Background: Family history is an important risk factor for CRC, but there is still confusion about the appropriate guidelines councilors should recommend to people depending on the specifics of their family history. Most previous studies that have estimated familial relative risk (FRR) of CRC have based this on first-degree relatives (FDRs), whereas information on second- or third-degree relatives (SDRs or TDRs) has been of poor quality or non-existent. The most notable exception to this is a publication by Taylor et al that utilized the Utah Population Database (UPDB), a population-based resource with a computerized genealogy linked to statewide cancer registry records.[1] They reported FRRs of CRC for probands by selected combinations of affected relatives, extending to third-degree. The aim of this study was to extend this work by developing a simple and clinically-useful model of familial CRC risk.

Methods: We restricted the analysis to people aged 30 years or older born between 1930 and 1985 (probands) from the UPDB. Data were collected on the proband's age, sex and history of CRC for FDRs, SDRs and TDRs. Unconditional multiple linear logistic regression was used to model the familial CRC risk for probands as a function of their family history measures. Various combinations of CRC status of relatives were considered, including categorizations by ages at diagnoses (<50, 50-59, 60-69, 70+), type of FDR (mother/father/brother/sister/child) and side of family for SDRs and TDRs (maternal or paternal). The best fitting model was determined by Akaike's Information Criterion.

Results: A total of 591,535 probands were extracted of whom 2,115 probands were identified as having a primary diagnosis of CRC.

The best-fitting model for CRC was FRR = exp(SUM/5)*0.8, where SUM equals:

4 points for each parent diagnosed with CRC

6 points for each sibling diagnosed with CRC

12 points for each child diagnosed with CRC

2 points for each SDR diagnosed with CRC

1 point for each TDR diagnosed with CRC.

Therefore, a doubling of risk would be 5 points, a tripling of risk would be 7 points, while a 5-fold increased risk would be 10 points. The model had good internal consistency. Additional information on ages at diagnoses of affected FDRs, SDRs or TDRs or whether diagnoses were confined to a particular side of the family did not improve the model fit.

Conclusions: This simple algorithm shows that knowing the total number of affected parents, siblings, children, SDRs and TDRs, irrespective of the age at diagnosis, is sufficient for accurate estimation of FRR. This model could be useful in the clinical and genetic counseling setting.

1. Taylor DP, et al. Population-based family history-specific risks for colorectal cancer: a constellation approach. Gastroenterology. 2010 Mar;138(3):877-85.

This abstract is also being presented as PosterB06.

Citation Format: Robert J. MacInnis, Mark A. Jenkins, John L. Hopper, Lisa A. Cannon-Albright. Utah familial colorectal cancer risk model. [abstract]. In: Proceedings of the AACR Special Conference: Improving Cancer Risk Prediction for Prevention and Early Detection; Nov 16-19, 2016; Orlando, FL. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2017;26(5 Suppl):Abstract nr PR11.