Estimating penetrance from multiple case families with predisposing mutations: extension of the ‘genotype-restricted likelihood’ (GRL) method

Abstract

Some diseases are due to germline mutations in predisposing genes, such as cancer family syndromes. Precise estimation of the age-specific cumulative risk (penetrance) for mutation carriers is essential for defining prevention strategies. The genotyperestricted likelihood (GRL) method is aimed at estimating penetrance from multiple case families with such a mutation. In this paper, we proposed an extension of the GRL to account for multiple trait disease and to allow for a parent-of-origin effect. Using simulations of pedigrees, we studied the properties of this method and the effect of departures from underlying hypotheses, misspecification of disease incidence in the general population or misspecification of the index case, and penetrance heterogeneity. In contrast with the previous version of the GRL, accounting for multiple trait disease allowed unbiased estimation of penetrance. We also showed that accounting for a parent-of-origin effect allowed a powerful test for detecting this effect. We found that the GRL method was robust to misspecification of disease incidence in the population, but that misspecification of the index case induced a bias in some situations for which we proposed efficient corrections. When ignoring heterogeneity, the penetrance estimate was biased toward that of the highest risk individuals. A homogeneity test performed by stratifying the families according to the number of affected members was shown to have low power and seems useless for detecting such heterogeneity. These extensions are essential to better estimate the risk of diseases and to provide valid recommendations for the management of patients.

Cancer Risks Associated With Germline Mutations in <emph type="ital">MLH1</emph>, <emph type="ital">MSH2</emph>, and <emph type="ital">MSH6</emph> Genes in Lynch Syndrome

“…This correction implies that families are informative only if at least 1 family member other than the proband has been tested for the mutation. 10 We applied the extended version of the GRL method taking into account the possibility of multiple trait phenotypes, as proposed by Bonaïti et al 15 using the GENERISK software.…”

“…13 (9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19) 14 (8)(9)(10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27) 20 ( outweigh the risks of inducing a premature menopause and definitive infertility. For MSH6 mutation carriers, our data suggest that the role of prophylactic gynecological surgery is more debatable because the risk of ovarian cancer by age 70 years was close to the lifetime risk estimated for the general population.…”

“…Furthermore, the extended version of the ascertainment-adjusted GRL method simultaneously takes into account various traits for the phenotype, ie, the different tumors of the Lynch syndrome spectrum, and eliminates the bias induced by analyzing each trait separately that should lead to an underestimate of the risks. 15 However, the trade off when using a retrospective likelihood is the rather wide CIs. This is especially apparent in the older age groups because many family members had not reached old age and when they did were seldom geno-typed.…”

New classification of endometrial cancers: the development and potential applications of genomic-based classification in research and clinical care

“…This may impact recommendations regarding the timing of screening or intervention e.g., lower lifetime risk and later average age of penetration for individuals with aberrant MSH6 [90–93] might enable delay of recommended risk reducing surgery as compared to other Lynch mutations.…”

A new scoring system in cancer genetics: application to criteria forBRCA1andBRCA2mutation screening

“…In cases where the risks at greater ages were not provided by these sources, we inferred the missing data from available data by fitting a mixture model with a subgroup of non-susceptible carrier individuals and a subgroup of susceptible carriers for which the risk is modelled by a Weibull function. This model is very similar to the ‘cured model’ of de Angelis,21 also used for penetrance estimation in the genotype-restricted likelihood method 22. Cumulative risks in the general population used were those published for France based on data of cancer registries 23…”