What are the questions that could be asked on medical biostatistics from a MD PHARMACOLOGY STUDENT. Give answers also.

Reading File

Finding Sources

Reading File

Finding Sources

Reading File

Finding Sources

Reading File

I now have extensive, authoritative content from the medical library. Let me compile a comprehensive Q&A guide for MD Pharmacology students on medical biostatistics.

Medical Biostatistics - Q&A for MD Pharmacology Students

This covers all high-yield topics examinable in MD Pharmacology - from study design and clinical trials (core to pharmacology) to statistical tests, error types, and diagnostic test evaluation.

SECTION 1: STUDY DESIGN

Q1. What is the PICO framework and why is it important in pharmacological research?

A: PICO is a structured method to formulate a clinical research question:

P - Patient/Population/Problem (e.g., hypertensive adults aged >60)
I - Intervention (e.g., ACE inhibitor)
C - Comparison (e.g., calcium channel blocker)
O - Outcome (e.g., reduction in major cardiovascular events)

It ensures the research question is clearly defined before designing a study, which is mandatory before searching for evidence or initiating a clinical trial. (Harriet Lane Handbook, 23rd ed.)

Q2. What are the major types of study designs? Classify them.

Category	Design	Features
Experimental	Randomized Controlled Trial (RCT)	Gold standard; random allocation to intervention/control
Experimental	Non-randomized/Quasi-experimental	No randomization
Observational - Analytical	Cohort study	Follows exposed vs. unexposed forward in time
Observational - Analytical	Case-control study	Compares cases vs. controls retrospectively
Observational - Analytical	Cross-sectional study	Prevalence snapshot at one point in time
Observational - Descriptive	Case report / Case series	Describes individual or grouped patient experiences
Secondary	Systematic review / Meta-analysis	Synthesizes multiple primary studies

For questions about drug interventions, RCTs are almost always the best study design because they eliminate residual confounding. (Harriet Lane Handbook; Rockwood & Green's Fractures, 10th ed.)

Q3. What is a Randomized Controlled Trial (RCT)? What are its key features?

A: An RCT is an experimental study in which participants are randomly assigned to an intervention group or a control group. Key features:

Randomization - eliminates both known and unknown confounders
Control group - provides a valid comparator (placebo or active comparator)
Blinding - reduces observer and participant bias
Prospective design - outcomes are measured after exposure/intervention

Key questions when appraising an RCT:

Were participants randomly allocated?
Was the control/comparator appropriate (beware "straw man" comparators)?
Were participants and investigators blinded?
Was an intention-to-treat (ITT) analysis performed?
Was the study adequately powered? (Harriet Lane Handbook, 23rd ed.)

Q4. What is the difference between a cohort study and a case-control study? Which is used in pharmacovigilance?

Feature	Cohort Study	Case-Control Study
Direction	Forward (prospective)	Backward (retrospective)
Starting point	Exposure (drug/factor)	Outcome (disease/ADR)
Measure of association	Relative Risk (RR)	Odds Ratio (OR)
Best for	Incidence, rare exposures	Rare diseases, ADR signals
Time & cost	Long, expensive	Shorter, cheaper

In pharmacovigilance, case-control studies are commonly used to detect rare adverse drug reactions (ADRs) because they are efficient for rare outcomes. Cohort studies are used for known drug exposures (e.g., post-marketing surveillance). (Miller's Review of Orthopaedics, 9th ed.)

Q5. What is the difference between internal validity and external validity?

Internal validity: The degree to which the study results are free from bias - the study was conducted in an unbiased fashion and results reflect the true effect within the study population.
External validity (generalizability): The degree to which the study results can be applied to the "real world" or other populations (e.g., can adult drug trial results be extrapolated to pediatric patients?).

A study with high internal validity may have low external validity if conducted in a highly selected population. (Harriet Lane Handbook, 23rd ed.)

Q6. What is blinding in clinical trials? Define single-blind, double-blind, and triple-blind.

Single-blind: Only the participant is unaware of treatment allocation
Double-blind: Both participant and investigator are unaware - this is the standard for most drug trials
Triple-blind: Participant, investigator, AND the data analyst/statistician are blinded

Related biases that blinding prevents:

Placebo effect - perceived treatment benefit due to belief in treatment
Nocebo effect - adverse effects due to perception of treatment
Hawthorne effect - participants change behavior because they know they are being studied
Observer-expectancy bias - researcher's belief in treatment efficacy affects their actions/assessment (Harriet Lane Handbook, 23rd ed.)

Q7. What is intention-to-treat (ITT) analysis? Why is it preferred over per-protocol (as-treated) analysis?

ITT analysis: Outcomes are analyzed based on the original group assignment, regardless of whether the participant actually received or completed the intervention.
Per-protocol (as-treated) analysis: Analyzes outcomes based on what treatment the participant actually received.

ITT is preferred because:

It preserves the benefits of randomization (comparability of groups)
It reflects real-world clinical effectiveness (not just efficacy under ideal conditions)
It avoids bias from non-random dropout (participant crossover is not random)

Per-protocol analysis is almost always incorrect for the primary outcome due to selection bias from crossover. (Harriet Lane Handbook, 23rd ed.)

SECTION 2: MEASURES OF ASSOCIATION AND EFFECT

Q8. Define and differentiate Relative Risk (RR) and Odds Ratio (OR).

Feature	Relative Risk (RR)	Odds Ratio (OR)
Definition	Ratio of incidence in exposed vs. unexposed	Ratio of odds of disease in cases vs. controls
Study design	Cohort studies, RCTs	Case-control studies
Formula	RR = [a/(a+b)] / [c/(c+d)]	OR = (a×d)/(b×c)
Interpretation	RR = 2 means 2× more likely to develop outcome	OR approximates RR when disease is rare
Null value	1.0	1.0

RR > 1: increased risk with exposure
RR < 1: protective effect (as seen with vaccines and many drugs)
OR approximates RR when the disease prevalence is low (<10%) - this is called the rare disease assumption

(Miller's Review of Orthopaedics, 9th ed.)

Q9. What is a confidence interval (CI)? How do you interpret a 95% CI?

A: A confidence interval quantifies the precision (uncertainty) of a statistical estimate (mean, RR, OR). A 95% CI means: if the study were repeated 100 times, 95 of those intervals would contain the true population value.

Interpretation for drug studies:

If the 95% CI for RR does NOT cross 1.0 → statistically significant
If the 95% CI for RR DOES cross 1.0 → not statistically significant
A narrow CI indicates high precision (large sample); a wide CI indicates imprecision (small sample)

Example: Drug A reduces MI risk, RR = 0.70 (95% CI: 0.55-0.89) → significant (CI does not cross 1.0).

(Miller's Review of Orthopaedics, 9th ed.)

Q10. What is Number Needed to Treat (NNT)? What is Number Needed to Harm (NNH)?

NNT = 1 / Absolute Risk Reduction (ARR)
- Represents how many patients need to receive the treatment for one additional patient to benefit
- Lower NNT = more effective treatment
NNH = 1 / Absolute Risk Increase (ARI)
- Represents how many patients need to receive the treatment for one additional patient to be harmed
- Higher NNH = safer treatment

Example:

Event rate in control = 20%, event rate in treatment group = 10%
ARR = 20% - 10% = 10% = 0.10
NNT = 1/0.10 = 10 (treat 10 patients to prevent 1 event)

NNT is a clinically intuitive measure widely used in pharmacology to communicate drug efficacy.

SECTION 3: HYPOTHESIS TESTING AND P-VALUES

Q11. What is the null hypothesis and alternative hypothesis?

Null hypothesis (H₀): States that there is NO difference between groups / no effect of the drug (e.g., "Drug A does not reduce blood pressure more than placebo"). This is the baseline assumption in every study.
Alternative hypothesis (H₁ or Hₐ): States that there IS a difference / effect (e.g., "Drug A reduces blood pressure more than placebo").

Statistical tests calculate the probability that the null hypothesis is true (the p-value). If p < 0.05, we reject H₀ and accept H₁. (Rockwood & Green's Fractures, 10th ed.)

Q12. What is a p-value? What does p < 0.05 mean?

A: The p-value is the probability of obtaining results at least as extreme as those observed, assuming the null hypothesis is true.

p < 0.05 (the conventional threshold): There is less than a 5% chance that the observed difference occurred by chance alone if the null hypothesis were true. We call this "statistically significant" and reject H₀.
p > 0.05: Results could plausibly have occurred by chance; we fail to reject H₀.
The 0.05 threshold is arbitrary (historically set by R.A. Fisher) but is the standard in medical literature.

Important caveat: Statistical significance ≠ clinical significance. A drug trial with a very large sample size may show p < 0.001 for a 1 mmHg reduction in BP - statistically significant but clinically meaningless. (Rockwood & Green's Fractures; Miller's Review of Orthopaedics)

Q13. What are Type I and Type II errors? Give examples relevant to pharmacology.

	Reality: H₀ True (no drug effect)	Reality: H₀ False (drug works)
Test says: Significant	Type I Error (α) - False positive	Correct decision (Power = 1-β)
Test says: Not significant	Correct decision (1-α)	Type II Error (β) - False negative

Type I error (α): Concluding a drug is effective when it is not. Conventionally set at 0.05 (5%).
- Example: Concluding Drug A reduces mortality when it truly does not.
Type II error (β): Concluding a drug has no effect when it truly does. Usually set at 0.10-0.20 (10-20%).
- Example: Missing a true antihypertensive effect because the sample was too small.

α and β are inversely related - reducing one increases the other (for a fixed sample size). (Creasy & Resnik's Maternal-Fetal Medicine; Rockwood & Green's Fractures, 10th ed.)

Q14. What is statistical power? How is it calculated and what factors affect it?

A: Statistical power = 1 - β = probability of correctly detecting a true effect (rejecting H₀ when it is indeed false).

Conventional power: 80% or 90% (β = 0.20 or 0.10)
A power of 80% means: if the drug truly works, the study will detect that 80% of the time.

Factors that INCREASE power:

Larger sample size (most important)
Larger effect size (bigger drug effect to detect)
Increasing α (raising the threshold, but this increases Type I error risk)
Lower variability in outcome data
One-tailed instead of two-tailed test (use with caution)

Clinical implication: An underpowered study may miss a clinically meaningful drug benefit or harm (especially uncommon ADRs in small studies). (Creasy & Resnik's Maternal-Fetal Medicine; Miller's Review of Orthopaedics, 9th ed.)

Q15. What components are needed for sample size calculation in a clinical trial?

A: For a cohort study or clinical trial, the following are required:

α error (Type I error threshold, usually 0.05)
β error (Type II error threshold, usually 0.10-0.20)
Incidence of outcome in unexposed/control subjects (from literature or pilot data)
Ratio of exposed to unexposed subjects (e.g., 1:1 randomization)
Minimum clinically important (detectable) relative risk or effect size

For a case-control study, you additionally need:

Prevalence of exposure in controls
Ratio of controls to cases
Minimum detectable odds ratio

Sample size estimates must be performed before the study begins (prospectively). Readers should scrutinize sample size in negative studies - an underpowered negative trial does not mean a drug is ineffective. (Creasy & Resnik's Maternal-Fetal Medicine)

SECTION 4: DESCRIPTIVE STATISTICS

Q16. Define mean, median, and mode. When is each used?

Mean: Sum of all values ÷ number of observations. Best for normally distributed (parametric) data.
Median: The middle value when data are arranged in order. Best for skewed data or ordinal variables.
Mode: The most frequently occurring value. Rarely used in clinical research.

Rule of thumb:

Normal distribution → use mean ± standard deviation (SD)
Skewed distribution → use median with interquartile range (IQR)
Drug concentration data (pharmacokinetics) is often right-skewed, so median/IQR is reported

(Rockwood & Green's Fractures, 10th ed.)

Q17. What is standard deviation (SD) vs. standard error of the mean (SEM)? Which should be reported?

Standard deviation (SD): Measures the spread/variability of individual data points around the mean. Describes the sample.
Standard error of the mean (SEM) = SD / √n. Measures the precision of the sample mean as an estimate of the population mean.

SEM is always smaller than SD - researchers sometimes misuse SEM in graphs to make data appear less variable than it is.

For reporting: SD should be used when describing sample characteristics (baseline demographics, PK parameters). SEM is used when reporting precision of an estimate.

(Rockwood & Green's Fractures, 10th ed.)

Q18. What is a normal distribution? What are its properties?

A: A normal (Gaussian) distribution is a symmetrical bell-shaped frequency distribution where:

Mean = Median = Mode
68% of values fall within ±1 SD of the mean
95% fall within ±1.96 SD of the mean
99.7% fall within ±3 SD of the mean

In pharmacology: Most PK parameters (Vd, Cl) are reported assuming log-normal distribution. When drug concentration data is log-transformed, it often becomes normally distributed (useful for dosing studies).

Departures from normality (skewed distributions) require non-parametric statistical tests.

SECTION 5: STATISTICAL TESTS

Q19. How do you choose the appropriate statistical test? Provide a framework.

Comparison	Data type	Test
2 groups (independent)	Continuous, normal	Independent samples t-test
2 groups (paired/matched)	Continuous, normal	Paired t-test
≥3 groups	Continuous, normal	ANOVA (followed by post-hoc tests)
2 groups	Continuous, non-normal/ordinal	Mann-Whitney U test
≥3 groups	Non-normal	Kruskal-Wallis test
2 groups	Categorical	Chi-square (χ²) test
Small samples / categorical	Categorical, small n	Fisher's exact test
Correlation between 2 continuous variables	Continuous, normal	Pearson's r
Correlation (non-normal/ordinal)	Ordinal/non-normal	Spearman's rank correlation
Predict outcome from one variable	Continuous outcome	Simple linear regression
Predict categorical outcome	Binary outcome	Logistic regression

Key rules:

Parametric tests require continuous data + normal distribution
Non-parametric tests for categorical or non-normally distributed data
Post-hoc testing is mandatory after ANOVA to find where specific differences lie

(Miller's Review of Orthopaedics, 9th ed.; Cummings Otolaryngology)

Q20. What is ANOVA? When is post-hoc testing needed?

A: Analysis of Variance (ANOVA) compares means across three or more groups simultaneously. It tests whether at least one group mean differs from the others.

One-way ANOVA: One independent variable with ≥3 groups (e.g., comparing 3 drug doses)
Two-way ANOVA: Two independent variables simultaneously (e.g., drug dose AND gender)
Repeated measures ANOVA: Same subjects measured multiple times (e.g., drug concentrations at T=0, 1, 2, 4 hours)

Post-hoc tests (e.g., Tukey's, Bonferroni, Scheffe) are mandatory after a significant ANOVA result to determine exactly which pairs of groups differ. Without post-hoc testing, you only know "at least one group differs" but not which one.

(Miller's Review of Orthopaedics, 9th ed.)

Q21. What is the chi-square (χ²) test? Give a pharmacological example.

A: The chi-square test compares the distribution of categorical (nominal or ordinal) data between two or more groups. It tests whether there is an association between two categorical variables.

Requirements: Expected frequency in each cell ≥ 5. For small sample sizes, use Fisher's exact test.

Pharmacological example: Comparing the proportion of patients experiencing nausea in Drug A group (15/100) vs. placebo group (6/100). Chi-square tests whether this difference in proportions is statistically significant.

(Miller's Review of Orthopaedics, 9th ed.)

Q22. What is logistic regression? When is it used in pharmacological research?

A: Logistic regression is used when the outcome variable is binary (categorical: yes/no, event/no event) and you want to predict the probability of that outcome from one or more predictor variables.

Output: Odds ratios (OR) with 95% CIs for each predictor.

Pharmacological examples:

Predicting the probability of a drug-induced adverse event (e.g., hepatotoxicity: yes/no) from dose, age, renal function
Identifying predictors of non-adherence to a drug regimen
Pharmacogenomics: predicting drug response (responder/non-responder) based on genetic variants

(Miller's Review of Orthopaedics, 9th ed.)

SECTION 6: DIAGNOSTIC TEST STATISTICS (HIGH-YIELD)

Q23. Define sensitivity and specificity with formulas. Which do you prioritize for screening vs. diagnosis?

Using the 2×2 table (TP = True Positive, FP = False Positive, TN = True Negative, FN = False Negative):

	Disease Present	Disease Absent
Test Positive	TP	FP
Test Negative	FN	TN

Sensitivity = TP / (TP + FN) - ability to detect true cases (avoids false negatives)
- Mnemonic: SnNout - a highly Sensitive test, when Negative, rules OUT disease
Specificity = TN / (TN + FP) - ability to correctly identify non-cases (avoids false positives)
- Mnemonic: SpPin - a highly Specific test, when Positive, rules IN disease

For SCREENING: Prioritize HIGH sensitivity (want to catch all cases, tolerate some false positives - e.g., HIV screening test) For DIAGNOSIS/CONFIRMATION: Prioritize HIGH specificity (want to confirm diagnosis, avoid false positives - e.g., Western blot for HIV)

(Henry's Clinical Diagnosis and Management by Laboratory Methods)

Q24. What are Positive Predictive Value (PPV) and Negative Predictive Value (NPV)? How does prevalence affect them?

PPV = TP / (TP + FP) = probability that a positive test result truly indicates disease
NPV = TN / (TN + FN) = probability that a negative test result truly indicates absence of disease

Effect of prevalence (pre-test probability):

As prevalence increases: PPV increases, NPV decreases
As prevalence decreases (e.g., general population screening): PPV decreases dramatically (more false positives), NPV increases

This is critical in pharmacology for biomarker-guided therapy:

A diagnostic test for a drug biomarker in a high-prevalence specialty clinic will have much higher PPV than the same test used in general population screening.

(Henry's Clinical Diagnosis and Management by Laboratory Methods)

Q25. What is the ROC curve? What is the area under the curve (AUC) and how is it interpreted?

A: The Receiver Operating Characteristic (ROC) curve plots Sensitivity (true positive rate) on the Y-axis vs. 1-Specificity (false positive rate) on the X-axis across all possible cut-off values for a test.

Area Under the Curve (AUC) / c-statistic:

AUC = 1.0: Perfect test
AUC = 0.9-1.0: Excellent
AUC = 0.8-0.9: Good
AUC = 0.7-0.8: Fair
AUC = 0.5: No discriminating ability (equivalent to chance/coin flip)

Uses in pharmacology:

Comparing two drug concentration biomarkers for predicting toxicity
Evaluating pharmacodynamic endpoints
Selecting optimal drug concentration cutoffs for therapeutic drug monitoring (TDM)

(Miller's Review of Orthopaedics, 9th ed.)

Q26. What is the likelihood ratio (LR)? How is it used clinically?

Positive LR (LR+) = Sensitivity / (1 - Specificity) = how much a positive test increases the probability of disease
Negative LR (LR-) = (1 - Sensitivity) / Specificity = how much a negative test decreases the probability of disease

Interpretation:

LR+ > 10: Large and conclusive increase in probability of disease
LR+ 5-10: Moderate increase
LR- < 0.1: Large and conclusive decrease in probability of disease
LR of 1.0: Test is useless (no change in probability)

LRs are better than sensitivity/specificity alone because they can be used with pre-test probability (Bayes' theorem / Fagan nomogram) to calculate post-test probability of disease.

SECTION 7: BIAS AND CONFOUNDING

Q27. Define bias. What are the major types relevant to drug trials?

A: Bias is a systematic error in a study that causes a deviation of results from the truth. Unlike random error, bias is directional and cannot be corrected by increasing sample size.

Major types:

Type	Description	Example in Pharmacology
Selection bias	Non-random selection of participants	Healthier patients selected for new drug arm
Recall bias	Cases recall exposure differently from controls	ADR case-control studies
Observer bias	Investigator's expectations influence measurement	Non-blinded drug assessment
Attrition/Withdrawal bias	Dropout not random across groups	Sicker patients drop from drug trial
Detection bias	Outcome measured differently between groups	More follow-up in drug group
Publication bias	Positive trials more likely to be published	Meta-analyses overestimate drug benefit

(Harriet Lane Handbook, 23rd ed.)

Q28. What is confounding? How is it controlled in drug studies?

A: A confounder is a variable that is associated with both the exposure (drug) and the outcome, and distorts the true association between them.

Example: In a study of aspirin and colorectal cancer - age is a confounder (older patients take more aspirin AND have higher cancer rates).

Methods to control confounding:

Design phase: Randomization (best - handles both known and unknown confounders), restriction, matching
Analysis phase: Stratification, multivariate regression (logistic or Cox regression), propensity score analysis

Randomization is the primary reason RCTs are considered the gold standard for drug efficacy testing.

SECTION 8: META-ANALYSIS AND SYSTEMATIC REVIEWS

Q29. What is a systematic review? How does it differ from a meta-analysis?

Systematic review: A structured, reproducible synthesis of all available evidence on a specific clinical question, using pre-defined inclusion/exclusion criteria and a thorough literature search. May or may not involve statistical pooling.
Meta-analysis: A statistical technique used within a systematic review to quantitatively pool results from multiple individual studies into a single summary estimate.

Not all systematic reviews contain meta-analyses (pooling is only appropriate when studies are sufficiently similar in design, population, and outcomes).

In evidence-based pharmacology, systematic reviews of RCTs represent the highest level of evidence for drug efficacy.

Q30. What is a forest plot? How do you read one?

A: A forest plot graphically displays individual study results and a pooled summary estimate in a meta-analysis.

How to read:

Each horizontal line = one study; the square = point estimate (RR/OR), width of the line = 95% CI
The size of the square reflects the weight of the study (usually based on sample size)
The diamond at the bottom = pooled summary estimate; its width = 95% CI of the pooled effect
The vertical line at 1.0 = line of no effect
- If the CI of a study or the diamond does NOT cross 1.0 → statistically significant
Heterogeneity (I² statistic): measures variability between studies
- I² > 50% suggests substantial heterogeneity (pooling may be inappropriate)

SECTION 9: PHARMACOLOGY-SPECIFIC STATISTICAL CONCEPTS

Q31. What is the difference between efficacy and effectiveness in clinical trials?

Efficacy: How well a drug works under ideal, controlled conditions (Phase III RCT - explanatory trial, per-protocol analysis)
Effectiveness: How well a drug works in real-world clinical practice (pragmatic trials, ITT analysis, post-marketing studies)

A drug may have high efficacy in a trial but low effectiveness in practice due to non-adherence, comorbidities, or use in different populations.

Q32. What are Phase I, II, III, and IV clinical trials?

Phase	Population	Primary Purpose	Key Statistics
Phase I	Healthy volunteers (20-80)	Safety, pharmacokinetics, dose escalation	PK parameters (Cmax, t½, AUC), MTD
Phase II	Patients with disease (100-300)	Efficacy signal, dose-finding, safety	Response rates, preliminary efficacy
Phase III	Large patient population (1000s)	Efficacy vs. standard care, safety	RR, OR, NNT, p-value, 95% CI
Phase IV	Post-marketing general population	Long-term safety, rare ADRs, pharmacovigilance	Cohort/case-control, signal detection

Q33. What is the difference between surrogate endpoints and clinical endpoints?

Surrogate endpoint: A laboratory or biomarker measurement used as a proxy for a clinically meaningful outcome (e.g., HbA1c for diabetes outcomes, LDL for cardiovascular events, BP for stroke)
Clinical (hard) endpoint: A direct measure of how a patient feels, functions, or survives (e.g., MI, death, hospitalization)

Caution: Not all surrogate-endpoint improvements translate to clinical benefit. Example: Several drugs that lower LDL or improve surrogate markers failed to reduce mortality in Phase III trials.

The use of composite outcomes to increase statistical power may also inflate apparent efficacy if driven by a non-clinically important component. (Harriet Lane Handbook, 23rd ed.)

Q34. What is multiplicity (multiple comparisons) and how is it corrected?

A: When multiple statistical tests are performed on the same dataset, the probability of a false-positive result (Type I error) increases with each additional test.

Testing 20 hypotheses at α = 0.05 → expect 1 false positive by chance alone

Corrections:

Bonferroni correction: Divide α by the number of comparisons (new α = 0.05/n). Conservative.
False Discovery Rate (FDR, Benjamini-Hochberg): Controls the expected proportion of false positives among significant results. Less conservative.
Pre-specifying primary vs. secondary outcomes in trial registration prevents post-hoc data dredging (p-hacking)

This is especially important in pharmacogenomics (GWAS studies involving thousands of SNPs) and Phase II basket/umbrella trials.

SECTION 10: EPIDEMIOLOGICAL MEASURES

Q35. Define prevalence and incidence. How are they related?

Prevalence = Number of existing cases / Total population (at a point in time). A snapshot.
- Used for chronic diseases; relevant for population-level drug prescribing burden
Incidence (rate) = Number of NEW cases / Population at risk over a specified time period. Requires follow-up.
- Measures disease occurrence/drug exposure risk dynamically

Relationship: Prevalence ≈ Incidence × Average duration of disease (for stable conditions)

A highly effective drug that shortens disease duration will decrease prevalence even without decreasing incidence
Incidence is more relevant in pharmacoepidemiology (measuring new ADR events)

(Miller's Review of Orthopaedics, 9th ed.)

Q36. What is absolute risk reduction (ARR) vs. relative risk reduction (RRR)?

ARR = Risk in control group - Risk in treatment group
- Example: 20% mortality in placebo, 14% in drug group → ARR = 6%
RRR = ARR / Risk in control group = (20% - 14%) / 20% = 30%
NNT = 1 / ARR = 1/0.06 = ~17

Key point: Drug trials often report RRR (sounds more impressive) rather than ARR (more clinically informative). A drug may reduce relative risk by 50% but the ARR may only be 1% if baseline risk is low (e.g., NNT = 100 - meaning 100 patients must be treated to prevent one event).

Critical appraisal: always convert RRR to ARR and NNT for clinical decision-making.

Sources: Harriet Lane Handbook 23rd ed.; Henry's Clinical Diagnosis & Management by Laboratory Methods; Rockwood & Green's Fractures in Adults 10th ed.; Creasy & Resnik's Maternal-Fetal Medicine; Miller's Review of Orthopaedics 9th ed.; Cummings Otolaryngology Head & Neck Surgery

Summary Table: High-Yield Mnemonics

Concept	Mnemonic
Sensitivity rules out	SnNout
Specificity rules in	SpPin
Type I error = False Positive	α = Alarm without fire
Type II error = False Negative	β = Blind to the truth
Power = 1 - β	The ability to find a real effect
NNT = 1/ARR	Lower is better (fewer to treat)
LR+ >10 or LR- <0.1	Clinically significant change

This is a shared conversation. Sign in to Orris to start your own chat.