---

Biostatistics for NEET PG 2026 - Complete Guide

---

MODULE 1: MEASURES OF CENTRAL TENDENCY & DISPERSION

Central Tendency

• Positively skewed (tail to the right): Mean > Median > Mode
• Negatively skewed (tail to the left): Mean < Median < Mode
• In a normal distribution: Mean = Median = Mode

Dispersion

• Range = Max - Min (crude, affected by outliers)
• Standard Deviation (SD) = average spread around the mean
• Variance = SD²
• Coefficient of Variation (CV) = (SD/Mean) × 100 -- used to compare variability across different units

• Mean ± 1 SD = 68.2% of data
• Mean ± 2 SD = 95.4% of data
• Mean ± 3 SD = 99.7% of data

---

MODULE 2: STUDY DESIGNS (Very High Yield)

Hierarchy of Evidence (strongest to weakest)

1. Systematic review / Meta-analysis
2. Randomized Controlled Trial (RCT) - Gold standard for causation
3. Cohort study
4. Case-control study
5. Cross-sectional study
6. Case series / Case reports
7. Expert opinion

Study Design Summary Table

Randomized Controlled Trial

• Gold standard for measuring causal relationships
• Randomization eliminates known AND unknown confounders - this is the key advantage over observational studies
• Blinding reduces bias: Single blind (patient), Double blind (patient + investigator), Triple blind (+ analyst)

Cohort Study

• Follow exposed vs. unexposed groups forward in time
• Can calculate: Incidence, RR, Attributable Risk
• Prospective cohort: More reliable, expensive, takes time
• Retrospective cohort: Uses existing records, faster

Case-Control Study

• Start with cases (have disease) and controls (no disease), look back at exposure
• Best for rare diseases and diseases with long latency
• Cannot calculate incidence or RR directly; uses Odds Ratio
• Recall bias is a major limitation

Cross-Sectional Study

• Snapshot in time; measures prevalence
• Cannot establish causality (chicken-egg problem)
• Cheap, fast, good for survey data

---

MODULE 3: THE 2×2 TABLE - Core of Biostatistics

                    DISEASE +      DISEASE -
TEST POSITIVE          a              b          a+b
TEST NEGATIVE          c              d          c+d
                      a+c            b+d         N

• High SeNsitivity = rules OUT disease (use for screening - don't miss cases)
• High SPecificity = rules IN disease (use for confirmation - don't over-diagnose)

• When disease prevalence rises: PPV rises, NPV falls
• Sensitivity and specificity are independent of prevalence
• PPV and NPV depend on prevalence

• Sensitivity = 196/200 = 98%
• Specificity = 180/200 = 90%
• PPV = 196/216 = 91%
• NPV = 180/184 = 98%

---

MODULE 4: MEASURES OF RISK

Relative Risk (RR)

• Used in cohort studies and RCTs
• RR = Incidence in exposed / Incidence in unexposed
• RR = 1: no association; RR > 1: positive association; RR < 1: protective

Odds Ratio (OR)

• Used in case-control studies
• OR = (a×d) / (b×c) from the 2×2 table
• When disease is rare, OR ≈ RR (rare disease assumption)

Attributable Risk (AR) = Risk Difference

• AR = Incidence (exposed) - Incidence (unexposed)
• Tells you the absolute excess risk due to exposure
• Example: 37 CHD events/10,000 in treated - 30/10,000 in placebo = 7 additional events/10,000

Number Needed to Treat (NNT)

• NNT = 1 / Attributable Risk
• Tells you how many patients you must treat to prevent 1 event
• Smaller NNT = more effective treatment

Number Needed to Harm (NNH)

• NNH = 1 / AR (when AR represents harm)
• Larger NNH = safer treatment

Relative Risk Reduction (RRR)

• RRR = AR / Incidence in control group
• Example: AR = 7/10,000, Incidence in control = 30/10,000 → RRR = 23%

---

MODULE 5: STATISTICAL TESTS

Choosing the Right Test

1. Is data categorical or continuous?
2. If continuous - is it normally distributed?
3. Same subjects (paired) or different subjects?
4. How many groups?

Correlation

• Pearson's r: continuous, normally distributed data (-1 to +1)
• Spearman's rho: non-parametric equivalent
• r = +1: perfect positive; r = -1: perfect negative; r = 0: no linear correlation
• Correlation ≠ causation

Regression

• Simple linear regression: predict continuous outcome from 1 predictor
• Multiple linear regression: continuous outcome from multiple predictors
• Logistic regression: categorical (binary) outcome - gives OR
• R² = proportion of variance explained by the model

---

MODULE 6: P-VALUE, CONFIDENCE INTERVALS & ERRORS

P-value

• Probability that the observed result occurred by chance alone, assuming the null hypothesis is true
• p < 0.05: result is statistically significant (less than 5% probability it's due to chance)
• p > 0.05: fail to reject the null hypothesis
• Statistical significance ≠ clinical significance - a tiny difference can be statistically significant with a large enough sample

Confidence Interval (CI)

• 95% CI means: if the study were repeated 100 times, 95 of those intervals would contain the true value
• If 95% CI for RR or OR includes 1.0 → result is NOT statistically significant
• If 95% CI for risk difference includes 0 → result is NOT statistically significant
• Narrow CI = more precise estimate (larger sample size)

Type I and Type II Errors

• Type I error (α): False positive - concluding there IS a difference when there isn't
  • Acceptable rate: α = 0.05 (5%)
• Type II error (β): False negative - concluding there is NO difference when there actually is
  • Acceptable rate: β = 0.20 (20%)
• Power = 1 - β = probability of detecting a true difference = ideally ≥ 80%

---

MODULE 7: BIAS & CONFOUNDING

Types of Bias

Confounding

• A confounder is a variable that is associated with both the exposure and the outcome
• The classic example: coffee drinking and lung cancer (confounded by smoking)
• RCTs eliminate confounding by randomization (both known and unknown confounders)
• Observational studies control confounding by: restriction, matching, stratification, multivariable regression

Internal vs. External Validity

• Internal validity: were study results unbiased? (did the study measure what it intended to?)
• External validity (generalizability): do results apply to real-world patients?

---

MODULE 8: SCREENING & DISEASE SURVEILLANCE

Criteria for a Good Screening Test (Wilson-Jungner)

1. Disease is an important health problem
2. Acceptable treatment is available
3. Facilities for diagnosis and treatment are available
4. There is a recognizable latent or early symptomatic stage
5. Suitable test or examination exists
6. Test is acceptable to the population
7. Natural history of disease is understood
8. Agreed policy on who to treat
9. Cost is economically balanced in relation to medical expenditure as a whole
10. Case-finding is a continuing process

ROC Curve

• Receiver Operating Characteristic curve plots Sensitivity (y-axis) vs. 1-Specificity (x-axis) at various cutoff values
• The better the test, the larger the Area Under the Curve (AUC)
• AUC = 0.5: no better than chance; AUC = 1.0: perfect test
• Used to compare accuracy of different diagnostic tests and to choose the optimal cutoff

---

MODULE 9: EVIDENCE-BASED MEDICINE (EBM)

PICO Framework

• P - Patient/Population/Problem
• I - Intervention
• C - Comparison
• O - Outcome

Levels of Evidence (high-yield for NEET PG)

1. Ia: Meta-analysis of RCTs
2. Ib: At least one RCT
3. IIa: At least one controlled study without randomization
4. IIb: At least one quasi-experimental study
5. III: Descriptive studies (case-control, cohort)
6. IV: Expert committee reports/opinion

Critical Appraisal

• Systematic review: exhaustive literature search + pre-defined inclusion criteria + pooled qualitative analysis
• Meta-analysis: statistical pooling of results from multiple studies to give a single quantitative estimate
• Forest plot: graphical display of individual study results and pooled result in meta-analysis; if the diamond crosses 1.0 (for RR/OR) → not significant

---

MODULE 10: SURVIVAL ANALYSIS

• Kaplan-Meier curve: plots survival probability over time; handles censored data
• Log-rank test: compares survival between two groups
• Hazard Ratio (HR): analogous to RR; used in survival analyses
• Cox regression (proportional hazards): adjusts for multiple covariates in survival analysis

---

HIGH-YIELD MNEMONICS & ONE-LINERS

---

QUICK-REVISION: COMMONLY TESTED NEET PG SCENARIOS

1. "Which study is best for rare disease?" → Case-control
2. "Best measure of central tendency for skewed data?" → Median
3. "Test for categorical data in 2×2 table?" → Chi-square (or Fisher's exact if n < 5 in any cell)
4. "Test comparing means of 3 groups?" → ANOVA
5. "Test comparing paired (before/after) data, non-normal?" → Wilcoxon signed-rank
6. "Screening test should be high in...?" → Sensitivity (rule out disease, don't miss cases)
7. "Confirmatory test should be high in...?" → Specificity (rule in disease, avoid false positives)
8. "Which bias affects case-control most?" → Recall bias
9. "Lead-time bias overestimates...?" → Survival benefit of screening
10. "CI for OR crosses 1 means...?" → Not statistically significant
11. "Power of a study = ?" → 1 - β (ideally ≥ 80%)
12. "Study that cannot prove causation?" → Cross-sectional, ecological

---

• Miller's Review of Orthopaedics 9th Edition (statistical tests, p-values, errors)
• Henry's Clinical Diagnosis and Management by Laboratory Methods (sensitivity, specificity, 2×2 table)
• Textbook of Family Medicine, 9th ed. (RR, AR, NNT)
• The Harriet Lane Handbook, 23rd ed. (EBM, PICO, study design)