1416

Risks of Cancer among Members of Families in the Gilda Radner Familial Ovarian Cancer Registry Mohannad F. Jishi, M.S.,* Jacqueline H . Itnyre, M.I.S.,t Ingrid A. Oakley-Girvan, M.P.H.,t M . Steven Piver, M.D.,* and Alice S. Whitternore, Ph.D.t

Background. Increasing scientific and public interest in hereditary cancer syndromes has created a need for estimates of lifetime cancer risks among members of fam- ilies with such syndromes.

Methods. Data from the Gilda Radner Familial Ovar- ian Cancer Registry were used to evaluate risk for can- cers of the breast, cervix, uterus, colorectum, and pros- tate in members of 143 families containing three or more reported cases of ovarian cancer among first- or second- degree relatives. These risks were compared with those that were expected based on general population rates ob- tained from the Connecticut Tumor Registry.

Results. Overall, family members' risk of cancer at any nonovarian site was 1.5 times that of the general pop- ulation (P < 0.001). Among female members, risk for can- cer of the breast was 2.5 times that of the general popula- tion. Risk for cancer of the uterus was 5 times that of the general population and increased with increasing num- ber of first-degree relatives with ovarian cancer. Among male family members having three or more first-degree relatives with ovarian cancer, prostate cancer risk was 4.5 times that of the general population. No excess risks were observed for cancer of the colorectum.

Conclusions. These data support previous reports of coaggregation of cancer of the breast, uterus and ovary, and suggest coaggregation between cancer of the ovary and prostate. Differences in cancer risk profiles observed in these families with multiple ovarian cancer and in car- riers of the gene BRCAl suggest that hereditary ovarian

From the *Department of Gynecologic Oncology, Roswell Park Cancer Center, New York State Department of Health, Buffalo, New York; and tDepartment of Health Research and Policy, Stanford Uni- versity School of Medicine Stanford, California.

Supported by NIH grants R35-CA47448 and R01-CA66190. The authors thank Anna Felberg for computing assistance. Address for reprints: Alice S. Whitternore, Ph.D., Department

of Health Research and Policy, Division of Epidemiology, Stanford University School of Medicine, Redwood Building, Room T204, Stan- ford, California 94305-5092.

Received February 15, 1995; revision received June 12, 1995.

cancer is genetically heterogeneous. Cancer 1995;76 1416-21.

Key words: breast cancer, hereditary ovarian cancer, prostate cancer, standardized incidence ratios, uterine cancer.

As more and more attention focuses on hereditary can- cer syndromes, there is increasing need for estimates of lifetime cancer risks among members of families exhib- iting such syndromes. Currently, there are relatively few data concerning risks for other malignancies among members of families with hereditary ovarian cancer syndromes. Such data are needed for the clinical man- agement and psychologic well being of individuals in these families. In addition, documentation of excess risks for other cancers in these families may help to elu- cidate the etiologies of the other cancers once the genes responsible for the syndromes are identified.

Data pertaining to familial incidence of cancers that coaggregate with ovarian cancer can be classified into three categories: (1) general population data, consisting largely of families with at most one ovarian cancer di- agnosis, which may have occurred by chancelw6; (2) data on families ascertained because of ovarian cancer syndromes, defined here as the occurrence of two or more ovarian cancers among first- or second-degree rel- ative~~,'; and (3) data on cancer risks among camers of mutated alleles of specific genes for ovarian cancer, such as BRCA1.9

Here, we estimated risks for cancer of the breast, cervix, uterus, large bowel, and prostate (nonovarian cancers) among members of 143 families ascertained because of ovarian cancer syndromes by the Gilda Radner Familial Ovarian Cancer Regstry. These fami- lies, which represent a small subset of families included in the Registry, were selected for analysis because they

Cancer Risk and Familial Ovarian Cancer/Jishi et al. 1417

contain three or more reported cases of ovarian cancer among first- or second-degree relatives, and, therefore, present examples of ovarian cancer syndromes that are unlikely to be due to chance. We evaluated risks of non- ovarian cancers among individuals classified according to their ”ovarian cancer risk category,” defined as the number of ovarian cancers among themselves and their first-degree relatives. Apart from its potential use to cli- nicians in counseling and screening members of ovarian cancer families, this classification of nonovarian cancer risks permits evaluation of the familial coaggregation of ovarian and nonovarian cancers and its genetic implica- tions.

Methods

The Gilda Radner Familial Ovarian Cancer Registry at the Roswell Park Cancer Institute (Buffalo, NY) was es- tablished in 1981 and as of December 31, 1992 con- tained 990 families with a total of 2376 reported cases of ovarian cancer. Individuals (who may or may not have been diagnosed with ovarian cancer) contact the Regis- try through their physician or through a toll-free tele- phone “hotline.” Those who report that their family contains two or more ovarian cancers among first- or second-degree relatives receive a brief mailed question- naire (Form A) informing them of the nature of the study and requesting information pertaining to all rela- tives with ovarian cancer and to incidence of all cancers among their first-degree relatives (parent-offspring and sibling relationships) and second-degree relatives (grandparent-grandchild, avuncular, and half-sibling relationships). Form A questionnaires also are sent to all relatives reported to have ovarian cancer. If a women with ovarian cancer is deceased, her questionnaire is completed by a surrogate (e.g., her next-of-kin). Re- ported ovarian cancer occurrence is verified by pathol- ogy and hospital records or death certificate whenever possible. However, we have not attempted to verify re- ported nonovarian cancer occurrence.

Individuals were eligible for inclusion in the pres- ent analysis if they fulfilled the following criteria: (1) they belonged to a family with at least three reported ovarian cancers among first- or second-degree relatives; (2) they were at least 20 years of age at the time of death or last contact; and (3) either (a) they (or a surrogate) completed a Form A questionnaire, or (b) they were the first-degree relative of someone who completed the questionnaire. Therefore, the study includes all first-de- gree relatives of individuals for whom a form A ques- tionnaire was completed, regardless of their ovarian cancer status, but only some first-degree relatives of affected women for whom a form A was not completed.

Each individual was classified into one of four ovar- ian cancer categories, according to the number of ovar- ian cancers (0, 1, 2, 3+) reported in the subgroup con- sisting of the individual and all his/her first-degree rel- atives. Reported numbers of cancers (breast, cervix, uterus not otherwise specified, colorectum, prostate, pancreas, other and unspecified cancers) were tabulated for each of the four categories,

Reported numbers of site-specific cancers were compared with the numbers expected based on popula- tion rates, specific for age and sex. To compute the ex- pected numbers of, for example, prostate cancer, we calculated each patient’s time at risk as the number of days from hi= date of birth to the date of either prostate cancer diagnosis, death, or last contact, whichever came first. We then cross-classified the total person-days at risk jointly by age (5-year categories) and calendar pe- riod (5-year periods from 1925 to 1990). Person-days at risk contributed before 1925 were assigned to the pe- riod of 1925-1929. We obtained the expected numbers of prostate cancers by multiplying the total person-days at risk in each age-calendar-year category by the age- and calendar-year-specific incidence rate for prostate cancer obtained from the Connecticut tumor registry. We then calculated standardized incidence ratios, i.e., ratios of observed to expected numbers of prostate can- cers. Similar calculations were applied to the other can- cer sites. To test the null hypothesis that a standardized incidence ratio equals one, we assumed that the ob- served number of cancers was distributed as a Poisson variable with the mean equal to the expected number of cancers. All P values were two-tailed.

Results

A total of 508 ovarian cancers were reported in the 143 pedigrees included in the current analysis; to date, 179 (35%) of these were verified medically. Among the 143 pedigrees, 75 contained at least one member with re- ported breast cancer, 13 contained 1 or more members with cervical cancer, 17 contained 1 or more members with uterine cancer not otherwise specified, 36 con- tained 1 or more members with colorectal cancer, and 21 contained 1 or more members with prostate cancer. Among the 75 breast/ovarian cancer pedigrees, 27 also contained at least 1 member with either uterine or colo- rectal cancer.

Table 1 shows the distribution of site-specific can- cer occurrence in the subset of eligible members of the selected high risk families, according to the number of reported first-degree relatives with ovarian cancer. Breast cancer was reported in 38 of the 747 women (5%). The median age at diagnosis was 50 years (data

1418 CANCER October 15,1995, Volume 76, No. 8

Table 1. Reported Cancer Occurrence Among Members* of the Gilda Radner Familial Ovarian Cancer Registry, by Number of First-Degree Relatives With Ovarian Cancer

Cancer site

Women (no.)

NO. of F-D Uterus, relatives Breast Colorectum Cervix unsaeci fied Othert N o cancer Total

0 1 0 1 6 1 2 12 3 3+ 19 4 Total 38 8

1 0 0 0 1 3 3 8 5 11

1 125 128 9 238 254 7 137 163

14 154 202 31 654 74 7

Breast Colorectum

Men(no.1

Prostate Other No cancer Total

0 0 0 0 12 171 183 1 0 2 1 7 169 179 2 0 1 2 7 70 81 3+ 0 2 6 14 56 78 Total 0 5 9 40 466 520

F-D: first-degree. * See text for eligibility criteria. t Includes unspecified sites.

not shown), versus a median age of 64 years for female patients with breast cancer in the general population." An additional diagnosis of ovarian cancer was reported in 18 of these 38 women. None of the male family mem- bers was reported to have had breast cancer. Cancers of the colon and rectum occurred in 13 (1 %) family mem- bers (median age, 52 years vs. 70 years" for men and 63 years vs. 73 years'" for women). Five of the eight women with colorectal cancer also had a reported diag- nosis of ovarian cancer. Cervical cancer was reported in 5 women (median age, 36 years vs. 47 years'"), and can- cer of the uterus not otherwise specified was reported in 11 (median age, 54 years vs. 66 years'"). Two women with reported cervical cancer and four women with re- ported uterine cancer also had a reported diagnosis of ovarian cancer. Prostate cancer was reported in nine (2%) of the men (median age, 70 years vs. 72 years"). The other cancers were too sparse and heterogeneous for analysis.

The crude prevalence estimates shown in Table 1 suggest that site-specific cancer prevalence increases with increasing number of first-degree relatives with ovarian cancer. However, these prevalence estimates are not adjusted for age or year of birth. Among the 520 men and 747 women included in the current analysis, data on times at risk of cancer and ages at cancer diag- nosis were available for a subset of 31 1 men (who con-

tributed a total of 15,400 person-years at risk) and 544 women (26,300 person-years at risk).

Table 2 shows the numbers of cancers expected among these individuals, assuming they developed cancer at the same rates as individuals of comparable age and year of birth in the Connecticut Tumor Regis- try. Comparison of observed to expected numbers shows that breast cancer risk is elevated approximately twofold among women with at least one affected first- degree relative, but there is little evidence of increasing risk with increasing number of such relatives. Uterine cancer risk is increased approximately fivefold overall, with the highest risk ratio (standardized incidence ratio = 7.7, P < 0.05) among women having three or more affected first-degree relatives. By contrast, there is no evidence for excess cervical cancer risk and little evi- dence for excess colorectal cancer risk among either male or female members of high risk ovarian cancer families, regardless of ovarian cancer incidence in their close relatives. Finally, although based on small num- bers (N = 6), men with three or more first-degree rela- tives with ovarian cancer were at elevated risk of pros- tate cancer (standardized incidence ratio = 4.5, P = 0.02). No excess risk was noted for men with fewer affected relatives. Statistically significant increases in risk were not found among family members having fewer than two first-degree relatives with ovarian can- cer, regardless of which cancer site was examined.

Cancer Risk and Familial Ovarian Cancer/Jishi et al. 1419

Table 2. Observed and Expected* Numbers of Site-Specific Cancers Among Members of the Gilda Radner Familial Ovarian Cancer Registry,t by Number of First-Degree Relatives With Ovarian Cancer

No. of first-degree relatives with ovarian cancer

0 1 2 3+ Total

Cancer site 0 E O / E 0 E O / E 0 E O / E 0 E O / E 0 E O / E

Breast (female) 1 0.6 1.7 6 2.6 2.3 11 4.0 2.73 17 7.1 2.48 35 14.3 2.511 CeNiX 1 0.1 8.7 0 0.5 0 1 0.9 1.1 2 1.7 1.2 4 3.3 1.2 Uterus,unspecified 0 0.1 0 0 0.4 0 3 0.6 4.9 8 1.0 7.7$ 11 2.2 5.0$ Colorectum

Male 0 1.9 0 2 1.8 1.1 1 0.9 1.1 2 1.8 1.1 5 6.4 0.8 Female 0 0.2 0 1 0.8 1.2 2 1.5 1.3 4 3.0 1.3 7 5.5 1.3

Prostate 0 1.5 0 1 1.5 0.7 0 0.7 0 6 1.3 4.5$ 7 5.1 1.4 All cancers 15 13.7 1.1 23 19.3 1.2 26 18.6 1.4 64 33.8 1.911 128 85.5 1.511 0: observed; E: expected. * Based on Connecticut tumor registry data, specific for age, sex, and calendar year. t See text for eligibility criteria. $ P < 0.051 g P < 0.01. /I P < 0.001.

Assuming lifetime risks in the general population of 12% for breast cancer’ and assuming (as suggested by the data in Table 2) a 2.5-fold increased risk for women with hereditary ovarian cancer syndromes, we found that these women had approximately a 2.5 X 12% =

30% lifetime risk of breast cancer. Similarly, the data in Table 2 suggest that such women have a lifetime risk of uterine cancer not otherwise specified of approximately 5 X 3% = 15%. The data contained too few men with prostate cancer for such risk estimation.

Discussion

We used family records contained in the Gilda Radner Familial Ovarian Cancer Registry to examine cancer in- cidence in members of families containing three of more women with reported ovarian cancer. Among female family members, breast cancer was reported at 2.5 times the rate expected, whereas uterine cancer was reported five times more often than expected. Among male fam- ily members having three or more mothers, sisters, or daughters with ovarian cancer, prostate cancer risk was 4.5 times that of the general population. No excess risk was noted for cancer of the colorectum.

Some limitations of the data must be considered when interpreting these findings. First, we cannot eval- uate ovarian cancer risks in these families because they came to our attention precisely because of multiple cases of the disease. Second, our classification of fami- lies according to number of members with ovarian can- cer was based on reported incidence; to date, only 35%

of all reported ovarian cancers have been verified by medical records. Similarly, incidence of other cancers in family members was based solely on reported inci- dence. Thus, there is the possibility of false negatives due to underreporting of relatives’ cancers. There is also the possibility that some ovarian cancers may have been reported as cancer of the uterus or colorectum, leading to overreporting of these cancers. Finally, fami- lies with ovarian cancer and other cancers may be more likely to contact the Registry than families without other cancers. Such selection bias seems unlikely because sev- eral of its predictions conflict with the data (e.g., excess cancers among family members with no affected first- degree relatives [not seen], and excess cancers at all sites [not seen]). Despite these possible limitations, it is of in- terest to compare the present findings with those based on findings of families with ovarian cancer in the gen- eral population, those of other families with ovarian cancer syndromes, and those of carriers of mutated al- leles of genes associated with ovarian cancer.

Studies pertaining to the general population have compared cancer incidence in relatives of ovarian can- cer patients to incidence in relatives of women of similar ages and races without the disease (control women). For example, data from a population-based U.S. case-con- trol study of ovarian cancer indicated that ovarian can- cer patients were more likely to report breast cancer in their first-degree relatives than were control women4 Indeed, breast cancer risk in mothers and sisters of pa- tients with ovarian cancer was 1.6 times that of such relatives of the control women,’ an increase similar to

1420 CANCER October 15,1995, Volume 76, No. 8

the 1 %fold breast cancer excess estimated by Houlston et al.’ in relatives of patients with ovarian cancer in Great Britain. Using the same population-based data as did Schildkraut et al., Claus et al.’ estimated lifetime breast cancer risks of 13% and 31%, respectively, for women having one and two first-degree relatives with ovarian cancer. These estimates of increased breast can- cer risk are consistent with the current estimated risk of 30% among members of families with three or more ovarian cancers. However, excess incidence of colorec- tal cancer in relatives of patients with ovarian cancer from the general population3c4 is inconsistent with the absence of reported colorectal cancer in the current data, although the excesses were small and failed to reach statistical significance. Moreover, some popula- tion-based studies have found no excess uterine cancer incidence in relatives of patients with ovarian which is inconsistent with the elevated uterine cancer risk observed here and in a geneology-based study in Utah.6 In agreement with the current findmgs, a cancer registry-based study of relatives of women with breast cancer in Iceland found elevated risks for cancer of the ovary, uterus, and prostate,” and a cohort study of postmenopausal women in Iowa found significant fa- milial aggregation of cancer of the ovary, breast, and prostate.”

Studies pertaining to cancer risks in members of families with ovarian cancer syndromes typically have been based on anecdotal reports of cancer incidence in specific pedigrees. Cancers of the breast, endometrium, and colon have been impli~ated.~.’~ Conversely, excess ovarian cancer incidence has been reported in members of families with hereditary nonpolyposis colon cancer syndromes.’ Although the present findings support the associations between ovarian cancer and cancer of the breast and uterus, they do not support a strong link be- tween cancer of the ovary and colon.

The most precise information pertaining to cancer risks associated with hereditary ovarian cancer syn- dromes is provided by estimates of lifetime risks among carriers of specific mutations of genes associated with the disease. Of the cancer susceptibility genes identified to date, only BRCAl has been associated with increased risk of ovarian cancer.14-16 Carriers of mutated alleles of the BRCAl gene have estimated lifetime risks of 87% and 44% for breast cancer and ovarian cancer, respec- tively. Carriers also have a fourfold increased risk of co- lon cancer, and a threefold increased risk of prostate cancer, but no excess of uterine ~ a n c e r . ~

These findings differ from those of the present study in two respects: risks for cancer of the breast and colorectum among BRCAl carriers are substantially higher than reported here, whereas uterine cancer risks

among carriers are lower than those reported here. These differences suggest that mutated alleles of BRCAl are inherited in only a fraction of these 143 multiplex families with ovarian cancer. Consistent with this inter- pretation is the fact that although BRCAl seems to be responsible for a large proportion of families with breast-ovarian cancer syndrome^,'^ the gene accounts for an estimated 78% of families with ovarian cancer without early breast cancer and an unknown fraction of families with no breast cancer. An absence of reported breast cancer characterizes some 48% of the 143 fami- lies analyzed here, all of whom reported at least three members with ovarian cancer.

The possibility that hereditary ovarian cancer is ge- netically heterogeneous also is supported by the vari- able trends in site-specific cancer risks with increasing numbers of affected first-degree relatives reported in the current data. The absence of trend in breast cancer risk is consistent with the hypothesis that all members of some families have a 50% chance of inheriting a sin- gle autosomal dominant gene such as BRCAI. In con- trast, the strong trends in prostate and uterine cancer risks with numbers of affected relatives suggests that some families are segregating other highly penetrant genes. Identification of these other genes is likely to help clarify the risks in all ovarian cancer families.

References

1.

2.

3.

4.

5.

6.

7.

8.

9.

Claus EB, Risch N, Thompson WD. The calculation of breast cancer risk for women with a first degree family history of ovar- ian cancer. Breast Cancer Res Treatment 1993;28:115-20. Houlston RS, Bourne TH, Collins WP, Whitehead MI, Campbell S, Slack J. Risk of ovarian cancer and genetic relationship to other cancers in families. Hum Hered 1993;43:111-5. Nelson CL, Sellers TA, Rich SS, Potter JD, McGovern PG, Kushi LH. Familial clustering of colon, breast, uterine and ovarian can- cers as assessed by family history. Genet Epidemiol 1993; 10:235- 44. Schildraut JM, Thompson WD. Relationship of epithelial ovar- ian cancer to other malignancies within families. Genet Epidemiol

Schildkraut JM, Risch N, Thompson WD. Evaluating genetic as- sociation among ovarian, breast and endometrial cancer: evi- dence for a breast/ovarian cancer relationship. A m J Hum Genet

Kerber RA, Slattery ML. The impact of family history on ovarian cancer risk: the Utah population data base. Arch Int Med

Lynch HT, Albano WA, Lynch JF, Lynch PM, Campbell A. Sur- veillance and management of patients of high genetic risk for ovarian carcinoma. Obstet Gynecol 1982;59:589-96. Lynch HT, Lanspa S, Smyrk T, Boman B, Watson P, Lynch J. Hereditary nonpolyposis colorectal cancer (Lynch syndromes 1 and 11). I: genetics, pathology, natural history and cancer control. Cancer Genet Cytogenet 1991;53:143-60. Ford D, Easton DF, Bishop DT, Narod SA, Goldgar DE, Breast

1988;5:355-67.

1989;45:521-9.

1995; 155~905-12.

Cancer Risk and Familial Ovarian Cancer/Jishi et al . 1421

Cancer Linkage Consortium. Risks of cancer in BRCAl-muta- tion carriers. Lancet 1994;343:692-5.

10. Ries LAG, Miller BA, Hankey BF, Kosary CL, Harras A, Edwards BK, editors. SEER Cancer Statistics Review, 1973-1991: tables and graphs. National lnstitutes of Health, National Cancer In- stitute; 1994 NIH Pub. No.: 94-2789. Tulinius H, Egilsson V, Olafsdottir GH, Sigvaldason H. Risk of prostate, ovarian, and endometrial cancer among relatives of women with breast cancer. BM] 1992;305:855-7.

12. Sellers TA, Potter JD, Rich SS, Drinhard CR, Bostick RM, et al. Familial clustering of breast and prostate cancers and risk of post- menopausal breast cancer. ] Nut2 Cancer lnst 1994; 86:1860-5. Lynch HT, Bewtra C, Lynch JF. Familial ovarian carcinoma. An2 ] Med 1985;81:1073-6.

11.

13.

14. Narod SA, Feunteun J, Lynch HT, Watson P, Conway T, Lynch J, et al. Familial breast-ovarian cancer locus on chromosome 17q12-23. Lancet 1991;338:82-3. Narod SA, Madlensky L, Bradley L, Cole D, Tonin P, Rosen B, et al. Hereditary and familial ovarian cancer in southern Ontario. Cancer 1994; 74:2341-6. Easton DF, Bishop DT, Ford D, Crockford GP, Breast cancer link- age consortium: genetic linkage analysis in familial breast and ovarian cancer: results from 214 families. Am H u m Genet 1993; 52:678-701. Steichen-Gersdorf E, Gallion HH, Ford D, Girodet C, Easton DF, DiCioccio RA, et al. Familial site-specific ovarian cancer is linked toBRCAl on 17q12-21.Arn]Hum Genet 1994;55:870-5.

15.

16.

17.