Q-Bank Breakdown: Relative risk vs odds ratio — Why Every Answer Choice Matters

You’re cruising through a biostats Q-bank and hit a classic trap: the stem screams “risk,” the answers whisper “odds,” and suddenly you’re second-guessing everything. Relative risk (RR) vs odds ratio (OR) is one of those concepts that’s easy in isolation but gets slippery in vignettes—especially when the study design is the real question.

Tag: Biostatistics > Study Design & Probability

---

The Clinical Vignette (Q-bank style)

A public health team investigates whether vaping is associated with chronic bronchitis. They identify 1,000 adults who vape and 1,000 adults who do not vape and follow both groups for 5 years.

At the end of follow-up:

• 120/1,000 vapers developed chronic bronchitis
• 60/1,000 non-vapers developed chronic bronchitis

Which of the following best estimates the association between vaping and chronic bronchitis?

A. Relative risk $= 2.0$
B. Odds ratio $\approx 2.14$
C. Incidence rate ratio
D. Attributable risk $= 6\%$
E. Positive predictive value

---

Step 1: Identify the Study Design (this chooses your measure)

This is a cohort study:

• Groups are defined by exposure (vape vs no vape)
• Followed over time to measure incidence (new cases)

Cohort studies can directly calculate risk, so the go-to association measure is:

• Relative risk (risk ratio, RR)

---

Step 2: Compute the Correct Answer (RR)

Risks

• Risk in exposed (vapers): $120/1000 = 0.12$
• Risk in unexposed: $60/1000 = 0.06$

Relative risk

$RR = \frac{0.12}{0.06} = 2.0$

✅ Correct answer: A. Relative risk = 2.0

Interpretation: Vapers have 2× the risk of developing chronic bronchitis over 5 years compared with non-vapers.

---

Why Every Answer Choice Matters (Systematic Distractor Breakdown)

Below is the same scenario—but each distractor is something students commonly reach for when they recognize numbers but miss the design or the definition.

---

A. Relative risk = 2.0 ✅ (Correct)

When you use RR:

• Cohort studies (prospective or retrospective cohort)
• Randomized controlled trials (RCTs)

High-yield:

• RR compares probabilities (risks), not odds.
• If RR > 1 → exposure associated with increased risk
• If RR < 1 → exposure is protective

---

B. Odds ratio ≈ 2.14 ❌ (Tempting, but not “best” here)

Let’s compute it to see why it’s tempting:

• Odds in exposed: $\frac{0.12}{0.88} = 0.13636$
• Odds in unexposed: $\frac{0.06}{0.94} = 0.06383$
• OR: $\frac{0.13636}{0.06383} \approx 2.14$

So yes, the OR is ~2.14, which is close to RR (2.0). But the question asks which best estimates the association.

Key point: In a cohort study, RR is the most direct and interpretable measure because you can compute incidence.

USMLE trap:

• OR is the standard for case-control studies, because you can’t compute incidence (no “risk” without a denominator representing the population at risk).

High-yield rule of thumb:

• When the outcome is rare (often <10%), OR $\approx$ RR
• When the outcome is not rare, OR overestimates RR (moves further away from 1)

Here the outcome risk in exposed is 12%, not super rare—so OR is already drifting upward.

---

C. Incidence rate ratio ❌ (Only if person-time is central)

Incidence rate ratio (IRR) compares incidence rates, typically: $\text{Incidence rate} = \frac{\text{new cases}}{\text{person-time}}$

You’d prefer IRR if:

• Follow-up time differs between people (loss to follow-up, varying entry times)
• The stem gives person-years (e.g., “120 cases per 5,000 person-years”)

In our vignette:

• Everyone is followed for the same 5 years
• We’re given clean risks (cases out of 1,000)

So RR is the cleanest “best estimate.”

High-yield: IRR commonly shows up in dynamic cohorts and open populations, and in studies emphasizing time-at-risk.

---

D. Attributable risk = 6% ❌ (That’s a difference, not a ratio)

Attributable risk (AR), also called risk difference, is:

$AR = Risk_{exposed} - Risk_{unexposed} = 0.12 - 0.06 = 0.06 = 6\%$

This is a real, meaningful number—but it answers a different question.

Interpretation: There are 6 extra cases per 100 people (over 5 years) attributable to vaping (assuming causality).

When AR is useful:

• Counseling and public health impact
• “How much disease could be prevented if we removed the exposure?”

USMLE distinction:

• RR = “How many times more likely?”
• AR = “How many more cases?”

---

E. Positive predictive value ❌ (Diagnostic test metric, not association)

PPV is:

$PPV = P(\text{disease} \mid \text{test positive})$

Nothing in this vignette is a diagnostic test with sensitivity/specificity, and we’re not asked to condition on a test result. So PPV doesn’t match the task.

High-yield: PPV and NPV depend on prevalence. Association measures (RR/OR) don’t depend on prevalence in the same way.

---

The 2×2 Table (How USMLE Wants You to Organize Your Brain)

From this table you can quickly compute:

• Risk exposed: $120/1000$
• Risk unexposed: $60/1000$
• RR: $\frac{120/1000}{60/1000}$
• Odds exposed: $120/880$
• Odds unexposed: $60/940$
• OR: $\frac{120/880}{60/940}$

---

Rapid-Fire High-Yield Rules (Test-Day Ready)

Choose RR vs OR based on study design

• Cohort / RCT → RR
• Case-control → OR (because incidence isn’t available)

Odds vs risk (don’t mix these up)

• Risk: $\frac{a}{a+b}$
• Odds: $\frac{a}{b}$

OR approximates RR when disease is rare

• Rare outcome → OR $\approx$ RR
• Not rare → OR exaggerates effect size away from 1

Attributable risk is about impact, not strength of association

• AR answers: “How much extra disease is due to exposure?”

---

How This Shows Up in Answer Choices (A Pattern to Recognize)

Q-banks often include both RR and OR on purpose:

• If you can compute RR, it’s usually the best choice in cohort/RCT stems
• OR distractors are there to punish “I can calculate something” without “I know what the question is asking”

If the stem screams followed over time, incidence, risk, exposed vs unexposed → think cohort → RR.