Population-Based Study of Changing Breast Cancer Risk in Icelandic BRCA2 Mutation Carriers, 1920–2000

Abstract

Background: Mutations in the BRCA genes increase the risk of breast cancer. Valid estimates of the magnitude of the lifetime risk of breast cancer in BRCA gene mutation carriers are needed for genetic counseling. Recent results suggest that penetrance has increased in recent birth cohorts. We examined the cumulative breast cancer incidence and mortality before age 70 over a diagnosis period of 80 years in Icelandic women who carried the BRCA2 founder mutation 999del5. Methods: Information on all breast cancers diagnosed in Iceland since 1911 was obtained from the Icelandic Cancer Registry. Mutation status was determined by molecular analysis of tissue samples for 847 breast cancer probands who were diagnosed from 1921 through 1985 and selected without knowledge of family history of breast cancer. We estimated the cumulative incidence and mortality from breast cancer before age 70 years in BRCA2 mutation carriers from the observed risks in first-degree relatives who were classified according to mutation status of probands and followed-up through 2002. Poisson modeling of these risks was also carried out. All statistical tests were two-sided. Results: Of the 847 probands, 88 carried the BRCA2 999del5 mutation and 759 did not. According to Poisson modeling, the cumulative incidence of breast cancer before age 70 years in mutation carriers increased from 18.6% (95% CI = 11.0% to 29.5%) in calendar year 1920 to 71.9% (95% CI = 45.9% to 100%) in 2002 ( P <.001); in relatives of probands who did not carry the BRCA2 mutation and in the general Icelandic population incidence increased over the same period from 2.6% to 10.7% and from 1.8% to 7.5%, respectively (all increases of approximately fourfold). During the same period, the cumulative risk of death from breast cancer before age 70 years for BRCA2 mutation carriers increased from 12.1% (95% CI = 5.3% to 23.9%) to 26.9% (95% CI = 10.9% to 55.5%) ( P = .08). However, because the probands were breast cancer patients and not a random sample from the population, some bias in the estimation of time trends in penetrance cannot be ruled out. Conclusions: The results indicate that the penetrance of the Icelandic BRCA2 founder mutation increased nearly fourfold in 80 years, whereas the risk of death from breast cancer before age 70 years increased only approximately twofold. Changes in penetrance with time should be considered when penetrance is estimated.

The online version of this article has been published under an Open Access model. Users are entitled to use, reproduce, disseminate, or display the Open Access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact: journals.permissions@oxfordjournals.org .

Breast cancer has the highest incidence and mortality of all cancers among women worldwide ( 1 , 2 ) . In most countries, the incidence of breast cancer has increased steadily for several decades ( 3 ) . In Iceland, the lifetime risk of breast cancer is currently 10%, a level similar to that in other Nordic countries ( 4 ) , and the incidence per 100 000 person-years (age-standardized according to the world standard population) has increased since 1921 ( 5 ) . It is not known to what extent the increase in breast cancer incidence is due to changes in the prevalence of breast cancer risk factors and to what degree it is due to a lower threshold for diagnosis of breast cancer related to increased awareness of breast cancer and the emergence of breast cancer screening programs.

Inherited mutations in the tumor suppressor genes BRCA1 and BRCA2 markedly increase the risks of breast and ovarian cancers ( 6 – 8 ) . In the Icelandic population, only one mutation has been detected in each of the BRCA genes. In the BRCA1 gene, there is a rare splice site mutation in exon 17 that leads to an amino acid change, D1692→N ( 9 ) . The BRCA2 mutation, 999del5, is much more frequent. It is a 5–base pair deletion in exon 9, starting at nucleotide 999, that leads to a premature stop codon at nucleotide 1047 and to a truncated version of the protein ( 8 , 10 ) . This BRCA2 mutation is carried by 6%–8% of all Icelandic women with breast cancer; among Icelandic women diagnosed with breast cancer before age 40, 24% carry this mutation ( 10 , 11 ) . The estimated prevalence of the 999del5 mutation in the general Icelandic population is 0.6% ( 11 , 12 ) . This founder mutation accounts for nearly 40% of familial breast cancers diagnosed in Iceland ( 13 ) . Thus, the high prevalence of this single mutation facilitates analyses of breast cancer risk in mutation carriers in the Icelandic population.

Valid estimates of the magnitude of the lifetime risk of breast cancer in carriers of BRCA gene mutations are of great importance because of their use as a basis for genetic counseling and because they can contribute to an increased understanding of the biologic consequences of these mutations. Penetrance of BRCA mutations is indicated by the lifetime risk, but reported estimates of this measure have varied greatly, in part because of the different definitions of the probands used as the basis for the estimation. Higher lifetime risks of breast cancer have been observed when probands are selected from among families at high risk of breast cancer than from among the general population ( 14 – 18 ) . The time factor, as reflected by year of birth or year of diagnosis, may be yet another source of variation in penetrance estimates, as suggested by a recent study of Ashkenazi Jewish patients from New York, which found that the cumulative breast cancer risk for mutation carriers born after 1940 was statistically significantly higher than that for mutation carriers born before 1940 ( 19 ) .

We used data from the Icelandic Cancer Registry, which contains information on all breast cancer cases diagnosed in Iceland since 1911, to examine population-based changes in breast cancer risk over a diagnosis period of 80 years. We specifically investigated whether the penetrance of the founder mutation BRCA2 999del5 had increased in the Icelandic population from 1921 through 2002 and estimated breast cancer risk in mutation carriers based on the cumulative incidence of breast cancer diagnosed in first-degree relatives of carriers and noncarriers ( 17 , 20 ) . We also estimated the risk of death from breast cancer before age 70 years in mutation carriers.

S UBJECTS AND M ETHODS

Study Population

We used the Icelandic Cancer Registry, a population-based registry that has registered all cancers in Iceland since 1955, as a source of breast cancer patients for this analysis. This registry also has information on all patients diagnosed with breast cancer in Iceland from 1911 to 1954 ( 21 , 22 ) . The Breast Cancer Family Collection of the Icelandic Cancer Registry consists of 994 population-based probands diagnosed with breast cancer from 1921 through 1985 and their first-, second-, and third-degree relatives, for a total of 28 806 female relatives and 29 603 male relatives. Of the 994 probands, 946 were diagnosed with invasive breast cancer and 48 were diagnosed with noninvasive breast cancer. The 994 probands were recruited to the Breast Cancer Family Collection on the basis of the year of their diagnosis and their age at diagnosis and without knowledge of their family history ( 5 ) : All breast cancer patients included in the Icelandic Cancer Registry who were born between 1834 and 1860 (55 probands); five randomly selected breast cancer patients born in 1865 and five randomly selected breast cancer patients born in 1875 (10 probands); every eighth breast cancer patient of those born between 1899 and 1916, ordered by date of diagnosis (46 probands); every eighth breast cancer patient diagnosed in 1954 and in 1972 (85 probands); and all breast cancer patients born after 1915 and diagnosed before 1986 (798 probands). Our study group consists of all probands with invasive breast cancer from this family collection for whom tissue samples were available and information on mutation status could be obtained and their female first-degree relatives—a total of 847 probands and 2836 relatives.

Information on breast cancer in first-degree relatives of probands was obtained by record linkage between the files of the 2836 first-degree relatives and the population-based Icelandic Cancer Registry, using precoded personal identifiers for both databases. All patient information and tissue samples, stripped of personal identifiers, were precoded in accordance with the requirements of the Data Protection Authorities and the National Bioethics Committee of Iceland. Institutional approval was obtained for this study.

Genetic Analysis

Tissue samples for the 847 probands diagnosed with invasive breast cancer were obtained from the University Hospital Department of Pathology and the Biological Specimen Bank of the Icelandic Cancer Society. The tissue samples were used to extract DNA as previously described ( 23 ) , and all DNA samples were then screened for the BRCA2 999del5 mutation by exon 9 amplification using a previously described reverse primer (5′-AAAGTCTGAAGAAAAATGATAGATTTA-3′) ( 8 ) and a forward primer (5′-AAAACCTGTAGTTCAACTAAACAG-3′), resulting in a polymerase chain reaction (PCR) product of 114 bp ( 24 ) . DNA was PCR amplified by using [α- ³² P]CTP labeling for mutation screening and haplotype analysis, and the products were run on 6% denaturing polyacrylamide gels and made visible by autoradiography ( 10 , 24 ) . Suspected carriers of the BRCA2 999del5 mutation were identified as having an additional band 5 bp smaller than the wild type PCR fragment.

Statistical Analysis

The cumulative incidence of breast cancer among BRCA2 999del5 mutation carriers was estimated as R2 × 2 − R1, where R2 is the cumulative incidence of breast cancer among first-degree relatives of carrier probands and R1 is the cumulative incidence of breast cancer among the first-degree relatives of noncarrier probands. This formula was derived by Wacholder et al. ( 20 ) for studies in which probands are a random sample of the population. However, because the probands in our study were breast cancer patients and not a random sample from the population, the estimated time trends in penetrance may be biased. The first-degree relatives were monitored with respect to their incidence of and death from breast cancer through December 31, 2002. Whenever two or more sisters from the same nuclear family were among the probands, the mother and the other sister(s) were included in the calculations only as relatives of the sister who was the first to be diagnosed with breast cancer. For the remaining sister proband(s), only daughters were included as relatives. We applied a similar rule to mother–daughter pairs (i.e., whenever a mother and daughter from the same nuclear family were among the probands, the daughter [or mother] and other sister[s] were included in the calculations only as relatives of the mother [or daughter], whoever was diagnosed first).

Breast cancer cases diagnosed from 1921 through 2002 among first-degree relatives of probands were stratified according to age at diagnosis (in 5-year intervals), calendar year of diagnosis (in 5-year intervals), and whether the proband carried the BRCA2 999del5 mutation. The risk years were stratified by the same criteria. We counted the observed number of cases in each stratum and calculated the person-years at risk by using the MANYRS computer program ( 25 ) . Breast cancer incidence in each stratum was calculated as the number of cases divided by the number of years at risk, and cumulative incidence was evaluated at age 70 years. When trying to calculate risk in first-degree relatives according to birth cohort, we had to exclude women born after 1950 because in this group no women older than 52 years had yet been diagnosed (last calendar year of follow-up is 2002) so it was not possible to calculate risk up to age 70 years for this group. The incidence data were also used to build Poisson regression models to examine associations among age and calendar time at diagnosis and breast cancer incidence. A previous study ( 5 ) of breast cancer incidence in Iceland has shown that the best fit of the models is attained when modeling separately for women younger than 45 years and for women 45 years or older. Details of the models for first-degree relatives of probands with and without the BRCA2 mutation and for the whole population are shown in the Appendix . Information on deaths from breast cancer was available for first-degree relatives for the entire study period (i.e., 1911–2002) but only from 1950 onward for the general population. Therefore, we built models for breast cancer mortality only for the first-degree relatives, using the same approach as that used to build the models for breast cancer incidence (see Appendix ), but because of limited data, especially for young women, a single model was used for all ages. We used logistic regression to examine associations among age and calendar year of diagnosis and the proportion of probands who were BRCA2 mutation carriers. The data were analyzed with the use of SPIDA software [version 6 ( 26 ) ] and the statistical significance level was set to P <.05; all statistical tests were two-sided.

The 95% confidence intervals (CIs) for estimates of cumulative incidence that were computed directly from the data were based on the assumption that the number of breast cancer cases in each stratum followed a Poisson distribution. Confidence intervals for values computed from the models were computed by a Monte Carlo simulation that generated random variables having the same covariance structure as the estimated coefficients from the Poisson regression and were based on the assumption that the logarithm of cumulative incidence was normally distributed. Goodness of fit of the models was tested by assuming the deviance to have a chi-square distribution with degrees of freedom equal to the number of strata with nonzero risk years minus the number of estimated parameters (three constants estimated by Poisson regression).

R ESULTS

Of the 847 probands whose BRCA2 mutation status could be determined, 88 had the BRCA2 999del5 mutation (BRCA2 mut) and 759 did not (BRCA2 wt). The proportion of BRCA2 mutation carriers among the probands (10.4%) did not change with calendar time ( P_trend = .76) but was strongly associated with age at diagnosis, decreasing statistically significantly from 15.4% (23/149) for probands aged 20–39 years at diagnosis to 4.4% (6/136) for probands aged 60 years or older at diagnosis (difference = 11.0% [95% CI = 4.3% to 17.8%; P =.002]).

Our calculations of cumulative incidence of breast cancer were based on 2567 first-degree female relatives (98 831 person-years) of mutation-negative probands and 269 first-degree female relatives (9323 person-years) of mutation-positive probands. Of the 165 first-degree female relatives who were diagnosed with breast cancer, 48 were related to the 88 probands who carried the mutation and 117 were related to the 759 probands who did not. The distributions of probands and first-degree relatives with breast cancer according to their year of diagnosis and BRCA2 mutation status are shown in Table 1 .

Table 2 shows the cumulative incidence of breast cancer before age 70 years for the Icelandic population and for first-degree relatives of the probands diagnosed during four consecutive periods and the estimated cumulative incidence for the mutation carriers during the same periods. Among the mutation carriers, estimated cumulative incidence increased from 17.9% (95% CI = 0% to 39.3%) during 1921–1944 to 65.5% (95% CI = 32.0% to 99.0%) during 1985–2002. The magnitude of this increase, almost fourfold, is similar to the magnitude of the increase in the cumulative incidence of breast cancer during the same periods among the general Icelandic population. For the overall study period, 1921–2002, the cumulative incidence of breast cancer before age 70 years was 4.8% (95% CI = 4.6% to 5.0%) for the general population and 6.5% (95% CI = 5.3% to 7.7%) for first-degree female relatives of probands with a wild-type BRCA2 gene, and the estimated value for mutation carriers was 46.3% (95% CI = 31.0% to 61.1%); the overall relative cumulative risk of breast cancer for mutation carriers compared with the general population was 9.6 (95% CI = 6.5 to 12.8). The relative cumulative risks for the four consecutive diagnosis periods in Table 2 were 7.8 (95% CI = 2.5 to 24.4), 6.8 (95% CI = 3.3 to 13.8), 11.1 (95% CI = 7.4 to 16.8), and 8.8 (95% CI = 5.9 to 13.4), respectively; there was no indication of a trend of changing relative risk during the period of observation.

Table 3 shows the results of Poisson modeling of the cumulative incidence of breast cancer before age 70 years by calendar year of diagnosis. The estimated incidence for BRCA2 mutation carriers increased from 18.6% (95% CI = 11.0% to 29.5%) for those diagnosed in 1920 to 71.9% (95% CI = 45.9% to 100%) for those diagnosed in 2000, a 3.9-fold increase (95% CI = 2.1-fold to 7.4-fold) over 80 years. In each age group modeled (i.e., <45 years and ≥45 years), the increase in the incidence of breast cancer was statistically significant ( P <.001). Goodness-of-fit tests of the models showed excellent fit for the relatives ( P ≥.95) but rather poor fit for the population ( P = .01). Details are shown in the Appendix .

We also analyzed the risk in first-degree relatives of probands according to birth cohorts instead of calendar year of diagnosis ( Table 4 ). The results indicated that the cumulative incidence of breast cancer before age 70 increased in successively later birth cohorts. However, these results should be interpreted with caution: The risk of breast cancer in the oldest cohorts is likely to be an underestimate because the first available risk period began in 1921, when the oldest women (born 1836) were already 85 years old. The analysis is also problematic for the youngest cohorts because many women have yet to attain the age of 70 years.

We next estimated the cumulative risk of death from breast cancer by calendar period at diagnosis. There were 42 breast cancer deaths among first-degree relatives of probands with a wild-type BRCA2 gene and 20 breast cancer deaths among first-degree relatives of probands with the BRCA2 mutation. When looking at the same four periods of diagnosis as in Table 2 , the number of deaths during individual periods was small, resulting in limited power for an analysis. Therefore, we instead estimated the cumulative risk of death from breast cancer for two longer periods of diagnosis, 1921–1964 and 1965–2002 ( Table 5 ). Among BRCA2 mutation carriers, the estimated cumulative risk of dying from breast cancer before the age of 70 years increased from 10.8% (95% CI = 0.6% to 21.0%) for women diagnosed during the period 1921–1964 to 23.8% (95% CI = 9.9% to 37.7%) for women diagnosed during the period 1965–2002. The results of Poisson modeling of the cumulative risk of death from breast cancer before age 70 years by calendar year of diagnosis are shown in Table 6 . Goodness-of-fit tests showed excellent fit ( P >.99). Among BRCA2 mutation carriers, the estimated risk of dying from breast cancer before age 70 years increased 2.2-fold (95% CI = 0.7-fold to 6.9-fold), from 12.1% in 1920 to 26.9% in 2000. That increase was considerably less than the estimated increase in the incidence of breast cancer during this period (i.e., 3.9-fold). However, the power for this analysis was low and the increase in the estimated risk of death from breast cancer was not statistically significant ( P = .08).

D ISCUSSION

Our results indicate that the risk of breast cancer before age 70 increased nearly fourfold from 1920 through 2000 in women who carried the BRCA2 999del5 mutation, in first-degree relatives of women with breast cancer who carried the wild-type allele, and in the general Icelandic population. The increase in each group was statistically significant. The risk estimate for the year 2000 was 72%. By contrast, during the same period, we observed only a twofold increase in the risk of death from breast cancer before age 70 years in BRCA2 mutation carriers. The risk estimate for the year 2000 was 27%.

Our results for BRCA2 mutation carriers, using unique Icelandic data, thus support findings from a New York study that reported increasing penetrance of BRCA mutations with time ( 19 ) . Our study differs from the New York study in several ways: We studied a different population; we based our risk estimates on the history of cancer in first-degree relatives of genotyped breast cancer probands instead of on the genotypes of first-degree relatives; and we applied a population-based study design, i.e., we could calculate directly the cumulative incidence during the overall period of 80 years in the first-degree relatives because they constituted a defined cohort that could be followed up by linking it with the population-based Cancer Registry. Also, we based our main conclusions on changes in risk over a defined calendar period of diagnosis instead of on changes in risk by birth cohorts. However, we also estimated cumulative incidence up to age 70 in the three successive birth cohorts—1836–1910, 1911–1930 and 1931–1950—and found that the risk had been increasing. Those results are likely to be less valid than the results based on year of birth, however, because the fixed follow-up period (1921–2002) means that young patients will be underrepresented in the first birth cohorts and older patients will be underrepresented in the latest birth cohorts. For example, a woman who was born in 1836 and diagnosed with breast cancer at the age of 40 years was diagnosed almost 40 years too early for her cancer to be listed in the Cancer Registry. On the other hand, a woman who was born in 1950 could be followed up only until she was 52 years old, when the study period ended.

An important question remains unanswered: Which has a stronger effect on the breast cancer risk, calendar year of diagnosis or year of birth? Presumably, breast cancer incidence by year of diagnosis will be strongly influenced by changes in awareness and screening behavior, whereas changes by birth cohorts will be more strongly influenced by changes in etiologic factors. However, we could not determine which was stronger, the effect of calendar year of diagnosis or the cohort effect. This was because Poisson models with log-linear associations of incidence rate (or death rate) with age, period, and birth cohort (higher-order associations were not statistically significant) have an identical fit to the data for age–period and age–cohort models because each of the three variables is a linear function of the other two ( 27 ) .

The marked increase in breast cancer risk in the Icelandic population is likely to be caused in part by the dramatic changes in lifestyle that occurred during the 20th century in Iceland. One manifestation of the biologic consequences of those changes is the increased life expectancy of 21-year-olds, which rose from 48 years to 61 years as estimated from death rates in 1921–30 and 1996–2000 respectively ( 28 ) ; another is the steady decline in age at menarche that has occurred during the 20th century ( 29 ) . Reproductive practices have also changed; the total fertility rate (i.e., the average number of children born alive per woman) decreased from 3.7 to 2.1 from 1921 through 2000, and the mean age of mothers at their first birth increased from 22 years in 1961 [the first year such data were available at Statistics Iceland ( 28 ) ] to 25 years in 2000.

Given these changes over time, it is interesting that neither the proportion of breast cancer probands with the BRCA2 999del5 mutation nor the relative cumulative incidence of breast cancer in mutation carriers compared with that in the general Icelandic population, appears to have changed statistically significantly with time. Moreover, Poisson modeling indicated that the increase in the risk of breast cancer before age 70 years during 1920–2000 was similar in all groups, i.e., 3.9-fold in the mutation carriers, 4.1-fold in relatives of noncarriers, and 4.2-fold in the general population. Thus, the factors that have influenced this increase appear to have had a similar overall effect on carriers of the BRCA2 mutation as they have had on the general population. However, results of other studies ( 30 – 34 ) indicate that the effects of pregnancies and breast feeding on breast cancer risk may differ between BRCA gene mutation carriers and women who do not carry BRCA gene mutations. Among BRCA gene mutation carriers, the use of oral contraceptives is associated with an increased risk of breast cancer ( 35 ) , whereas exercise and parity are associated with a reduced risk ( 19 ) and smoking history is not associated with risk ( 36 ) . It may seem surprising that we found that the increase in risk with time was similar for mutation carriers and the population, considering the different effects of pregnancies and breast feeding on mutation carriers and noncarriers. The 95% confidence intervals for our estimates were wide, however, and there may be some bias in the estimation of the time trends in penetrance because the probands were breast cancer patients and not a random sample from the population. Thus, there may still be some undetected differences in risk between BRCA2 999del5 mutation carriers and noncarriers. A better understanding of the effects of environmental factors on the breast cancer risk in mutation carriers would likely lead to both new preventive strategies and increased knowledge of the functions of the products of the BRCA genes. We are currently studying the effects of potential modifying factors, both environmental and genetic, on breast cancer risk.

The increasing breast cancer incidence over time that we observed may also, to some extent, be associated with increasing awareness of breast cancer and increasing diagnostic activity with time, which could lead to the diagnosis of lesions that might not have become clinically meaningful. Mammography screening may have this effect ( 37 ) . However, results of a recent Nordic study ( 38 ) indicated that mammography mainly advances the date of diagnosis of breast cancer. This time–trend analysis, which used age–period–cohort models ( 38 ) , showed that women who were routinely offered screening had a higher risk of breast cancer than women who were not offered screening, but the incidence rates decreased again when the women who were routinely offered screening reached the age of 70 years, when they left the screening program. Therefore, at least part of the increase in breast cancer incidence in Iceland that has occurred since 1987 (when a population-based mammography screening program started) can be explained by the effects of screening. The discrepancy between the magnitude of the increases in the cumulative incidence of breast cancer and the cumulative risk of death from breast cancer we observed can thus partly be explained by advancement of diagnosis (and thus the temporary increase in incidence) due to mammography screening, but this explanation applies only for the last 15 years of diagnosis. Other explanations are improved prognosis of breast cancer patients due to diagnosis at earlier clinical stages and advances in treatment of all stages of breast cancer.

Our study has several potential limitations. First, our results are based on one specific BRCA2 mutation. However, because the mutation leads to a premature truncation of protein translation and no mutant protein is detectable ( 39 ) , our results are likely to apply to other mutations that lead to complete inactivation of the BRCA2 gene. Second, selection bias might be a concern because the analyses were partly based on stored pathology tissue samples. However, all pathology tissue samples (i.e., paraffin blocks and microscope slides) obtained from cancer patients in Iceland, including samples dating back to the first half of the 20th century, are stored in a tissue bank in the Department of Pathology at the University Hospital. No samples have been deliberately discarded; thus any missing samples are missing at random. Third, the mean age at diagnosis of the probands was lower than the overall mean age at diagnosis in the Icelandic Cancer Registry, i.e., around 50 years instead of 60 years (data not shown). This lower age distribution was a consequence of the criteria that we used for genealogic tracing, in which both birth and diagnosis years were predefined for most of the probands ( 5 ) . Because the prevalence of the BRCA2 mutation is increased in young breast cancer patients ( 10 – 12 ) , this age truncation may account for the somewhat higher prevalence of the BRCA2 mutation among the probands (10.7%) than the 6%–8% prevalence observed in Icelandic breast cancer patients in general ( 11 ) and therefore it is possible that high-risk families are to some extent overrepresented in our data. Furthermore, it has been pointed out ( 40 ) that in all studies on risk in mutation carriers that use probands who are breast cancer patients, it can be expected that high-risk families will be more heavily represented than low risk families. This overrepresentation of high-risk families occurs because all risk factors, both genetic and environmental, are likely to be more highly presented in breast cancer patients than in the rest of the population. This bias is likely to result in some overestimation of the cumulative incidence. Thus, the risk estimate for BRCA2 mutation carriers of 72% in the year 2000 is likely to be high, as may be the overall risk estimate of 46% for the entire diagnosis period 1921–2002.

Our results indicate that the penetrance of the Icelandic BRCA2 founder mutation has increased dramatically over time and thus that the penetrance can be heterogenous, even between carriers of the same BRCA mutation. The increase in the probability of death from breast cancer before age 70 years was considerably lower than the increase in the incidence, with a current estimate of 27%. Changes in penetrance with time should be taken into account when the effects of BRCA mutations are being estimated for genetic counseling.

Appendix

Poisson model for incidence rate of breast cancer (first primary)—Iceland 1921–2002.

Ixi = exp(C0i + C1i × A + C2i × T ) × 100 000, where

  Ixi = incidence rate per 100 000 per year for x = Population,

   first-degree relatives of BRCA2 wild-type probands, or

   first-degree relatives of BRCA2 mutation probands.

  i = 1 for A < 45 y

   2 for A ≥ 45 y.

A = age (y).

T = calendar year – 1900.

Goodness-of-fit tests for the overall fit of the models.

 Population, age 20–44 y: Deviance (Dev) = 122.2,

 Degrees of freedom (DF) = 82, P = .003.

 Population, age ≥ 45 y: Dev = 192.0, DF = 150, P = .01.

 Relatives, age 20–44 y: Dev = 74.3, DF = 166, P >.99.

 Relatives, age ≥ 45 y: Dev = 244.6, DF = 281, P = .95.

Poisson model for rate of death from breast cancer—Iceland 1921–2002.

Dx = exp(C0 + C1 × A + C2 × T ) × 100 000, where

 Dx = rate of death per 100 000 per year for

   x = first-degree relatives of BRCA2 wild-type probands or

   first-degree relatives of BRCA2 mutation probands.

A = age (y).

T = calendar year − 1900.

Goodness-of-fit test for the overall fit of the model.

Relatives, age ≥ 20 y: Dev = 253.3, DF = 451, P >.99

References