An observational cohort study utilizes a national healthcare database to evaluate the effi… — USMLE Step 2 CK

An observational cohort study utilizes a national healthcare database to evaluate the efficacy of a newly approved intravenous antifungal agent compared to standard amphotericin B for the treatment of invasive mucormycosis. The primary outcome is 30-day all-cause mortality.

The initial analysis yields a crude relative risk ( $RR$ ) of $1.85$ ( $95\% \text{ CI}: 1.40-2.45$ ), suggesting that the new agent is associated with a significantly higher mortality risk compared to standard therapy.

Concerned about these findings, the investigators perform a stratified analysis based on the presence or absence of cavernous sinus thrombosis at the time of diagnosis. The results are shown below:

Stratum	Relative Risk ( $95\% \text{ CI}$ )
Cavernous sinus thrombosis present	$0.65$ ( $0.50-0.85$ )
Cavernous sinus thrombosis absent	$0.70$ ( $0.55-0.90$ )

Based on these results, which of the following epidemiologic phenomena best explains the observed discrepancy between the crude and stratified relative risks?

Preferential administration of the new agent to patients with more advanced disease at baselineCevap
B
Differential magnitude of the new agent's efficacy depending on the presence of cavernous sinus thrombosis
C
Cavernous sinus thrombosis acting as an intermediate step on the causal pathway between treatment and mortality
D
Selection of study participants based primarily on their mortality outcome rather than their exposure status
E
Disproportionate loss to follow-up among patients who received standard therapy and had severe disease

Cevap

Preferential administration of the new agent to patients with more advanced disease at baseline

The data demonstrates a classic case of confounding, specifically Simpson's paradox, where the crude association (

RR = 1.85

, showing harm) is completely reversed upon stratification by disease severity (

RR = 0.65

and

0.70

, showing benefit). The stratified relative risks are similar to each other, confirming confounding rather than effect modification. In observational studies evaluating treatments, this pattern is almost always caused by 'confounding by indication.' This means the new drug was not randomly assigned; instead, physicians preferentially administered the new drug to patients who already had more advanced disease (e.g., cavernous sinus thrombosis) and therefore had a much higher baseline risk of mortality. Once the analysis adjusts for this baseline severity via stratification, the true protective effect of the drug is revealed.

Adım Adım Çözüm

Analyze the crude relative risk (RR).

The crude

RR

1.85

, which is greater than

1.0

This initially suggests that the new antifungal agent increases the risk of mortality compared to standard therapy.

Analyze the stratified relative risks.

The

RR

0.65

in the stratum with thrombosis and

0.70

in the stratum without thrombosis.

Both stratified

RR

s are less than

1.0

, indicating that within each severity group, the new agent actually reduces mortality.

Compare the crude and stratified estimates to identify the epidemiological phenomenon.

The stratified estimates are similar to each other (

0.65

and

0.70

) but differ drastically from the crude estimate (

1.85

), crossing the null value of

1.0

When stratified estimates are similar to each other but different from the crude estimate, confounding is present. The complete reversal of the association is known as Simpson's paradox.

Determine the specific clinical mechanism of confounding in this study design.

The correct mechanism is confounding by indication.

In observational studies of medical treatments, providers often prescribe newer, stronger, or salvage therapies preferentially to patients with more severe disease. Because these patients have a higher baseline risk of death, the drug falsely appears to cause higher mortality in the crude analysis.

Anahtar Kavram

Confounding by indication (Simpson's paradox) occurs when a treatment appears harmful in crude analysis because it is preferentially given to sicker patients, but proves protective once stratified by disease severity.

Alternatif Yöntem

A quick visual heuristic for USMLE questions: If Crude RR is outside the range of the Stratified RRs, the answer is Confounding. If the Crude RR is roughly in the middle of two widely different Stratified RRs, the answer is Effect Modification.

Tahmini Süre:2m 0s

Soru

Cevap

Adım Adım Çözüm

Anahtar Kavram

Sık Yapılan Hatalar

Alternatif Yöntem

Konular

Bu konudan daha fazla soru