Master Biostatistics & Epidemiology
for USMLE Step 1

What the USMLE Step 1 Tests in Biostatistics & Epidemiology

USMLE Step 1 Biostatistics & Epidemiology tests the application of statistical reasoning to clinical decision-making. Candidates must interpret study designs (RCT, cohort, case-control, cross-sectional) and their biases (selection, information, confounding). They calculate and apply sensitivity, specificity, predictive values, likelihood ratios, and number needed to treat/harm. They identify measures of disease frequency (incidence, prevalence) and association (relative risk, odds ratio, attributable risk). They understand statistical significance (p-value, confidence intervals), power, and Type I/II errors. Candidates also evaluate diagnostic test performance using ROC curves and Bayes' theorem. Clinical scenarios involve screening (e.g., mammography for breast cancer, PSA for prostate cancer), outbreak investigations, and therapeutic trials (e.g., statins for MI prevention). They must apply intention-to-treat analysis, survival curves, and Kaplan-Meier estimates. Knowledge of ethical principles (informed consent, IRB) and bias reduction (randomization, blinding, stratification) is essential.

High-Yield Concepts

• Sensitivity and Specificity: Sensitivity = TP/(TP+FN) – the ability to detect disease when present; high sensitivity rules out disease (SnNOut). Specificity = TN/(TN+FP) – the ability to rule out disease when absent; high specificity rules in disease (SpPIn). Example: D-dimer for DVT – high sensitivity (low D-dimer rules out DVT), but poor specificity. Mammography for breast cancer: sensitivity ~85%, specificity ~90%.
• Positive and Negative Predictive Values: PPV = TP/(TP+FP) – probability disease is present given a positive test; depends on disease prevalence. NPV = TN/(TN+FN) – probability disease is absent given a negative test. For a test with 95% sensitivity and 95% specificity in a population with 1% prevalence (e.g., HIV screening), PPV is only ~16%, meaning many positives are false. NPV remains high (~99.9%).
• Likelihood Ratios: LR+ = sensitivity/(1-specificity) – how much a positive test increases the odds of disease. LR- = (1-sensitivity)/specificity – how much a negative test decreases odds. LR+ >10 or LR- <0.1 provide strong diagnostic evidence. Example: LR+ of 10 for elevated troponin in MI means post-test odds are 10 times pre-test odds. Use Fagan nomogram for quick calculation.
• Number Needed to Treat (NNT) and Number Needed to Harm (NNH): NNT = 1/absolute risk reduction (ARR). Example: Statin therapy for primary prevention of MI: 5-year risk of MI 10% in placebo vs 7% in treatment → ARR = 3% → NNT = 33 (treat 33 patients for 5 years to prevent 1 MI). NNH = 1/attributable risk increase (e.g., major bleeding with warfarin). Lower NNT indicates more effective treatment.
• Relative Risk and Odds Ratio: Relative risk (RR) = risk in exposed/risk in unexposed – used in cohort studies. RR = 1 (no association), >1 (risk factor), <1 (protective). Odds ratio (OR) = odds of exposure in cases/odds in controls – used in case-control studies. For rare diseases (<10%), OR approximates RR. Example: Smoking and lung cancer – OR ~20 in case-control studies, RR ~20 in cohort studies.
• Confounding and Bias: Confounding: a third variable associated with both exposure and outcome (e.g., age confounding the link between grey hair and heart disease). Control via randomization, stratification, or multivariable analysis. Selection bias: e.g., Berkson's bias in hospital-based case-control studies. Information bias: recall bias (cases remember exposures better) or observer bias. Minimize with blinding and standardized data collection.
• Study Designs and Hierarchy of Evidence: RCT (gold standard for efficacy) – double-blind, placebo-controlled, e.g., randomized trial of ACE inhibitors in heart failure. Cohort study – follow exposed and unexposed over time (e.g., Framingham Heart Study). Case-control – compare past exposures (e.g., thalidomide and phocomelia). Cross-sectional – prevalence survey (e.g., NHANES). Systematic reviews/meta-analyses (e.g., Cochrane reviews) sit atop the hierarchy.
• Survival Analysis and Kaplan-Meier Curves: Kaplan-Meier estimates survival probability over time, accounting for censoring (e.g., patients lost to follow-up). Log-rank test compares survival curves between groups (e.g., chemotherapy vs placebo for ovarian cancer). Hazard ratio = ratio of hazard rates (instantaneous risk of event). Example: HR of 0.7 for new drug means 30% reduction in hazard of death over study period.

Common Traps in Biostatistics & Epidemiology Questions

• Confusing prevalence with incidence: prevalence = existing cases/total population at a point in time; incidence = new cases/person-time at risk.
• Assuming PPV is constant: PPV decreases as disease prevalence decreases, even if sensitivity and specificity are high.
• Mistaking odds ratio for relative risk: OR overestimates RR when disease prevalence >10%.
• Thinking p-value <0.05 means the null hypothesis is false: it only indicates probability of observed data if null were true; it does not measure clinical significance.
• Forgetting that intention-to-treat analysis includes all randomized patients regardless of protocol adherence; per-protocol analysis may overestimate treatment effect.
• Believing a test with high sensitivity is always good for screening: it must also have reasonable specificity to avoid excessive false positives (e.g., low-specificity tests cause harm from overdiagnosis).

How to Revise Biostatistics & Epidemiology for the USMLE Step 1

Focus on applying formulas to clinical vignettes: calculate sensitivity, specificity, PPV, NPV, NNT, ARR, RR, OR from 2x2 tables. Practice interpreting confidence intervals (if CI includes 1 for RR/OR, not significant; if includes 0 for risk difference, not significant). Know the hierarchy of evidence and common biases (e.g., recall bias in case-control, lead-time bias in screening). Questions often present a study design and ask to identify the type or a flaw. Memorize key cut-offs: p<0.05, CI 95%, alpha=0.05, beta=0.20 (power=80%). Use the mnemonic 'Definitive Diagnosis' for SpPIn/SnNOut. Review screening criteria (Wilson & Jungner) and ethical principles (Belmont Report). Practice with USMLE-style vignettes that mix epidemiology with clinical decision-making (e.g., what test to order next based on LR). Prioritize understanding Bayes' theorem and how pre-test probability shifts with test results.

Practise it: MedLumen has 30 Biostatistics & Epidemiology questions for the USMLE Step 1, each with a full explanation and references.