Glossary

Models taking a comprehensive approach postulate explicit processes for phenomena that cannot be directly observed. For example, growth rates for untreated pre-cancerous conditions are not typically observed but may be estimated (assuming a particular growth model) from autopsy data.

Observation focused models are constructed using only observed data with no explicit modeling of unobserved processes behind the observed data.

While models are often multi-purpose, this property describes the effort's initial primary goal. Not subsequent or ancillary uses.

Screening evaluation includes both program evaluation (i.e.: selective screening targets, intervals, and related protocols) and evaluation of screening tests in the context of assumptions regarding natural history, incidence, prevalence and test dissemination.

Models with this primary purpose are focused on causes and effects of disease in individuals and populations and in forming public heath decisions via simulation that synthesizes and extends empirical data.

Models with this primary purpose are concerned with the evaluation of actual or proposed public health policies and their impact on a population's health.

Models with this primary purpose typically synthesize and explain surveillance data and focus on macro-scale trends seen in populations.

This property group addresses specific features modeling teams have chosen to include in models.

Features concerned with evaluation of methods and effects of approaches to cancer prevention.

Features concerned with detection of pre-clinical cancer in individuals or sub-populations.

Features involving the treatment of detected disease, often differentiated by temporal variables (e.g.: calendar year, age) and/or mode of detection.

Features that account for new cancer in non-primary locations after initial diagnosis.

Features that explicitly allow for cancer to reoccur in the same individual after initial detection and possible treatment. Does not include models that implicitly account for recurrence via long-term survival statistics.

Models that include the concept of a defined tumor with a defined size that can change over time, possibly influencing detectability and treatability.

Models that include generation of marker levels such as tumor-specific antigens separate from tests for those biomarkers.

Features that account for precancer

Features that account for a virus (such as HPV) after initial diagnosis

Features other than dysplasia and virus that account for precancer after initial diagnosis

Computer models that operate at the level of individuals or smaller entities such as tumors or cells are considered 'micro simulations'. Micro simulations may involve sub-components that are analytical in nature. That is, micro simulations may involve sub-components that are strictly analytical, but the complete model cannot be represented as a closed form equation.

Computer models that operate at the level of populations (i.e. not individuals). Micro simulations may involve sub-components that are strictly analytical, but the complete model cannot be represented as a closed form equation.

Analytic models are mathematical models based on closed form equations that represents a system in terms of logical and quantitative relationships.Software may be employed in automation of calibration and other result generation, but the core model is described in terms of equations not software.

An agent-based model is a micro simulation that allows for interaction between simulated entities.For example, smoking behavior in a simulated individual may be influenced by the behavior of other simulated individuals.

Dynamic Transmission models direct and indirect effects from communicable diseases that can't be effectively modeled through other methods. Dynamic Transmission models in the cervical sites model the risk of HPV infection depending on the number of new partners, the prevalence of HPV in the population, and the probability of HPV transmission per susceptible-infected partnership.

Longitudinal models represent sequences of individual clinical observations or events, typically with time as independent variable, often with event probabilities conditional on previous event outcomes.

Models using this approach estimate parameters by maximizing a defined likelihood function. That is, maximizes the agreement of the model's outputs with observed data.

Models involving stochastic processes that employ probability distributions that change over time.

State transition models posit a set of discrete states that the unit of analysis (typically an individual person) can be in at any time. The conditional probabilities of transitioning between states over a given interval of time are established as model parameters. An individual's longitudinal histories can be generated. Variants of Markov models fall into this category.

Time to event models utilize time-based event probability distributions to determine the time (typically indexed by age) of various events. These distributions may be conditional. A variant of Monte-Carlo methodology is typically employed in these models.

This property defines the smallest entity the model is concerned with. While models may aggregate to person or population levels, those aggregations may be generated on the basis of explicit simulation of individual tumors or cells. This property clarifies the lowest level of explicit simulation.

Breast Cancer Surveillance Consortium 
http://breastscreening.cancer.gov/

Two County Trial Data (Sweden)

Mammographic screening and mortality from breast cancer: The Malmo mammographic screening trial

Canadian National Breast Cancer Screening Studies

Wisconsin Cancer Reporting System

Nurses' Health Studies
http://www.nurseshealthstudy.org/

Health Professionals' Follow-up Study
https://www.hsph.harvard.edu/hpfs/

Cancer Prevention Study I

Cancer Prevention Study II
http://www.cancer.org/research/researchtopreventcancer/currentcancerpreventionstudies/cancer-prevention-study

Nationwide network and registry of histo- and cytopathology in the Netherlands

Population-based Research Optimizing Screening through Personalized Regimens
http://healthcaredelivery.cancer.gov/prospr/

Kaiser Permanente Northern California

Surveillance, Epidemiology, and End Results
http://seer.cancer.gov/

New Mexico HPV Pap Registry
http://hpvprevention.unm.edu/NMHPVPR/

Finnish Cancer Registry
http://www.cancer.fi/syoparekisteri/en/

New South Whales (NSW) Central Cancer Registry (CCR)
http://www.cancerinstitute.org.au/data-and-statistics/cancer-registries/nsw-cancer-registry

Australian Cancer Database
http://www.aihw.gov.au/australian-cancer-database/

Netherlands Cancer Registry

National Mortality Database
http://www.aihw.gov.au/deaths/aihw-deaths-data/#nmd

Victorian Cervical Cytology Register

Other Cancer Registry not listed

A linked data set is a combination of otherwise separate data sets whereby records have been linked via matching of attributes or common identifiers.

The SEER Medicare data consists of a linkage of the clinical data collected by the SEER registries with claims for health services collected by Medicare for its beneficiaries.

Other Linked data source not listed

European Randomized Study of Screening for Prostate Cancer
http://www.erspc-media.org/

Prostate Cancer Prevention Trial
http://www.cancer.gov/clinicaltrials/noteworthy-trials/pcpt/Page1

Prostate, Lung, Colorectal, and Ovarian trial
http://prevention.cancer.gov/plco

Scandinavian Prostate Cancer Group-4 randomized trial
http://www.ncbi.nlm.nih.gov/pubmed/23271778

Health Insurance Plan Study

Women’s health Initiative (WHI) Postmenopausal Hormone Therapy Trials
http://www.nhlbi.nih.gov/whi/background.htm

National Polyp Study

National Lung Screening Trial
http://www.cancer.gov/clinicaltrials/noteworthy-trials/nlst

Population Based Screening Study Amsterdam
http://www.ncbi.nlm.nih.gov/pubmed/15054873

New Technologies for Cervical Cancer screening trial

Other Clinical trial not listed

National Comprehensive Cancer Network
http://www.nccn.org/

NCI SEER Patterns of Care
http://appliedresearch.cancer.gov/poc/

National Health Interview Surveys
http://www.cdc.gov/nchs/nhis.htm

National Health and Nutrition Examination Surveys
http://www.cdc.gov/nchs/nhanes.htm

Behavioral Risk Factor Surveillance System
http://www.cdc.gov/brfss/

National Hospital Discharge Survey
http://www.cdc.gov/nchs/nhds/

Australian Study of Health and Relationships
http://www.ashr.edu.au/

Other survey not listed

Early Breast Cancer Trialists' Collaborative Group
http://ipdmamg.cochrane.org/early-breast-cancer-trialists-collaborative-group

Benefit from early detection is realized, in part or in whole, by reduction of stage at time of detection and treatment.

Benefit from early detection is conveyed implicitly based on reduced tumor size at the time of detection and treatment independent from that which may already be accounted for in stage shift benefit.

Benefits accrued from screen detection may be affected by one or more temporal trends including calendar year, age and/or birth cohort.

Benefit is attributed based on a cure/no-cure rate that is advantageous for screen-detected cancers

Lead-time is included as a survival benefit

Benefit varies depending on method of screening, separate from screening performance characteristics. This factor can be used to capture benefits from within-stage shifts.

Benefit is accrued due to prevention of cancer as a result of screening.

Treatment benefit is allocated based on estrogen receptor (ER) status

Treatment benefit may be affected by one or more temporal trends including calendar year, age and/or birth cohort.

Benefit varies depending on treatment method, which may in turn depend on dissemination of treatment modality.

Benefit is accrued due to treatment of precancer lesions

Benefit is accrued due to prevention of cancer as a result of vaccination

Cancer incidence

Typically non-cancerous conditions that affect screening. For example, gastroesophageal reflux disease and Barrett's esophagus.

Model inputs concerned with screening attendance and the uptake of disseminated screening opportunity. Models that simulate this aspect of screening often employ economic aspects of promotion and human behavior.

Inputs concerned with the availability of screening and possibly specific screening modality.

Inputs concerned with the performance of screening at the individual test level.

Inputs concerned with the effects of screen detection on an individual's treatment and survival.

Inputs to inform modification of regular screen schedules (test interval, thresholds etc) based on test results (eg marker levels).

Certain pre-cancerous findings during screening may trigger non-routine surveillance for which attendance parameters may be an input to the model. For example, post-adenoma follow-up for Colorectal cancer, or follow-up for Esophageal Adenocarcinoma (EAC) after diagnosis of Barrett's Esophagus (BE).

e.g. Diagnosis of CIN 1, CIN2/3

Inputs associated with attendence for CIN diagnosis and update of treatment

Inputs associated with test performance for CIN diagnosis

Model inputs concerned with the availability of specific treatments. For example, treatment availability may depend on calendar year and/or geographic region.

Model inputs concerned with the capacity for beneficial change in laboratory studies or clinical trials.

Model inputs concerned with the capacity for beneficial change in clinical practice.

Model inputs associated with the treatment of precancer lesions

Model inputs associated with the attendance of precancer treatment appointments

Model inputs associated with the efficacy of treatment of precancer lesions

The actual percentage of patients still alive at some specified time after diagnosis of cancer. It considers deaths from all causes, cancer or otherwise.

An estimate of the percentage of patients who would be expected to survive the effects of their cancer.

Inputs concerned with the probability of death at a certain age from causes other than the disease under investigation.

Inputs concerned with the probability of cause of death being due to the disease under investigation vs other general causes.

Inputs concerned with distribution of certain attributes of tumors. For example, growth rates, diffusion, and treatability.

Inputs concerned with factors that may influence incidence or treatment of disease. For example, smoking history, obesity, diet.

Inputs concerned with risk factors the model considers specifically as opposed to implicitly in other model inputs.

Inputs concerned with obesity possibly measured by any standard index (e.g. Body Mass Index)

Inputs that drive age-based risk in addition to typical age-specific parameters such as incidence and survival

Inputs relating to the medical history of the individual's family (not the individual themselves)

Inputs relating only to the individual (not the individual's family)

The number of times an individual has given birth

Oral contraceptives use

Other inputs concerning factors not covered above. For example: Hereditary nonpolyposis colorectal cancer (HNPCC), Irritable Bowel Syndrome (IBS) and workplace exposure to carcinogens.

Demographic inputs concerned with the simulated population.

Inputs concerned with the underlying natural history of the disease under investigation.

Inputs concerned with tracking costs of screening, vaccination, treatment. Costs may be relative to society, provider, or patient, and may be nominal or adjusted.

Quadrivalent (4v) vaccine

Bivalent (2v) vaccine

Nonavalent (9v) vaccine

Level and duration of vaccination

Outputs concerned with reporting costs of alternative interventions.

Outputs involving measurement of tumor size over time or at specific time points such as clinical or screen diagnosis

Outputs providing data on simulated cells. For example: cell kinetics, Temporal trends on cell kinetics, spatial growth of High grade dysplasia (HGD) and Esophageal Adenocarcinoma (EAC) clones

Outputs concerning in incidence of disease

Presence of Human Papillomavirus (HPV) Infection

Other conditions other than precancer and cancer (such as HPV Infection)

Outputs regarding the effect of treatments, often conditioned on age, period, and state of disease progression at the time of treatment.

Outputs regarding the effect of the treatment of precancer lesions

Outputs regarding the effect of screening

Outputs tracking the test performance as generated by the model. Can be conditioned on follow up time and censoring, or represent actual (often unobservable for negative tests) performance metrics.

Outpus concerned with distribution of certain attributes of tumors. For example, growth rates, diffusion, and treatability.

Outputs pertaining to the patterns of smoking in the simulated population, possibly in the context of simulated interventions.

Results drawn from simulated individual's disease including, onset, transitions to malignancy, stage transitions, metastasis, growth rates etc.

Outcomes are the population effects deemed interesting in the context of the interventions or exposures being modeled. Typical outcomes are survival, morbidity, mortality and other such results one would expect to be affected by interventions and trends in risk.

Measurements of time between diagnosis and death.

Measurements of difference of life years

Quality adjusted years of life.

Cause of death is accounted for, at least in terms of target disease vs other causes.

Mortality from any cause is provided

Temporal trends of disease

Effect from vaccination on population, including proportion of cervical disease attributable to vaccination HPV types, vaccination coverage, and duration of vaccine-induced immunity

Allows for an outcome in which the screening event indicates a positive disease state in the absence of the disease.

Allows for an outcome in which the screening event indicates a positive disease state in the presence of the disease.

Note that negative disease state may be expressed in terms of a defined follow-up time (observed) and/or be based on simulation data typically unobserved in real world studies. See specific model notes for details.

Note that negative disease state may be expressed in terms of a defined follow-up time (observed) and/or be based on simulation data typically unobserved in real world studies. See specific model notes for details.

Provides measures of the event (which typically cannot be observed) where screening detects cancer that would have otherwise gone undetected with lifetime follow up. That is, would not have surfaced in the person's lifetime.

The screening history of individuals and/or sub-populations

The treatment history of individuals and/or sub-populations

Pregnancy outcomes from Cervical intraepithelial neoplasia (CIN) treatment