A randomized study assigned patients with sickle cell anemia and silent stroke to either monthly transfusion therapy (transfusion group of 110 patients) or usual care (observation group of 105 patients). An excerpt of the study results is provided below: Which of the following adverse event estimates is most subject to selection bias?

Question

A randomized study assigned patients with sickle cell anemia and silent stroke to either monthly transfusion therapy (transfusion group of 110 patients) or usual care (observation group of 105 patients).  An excerpt of the study results is provided below: Which of the following adverse event estimates is most subject to selection bias?

Quizplus · Accepted Answer

Explanation: 11ec2b69_bde6_26ac_8f63_69fe65e5c35c_MD0005_00 Selection bias occurs when the sample is unrepresentative of the target population. This may lead to a different measure of association (eg, rate ratio, odds ratio) in the sample than in the target population. In this example, for 4 out of 5 adverse events studied (acute chest syndrome, avascular hip necrosis, headache, vasooclusive crisis), data are provided for all 105 participants in the observation group and 110 participants in the transfusion group. However, only 35 out of 105 participants (= 33%) in the observation group had their ferritin level measured, so the results may be unrepresentative of the entire group. This makes it difficult to determine whether the calculated incidence rate ratio is reflective of the true difference between the observation and transfusion groups. One possible explanation for the low proportion compared to the transfusion group (91 / 110 = 83%) is that participants who received transfusions were more likely to have iron levels measured. Selection bias is often seen in surveys and polls. A high rate of nonresponse may lead to selection bias as the nonresponders may differ in some way from the responders. Other types of selection bias are seen in the table. (Choices A, B, C, and E) The entire sample (patients in both observation [N = 105] and transfusion groups [N = 110]) was analyzed for the other adverse effects (acute chest syndrome, avascular necrosis, headache, and vasoocclusive crisis); this minimized the risk of selection bias. Educational objective: Selection bias occurs when the sample is unrepresentative of the target population and may lead to incorrect measures of association.

A Randomized Study Assigned Patients with Sickle Cell Anemia and Silent