out of totals: 21/364 in the surgery group and 31/367 in the observation group. With such large sample sizes, the difference between these two proportions has a roughly normal distribution, so you can test whether the proportions are different using a two-proportion z-test. This test specifically assesses if p1 equals p2 or if there’s a statistically meaningful difference between the two independent proportions.

Paired t-test isn’t appropriate because the data are binary outcomes, not continuous measurements from paired observations. ANOVA is for comparing means of a continuous variable across three or more groups, not for proportions. Fisher’s exact test could be used for a 2x2 table, especially with small counts, but with these sample sizes the two-proportion z-test is the standard and more straightforward choice.