A multi-center research group is conducting a randomized, double-blind, placebo-controlled… — USMLE Step 3

A multi-center research group is conducting a randomized, double-blind, placebo-controlled trial to evaluate the efficacy of a novel immunomodulator for the prevention of acute exacerbations in patients with severe, uncontrolled asthma. The original study design calculated that $400$ patients would be required to achieve $80\%$ statistical power to detect a $30\%$ relative reduction in exacerbation rates, with a two-sided Type I error rate ( $\alpha$ ) of $0.05$ .

During a planned interim analysis, the data safety monitoring board notes that the true relative reduction in exacerbations appears to be closer to $15\%$ . While this smaller effect size remains clinically significant, the study is now substantially underpowered to detect it by the end of the trial. The board recommends modifying the trial protocol to restore $80\%$ statistical power to detect this $15\%$ difference, strictly without increasing the pre-specified probability of a Type I error.

Which of the following trial modifications would appropriately achieve this goal? (Select all that apply)

Increase the total target patient enrollment size beyond the original calculation of 400Cevap
Extend the follow-up duration of the trial to allow a higher number of exacerbation events to accrueCevap
Change the primary endpoint analysis from a dichotomous outcome (presence/absence of exacerbations) to a continuous measure (mean exacerbation rate per patient-year)Cevap
D
Increase the pre-specified significance threshold ( $\alpha$ ) from 0.05 to 0.10
E
Restrict the final analysis to evaluate effect modification by exclusively analyzing the subgroup of patients with the highest baseline eosinophil counts
F
Perform multiple additional interim data analyses using unadjusted $p$ -values to increase the likelihood of detecting early significance

Cevap

The correct modifications to increase statistical power without inflating Type I error include increasing the total sample size, extending the follow-up duration to accrue more events, and changing the primary endpoint from a dichotomous outcome to a continuous measure.

Statistical power (the probability of correctly rejecting a false null hypothesis) is mathematically dependent on sample size, effect size, variance, and the significance level (

\alpha

). When the true effect size is smaller than anticipated, power drops. To restore power without increasing

\alpha

(which is forbidden by the scenario), researchers can: 1) Increase the total sample size to reduce standard error; 2) Extend the follow-up period to accrue a higher number of events, which serves as the effective sample size in event-driven analyses; and 3) Utilize continuous outcome measures rather than dichotomous ones, as continuous data inherently retains more statistical information and reduces error variance, thereby increasing power.

Adım Adım Çözüm

Analyze the relationship between effect size and statistical power.

A smaller true effect size (

15\%

30\%

) means the original sample size is insufficient, causing statistical power to drop below the

80\%

target.

Effect size and power are directly proportional; smaller differences are harder to detect.

Identify general statistical methods to increase power.

Power (

1-\beta

) can be augmented by increasing sample size (

N

), increasing the number of events, decreasing data variance, or increasing the

\alpha

level.

These variables are mathematically linked in power calculations.

Apply the stem's explicit constraints.

The scenario explicitly forbids increasing the Type I error rate (

\alpha

). Any options that inflate false positives (like raising

\alpha

or performing unadjusted multiple comparisons) must be eliminated.

Protocol amendments cannot violate the core integrity of the trial's safety constraints.

Evaluate the impact of data types on power.

Recognize that continuous data preserves more information than dichotomized data, inherently yielding greater statistical power for the same sample size.

Dichotomization throws away variance data, reducing the sensitivity of statistical tests.

Anahtar Kavram

Statistical power can be augmented by increasing sample size, accruing more clinical events (extending follow-up), or utilizing continuous rather than categorical outcome measures, provided Type I error constraints are maintained.

Soru

Cevap

Adım Adım Çözüm

Anahtar Kavram

Sık Yapılan Hatalar

Konular

Bu konudan daha fazla soru