Absolute Risk

Absolute risk is fundamentally defined by a specific time horizon—typically 5-year, 10-year, or lifetime windows. Unlike relative risk, which expresses fold-change in odds, absolute risk answers the clinically relevant question: 'What is the probability this individual develops the disease within the next decade?' This temporal constraint makes it directly usable for clinical decision-making and screening guideline adherence.

• A 10-year absolute risk of 8% for breast cancer triggers different screening protocols than a lifetime risk of 12%
• Time horizons are calibrated to disease-specific latency periods and intervention windows
• Competing mortality risks must be accounted for in longer time horizons

Absolute risk anchors polygenic risk scores to real-world epidemiology by multiplying genetic relative risk against population-level baseline incidence rates. Without this calibration, a PRS provides only a rank ordering of genetic susceptibility. The baseline incidence—derived from cancer registries, national health databases, or cohort studies—converts the relative hazard into an interpretable probability.

• Baseline rates are stratified by age, sex, and ancestry to prevent miscalibration
• Incidence data sources include SEER, UK Biobank, and national health registries
• Failure to calibrate properly leads to systematic overestimation or underestimation of risk

Comprehensive absolute risk models combine the polygenic risk score with established epidemiological risk factors into a single unified estimate. These include family history, lifestyle exposures, clinical biomarkers, and demographic variables. The integration is typically performed through a Cox proportional hazards model or a Gail-type absolute risk model, which jointly estimates the hazard ratio from all covariates.

• Family history often contributes independently from common-variant PRS, capturing rare variant effects
• Modifiable risk factors like BMI or smoking status allow for dynamic risk re-estimation
• Interaction terms between genetic and environmental factors can be incorporated for more precise stratification

Absolute risk models are evaluated on two orthogonal dimensions: discrimination (how well the model separates cases from controls, measured by AUC-ROC) and calibration (how closely predicted probabilities match observed event rates). A well-calibrated model ensures that among individuals assigned a 10% risk, approximately 10% actually develop the disease. Calibration plots and the expected-to-observed ratio are standard validation tools.

• Good discrimination does not guarantee good calibration—systematic bias can inflate all predictions
• Calibration drift occurs when models are applied to populations with different baseline incidence
• Hosmer-Lemeshow and Greenwood-Nam-D'Agostino tests formally assess calibration quality

Absolute risk estimates are translated into clinical utility through risk stratification thresholds that trigger specific interventions. These thresholds are determined by cost-benefit analyses balancing the harms of over-screening against the benefits of early detection. For example, a 10-year absolute risk exceeding 2.75% for breast cancer may trigger MRI screening eligibility under certain guidelines.

• Thresholds vary by disease severity, screening modality risks, and healthcare system resources
• Net benefit analysis using decision curves evaluates clinical utility across threshold ranges
• Risk reclassification tables quantify how many individuals cross clinical thresholds when PRS is added to baseline models

Absolute risk estimates derived from European-ancestry GWAS and European-ancestry incidence rates produce systematically biased predictions when applied to non-European populations. This arises from differences in allele frequencies, linkage disequilibrium patterns, and baseline disease incidence. Proper cross-ancestry deployment requires recalibrating the baseline hazard and, ideally, using trans-ancestry PRS methods that incorporate diverse training data.

• Recalibration adjusts the baseline survival function to match the target population's incidence
• Calibration-in-the-large corrects for systematic over- or under-prediction across populations
• Without recalibration, absolute risk models exacerbate health disparities by providing inaccurate guidance to underrepresented groups

The ratio of the probability of an event occurring in an exposed group to the probability in a non-exposed group. Unlike absolute risk, it does not convey the baseline frequency of the outcome.

• Formula: Risk in exposed / Risk in unexposed
• Key Distinction: A high relative risk (e.g., 2.0) can correspond to a very small absolute risk increase if the baseline incidence is rare.
• Example: A drug doubling the risk of a condition that affects 1 in 10,000 people changes absolute risk from 0.01% to 0.02%.

The number of new cases of a disease occurring in a defined population during a specified time period. This rate serves as the baseline hazard in absolute risk models.

• Calculation: (New cases / Person-time at risk) × multiplier
• Role in PRS: Combined with a polygenic risk score to calibrate genetic risk against real-world disease frequency.
• Example: A 10-year incidence rate of 5% for a cancer type provides the foundation upon which genetic risk multipliers operate.

An epidemiological measure derived from absolute risk that indicates how many patients must receive a specific intervention to prevent one additional adverse outcome.

• Direct Dependence: NNT is the inverse of the Absolute Risk Reduction (ARR).
• Clinical Utility: Provides a more intuitive measure of treatment impact than relative risk reduction for clinical decision-making.
• Example: If a statin reduces 10-year cardiovascular absolute risk from 10% to 8%, the ARR is 2% and the NNT is 50.

The process of categorizing a population into distinct groups based on their predicted probability of developing a disease, using absolute risk thresholds to guide clinical intervention.

• Thresholds: Commonly use categories like 'low' (<5%), 'intermediate' (5-20%), and 'high' (>20%) 10-year risk.
• PRS Integration: Polygenic risk scores refine stratification beyond traditional clinical factors alone.
• Example: Breast cancer screening guidelines increasingly incorporate absolute risk models combining genetic, reproductive, and familial factors.

The proportion of disease cases in a population that would be eliminated if a specific risk factor were removed. It depends on both the relative risk of the factor and its prevalence in the population.

• Formula: [Prevalence × (RR - 1)] / [Prevalence × (RR - 1) + 1]
• Public Health Context: Guides resource allocation by quantifying the population-level impact of modifying a risk factor.
• Example: A common genetic variant with a small effect size can have a larger attributable fraction than a rare, high-penetrance mutation.

Events that preclude the occurrence of the primary event of interest. In absolute risk estimation, failing to account for competing risks leads to overestimation of the cumulative incidence.

• Method: The Fine-Gray subdistribution hazard model or cause-specific cumulative incidence functions (CIF) are used for correct estimation.
• Relevance: Critical in older populations where mortality from other causes is high.
• Example: A 10-year absolute risk model for prostate cancer must account for the competing risk of death from cardiovascular disease.