# C.O.S. Résumé de rapport

Project ID:
QLK1-CT-2000-00466

Financé au titre de:
FP5-LIFE QUALITY

Pays:
Italy

## New software for estimating the probability of mutation of BRCA genes on the basis of the family history of breast and ovarian cancer

Several methods have been used to estimate the probability of carrying a BRCA mutation, based on the number of breast and ovarian cancer among 1st and 2nd degree relatives, the age of occurrence, the presence of bilateral breast cancers and, when relevant, ethnicity. Most methods, however, do not take into account the size of the family (in particular the number and ages of healthy relatives), the lifespan of each family member that has actually been observed, the structure of the pedigree, or the change of cancer risk over subsequent generations. G. Parmigiani developed a method that properly incorporates the whole information available from each family member.

The probability that a given family member is a carrier is obtained comparing the probabilities of observing his/her family history assuming either the incidence of the general population or the gene penetrance. The computation is based on Bayes�s theorem, using the mutation prevalence in the population as the prior distribution and the family history as the evidence. The method takes fully into account mendelian transmission pathways. The weight of the maternal grandmother or of a aunt with breast cancer, for instance, will be higher if the mother had breast cancer too (or if she died when still too young to have breast cancer) than if the mother reached adult or old age without developing the disease.

The results of the Parmigiani method are quite sensitive to varying assumptions about the age specific incidence function of breast and ovarian cancer in mutation carriers and in the general population. In the C.O.S. project the method has been improved using generation specific incidence functions estimated from age-period-cohort models, i.e. a different function for each family member. As a consequence it provides more accurate estimates of the breast and ovarian cancer cases that one would expect in the family, either if the family harbour a mutation or not. The lifetime cumulative breast cancer risk of a European woman born in the fifties, in fact, is about three times higher than the risk of her grandmother born at the beginning of the 20th century. A similar cohort effect has also been shown for mutation carriers.

The specific software developed for the C.O.S. project provides an estimate of the probability that a breast cancer patient carry an high penetrance BRCA 1 or 2 mutation on the basis of her family history.

It requires the following assumptions:

- The prevalence of mutation in the population (more specifically the software incorporates the allele frequency),

- Birth cohort specific incidence curves of breast and ovarian cancer in the general population, and

- Birth cohort specific incidence curves in carriers of BRCA mutation, i.e. the penetrance. It requires entering dates of birth, death and diagnosis of breast and ovarian cancer for all members of the family. Country and cohort specific incidence estimates have been computed on the basis of mathematical modelling of mortality and survival trends.

The validity of the estimates has been checked plotting estimated incidence trends against the trends observed by cancer registries operating since several decades. Country specific estimates have been used for Estonia, France, Germany, Italy, Scotland and Slovenia.

The same C.O.S. software has been used to estimate the penetance function of BRCA1 and BRCA2 mutations. Instead of estimating the probability of mutation given the family history, in fact, it can estimate the best penetrance curves to explain the observed incidence in a series of families with proven mutation. Based on a numerical maximisation method, the model identifies the curves that maximise the probability of observing the clinical history of the families.

Several authors have estimated BRCA allelic frequencies in western populations. We have used the following estimates:0.00120 for BRCA1 and 0.00044 for BRCA2 for all countries except Israel, for which we have used the allelic frequencies estimated for the Ashkenazi population, the most prevalent ethnic group in Israel: 0.0100 for BRCA1 and 0.0036 for BRCA2.

The probability that a given family member is a carrier is obtained comparing the probabilities of observing his/her family history assuming either the incidence of the general population or the gene penetrance. The computation is based on Bayes�s theorem, using the mutation prevalence in the population as the prior distribution and the family history as the evidence. The method takes fully into account mendelian transmission pathways. The weight of the maternal grandmother or of a aunt with breast cancer, for instance, will be higher if the mother had breast cancer too (or if she died when still too young to have breast cancer) than if the mother reached adult or old age without developing the disease.

The results of the Parmigiani method are quite sensitive to varying assumptions about the age specific incidence function of breast and ovarian cancer in mutation carriers and in the general population. In the C.O.S. project the method has been improved using generation specific incidence functions estimated from age-period-cohort models, i.e. a different function for each family member. As a consequence it provides more accurate estimates of the breast and ovarian cancer cases that one would expect in the family, either if the family harbour a mutation or not. The lifetime cumulative breast cancer risk of a European woman born in the fifties, in fact, is about three times higher than the risk of her grandmother born at the beginning of the 20th century. A similar cohort effect has also been shown for mutation carriers.

The specific software developed for the C.O.S. project provides an estimate of the probability that a breast cancer patient carry an high penetrance BRCA 1 or 2 mutation on the basis of her family history.

It requires the following assumptions:

- The prevalence of mutation in the population (more specifically the software incorporates the allele frequency),

- Birth cohort specific incidence curves of breast and ovarian cancer in the general population, and

- Birth cohort specific incidence curves in carriers of BRCA mutation, i.e. the penetrance. It requires entering dates of birth, death and diagnosis of breast and ovarian cancer for all members of the family. Country and cohort specific incidence estimates have been computed on the basis of mathematical modelling of mortality and survival trends.

The validity of the estimates has been checked plotting estimated incidence trends against the trends observed by cancer registries operating since several decades. Country specific estimates have been used for Estonia, France, Germany, Italy, Scotland and Slovenia.

The same C.O.S. software has been used to estimate the penetance function of BRCA1 and BRCA2 mutations. Instead of estimating the probability of mutation given the family history, in fact, it can estimate the best penetrance curves to explain the observed incidence in a series of families with proven mutation. Based on a numerical maximisation method, the model identifies the curves that maximise the probability of observing the clinical history of the families.

Several authors have estimated BRCA allelic frequencies in western populations. We have used the following estimates:0.00120 for BRCA1 and 0.00044 for BRCA2 for all countries except Israel, for which we have used the allelic frequencies estimated for the Ashkenazi population, the most prevalent ethnic group in Israel: 0.0100 for BRCA1 and 0.0036 for BRCA2.