Dataset Information

Comparison of approaches for incorporating new information into existing risk prediction models.

ABSTRACT: We compare the calibration and variability of risk prediction models that were estimated using various approaches for combining information on new predictors, termed 'markers', with parameter information available for other variables from an earlier model, which was estimated from a large data source. We assess the performance of risk prediction models updated based on likelihood ratio (LR) approaches that incorporate dependence between new and old risk factors as well as approaches that assume independence ('naive Bayes' methods). We study the impact of estimating the LR by (i) fitting a single model to cases and non-cases when the distribution of the new markers is in the exponential family or (ii) fitting separate models to cases and non-cases. We also evaluate a new constrained maximum likelihood method. We study updating the risk prediction model when the new data arise from a cohort and extend available methods to accommodate updating when the new data source is a case-control study. To create realistic correlations between predictors, we also based simulations on real data on response to antiviral therapy for hepatitis C. From these studies, we recommend the LR method fit using a single model or constrained maximum likelihood. Copyright © 2016 John Wiley & Sons, Ltd.

SUBMITTER: Grill S

PROVIDER: S-EPMC8182952 | biostudies-literature | 2017 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Comparison of approaches for incorporating new information into existing risk prediction models.

Grill Sonja S Ankerst Donna P DP Gail Mitchell H MH Chatterjee Nilanjan N Pfeiffer Ruth M RM

Statistics in medicine 20161211 7

We compare the calibration and variability of risk prediction models that were estimated using various approaches for combining information on new predictors, termed 'markers', with parameter information available for other variables from an earlier model, which was estimated from a large data source. We assess the performance of risk prediction models updated based on likelihood ratio (LR) approaches that incorporate dependence between new and old risk factors as well as approaches that assume ...[more]

PMID: 27943382

Similar Datasets

Project description:IntroductionMutations in BRCA1 and BRCA2 confer high risks of breast cancer and ovarian cancer. The risk prediction algorithm BOADICEA (Breast and Ovarian Analysis of Disease Incidence and Carrier Estimation Algorithm) may be used to compute the probabilities of carrying mutations in BRCA1 and BRCA2 and help to target mutation screening. Tumours from BRCA1 and BRCA2 mutation carriers display distinctive pathological features that could be used to better discriminate between BRCA1 mutation carriers, BRCA2 mutation carriers and noncarriers. In particular, oestrogen receptor (ER)-negative status, triple-negative (TN) status, and expression of basal markers are predictive of BRCA1 mutation carrier status.MethodsWe extended BOADICEA by treating breast cancer subtypes as distinct disease end points. Age-specific expression of phenotypic markers in a series of tumours from 182 BRCA1 mutation carriers, 62 BRCA2 mutation carriers and 109 controls from the Breast Cancer Linkage Consortium, and over 300,000 tumours from the general population obtained from the Surveillance Epidemiology, and End Results database, were used to calculate age-specific and genotype-specific incidences of each disease end point. The probability that an individual carries a BRCA1 or BRCA2 mutation given their family history and tumour marker status of family members was computed in sample pedigrees.ResultsThe cumulative risk of ER-negative breast cancer by age 70 for BRCA1 mutation carriers was estimated to be 55% and the risk of ER-positive disease was 18%. The corresponding risks for BRCA2 mutation carriers were 21% and 44% for ER-negative and ER-positive disease, respectively. The predicted BRCA1 carrier probabilities among ER-positive breast cancer cases were less than 1% at all ages. For women diagnosed with breast cancer below age 50 years, these probabilities rose to more than 5% in ER-negative breast cancer, 7% in TN disease and 24% in TN breast cancer expressing both CK5/6 and CK14 cytokeratins. Large differences in mutation probabilities were observed by combining ER status and other informative markers with family history.ConclusionsThis approach combines both full pedigree and tumour subtype data to predict BRCA1/2 carrier probabilities. Prediction of BRCA1/2 carrier status, and hence selection of women for mutation screening, may be substantially improved by combining tumour pathology with family history of cancer.

Dataset Information

Comparison of approaches for incorporating new information into existing risk prediction models.

Publications

Comparison of approaches for incorporating new information into existing risk prediction models.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets