What is a risk factor?
Factor associated with a disease; can be causal (directly causes disease) or non-causal (marker — related to something else that causes it)
2 variables are associated when…
Change in one corresponds to change in the other (association ≠ causation)
Positive vs negative association
Positive: ↑ exposure → ↑ disease | Negative: ↑ exposure → ↓ disease
3 measures of association
RR (risk ratio/rate ratio), AR (attributable risk), OR (odds ratio)
RR formula
CI in exposed ÷ CI in non-exposed (or incidence rate in exposed ÷ incidence rate in non-exposed for rate ratio)
RR interpretation: =1, >1, <1
=1: no association | >1: positive association (possible causal) | <1: negative association (possible protective)
RR scenario — 5-yr DM risk: obese 8/99, non-obese 4/109. Calculate & interpret RR
RR = (8/99)/(4/109) = 0.081/0.037 = 2.2 → obese have 2.2× the risk of DM vs non-obese
RR scenario — CAD incidence: LDL≥125 = 55/1000 p-y, LDL<125 = 25/1000 p-y. RR?
RR = 0.055/0.025 = 2.2 → high LDL group has 2.2× the risk of CAD vs low LDL group
When to use risk ratio vs rate ratio
Risk ratio: uses cumulative incidence (proportion) | Rate ratio: uses incidence rate (events/person-time)
Multi-level exposure: what changes about RR calculation?
Instead of exposed vs unexposed, choose one level as reference (non-exposed) and calculate RR for each other level against it
Multi-level exposure scenario — CHD risk in women: Normal 3%, Mild HC 7%, Severe HC 11%. Using normal as reference, calculate RR for Mild and Severe HC
Mild HC RR = 7/3 = 2.3 | Severe HC RR = 11/3 = 3.7 → dose-response: higher cholesterol = higher CHD risk
Multi-level exposure scenario — CHD risk in men: Normal 8%, Mild HC 12%, Severe HC 18%. Using normal as reference, calculate RR for Mild and Severe HC
Mild HC RR = 12/8 = 1.5 | Severe HC RR = 18/8 = 2.2
Multi-level exposure scenario — 10-yr CHD incidence: Women: Normal 3%, Mild HC 7%, Severe HC 11% | Men: Normal 8%, Mild HC 12%, Severe HC 18%. Compare CHD risk between sexes at each cholesterol level (women as reference) Normal: 8/3 = 2.7 | Mild HC: 12/7 = 1.7 | Severe HC: 18/11 = 1.6 → men consistently higher risk, but gap narrows as HC worsens
Normal: 8/3 = 2.7 | Mild HC: 12/7 = 1.7 | Severe HC: 18/11 = 1.6 → men consistently higher risk, but gap narrows as HC worsens
In multi-level RR, what does a dose-response pattern suggest?
Increasing RR with increasing exposure level suggests a causal relationship
Attributable risk (AR) — what does it measure?
Absolute difference in disease probability between exposed and non-exposed; measures excess risk due to exposure (also called public health risk)
AR formula
AR = Risk(exposed) − Risk(unexposed)
AR vs RR — key difference in operation
AR = subtract | RR = divide (same data, different operation)
AR scenario — 5-yr DM risk: obese 8/99, non-obese 4/109. Calculate AR & interpret
AR = 8/99 − 4/109 = 0.081 − 0.037 = 0.044 → obese have 0.044 (4.4%) HIGHER absolute risk of DM than non-obese
AR from RR only (no raw data) — formula & when used
AR = (RR−1)/RR; used when only RR is given, gives proportion of exposed group’s risk attributable to exposure
AR scenario — esophageal cancer: heavy smokers have RR=5 vs non-smokers. What proportion of cancer risk in smokers is attributed to smoking?
AR = (5−1)/5 = 4/5 = 0.8 → 80% of esophageal cancer risk in heavy smokers is attributable to smoking
When is OR used instead of RR?
Example ? (In which study OR is used)
When RR cannot be calculated — especially in case-control studies; outcome is dichotomous (case vs control); exposure can be categorical or continuous
How is OR interpreted?
Odds of disease in exposed vs non-exposed (NOT odds of exposure in diseased vs non-diseased)
OR formula (2×2 table)
OR = ad/bc (cross-product rule) | a=exposed cases, b=exposed controls, c=unexposed cases, d=unexposed controls
OR scenario — DM case-control: overweight 30 cases/15 controls, normal weight 70 cases/85 controls. Calculate & interpret OR
OR = (30×85)/(70×15) = 2550/1050 = 2.4 → odds of DM in obese are 2.4× the odds in non-obese
