Prediction Models for Hereditary Colon Cancer Risk

At Simmons Comprehensive Cancer Center, we use a variety of assessment tools designed to predict an individual’s likelihood of Lynch syndrome (LS). Statistical models have not been developed to predict risk for other hereditary colon cancer syndromes at this time. Not all tools can be appropriately applied to all patients. Each tool is most effective when the patient’s characteristics and family history are similar to those of the study population on which the tool was based. All have limitations, and the risk estimates derived from the tools may differ for an individual patient.

MMRPRO

This tool is available within the CancerGene risk assessment software that also harbors the BRCAPRO model. It utilizes personal/family history of colon and/or endometrial cancer within first- and second-degree relatives, ancestry, location of the colon cancer, age of onset, age of unaffected family members, results of MSI/IHC testing of colon tumor (MLH1, MSH2, MSH6 only) and LS-related mutation carrier status to determine LS-related mutation risk in MLH1, MSH2, and MSH6 only. MMPRO has a few important limitations to keep in mind. Mutation risk calculation does not include LS-related cancers outside of colon and endometrium, MSI in sporadic setting (i.e., BRAF and MLH1 hypermethylation testing of tumor), nor colorectal adenomas/polyps or prophylactic hysterectomy (TAH). Additionally, risk of a PMS2 gene mutation is not calculated with this model.

Myriad Prevalence Tables

Another tool utilized commonly that is also in CancerGene software is the Myriad Prevalence Tables. It is not technically a model of LS mutation risk, but a compilation of data placed in a categorical table from Myriad Genetics’ sequencing of MLH1/MSH2 genes. The information to be gathered from this data is only for likelihood of MLH1 and MSH2 mutation (MSH6 and PMS2 not included). Data is based on the proband with a history of cancer and includes first- and second-degree relatives with colon/endometrial cancer, as well as extra-colonic cancers associated with LS. An important limitation is possible sample bias that may not be representative of general population being tested. Also, the calculations were based on what was recorded on test requisition form, which could possibly be erroneous or incomplete.

PREMM1,2,6 Model

This model predicts likelihood for MLH1/MSH2/MSH6 mutations in an affected proband. The risk calculation is based on the proband’s personal history of a LS-related cancer (colon, endometrial or extra-colonic), age at diagnosis, and gender. Family history considerations in this model include first- or second-degree relatives with LS-related cancers and ages at diagnosis. An important limitation is the exclusion of MSH6 rearrangement and PMS2 mutation data in the study cohort from which this data is based. It also does not take family size or unaffected individuals into account and does not incorporate MSI/IHC data or history of adenomas.

MMR Predict Model

This model utilizes age at diagnosis of colon cancer, multiple primary colon cancers (synchronous/metachronous), family history of colon/endometrial cancer in first-degree relatives only and presence or absence of endometrial cancer. Of note, this model was validated in a Scottish cohort of affected individuals diagnosed before age 55, so it is unclear if this model performs as well in older populations or other ethnicities.