Table 1. Summary of features of existing R packages for risk prediction.

✓ denotes the presence of the feature and × denotes absence of the feature.

Packages	Model building				Model validation
	Calibration to population incidence	Detailed family history	Special option for SNP markers	Imputation of missing risk-factors	Full cohort	Two-phase study	Imputation of missing risk-factors
riskRegressiona	×	×	×	×	✓	×	×
predictABELa	×	×	×	×	✓	×	×
BCRAb	✓c	×	×	×	×	×	×
BayesMendelb	×	✓	×	✓	×	×	×f
rmap	×	×	×	×	✓	✓e	×
iCARE b	✓c	×	✓d	✓	✓	✓e	✓f

^a These packages include some functions for model building (see Section), but those approaches do not demonstrate the key features shown in the above table.

^b Capability to use information from multiple data sources, e.g., BCRA and iCARE can use relative risk parameters from cohort or case-control studies and disease incidence and mortality rates from population registries.

^c BCRA estimates baseline hazard and calibrates the model to the underlying population incidence rates using distribution of risk-factors from cases in a specific study that may not be representative of the general population. This step is implemented in iCARE using a reference dataset that provides information on the distribution of risk-factors in the general population.

^d iCARE includes the special option in which independent SNP markers can be included using published estimates of odds ratios and allele frequencies.

^e Inverse probability weighted estimators of model validation statistics are implemented, accounting for bias due to non-random sampling using sampling weights.

^f BayesMendel incorporates imputation methods for certain risk-factors (e.g., age), but they do not implement any method of validating risk prediction models. iCARE implements an inbuilt imputation approach to deal with missing risk-factors using a reference risk-factor dataset representative of the underlying population. The standardized model validation methods implemented in iCARE can take advantage of this inbuilt feature to impute missing risk-factors in the validation study.