1. Study Types

•    Cohort study

o    Starts with exposure → follows over time to see outcome.

o    Measures relative risk (RR).

o    Prospective or retrospective.

•    Case-control study

o    Starts with outcome → looks back for exposure.

o    Measures odds ratio (OR).

o    Useful for rare diseases.

•    Cross-sectional study

o    Measures exposure and outcome at a single point in time.

o    Provides prevalence data.

•    Randomised controlled trial (RCT)

o    Participants randomly allocated to intervention or control.

o    Measures efficacy, minimizes confounders.

2. Bias Types

•    Selection bias: differences in baseline characteristics between groups (e.g., healthy volunteer effect).

•    Recall bias: cases recall exposure differently than controls (common in case-control studies).

•    Measurement (information) bias: systematic errors in data collection (e.g., poorly calibrated instrument).

•    Publication bias: positive studies more likely to be published → affects meta-analyses.

3. Measures of Disease Frequency

•    Incidence: new cases over a specified time period / population at risk.

•    Prevalence: total cases (new + existing) at a point in time / population.

•    Relative risk (RR): risk of outcome in exposed / unexposed.

•    Odds ratio (OR): odds of exposure among cases / controls; approximates RR when disease is rare.

4. Diagnostic Test Characteristics

Metric            Meaning                                                                    Formula

Sensitivity       Proportion of true positives detected                         TP / (TP + FN)

Specificity       Proportion of true negatives correctly identified        TN / (TN + FP)

PPV                Probability disease present if test positive                TP / (TP + FP)

NPV                Probability no disease if test negative                      TN / (TN + FN)

Sensitivity and specificity are intrinsic to the test.

PPV and NPV depend on disease prevalence.

5. Statistical Significance

•    p-value: probability that observed result is due to chance if null hypothesis true.

o    p < 0.05 → typically considered significant.

•    Confidence interval (CI):

o    Range in which true parameter is likely to lie (usually 95% CI).

o    If CI for RR or OR does not include 1 → statistically significant.

6. Interpreting Plots in Meta-Analyses

•    Forest plot:

o    Shows individual study effects and pooled effect.

o    Squares = individual study effect size; line = CI.

o    Diamond = pooled effect estimate.

•    Funnel plot:

o    Assesses publication bias.

o    Symmetry suggests low bias; asymmetry suggests potential publication bias.

Extra Revision Pearls

•    Number needed to treat (NNT) = 1 / absolute risk reduction (ARR).

•    Lead-time bias: apparent survival benefit due to earlier detection, not actual improvement.

•    Length-time bias: screening more likely to detect slowly progressing disease.

•    Intention-to-treat analysis: includes all randomised participants; preserves randomisation.