During a monthly departmental conference, a cardiology fellow presents data on a novel dir… — USMLE Step 3

During a monthly departmental conference, a cardiology fellow presents data on a novel direct oral anticoagulant used for stroke prevention in atrial fibrillation. The fellow displays a summary slide provided by the drug's manufacturer, which highlights a relative risk (RR) of $0.45$ ( $95\%$ CI: $0.38 - 0.53$ , $p < 0.01$ ) for major gastrointestinal bleeding when compared to standard warfarin therapy. The data is derived from a recently published retrospective cohort study of $2,400$ patients.

A faculty member questions the validity of the manufacturer's bold safety claim and projects the study's unadjusted baseline characteristics table for the audience:

Characteristic	Novel DOAC ( $n=1200$ )	Warfarin ( $n=1200$ )
Mean Age (years)	$62$	$79$
GFR $< 30 \text{ mL/min}$	$11\%$	$46\%$
History of Peptic Ulcer	$2\%$	$9\%$

The faculty member notes that the reported relative risk was calculated directly from the raw event rates without any multivariable regression or propensity score matching.

Which of the following is the most accurate epidemiological critique of the safety claim presented on the manufacturer's slide?

The apparent safety benefit is primarily driven by confounding by indication rather than true drug superiority.Answer
B
The baseline demographic differences represent effect modification, which inherently invalidates the unadjusted relative risk.
C
The relative risk is an invalid metric here because the positive predictive value of bleeding depends on population prevalence.
D
The analysis utilized an inappropriate inferential statistical test for evaluating categorical adverse events across two cohorts.
E
The outcome disparity simply reflects expected pharmacokinetic differences in the intrinsic versus extrinsic coagulation cascades.

Answer

The apparent safety benefit is primarily driven by confounding by indication rather than true drug superiority.

The correct answer correctly identifies that the baseline imbalances in age and comorbidities act as confounders. In observational studies, physicians do not assign medications randomly; they prescribe based on patient characteristics. Here, sicker and older patients were preferentially given warfarin (likely due to provider familiarity or specific guidelines for advanced renal disease). Because older age and poor renal function independently increase bleeding risk, the unadjusted relative risk makes the novel drug look much safer than it actually might be. This phenomenon is known as confounding by indication.

Step-by-Step Solution

Analyze the study design and the data provided in the stem.

The study is a retrospective cohort study without multivariable adjustment. The table shows massive differences in baseline risk factors (age, renal function, ulcer history) between the two groups.

Identifying the study type and baseline differences is crucial for evaluating the validity of the reported outcomes.

Evaluate the impact of the baseline characteristics on the outcome.

The warfarin group is significantly older, has worse renal function, and has a higher history of peptic ulcers. All of these independently increase the risk of the primary outcome (gastrointestinal bleeding).

If the baseline risk of bleeding is already much higher in the warfarin group, a higher incidence of bleeding in that group cannot be reliably attributed to the drug itself.

Synthesize the methodological flaw.

Because providers likely chose warfarin for sicker patients based on established familiarity, the indication for the specific drug choice is linked to the outcome. This is classic confounding by indication.

Observational studies without matching or regression cannot account for non-randomized treatment allocation, invalidating raw efficacy or safety claims.

Key Concept

Confounding by indication in observational studies

Alternative Method

Whenever evaluating a pharmaceutical advertisement based on an observational study, immediately look for 'Table 1' (Baseline Characteristics). If the groups are not well-matched and the headline data is 'unadjusted', the results are almost always confounded.

Estimated Time:1m 15s

Question

Answer

Step-by-Step Solution

Key Concept

Common Mistakes

Alternative Method

Topics

More questions from this topic