Researchers are conducting a clinical trial to assess the effect of erythropoietin on the survival of patients with end-stage renal disease receiving dialysis. Statistical analysis of the data shows that, after adjustment for age, sex, and duration of disease, patients who did not receive erythropoietin were at increased risk of mortality compared with those who did; the calculated relative risk (RR) of survival and corresponding p-value for patients treated with erythropoietin were RR = 2.4 with p = 0.005. Among patients with diabetes, erythropoietin use was associated with higher survival rates, with an RR of 6.2 and p

Question

Researchers are conducting a clinical trial to assess the effect of erythropoietin on the survival of patients with end-stage renal disease receiving dialysis.  Statistical analysis of the data shows that, after adjustment for age, sex, and duration of disease, patients who did not receive erythropoietin were at increased risk of mortality compared with those who did; the calculated relative risk (RR) of survival and corresponding p-value for patients treated with erythropoietin were RR = 2.4 with p = 0.005.  Among patients with diabetes, erythropoietin use was associated with higher survival rates, with an RR of 6.2 and p < 0.001.  Among patients without diabetes, however, the corresponding RR was 0.94 with p = 0.56.  The researchers are preparing a manuscript for publication of their findings.  Which of the following recommendations is most appropriate for the manuscript?

Quizplus · Accepted Answer

Explanation: Effect modification results when an external variable (effect modifier) has a positive or negative impact on the observed effect of a risk factor (exposure) on disease status (outcome). For instance, aspirin use is associated with Reye syndrome in children but not adults; therefore, age (effect modifier) modifies the effect of aspirin (exposure) on Reye syndrome development (outcome). Stratification based on the modifier (eg, analyzing the data by age) can help detect effect modification as it typically shows different effects in each stratum (eg, children, adults); as a result, separate measures of outcome should be reported for each stratum. This study analyzed the association between erythropoietin (exposure) and survival (outcome) among patients with end-stage renal disease on dialysis. Based on the p-values, diabetes appeared to modify the effect of erythropoietin on survival: Among patients with diabetes, erythropoietin was significantly associated with higher survival rates, as the relative risk (RR) was 6.2 with a statistically significant p-value (p < 0.05). In contrast, among patients without diabetes, there was no association between erythropoietin and survival, as the p-value (p = 0.56) was nonstatistically significant (p > 0.05). Therefore, the manuscript should report separate RRs (with their corresponding p-values) for patients with and without diabetes. Effect modification is not due to flaws in the design or analysis phases of the study and should not be controlled or adjusted (Choice D). The RRs and their corresponding p-values should be reported for each stratum of the effect modifier (ie, with and without diabetes) (Choice B). Ignoring the effect modification and reporting only the overall RR would provide a false estimate of the true association between exposure and outcome (Choices A and C). Stratified analysis can help distinguish between effect modification and confounding. With confounding, there is usually no significant difference between the strata. An example of confounding is shown in the exhibit: In a primary school, bigger shoe size may appear to be associated with higher intelligence quotient on crude analysis, but with stratification into younger (eg, grade 1) and older (eg, grade 6) children, the association is no longer statistically significant. In this example, age is a confounder, a variable that is associated with both exposure (shoe size) and outcome (intelligence quotient) and that muddles the exposure-outcome relationship. exhibit Educational objective: Effect modification results when an external variable has a positive or negative impact on the observed effect of a risk factor on disease status. Stratification can help detect effect modification. Separate measures of outcome should be reported for each level of an effect modifier.

Researchers Are Conducting a Clinical Trial to Assess the Effect