1st Edition
Absolute Risk Methods and Applications in Clinical Management and Public Health
Absolute Risk: Methods and Applications in Clinical Management and Public Health provides theory and examples to demonstrate the importance of absolute risk in counseling patients, devising public health strategies, and clinical management. The book provides sufficient technical detail to allow statisticians, epidemiologists, and clinicians to build, test, and apply models of absolute risk.
Features:
- Provides theoretical basis for modeling absolute risk, including competing risks and cause-specific and cumulative incidence regression
- Discusses various sampling designs for estimating absolute risk and criteria to evaluate models
- Provides details on statistical inference for the various sampling designs
- Discusses criteria for evaluating risk models and comparing risk models, including both general criteria and problem-specific expected losses in well-defined clinical and public health applications
- Describes many applications encompassing both disease prevention and prognosis, and ranging from counseling individual patients, to clinical decision making, to assessing the impact of risk-based public health strategies
- Discusses model updating, family-based designs, dynamic projections, and other topics
Ruth M. Pfeiffer is a mathematical statistician and Fellow of the American Statistical Association, with interests in risk modeling, dimension reduction, and applications in epidemiology. She developed absolute risk models for breast cancer, colon cancer, melanoma, and second primary thyroid cancer following a childhood cancer diagnosis.
Mitchell H. Gail developed the widely used "Gail model" for projecting the absolute risk of invasive breast cancer. He is a medical statistician with interests in statistical methods and applications in epidemiology and molecular medicine. He is a member of the National Academy of Medicine and former President of the American Statistical Association.
Both are Senior Investigators in the Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health.
Introduction
Examples of risk models for disease incidence
Breast cancer incidence
A brief survey of models
The National Cancer Institute’s (NCI’s) Breast Cancer Risk
Assessment Tool, BCRAT
Other models of cancer incidence
Framingham Model for incidence of coronary heart disease
Applications of risk models for disease incidence
Prognosis after disease diagnosis
Contents of book
Definitions and basic concepts for survival data in a cohort without covariates
Basic survival concepts
Choice of time scale: age, time since diagnosis, time since accrual or counseling
Censoring
Right censoring
Truncation
Life-table estimator
Kaplan–Meier survival estimate
Counting processes and Markov methods
Competing risks
Concepts and definitions
Pure versus cause-specific hazard functions
Non-parametric estimation of absolute risk
Regression models for absolute risk estimated from cohort data
Cause-specific hazard regression
Estimation of the hazard ratio parameters
Non-parametric estimation of the baseline hazard
Semi-parametric estimation of absolute risk rm
Estimation of a piecewise exponential baseline hazard model
Alternative hazard models
Cumulative incidence regression
Proportional sub-distribution hazards model
Other cumulative incidence regression models
Relationship between the cause-specific and the proportional sub-distribution hazards models
Examples
Absolute risk of breast cancer incidence
Absolute risk of second primary thyroid cancer (SPTC) incidence
Estimating cause-specific hazard functions from sub-samples from cohorts
Case-cohort design
Nested case-control design
Estimating cause specific hazard functions from cohorts with complex survey designs
Example of survey design
Data
Estimation of hazard ratio parameters and the baseline hazard function
Example: absolute risk of cause-specific deaths from the NHANES I and II
Variance estimation
Approaches to variance estimation
Influence function based variance of the absolute risk estimate from cohort data
Estimating absolute risk by combining case-control or cohort data with disease registry data
Relationship between attributable risk, composite age-specific incidence, and baseline hazard
Estimating relative risk and attributable risk from case-control data
Estimating relative risk and attributable risk from cohort data
Estimating the cause-specific hazard of the competing causes of mortality, λ(t; z)
Some strengths and limitations of using registry data
Absolute risk estimate
Example: estimating absolute risk of breast cancer incidence by combining cohort data with registry data
Variance computations
Relative risk parameters and attributable risk estimated from a case-control study
Relative risk parameters and attributable risk estimated from a cohort study
Variance computation when an external reference survey is used to obtain the risk factor distribution
Resampling methods to estimate variance
Assessment of risk model performance
Introduction
The risk distribution
The predictiveness curve
Calibration
Definition of calibration and tests of calibration
Reasons for poor calibration and approaches to recalibration
Assessing calibration with right censored data
Assessing calibration on the training data, that is, internal validation
Accuracy measures
Predictive accuracy: the Brier score and the logarithmic score
Classification accuracy
Distribution of risk in cases and non-cases
Accuracy criteria
Discriminatory accuracy
Extensions of accuracy measures to functions of time and allowance for censoring
Criteria for applications of risk models for screening or high-risk interventions
Proportion of cases followed and proportion needed to follow
Estimation of PCF and PNF
Model assessment based on expected costs or expected utility specialized for a particular application
Two health states and two intervention options
More complex outcomes and interventions
Example with four intervention choices
Multiple outcomes in prevention trials
Expected cost calculations for outcomes following disease diagnosis
Comparing the performance of two models
Use of external validation data for model comparison
Data example
Comparison of model calibration
Model comparisons based on the difference in separate model-specific estimates of a criterion
Comparisons of predictive accuracy using the Brier and logarithmic scores
Classification accuracy criteria based on single risk threshold
Comparisons based on the receiver operating characteristic (ROC) curve
Integrated discrimination improvement (IDI) and mean risk difference
Comparing two risk models based on PCF, PNF, iPCF, or iPNF
Comparisons based on expected loss or expected benefit
Joint distributions of risk
Risk stratification tables and reclassification indices
Net reclassification improvement (NRI)
Extensions of NRI
Concluding remarks
Building and updating relative risk models
Introductory remarks
Selection of covariates
Missing data
Types of missing data
Approaches to handling missing data
Updating risk models with new risk factors
Estimating an updated relative risk model, rr(X,Z), from case control data
Estimating rr(X,Z) from new data only
Incorporating information on rr(X) into rr(X,Z) via likelihood ratio (LR) updating
Joint estimation of LRD(Z|X)
Estimating LRD(Z|X) based on fitting separate models for cases (D = ) and non-cases (D = )
LR updating assuming independence of Z and X (independence Bayes)
LR updating with multiple markers
Joint estimation, logistic model with offset
Independence Bayes with shrinkage
Updating using constrained maximum likelihood estimation (CML)
Simulations
Summary
Risk estimates based on genetic variants and family studies
Introduction
Mendelian models: the autosomal dominant model for pure breast cancer risk
Models that allow for residual familial aggregation to estimate pure breast cancer risk
Polygenic risk
Models with latent genetic effects: BOADICEA and IBIS
Copula models
Estimating genotype-specific absolute risk from family-based designs
General considerations
Combining relative-risks from family-based case-control studies with population-based incidence rates
Kin-cohort design
Families with several affected members (multiplex pedigrees)
Comparisons of some models for projecting breast cancer risk
Discussion
Related topics
Introduction
Prognosis following disease onset
Missing or misclassified information on event type
Time varying covariates
Fixed versus time-varying covariates and internal versus external time-varying covariates
Joint modeling of covariates and health outcomes, including multistate models
Landmark analysis
Risk model applications for counseling individuals and for public health strategies for disease prevention
Use of risk models in counseling individuals
Providing realistic risk estimates and perspective
More formal risk-benefit analysis for individual counseling
Use of risk models in public health prevention
Designing intervention trials to prevent disease
Assessing absolute risk reduction in a population from interventions on modifiable risk factors
Implementing a “high risk” intervention strategy for disease prevention
Allocating preventive interventions under cost constraints
Biography
Ruth M. Pfeiffer is a mathematical statistician and Fellow of the American Statistical Association, with interests in risk modeling, dimension reduction, and applications in epidemiology. She developed absolute risk models for breast cancer, colon cancer, melanoma, and second primary thyroid cancer following a childhood cancer diagnosis.
Mitchell H. Gail developed the widely used "Gail model" for projecting the absolute risk of invasive breast cancer. He is a medical statistician with interests in statistical methods and applications in epidemiology and molecular medicine. He is a member of the National Academy of Medicine and former President of the American Statistical Association.
Both are Senior Investigators in the Division of Cancer Epidemiology and Genetics, National Cancer Institute, National Institutes of Health.
"Written by two leading experts in the field, this book provides a comprehensive overview of absolute risk, including both theoretical basis and clinical implications before and after the disease diagnosis. Equipped with sufficient technical details on the estimation and inference of absolute risk aswell as a range of real examples, this book is targeted toward a broad audience, including epidemiologists, clinicians, and statisticians. While a few other books on theoretical aspects of absolute risk are available in the literature, the book by Pfeiffer and Gail treats absolute risk from several new angles . . ."
~Journal of the American Statistical Association"The book by Pfeiffer and Gail leads us into the higher statistical levels of predicting the medical future. The main focus is on the concept of the absolute risk of an event because this has a clinically meaningful interpretation for the individual person. The much more commonly reported hazard ratios of health research do not provide a directly useful number for the single subject...The examples are about the real world (mostly cancer research), and the mathematics provide all the formula for building a well‐calibrated absolute risk model and the validation study...The book contains a lot of material which is very difficult to find elsewhere, for example, on family studies, handling of missing data, and landmark analysis with time-dependent covariates. Overall, I found the book to provide a very complete documentation of a highly important subject. The authors are to be thanked for their thoroughness and congratulated for their work, which should be useful for many real‐world applications of absolute risk."
~Biometrics