An investigator wants to compare mean HbA1c levels among three groups.
a. Suppose that the three groups being compared involve a control and two experimental therapies. The purpose of the study is exploratory to see if the therapies impact HbA1c. The investigator uses the Bonferroni adjustment to reduce the chance of making a type 1 error. Is this appropriate? What are the implications of making the adjustment?
b. Suppose that the study is an observational one and is the first investigation of differences in HbA1c for these three groups. The purpose of the study is exploratory to see if a clinical trial investigating these three groups is warranted. Without adjusting for multiple comparisons, the investigator finds that there is a difference between groups 1 and 2, but after adjusting for multiple comparisons, the p-value is no longer significant. Advise the investigator.
c. Suppose that the study is a confirmatory clinical trial to investigate the impact of a no-treatment control, an active control, and a treatment on HbA1c. When making comparisons among the three groups, the investigator decides to implement an adjustment to protect against making a type 1 error. Is this appropriate? What are the implications of making the adjustment?

Question

An investigator wants to compare mean HbA1c levels among three groups. 
a.      Suppose that the three groups being compared involve a control and two experimental therapies. The purpose of the study is exploratory to see if the therapies impact HbA1c. The investigator uses the Bonferroni adjustment to reduce the chance of making a type 1 error. Is this appropriate? What are the implications of making the adjustment?
b.      Suppose that the study is an observational one and is the first investigation of differences in HbA1c for these three groups. The purpose of the study is exploratory to see if a clinical trial investigating these three groups is warranted. Without adjusting for multiple comparisons, the investigator finds that there is a difference between groups 1 and 2, but after adjusting for multiple comparisons, the p-value is no longer significant. Advise the investigator.
c.      Suppose that the study is a confirmatory clinical trial to investigate the impact of a no-treatment control, an active control, and a treatment on HbA1c. When making comparisons among the three groups, the investigator decides to implement an adjustment to protect against making a type 1 error. Is this appropriate? What are the implications of making the adjustment?

Quizplus · Accepted Answer

The Answer is a. Performing a Bonferroni adjustment would reduce the chance of making a type 1 error, but since the purpose of the study is exploratory, a multiple comparison adjustment may not be necessary. There are only three groups and a Bonferroni adjustment may inflate the type 2 error (a more serious error in this case).
b. If the investigator is only interested in understanding the differences between groups 1 and 2, I would advise the investigator to only focus on those two groups. If the investigator just focuses on groups 1 and 2, then a multiple comparison adjustment will not need to be made.
c. A multiple comparison adjustment may be appropriate. In confirmatory trials, type 1 errors tend to be more serious than type 2 errors. For this reason, adjustment for multiple comparisons would be justified.

An Investigator Wants to Compare Mean HbA1c Levels Among Three