Am.J. Hum. Genet. 52:678-701, 1993
`
`Genetic Linkage Analysis in Familial Breast and Ovarian
`Cancer: Results from 214 Families
`
`D. F. Easton,* D. T. Bishop,t D. Ford,* G. P. Crockford,t and the Breast Cancer Linkage
`Consortium'
`*Cancer Research Campaign and Institute of Cancer Research, Section of Epidemiology, Belmont, Surrey, England; and timperial Cancer Research
`Fund, Genetic Epidemiology Laboratory, St. James's Hospital, Leeds
`
`Summary
`Breast cancer is known to have an inherited component, consistent in some families with autosomal dominant
`inheritance; in such families the disease often occurs in association with ovarian cancer. Previous genetic linkage
`studies have established that in some such families disease occurrence is linked to markers on chromosome 17q.
`This paper reports the results of a collaborative linkage study involving 214 breast cancer families, including
`57 breast-ovarian cancer families; this represents almost all the known families with 17q linkage data. Six
`markers on 17q, spanning approximately 30 cM, were typed in the families. The aims of the study were to define
`more precisely the localization of the disease gene, the extent of genetic heterogeneity and the characteristics of
`linked families and to estimate the penetrance of the 17q gene. Under the assumption of no genetic
`heterogeneity, the strongest linkage evidence was obtained with D17S588 (maximum LOD score [Zm.] = 21.68
`at female recombination fraction [Of] = .13) and D17S579 (Zmax = 13.02 at Of = .16). Multipoint linkage analysis
`allowing for genetic heterogeneity provided evidence that the predisposing gene lies between the markers
`D17S588 and D17S250, an interval whose genetic length is estimated to be 8.3 cM in males and 18.0 cM in
`females. This position was supported over other intervals by odds of 66:1. The location of the gene with respect
`to D17S579 could not be determined unequivocally. Under the genetic model used in the analysis, the best
`estimate of the proportion of linked breast-ovarian cancer families was 1.0 (lower LOD - 1 limit 0.79). In
`contrast, there was significant evidence of genetic heterogeneity among the families without ovarian cancer, with
`an estimated 45% being linked. These results suggest that a gene(s) on chromosome 17q accounts for the majority
`of families in which both early-onset breast cancer and ovarian cancer occur but that other genes predisposing
`to breast cancer exist. By examining the fit of the linkage data to different penetrance functions, the cumulative
`risk associated with the 17q gene was estimated to be 59% by age 50 years and 82% by age 70 years. The
`corresponding estimates for the breast-ovary families were 67% and 76%, and those for the families without
`ovarian cancer were 49% and 90%; these penetrance functions did not differ significantly from one another.
`
`Received June 15, 1992; revision received October 30, 1992.
`Address for correspondence and reprints: D. Timothy Bishop,
`ICRF Genetic Epidemiology Laboratory, Ashley Wing, St. James's
`Hospital, Beckett Street, Leeds LS9 7TF, England.
`1. Collaborating groups' abbreviations, used throughout this re-
`port, and members of the Breast Cancer Linkage Consortium are as
`follows (for a full list of affiliations, see accompanying papers): Aber-
`deen = University of Aberdeen, Aberdeen-N. Haites, B. Milner,
`and L. Allan; Berkeley = University of California, Berkeley-M.-C.
`King, A. Bowcock, and L. Anderson; CRC = Cancer Research Cam-
`paign and Institute of Cancer Research, London and Cambridge-
`D. F. Easton, B. A. J. Ponder, J. Peto, S. Smith, K. Anderson, D. Ford,
`and M. Stratton; FCN = French Cooperative Network, Lyon, Paris,
`and Clermont-Ferrand-H.
`Sobol,
`S. Mazoyer,
`Lyonnet,
`D.
`Y. J. Bignon, and P. Lalle; IARC = International Agency for Research
`on Cancer, Lyon-S. A. Narod, G. M. Lenoir, J. Feunteun, and H.
`
`678
`
`Lynch; Iceland = University of Iceland, Reykjavik-A. Arason, R.
`Barkardottir, and V. Egilsson; ICRF = Imperial Cancer Research
`Fund, Leeds and London-D. T. Bishop, D. M. Black, D. Kelsell,
`V. A. Murday, E. Solomon, N. Spurr, and G. Turner; Leiden = Lei-
`den University, Leiden-P. Devilee; Manchester = Christie Hospital
`and Holt Radium Institute, Manchester-J. M. Birch, M. S. San-
`tibaiiez-Koref, and M. D. Teare; MDC = Max-Delbriick-Centrum
`fur Molekulare Medizin, Berlin-S. Scherneck and W. Zimmerman;
`MRC = Medical Research Council Human Genetics Unit, Edin-
`burgh-M. Steel, D. Porter, B. B. Cohen, and A. Carothers; Stock-
`holm = Karolinska Hospital, Stockholm-A. Lindblom and C. Lars-
`son; and Utah = University of Utah, Salt Lake City-L. A.
`Cannon-Albright, D. Goldgar, and M. Skolnick.
`C 1993 by The American Society of Human Genetics. All rights reserved.
`0002-9297/93/5204-0005$02.00
`
`GeneDX 1008, pg. 1
`
`

`
`Breast Cancer Linkage Analysis
`
`Introduction
`
`A family history of breast cancer is a now well-estab-
`lished risk factor for the disease (Kelsey 1979). Many
`previous epidemiological studies have shown a signifi-
`cantly increased risk of breast cancer in the relatives of
`breast cancer cases as compared with the general popu-
`lation (e.g., see Adami et al. 1980; Claus et al. 1990).
`With some exceptions (e.g., Mettlin et al. 1990), these
`studies generally show that the risk of breast cancer in
`relatives increases with decreasing age of the index case
`and is greater for women with several affected relatives
`than for women with one affected relative (Claus et al.
`1990, 1991). In their analysis of the CASH study, based
`on the families of over 4,500 breast cancer cases and
`matched controls, Claus et al. (1991) found that familial
`clustering of breast cancer could be best explained by a
`model in which there are one or more autosomal domi-
`nant predisposing genes conferring a high risk of the
`disease; germ-line mutations in these genes must, how-
`ever, only account for a minority of cases of the disease,
`the remaining cases occurring sporadically; but the pro-
`portion of cases occurring in predisposed individuals
`increases with decreasing age at onset. Similar models
`have been obtained by several other authors (Williams
`and Anderson 1984; Bishop et al. 1988; Newman et al.
`1988; Iselius et al. 1991).
`Ovarian cancer is also known to have a familial com-
`ponent, and there is evidence that this is also at least
`partly the result of an autosomal dominant gene(s) with
`high penetrance (Schildkraut and Thompson 1988;
`D. F. Easton, unpublished data). Moreover, the risk of
`breast cancer is increased in ovarian cancer relatives,
`and vice versa (Schildkraut et al. 1989; D. F. Easton, F.
`Matthews, A. J. Swerdlow, and J. Peto, unpublished
`data). This suggests the existence of a gene(s) predispos-
`ing to both breast cancer and ovarian cancer. In addi-
`tion, many impressive pedigrees supporting the exis-
`tence of such dominant highly penetrant breast and/or
`ovarian cancer genes have been published (e.g., see
`Gardner and Stephens 1950; Fraumeni et al. 1975;
`Lynch et al. 1978).
`The first convincing localization of a breast cancer
`gene by genetic linkage was achieved by Hall et al.
`(1990), who obtained a LOD score of 2.35 at 0 = .20
`with D17S74 located in chromosomal region 17q21.
`There was significant evidence of heterogeneity, with
`an estimated 40% of families being linked. Analysis by
`age showed that linkage evidence was clearest in those
`families with earliest age at onset (Merette et al. 1992).
`Among families with an age at onset of less than 46
`
`679
`
`years, the maximum LOD score was 5.98 at 0 = .001
`(Hall et al. 1990). Families with later onset showed evi-
`dence against linkage. However, subsequent reanalysis
`(using analytical methods similar to those in the present
`paper) has shown that the evidence against linkage in
`later-onset families disappears if sporadic cases are al-
`lowed for appropriately, since these families contain a
`high proportion of sporadic cases (Margaritte et al.
`1992).
`Confirmation of this linkage result was provided by
`Narod et al. (1991), who obtained a LOD score of 2.20
`at 0 = .20 with D17S74 in five breast-ovarian cancer
`families. That study also found significant evidence of
`genetic heterogeneity, linkage being apparently re-
`stricted to three of the five families tested. There was
`no apparent difference in average age at onset between
`linked and unlinked families in that study. This breast-
`ovarian
`has been formally
`cancer locus
`labeled
`"BRCA1" (Solomon and Ledbetter 1991).
`The aims of this collaborative study were fourfold:
`first, to attempt to confirm the previously published
`linkage results on a much larger set of breast cancer
`families; second, to localize the gene to a particular
`chromosomal interval, by using several markers span-
`ning the region of interest; third, to define the extent of
`genetic heterogeneity and to attempt to characterize
`linked and unlinked families; and, fourth, to use the
`marker information to provide an estimate of the pene-
`trance of the 17q gene.
`In order to provide as unbiased an assessment as pos-
`sible of the extent of heterogeneity, the aim was to
`encourage participation by as many groups as possible,
`worldwide, who are conducting breast cancer linkage
`studies. Groups were asked to submit linkage informa-
`tion on all families studied, regardless of the strength of
`their evidence for or against linkage. The majority of
`the groups have been collaborating over the past 3 years
`by participating in biennial meetings at which linkage
`results have been shared. When the linkage to D17S74
`was reported by Hall et al. (1990) and replicated by
`Narod et al. (1991), the groups all typed D17S74. Heter-
`ogeneity analysis for the combined set of families was
`significant, but there was no resolution of the location
`of the breast cancer locus (data not shown). In brief, the
`data were consistent with, at one extreme, the disease
`being tightly linked to D17S74 in 20% of the families
`and, at the other extreme, 90% percent being linked
`but with the disease gene 20 cM from D17S74.
`To resolve this and to facilitate a joint multipoint
`linkage analysis, a common set of highly polymorphic
`markers was recommended for typing across the family
`
`GeneDX 1008, pg. 2
`
`

`
`680
`
`set. Initially, these markers were chosen to be up to 20
`cM either side of D17S74, but early results suggested
`that only markers proximal to D17S74 were relevant. A
`modified set of six markers focusing on proximal loca-
`tions was drawn up, and we report here on the results
`of these analyses.
`
`Families, Material, and Methods
`Families
`Thirteen groups contributed a total of 214 families
`to this study. Accompanying papers by the collaborat-
`ing groups describe in detail the methods used for se-
`lecting families, data collection, and confirmation of
`diagnoses. Criteria for family selection differed be-
`tween groups. Two of the groups (Utah and Iceland)
`ascertained their families by using record linkage of
`population genealogies to cancer registries (Arason et
`al. 1993; Goldgar et al. 1993). Most of the remaining
`families were ascertained through collaborating physi-
`cians. The majority of families were ascertained pre-
`dominantly through a family history of breast cancer,
`though one group (CRC) selected some families on the
`basis of ovarian cancers in the family (Smith et al. 1993)
`
`Easton et al.
`
`while another group (IARC) selected most of their fami-
`lies for both breast and ovarian cancer (Feunteun et al.
`1993). One family was ascertained and typed indepen-
`dently by two groups (reference numbers IARC 2775
`and Utah 1901). All the analyses for this family are
`based on the IARC data.
`In 10 of the families one breast cancer occurred in an
`individual on the opposite side of the family from the
`remaining cancers; these breast cancer cases are ignored
`in the summaries of numbers and ages of cases in the
`families (but not in the linkage analysis). In one family
`(see Arason et al. 1993) there were three breast cancer
`cases on the opposite side of the family. In the remain-
`ing 203 families all cases of breast and ovarian cancer
`occurred in a pattern consistent with segregation of a
`single rare autosomal dominant gene.
`Table 1 summarizes basic details about the families
`studied. All the families except three contained at least
`two breast cancer cases (one family contained four
`ovarian cancers and no breast cancers, one contained
`three ovarian cancers and one breast cancer, and one
`contained seven ovarian cancers and one breast cancer).
`One hundred twenty-four (58%) contained at least four
`cases, and 77 (36%) contained at least three cases under
`
`Table I
`Summary of Types of Family in the Study
`
`No. OF FAMILIES
`
`Average Age at
`Diagnosis of
`Breast Cancer
`(years)
`
`No. of Ovarian
`Cancers
`
`No. OF PEDIGREE MEMBERS
`
`Male
`Breast
`2 >2 Cancers Total Total Bilateral
`
`Breast Cancer
`
`Diagnosis at Ovarian
`<45 Years
`Cancer Total
`
`Affected
`
`Typed
`
`GRouPa
`
`Total <45 45-60 >60
`
`0
`
`1
`
`50
`4
`1
`Aberdeen ........
`1
`1
`141
`356
`20
`2
`1
`Berkeley .........
`17
`23
`5
`1
`599 192
`1 31 2 2 8
`19
`43 23
`CRC ...........
`570
`119
`7
`2
`6
`4
`6
`13
`19
`IARC ...........
`261
`48
`3
`1
`3
`Iceland ..........
`5
`2
`7
`330
`64
`2
`12
`2
`7
`ICRF ...........
`9
`16
`353
`61
`2
`11
`Leiden ...........
`6
`13
`5
`303 76
`14 3 1 1
`10
`9
`19
`FCN ...........
`92
`30
`3
`6
`3
`9
`6
`MDC ...........
`9
`426
`72
`10
`2
`1
`9
`5
`15
`MRC ...........
`3
`59
`26
`6
`1
`2
`3
`Manchester
`2
`7
`......
`6
`82
`205
`29
`13
`Stockholm .......
`11
`29
`5
`351 72 _
`11 I
`1
`4
`8
`13
`Utah ...........
`3,955
`101
`987
`157 26 10
`19
`102
`93
`Total ..........
`214
`a One family was independently identified by both the IARC group and the Utah group. In this table, the family is only listed under the IARC
`group, as the latter had sampled more extensive branches of the pedigree.
`
`7
`28
`19
`7
`12
`2
`8
`
`3
`56
`99
`76
`20
`34
`23
`35
`16
`34
`8
`15
`39
`458
`
`8
`7
`49
`42
`7
`6
`4
`8
`3
`10
`1
`
`8
`153
`
`28
`256
`245
`356
`156
`157
`246
`192
`62
`239
`25
`89
`209
`2,260
`
`4
`117
`153
`85
`37
`45
`39
`69
`25
`47
`16
`68
`44
`749
`
`2
`
`2
`1
`
`1
`
`1
`
`5
`
`1
`
`2
`
`1
`21
`
`GeneDX 1008, pg. 3
`
`

`
`Breast Cancer Linkage Analysis
`
`the age of 45 years. Consequently, in light of the known
`epidemiological data, few of the families could be
`chance aggregations. Fifty-seven (27%) of the families
`contained at least one case of ovarian cancer (These
`families are referred to hereafter as "breast-ovary fami-
`lies."), and 31 (14%) contained two cases. Of the re-
`maining families, four contained one or more cases of
`male breast cancer and are referred to as "male breast
`cancer families." (One of the breast-ovary families also
`contained one case of male breast cancer.) The remain-
`ing 153 families which contained only female breast
`cancer are referred to as "breast-only families." Seven
`hundred forty-nine (66%) of the 1,140 breast and ovar-
`ian cancer cases in the families were typed for at least
`one marker.
`The numbers in table 1 summarize the status of the
`families as recorded at the time of this analysis (in
`March 1992). Some collaborating groups have accumu-
`lated further information since this date, and the totals
`in the accompanying papers therefore differ slightly
`from those given in table 1.
`There are clear differences between the ages at onset
`of the breast and ovarian cancers in the families ascer-
`tained. Among the diagnoses of breast and ovarian
`cancers, 4% of those made before the age of 30 years
`are ovarian cancer, compared with 11% of diagnoses
`made between 30 and 59 years of age and 15% of those
`made at age 60 years or older. For the CRC and IARC
`families which focused attention for ascertainment on
`both ovarian and early-onset breast cancer, the corre-
`sponding figures were 7%, 18%, and 25%.
`
`Markers
`Six polymorphic markers (GH, D17S74, NME1,
`D17S588, D17S579, and D17S250) were typed in the
`families, although not all of the families were typed for
`all of the markers. All of these markers were CA-repeat
`polymorphisms, with the exception of the VNTR
`cMM86 (D17S74) used in the original reports of link-
`age. These markers were chosen because of both their
`known high levels of polymorphism and their locations.
`Details on the markers, as well as on others subse-
`quently used by consortium members, are summarized
`in the Appendix, which is supplied by Dr. Nigel Spurr.
`The laboratory methods used by individual groups to
`type these markers are described in the accompanying
`papers. In total, 194 families (91%) were typed for
`D17S588, 189 (88%) for D17S74, 189 (88%) for
`D17S579, 172 (80%) for D17S250, 163 (76%) for
`NME1, and 141 (66%) for GH.
`
`681
`
`Marker Allele Frequencies
`One disadvantage of using microsatellite and VNTR
`markers was that consistent scoring of marker alleles,
`between groups and even between families being stud-
`ied by the same group, could not be guaranteed. This is
`an important issue, since misspecifying allele frequen-
`cies can lead to exaggerated LOD scores and, in excep-
`tional cases, to a biased recombination fraction esti-
`mate (Ott 1992). Some of the groups did attempt to
`score D17S588 and NME1 consistently, but we found
`discrepancies in the sizes of the most common alleles
`reported. For these reasons, our analyses assume equal
`allele frequencies for each marker. The markers GH
`and D17S74 are known to be highly polymorphic (some
`groups scored 12 or more distinct alleles within some
`families), so we assumed that all alleles had frequency
`.05. The remaining markers were less polymorphic, and
`we chose allele frequencies to reflect their observed
`polymorphism. For D17S579 we assumed 10 equally
`frequent alleles, for D17S250 and D17S588 6 alleles,
`and for NME1 5 alleles. In some families, more alleles
`were recognized, and recoding of the marker alleles
`within such families was necessary (see Statistical
`Methods section, below). This solution is clearly not
`ideal, but we made these assumptions to avoid the
`greatly exaggerated LOD scores which would result
`from underestimating the frequency of an allele ob-
`served in two distantly related affected individuals. As a
`check on the robustness of the results to misspecifica-
`tion of the allele frequencies, we also repeated the two-
`point analyses for D17S588 and D17S579, assuming
`different allele frequencies. Reducing the allele fre-
`quencies increased the maximum LOD scores, but the
`recombination
`estimates remained essentially
`un-
`changed.
`Statistical Methods
`The two-point and multipoint linkage analyses were
`carried out using the LINKAGE program (Lathrop and
`Lalouel 1984; Lathrop et al. 1984). As a standard ge-
`netic model for breast cancer, for use in the linkage
`analysis, we used the model derived by Claus et al.
`(1991) from the Cancer and Steroid Hormone Study. In
`this model, breast cancer susceptibility is conferred by
`an autosomal dominant allele, with population fre-
`quency .0033, such that the breast cancer risk is 67% by
`age 70 years in carriers and is 5% in noncarriers. A
`minor modification to this model was made, in order to
`remove the large increase in risk to gene carriers over
`the age of 70 years as compared with the risk at age
`60-69 years. In our model, the age-specific risk of
`
`GeneDX 1008, pg. 4
`
`

`
`682
`
`breast cancer after age 70 years in gene carriers was
`taken to be the same as the risk at age 60-69 years. This
`model is referred to as the "CASH" model throughout.
`To implement this model, individuals were assigned to
`one of seven age groups (less than 30 years, 30-39 years,
`40-49 years, 50-59 years, 60-69 years, 70-79 years,
`and 80 or more years), on the basis of either the age at
`diagnosis of breast cancer, if affected, or the age at
`death or last observation, if unaffected. Because the
`disease status is censored at the age of last observation
`for unaffected individuals, the likelihood must be cal-
`culated according to survival analysis methods (Elston
`1973). This was achieved by assigning affected and un-
`affected individuals to different liability classes, so that
`14 liability classes (7 age groups X 2 disease classifica-
`tions) were used in total. The "penetrance" probabili-
`ties assigned to these classes were the genotype-specific
`cumulative risk, for unaffecteds, and the genotype-spe-
`cific density, for affecteds, appropriate to each age
`group (with the values at ages 25, 35, 45, 55, 65, 75 and
`85 years being used, assuming a constant incidence rate
`over the decade). These values are shown in table 2.
`Since male breast cancer is so rare in the general pop-
`ulation, male cases in these families are likely to be the
`result of genetic predisposition. We therefore assigned
`
`Table 2
`Penetrance Values Used in the LINKAGE Program
`
`AGE GROUP
`(years)
`
`Unaffected females:
`<30 ............
`30-39 ..........
`40-49 ..........
`50-59 ..........
`60-69 ..........
`70-79 ..........
`>80 ............
`Affected females:
`<30 ............
`30-39 ..........
`40-49 ..........
`50-59 ..........
`60-69 ..
`..
`70-79 ..........
`>80 ..........
`
`PENETRANCE OF GENOTYPEa
`
`dd
`
`.00009
`.00146
`.0083
`.021
`.039
`.061
`.082
`
`.00002
`.00026
`.00112
`.00137
`.00226
`.00218
`.00213
`
`Dd
`
`.008
`.083
`.269
`.469
`.616
`.724
`.801
`
`.00167
`.01276
`.02305
`.01711
`.01260
`.00908
`.00654
`
`;
`
`DD
`
`.008
`.083
`.269
`.469
`.616
`724
`.801
`
`.00167
`.01276
`.02305
`.01711
`.01260
`.00908
`.00654
`
`NOTE.-Male breast cancers and ovarian cancers were assigned to
`the youngest age group.
`a D refers to the disease allele conferring susceptibility to breast
`and ovarian cancers, and d is the "normal" allele.
`
`Easton et al.
`
`affected males to the "female affected before age 30"
`liability class, to maximize their probability of being
`carriers. Unaffected males were assigned to the "female
`unaffected at age 30" liability class, approximately
`equivalent to being of unknown disease status.
`Ovarian cancers presented a problem, since no ge-
`netic model specifically applying to breast-ovary fami-
`lies has been derived. We considered that ovarian
`cancers should be treated differently than breast
`cancers, first, because the disease is much rarer in the
`general population (and therefore cases in the families
`are more likely to be genetic) and, second, because ex-
`amination of these families and of those previously
`published suggests that (in contrast to breast cancer)
`ovarian cancers occurring at older ages are also part of
`the syndrome. We therefore assigned ovarian cancer
`cases to the "female affected before age 30" liability
`class. However, we also carried out analyses in which
`women with ovarian cancer were assigned risks based
`on the age at diagnosis, in the same manner as for breast
`cancer. If both cancers occurred in the same woman,
`the age of the first cancer was taken. The differences in
`results obtained by using these two approaches were
`very minor (see Results). All other cancers were ignored
`in the analyses. There is no clear evidence that any
`other cancers should be regarded as part of the syn-
`drome, and, consequently, information on other
`cancers was not consistently reported. Excesses of
`other cancers could not therefore be reliably examined
`in this data set.
`Linkage analyses in these families and in the CEPH
`families (Fain et al. 1991) shows that the female genetic
`map is considerably longer than the male map, over this
`region. Over the whole region, the distance ratio is ap-
`proximately twofold, although there is variation across
`the region. For this reason, all two-point LOD scores
`have been calculated on the assumption that the female
`distance is twice the male distance. Results are summa-
`rized in the tables, according to the female recombina-
`tion fraction (Of).
`Given both the large number of alleles present for
`each marker and the complexity of many of the fami-
`lies, a full multipoint analysis including all markers and
`BRCA1 was not feasible. Instead, we performed a series
`of multipoint analyses. The first involved D17S250,
`D17S588, and BRCA1-except for the IARC families,
`for which no D17S250 data were available. In the IARC
`families, the analysis involved D17S74, D17S588, and
`BRCA1, with the LOD scores being evaluated at the
`equivalent locations on the genetic map, to allow a
`combined analysis of all families. In subsequent analy-
`
`GeneDX 1008, pg. 5
`
`

`
`Breast Cancer Linkage Analysis
`
`ses to further resolve the location, we analyzed
`D17S588, D17S579, and BRCA1. The position of the
`disease locus was also assessed by a more informal ex-
`amination of recombinant events.
`Multipoint analyses were performed using LINK-
`MAP (Lathrop et al. 1984), again by assuming a two-
`fold-increased map distance for females versus males.
`The computational efficiency of multipoint linkage
`analysis depends on the number of alleles at each of the
`loci. Analyses take progressively longer, the larger the
`number of alleles-while ignoring the phenotypic dif-
`ferences between alleles loses linkage information. The
`compromise which we adopted involved recoding the
`markers to six alleles for D17S588 and D17S250, to
`seven alleles for D17S579 (six "observed" alleles each
`with frequency .1 and a seventh "unobserved" allele
`with frequency .4), and to seven alleles for D17S74 (six
`"observed" alleles each with frequency .05 and a sev-
`enth "unobserved" allele with frequency .70). In total,
`11 families show more than six alleles for D17S588, as
`do 19 for D17S250 and 16 for D17S579. For all except
`two of these families, recoding could be performed
`without change of linkage information, by considering
`those nuclear families in which the allele transmitted
`from each parent to child could be unequivocally de-
`termined. Alleles carried by individuals marrying into
`such nuclear families were then renumbered while
`maintaining evidence of transmission. As in the paper
`by Ott (1978), such recoding changes the absolute value
`of the likelihood but does not change the LOD score.
`In the two exceptions, some linkage information was
`lost, but this was shown to be trivial by comparing
`pairwise LOD scores before and after recoding. For
`D17S74, the large number of alleles recognized made
`recoding necessary in 35 families. In this case, recoding
`was conducted in such a way as to preserve as much
`linkage information as possible; consequently, some of
`the marker phase information from spouses of pre-
`sumed carriers was lost. Specifically, in some families a
`noncarrier spouse heterozygous for D17S74 was re-
`coded as being homozygous. Even with this recoding,
`each multipoint analysis took approximately 3 wk on a
`dedicated SUN 4 workstation.
`Heterogeneity of linkage for various subgroups of
`families was examined by comparing log likelihoods
`calculated separately under the assumptions of hetero-
`geneity and homogeneity. The subgroups considered
`were based on considerations of the genetic epidemiol-
`ogy of breast and ovarian cancer and so included an
`investigation both of age-at-onset effects for breast
`cancer and of the relevance of ovarian cancer cases
`
`683
`
`within a family. Following preliminary evidence that
`different proportions of breast-ovary families and
`breast-only families were linked to 17q, we estimated
`the proportion of each set of families linked, assuming
`that a single gene was responsible for both sets of fami-
`lies. This analysis was performed by computing the log
`likelihood over all families, postulating separate linked
`proportions for the two classifications of families but
`the same 17q susceptibility locus.
`The linkage data were used to provide an estimate of
`the age-specific penetrance of the 17q gene, by maxi-
`mizing the LOD score over different values of the pene-
`trance function, which is equivalent to maximizing the
`likelihood conditional on all the disease phenotypes
`(Risch 1984). This method allows the penetrance to be
`estimated free of bias due to ascertainment of families
`on the basis of multiple affected individuals. This analy-
`sis was carried out using a modified version of the
`ILINK program (Lathrop et al. 1984). The analyses
`were based on the linkage results with D17S588, the
`marker showing the strongest linkage evidence-ex-
`cept for the Berkeley families, for which D17S588 was
`not completely typed; D1 7S579 was used for these fami-
`lies (For other groups, D17S588 was more completely
`typed than D17S579). For this analysis the recombina-
`tion fraction between BRCA1 and D17S588 was fixed
`at .04 in males and .08 in females, as predicted by the
`multipoint analysis. For the breast-ovary families, pene-
`trance was modeled assuming a separate parameter for
`the ratio of the incidence rates in gene carriers versus
`that in noncarriers, for each of the seven 10-year age
`groups defined previously. The rates in noncarriers
`were fixed at the rates in noncarriers in the CASH
`model. Since the aim of this analysis is to estimate the
`overall penetrance, age at onset was defined by the age
`at diagnosis of the first cancer (either breast or ovary).
`This assumption is in contrast to the previous linkage
`analyses, where ovarian cancers were distinguished
`from breast cancers (see above).
`For the breast cancer families without ovarian cancer
`there was evidence of genetic heterogeneity. In order to
`obtain an estimate of the penetrance, we assumed, for
`simplicity, that the families resulted from the segrega-
`tion of either BRCA1 or a second, unlinked autosomal
`dominant gene; these two genes were allowed to confer
`different penetrances. The penetrance of the linked
`gene was modeled in terms of seven age-specific risks,
`as for the breast-ovary families. The penetrance con-
`ferred by the unlinked gene was initially assumed to be
`equal to the penetrances given by Claus et al. (1991) and
`used in the previous analyses, but different penetrances
`
`GeneDX 1008, pg. 6
`
`

`
`684
`
`were also considered. It was also necessary to estimate a
`parameter y = p1/(p1 + P2), where p1 and P2 are the
`gene frequencies of the linked and unlinked genes, re-
`spectively; y is not, of course, the same as the propor-
`tion of linked families, since the two genes can confer
`different penetrances, and the probability of a family
`being linked will therefore depend on the number and
`ages at onset in affected individuals.
`
`Results
`Linkage Map
`As a basis for the multipoint linkage analysis, we first
`constructed a multipoint linkage map of the six
`markers, based on the marker data in the breast cancer
`families and using the program CRIMAP (Barker et al.
`1987; Lander and Green 1987). The preferred order for
`the markers, together with the corresponding sex-spe-
`cific genetic distances between markers, is shown in
`figure 1. This order is preferred to all other orders with
`odds of more than 1014:1. It also corresponds to the
`order generated by analysis of recombinants in CEPH
`families (Fain et al. 1991), and is consistent with the
`order of loci generated by somatic cell and radiation
`hybrids (Black et al. 1993). The corresponding two-
`point marker-marker LOD scores and recombination
`fraction estimates are shown in table 3.
`
`Two-Point Linkage Results under Homogeneity
`Table 4 summarizes the overall two-point LOD
`scores for linkage between BRCA1 and the six chromo-
`some 17q markers, by the female recombination frac-
`tion. Significant evidence of linkage is seen with all
`markers except growth hormone (GH). The strongest
`evidence of linkage is with D17S588 (Zmax = 21.68 at Of
`= .13) and D17S579 (Zmrx = 13.02 at Of = .16). In terms
`of the maximum LOD score, this represents more than
`
`Easton et al.
`
`a fourfold increase in the linkage evidence over that
`previously reported (Hall et al. 1990; Narod et al.
`1991).
`Table 5 presents a more detailed breakdown of the
`results for D17S588 and D17S579, according to the
`types of cancer occurring in the families. The majority
`of the linkage evidence comes from the breast-ovary
`families (Zna = 18.44 at Of = .08, for D17S588; and
`Zma = 11.24 at of = .08, for D17S579). There is also
`significant linkage evidence within the families contain-
`ing only female breast cancer cases, but the LOD scores
`are maximized at significantly greater distances (3.60 at
`Of = .22, for D17S588; and 3.74 at Of = .22, for
`D17S579). This is suggestive of genetic heterogeneity,
`which is examined in more detail below.
`Of the four families containing male breast cancer
`(but not ovarian cancer), there is some evidence of link-
`age with D17S588, though not with D17S579. One fam-
`ily shows evidence of linkage (see Cohen et al. 1993),
`and the male breast cancer case shares the disease-bear-
`ing chromosome, while one particularly impressive fam-
`ily (see Hall et al. 1990) shows strong evidence against
`linkage. This latter family was not typed with D17S588,
`which explains the discrepancy in results between the
`two markers. These male breast cancer families have
`been excluded from all subsequent analyses.
`Among the breast-ovary families, no difference is ap-
`parent between families with just one ovarian cancer
`case and those with multiple ovarian cancer cases. We
`repeated the analysis for the breast-ovary families, al-
`tering the liability class for the ovarian cancer cases to
`that appropriate to the age at diagnosis, as f