Statistics and Epidemiology for MRCOG

Diagnostic Test Measures

| Measure | Formula | Interpretation | |---------|---------|---------------| | Sensitivity | TP / (TP + FN) | Ability to detect disease (true positive rate); rules OUT disease when negative (SnNOUT) | | Specificity | TN / (TN + FP) | Ability to exclude disease (true negative rate); rules IN disease when positive (SpPIN) | | PPV | TP / (TP + FP) | Probability of disease given positive test; INCREASES with prevalence | | NPV | TN / (TN + FN) | Probability of no disease given negative test; DECREASES with prevalence | | LR+ | Sensitivity / (1 - Specificity) | How much a positive result increases the probability of disease | | LR- | (1 - Sensitivity) / Specificity | How much a negative result decreases probability |

• PPV and NPV are affected by disease prevalence; sensitivity and specificity are NOT
• SnNOUT: high Sensitivity → Negative result rules OUT disease
• SpPIN: high Specificity → Positive result rules IN disease

Study Designs

Hierarchy of Evidence (strongest to weakest)

1. Systematic reviews and meta-analyses
2. Randomised controlled trials (RCTs)
3. Cohort studies (prospective > retrospective)
4. Case-control studies
5. Cross-sectional studies
6. Case series / case reports
7. Expert opinion

Key Study Types

• RCT: gold standard for interventions; randomisation reduces confounding; blinding reduces observer bias
• Cohort study: follows exposed/unexposed groups over time; measures relative risk (RR) and incidence
• Case-control study: compares cases (disease) with controls (no disease); looks BACKWARDS at exposure; measures odds ratio (OR); efficient for rare diseases
• Cross-sectional study: snapshot in time; measures prevalence; cannot determine causation

Measures of Association

| Measure | Formula | Used In | |---------|---------|---------| | Relative Risk (RR) | Risk in exposed / Risk in unexposed | Cohort studies, RCTs | | Odds Ratio (OR) | (a×d) / (b×c) from 2×2 table | Case-control studies; approximates RR when disease is rare | | Absolute Risk Reduction (ARR) | Risk in control - Risk in treatment | Treatment benefit | | Number Needed to Treat (NNT) | 1 / ARR | Clinical utility of treatment | | Number Needed to Harm (NNH) | 1 / (Risk in exposed - Risk in unexposed) | Risk assessment |

• RR = 1: no association; RR > 1: increased risk; RR < 1: protective
• NNT of 10 means 10 patients treated for 1 to benefit

Types of Bias

• Selection bias: non-random allocation or loss to follow-up distorts groups
• Recall bias: participants with disease remember exposures differently (common in case-control studies)
• Observer (detection) bias: awareness of group allocation influences assessment (prevented by blinding)
• Publication bias: positive results more likely published; detected by funnel plot
• Lead-time bias: screening appears to prolong survival by detecting disease earlier, not actually improving outcomes
• Length-time bias: screening detects slower-growing, less aggressive disease preferentially
• Confounding: a third variable associated with both exposure and outcome; controlled by randomisation, stratification, multivariate analysis, matching

Statistical Concepts

• P-value: probability of observing results as extreme as obtained if null hypothesis is true; p < 0.05 = statistically significant
• Type I error (alpha): rejecting null hypothesis when it's true (false positive); probability = significance level (usually 0.05)
• Type II error (beta): failing to reject null hypothesis when it's false (false negative)
• Power: 1 - beta; probability of detecting a true effect; usually aim for ≥80%
• Confidence interval (CI): range within which the true value lies with stated probability (usually 95%); if 95% CI for RR crosses 1, result is NOT statistically significant

Screening Criteria (Wilson-Jungner, 1968)

1. Important health problem
2. Accepted treatment available
3. Facilities for diagnosis and treatment available
4. Recognisable latent or early symptomatic stage
5. Suitable screening test available
6. Test acceptable to the population
7. Natural history of condition understood
8. Agreed policy on whom to treat
9. Cost-effective
10. Screening should be a continuous process

Clinical Trial Concepts

• Intention-to-treat (ITT): analyses all participants in the group they were randomised to, regardless of compliance; preserves randomisation; reduces bias; may underestimate treatment effect
• Per-protocol analysis: analyses only participants who completed the protocol; may overestimate treatment effect; introduces bias
• Blinding: single-blind (participant), double-blind (participant + investigator), triple-blind (+ analyst)
• Allocation concealment: different from blinding; ensures the person enrolling participants cannot predict group assignment
• Superiority trial: aims to show one treatment is better
• Non-inferiority trial: aims to show new treatment is not worse by a defined margin
• Equivalence trial: aims to show treatments are similar

Important Facts for MRCOG

• Sensitivity and specificity: NOT affected by prevalence; PPV and NPV ARE
• Case-control: measures odds ratio; efficient for rare diseases; prone to recall bias
• NNT = 1/ARR
• Lead-time bias: makes screening look effective when it may not be
• Type I error: false positive (rejecting true null); Type II: false negative
• ITT preserves randomisation; per-protocol may overestimate effect
• If 95% CI for RR crosses 1: not statistically significant
• Wilson-Jungner criteria: 10 criteria for screening programmes

Clinically reviewed by NHS obstetricians and gynaecologists

MRCOG AI is an independent educational tool. It is not affiliated with, endorsed by, or connected to the Royal College of Obstetricians and Gynaecologists (RCOG).