Ovarian Cancer Survival McGuire et al. Survival in Epithelial Ovarian Cancer Patients with Prior Breast Cancer

Transcription

1 American Journal Eidemiology Coyright 2000 by The Johns Hokins University School Hygiene and Public Health All rights reserved Vol. 152, 6 Printed in U.S.A. Ovarian Cancer Survival McGuire et al. Survival in Eithelial Ovarian Cancer Patients with Prior Breast Cancer Valerie McGuire, Alice S. Whittemore, Robin Norris, and Ingrid Oakley-Girvan Ovarian cancer atients who carry germ-line BRCA1 mutations may have imroved survival comared with ovarian cancer atients without these mutations. To evaluate this hyothesis, the authors comared survival in ovarian cancer atients who had a history rior breast cancer with that atients without such a history. Secifically, they used data from the oulation-based US Surveillance, Eidemiology, and End Results (SEER) Program to assess time to death from ovarian cancer among ovarian cancer atients with and without a rior breast cancer. All 25,637 White women diagnosed with invasive eithelial ovarian cancer in SEER registries between 1973 and 1995 were included. Of these, women had had a rior breast cancer diagnosis. The ovarian cancer death rate among women with rior breast cancer was significantly lower than that women with ovarian cancer only, adjusted for age and stage at ovarian cancer diagnosis. The survival advantage was most ronounced among older women and among those whose ovarian cancers were more advanced at the time diagnosis. These results lend indirect suort to rior findings imroved ovarian cancer survival in BRCA1 mutation carriers. Am J Eidemiol 2000;152: breast neolasms; genes, BRCA1; ovarian neolasms; survival Eithelial ovarian cancer is the sixth most common cancer among women and is the leading cause death from gynecologic malignancies (1). The overall 5-year survival rate for ovarian cancer in the United States is 46 ercent (2). Over 65 ercent atients resent with advanced disease, and their 5-year survival is 20 ercent (2). Carriers germline mutations the gene BRCA1 have an increased risk ovarian cancer. The risk by age 70 years among BRCA1 heterozygotes has been estimated as 16 ercent (3), as 21 ercent (4), and as 40 ercent (5). These risks are considerably higher than the corresonding risk 1 ercent in the general oulation (4). Several studies have investigated the relation between BRCA1 mutation status and survival from ovarian cancer, with conflicting results (6 11). Data from some them suggest that ovarian cancer atients who carry germ-line BRCA1 mutations have imroved survival comared with ovarian cancer atients without these mutations (6 8). To account for these observations, it has been seculated that the survival advantage for BRCA1 mutation carriers reflects better resonse to chemotheray (8). However, other studies have reorted similar survival ovarian cancer atients with and Received for ublication June 3, 1999, and acceted for ublication December 16, Abbreviations: CI, confidence interval; SEER, Surveillance, Eidemiology, and End Results. From the Deartment Health Research and Policy, Stanford University School Medicine, Stanford, CA. Corresondence to Dr. Valerie McGuire, Deartment Health Research and Policy, HRP Redwood Building, Room T213C, Stanford University School Medicine, Stanford, CA ( vmcguire@leland.stanford.edu). without BRCA1 mutations (9 11). The majority these studies used clinic-based samles atients with and without mutations, who may have been selected for factors related to survival. To further evaluate the hyothesis imroved survival without such selection bias, we comared survival in a oulation-based samle women who had a history rior breast cancer with that women without such a history. This comarison is relevant because it has been estimated that some 88 ercent women with both breast and ovarian cancer carry BRCA1 mutations (12), while less than 5 ercent all ovarian cancer atients do so (4, 13). (The 88 ercent estimate was obtained from women attending high-risk clinics and, thus, may be somewhat higher than would be seen in a samle such atients from the general oulation.) MATERIALS AND METHODS Study oulation We analyzed data from nine oulation-based cancer registries oerated by the Surveillance, Eidemiology, and End Results (SEER) Program the National Cancer Institute. The registries are located in Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, northern California, Seattle, and Utah. All 25,637 White women who were diagnosed with invasive eithelial ovarian cancer between 1973 and 1995 were included. Of these, women were reorted in a SEER registry as having had a rior breast cancer diagnosis. We excluded women who were diagnosed with ovarian cancer low malignant otential (n = 1,591). We also excluded non-white women with eithelial ovarian cancer (n 2,822), because too few these women had reorted rior breast cancer diagnoses to ermit searate analysis. 528

2 Ovarian Cancer Survival 529 Statistical analysis The event interest was death from ovarian cancer as determined by information from death certificates. The SEER registries obtain cause death information from the state Vital Statistics Deartments, which determine whether the conditions listed on the death certificates are immediate or underlying causes death. If a woman died from causes other than ovarian cancer, her time to death was treated as censored. We comuted Kalan-Meier estimates ovarian cancer survival robability versus time since ovarian cancer diagnosis (14). We used Poisson regression methods (15, 16) to calculate exected numbers deaths among eithelial ovarian cancer atients with a rior history breast cancer, based on rates in women without such a history, adjusted for age at diagnosis, stage disease at diagnosis, and geograhic location the SEER registry. To test the null hyothesis that atients with and without breast cancer were similar with resect to various characteristics (age at diagnosis, stage disease at diagnosis, histologic subtye ovarian cancer), we used the standard chi-squared test for homogeneity in a 2 k table, where k is the number categories a given characteristic (17). We evaluated survival secific for age at ovarian cancer diagnosis (<35, 35 44, 45 54, 55 64, 65 74, 75 84, 85 years), stage disease at ovarian cancer diagnosis (localized, regional, distant, unstaged), and SEER registry (Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, northern California, Seattle, Utah). The SEER Program classifies an ovarian cancer as localized (cancer is confined to the ovary), regional (cancer extends beyond the ovary into surrounding organs and tissues or regional lymh nodes), or distant (cancer has sread to sites remote from the ovary). Other otential confounders, such as family history, arity, and oral contracetive use, are not available in the SEER data. RESULTS Table 1 summarizes the clinical and athologic features the two grous women. Women with a rior breast cancer were older at the time their ovarian cancer diagnoses ( ) and were less likely to have been diagnosed with localized disease ( 0.05). The tumor histologies were similar for the two grous women ( 0.10). Table 2 shows observed and exected numbers deaths for the White women diagnosed with ovarian cancer and rior breast cancer, by age at ovarian cancer diagnosis. The exected numbers are based on the death rates in White women with ovarian cancer only, adjusted for ovarian cancer stage. Overall, the observed number deaths from ovarian cancer among women with rior breast cancer was significantly less than exected ( ). This deficit was statistically significant in each the three age grous sanning the age range from 55 to 84 years. Table 3 shows observed and exected numbers ovarian cancer deaths among the White women diagnosed with ovarian cancer and rior breast cancer, by ovarian cancer stage. TABLE 1. Characteristics White eithelial ovarian cancer atients with and without rior breast cancer, SEER* registries, Characteristic Age (years) < Stage Localized Regional Distant Unstaged Histology Serous Mucinous Endometrioid Clear cell Other Ovarian cancer only 1,098 2,160 4,329 6,099 6,018 3,907 1,202 5,665 3,099 14,543 1,506 9,819 3,284 3,139 1,009 7, Ovarian cancer and rior breast cancer % % ,813 * SEER, Surveillance, Eidemiology, and End Results The exected numbers are based on the death rates in White women with ovarian cancer only, adjusted for age at ovarian cancer diagnosis. The observed number deaths was significantly lower than exected, in all women and in women diagnosed with advanced ovarian cancer. Differences between the number observed and exected deaths did not achieve statistical significance at the other stages. Figure 1 shows stage-secific survival robabilities for the two grous women. Overall, the estimated 5-year survival robability for women with a rior breast cancer was 49 ercent (95 ercent confidence interval (CI): 45, 56), comared with a corresonding robability 45 ercent among women without rior breast cancer (95 ercent CI: 44, 46). The 5-year survival advantage for atients with a rior breast cancer was most ronounced in women diagnosed with distant disease at ages 55 years and older. The 5-year survival for women with breast cancer was 32 ercent (95 ercent CI: 26, 38) comared with a corresonding robability 20 ercent for women without breast cancer (95 ercent CI: 19, 21). Adjustment for age and stage disease at ovarian cancer diagnosis suorted this survival advantage: the age- and stage-adjusted death rate ratio comaring women with and without rior breast cancer was 0.75 (95 ercent CI: 0.68, 0.84). Adjustment for geograhic area did not significantly affect the death rate ratios.

3 530 McGuire et al. TABLE 2. Observed and exected numbers deaths among White eithelial ovarian cancer atients with rior history breast cancer, by age at ovarian cancer diagnosis, SEER* registries, Age at diagnosis (years) subjects Personyears follow-u Observed deaths Exected deaths Death rate ratio % CI* < , , , , , , , 1.18 DISCUSSION 2, * SEER, Surveillance, Eidemiology, and End Results; CI, confidence interval. Due to ovarian cancer, based on death rates in ovarian cancer atients without a rior breast cancer, adjusted for stage at ovarian cancer diagnosis. Adjusted for stage disease at diagnosis. Adjusted for age and stage disease at diagnosis , 0.84 These oulation-based data suggest that, among White women diagnosed with invasive eithelial ovarian cancer, those with a rior history breast cancer survive their ovarian cancer longer than do women without such a history. Although BRCA1 mutation status was not available for these women, it has been estimated that 88 ercent women with both breast and ovarian cancers are BRCA1 mutation carriers (12) comared with 5 ercent all ovarian cancer atients (4, 13). Thus, the data rovide indirect suort for the hyothesis that BRCA1 heterozygote ovarian cancer atients have better rognosis than do other ovarian cancer atients. Several strengths and limitations this analysis warrant discussion. One strength is that the comarison is based on incident eithelial ovarian cancers identified through a oulation-based surveillance system and, thus, is free from the otential selection bias ertinent to some the other studies addressing the issue. These latter studies comared survival BRCA1 heterozygote atients who were identified because they had been attending a high-risk clinic with survival atients identified in some other setting. Individuals attending high-risk clinics tend to have closer surveillance, which could lead to earlier detection and suriously imroved survival (i.e., lead time bias ). A second strength arises because, unlike most the revious analyses, we did not rely on mutation testing archived araffin-embedded tissue, a rocedure that can lead to surious BRCA1 mutations (18). Such surious mutations would blur the distinction between BRCA1 heterozygotes and nonheterozygotes, leading to failure to detect a survival difference. This strength is counterbalanced, however, by the otential for a different form misclassification in the resent analysis. Some the atients classified as having no rior breast cancer may have been diagnosed with the disease while residing outside the SEER catchment area. Such misclassification would decrease the chance finding a survival difference and would bias the death rate ratio toward unity. However, the death rate ratio was significantly reduced and the magnitude the effect may have been underestimated. TABLE 3. Observed and exected numbers deaths among White eithelial ovarian cancer atients with a rior history breast cancer, by stage at ovarian cancer diagnosis, SEER* registries, Stage at diagnosis subjects Personyears follow-u Observed deaths Exected deaths Death rate ratio % CI* Localized Regional Distant Unstaged , , , , , , * SEER, Surveillance, Eidemiology, and End Results; CI, confidence interval. Due to ovarian cancer, based on death rates in ovarian cancer atients without a rior breast cancer, adjusted for age at ovarian cancer diagnosis. Adjusted for age at diagnosis. Adjusted for age and stage disease at diagnosis , 0.84

4 Ovarian Cancer Survival 531 FIGURE 1. Kalan-Meier robability survival from invasive eithelial ovarian cancer among atients with and without rior breast cancer, by stage disease at ovarian cancer diagnosis, Surveillance, Eidemiology, and End Results (SEER) registries, A second limitation the resent analysis is the absence data on clinical rognostic factors for ovarian cancer survival, such as the extent tumor before and after surgery, and the tye chemotheray. If atients with a rior breast cancer receive more otimal debulking or more effective chemotheray than do atients without such a rior cancer, our inability to adjust for the differences could mislead us to erroneously interret the observed differences as biologically based. A third limitation the study is the lack information on ovarian cancer screening history. Breast cancer atients may be screened more aggressively for ovarian cancer than women in the general oulation, which might cause earlier than usual detection the disease and imroved survival. Three other studies have reorted imroved ovarian cancer survival in BRCA1 carriers. Rubin et al. (6) estimated the median survival 43 BRCA1-heterozygote atients with advanced disease to be 77 months, comared with 29 months in control atients matched for age, tumor stage, histology, and grade. In a second study, the 5-year survival rate for 13 Jaanese BRCA1 heterozygotes with stage III ovarian cancer was 79 ercent, comared with a rate 30 ercent for 29 stage III atients who were deemed soradic (criteria unknown) and who were matched to BRCA1-ositive atients on age and treatment tye at diagnosis (7). These two studies have several limitations. BRCA1 mutation status was determined using archived tissue, although the Jaanese study did identify some BRCA1 carriers with DNA from blood lymhocytes. The BRCA1 heterozygotes were identified through high-risk clinics, leading to ossible selection bias. Rubin et al. (6) were unable to control for the extent disease and tye treatment, since several the BRCA1 heterozygotes had been diagnosed many years ago. The third study reorting imroved survival in BRCA1 carriers included 933 consecutive cases diagnosed with eithelial ovarian cancer from a single institution (8). The median 5-year survival for 67 advanced stage BRCA1 ovarian cancer atients was significantly longer than that matched ovarian cancer atients without BRCA1 mutations. Interestingly, the authors found that BRCA1-heterozygote atients received otimal debulking more frequently than did atients without BRCA1 mutations. The strengths this study include selection consecutive cases from a single institution and more detailed information on the extensiveness disease and the tye chemotheray. As with the two revious studies, BRCA mutation status was determined using archived tissue. In contrast to these results, three studies have found no survival differences in BRCA1 mutation carriers comared with either atients who tested BRCA1 negative or untested atients in the general oulation. Pharaoh et al. (9) reorted that the 5-year survival in 127 ovarian cancer atients from multile-case families segregating BRCA1 mutations was similar to that 119 atients from families testing negative for BRCA1 mutations. A strength this study is its use DNA from eriheral lymhocytes to determine BRCA1 mutation status. The survival similarity is noteworthy in view the fact that the atients from BRCA1-ositive families were more likely than those from BRCA1-negative families to be diagnosed with stage III/IV disease (89 ercent vs. 65 ercent). In a second study, Johannsson et al. (10) comared survival among 38 Swedish BRCA1-ositive ovarian cancer

5 532 McGuire et al. atients with that 97 ovarian cancer atients from the general Swedish oulation, matched to the BRCA1 heterozygotes on age, stage disease at diagnosis, and calendar year diagnosis. BRCA1 carriers were those who tested ositive in analysis either lymhocytes or archived tissue, or those deemed to be obligate carriers. The death rate ratio was 1.2 (95 ercent CI: 0.5, 2.8). Finally, Lee et al. (11) comared ovarian cancer survival among 10 first-degree relatives US Ashkenazi Jewish carriers BRCA1 or BRCA2 mutations with that 116 ovarian cancer atients who were first-degree relatives Ashkenazi Jewish noncarriers. They found similar survival in the two grous relatives. These latter three studies share some the limitations discussed earlier. In articular, results from the first two studies are based on BRCA1-heterozygote atients articiating in high-risk clinics. None them controlled for extent disease or tye treatment. It is also ossible that inconsistencies among the ositive and negative findings reflect survival differences in carriers different sitesecific BRCA1 mutations in different oulations. A survival advantage for BRCA1-heterozygote ovarian cancer atients, if not due to chance or bias, may reflect their better resonse to chemotheray (8). If the BRCA1 rotein lays a role in DNA reair (19), then BRCA1-deficient cancer cells may be unable to reair the DNA damage induced by chemotheraeutic agents and thus are killed by the treatment. Further studies are needed to address this issue. REFERENCES 1. Parkin DM, Pisani P, Ferlay J. Global cancer statistics. CA Cancer J Clin 1999;49: Ries LAG, Kosary CL, Hankey BF, et al, eds. SEER cancer statistics review, Bethesda, MD: National Cancer Institute, Struewing JP, Hartge P, Wacholder S, et al. The risk cancer associated with secific mutations BRCA1 and BRCA2 among Ashkenazi Jews. N Engl J Med 1997;336: Whittemore AS, Gong G, Itnyre J. Prevalence and contribution BRCA1 mutations in breast cancer and ovarian cancer: results three US oulation-based case-control studies ovarian cancer. Am J Hum Genet 1997;60: Easton D, Ford D, Bisho DT, et al. Breast and ovarian cancer incidence in BRCA1 mutation carriers. Am J Hum Genet 1995; 56: Rubin SC, Benjamin I, Behbakht K, et al. Clinical and athological features ovarian cancer in women with germ-line mutations BRCA1. N Engl J Med 1996;335: Aida H, Takakuwa K, Nagata H, et al. Clinical features ovarian cancer in Jaanese women with germ-line mutations BRCA1. Clin Cancer Res 1998;4: Boyd J, Sonoda Y, Federici MG, et al. Clinicoathologic features BRCA-linked and soradic ovarian cancer. JAMA 2000;283: Pharaoh PD, Easton DF, Stockton DL, et al. Survival in familial BRCA1-associated and BRCA2-associated eithelial ovarian cancer. Cancer Res 1999;59: Johannsson OT, Ranstam J, Borg A, et al. Survival BRCA1 breast and ovarian cancer atients: a oulation-based study from southern Sweden. J Clin Oncol 1998;16: Lee JS, Wacholder S, Struewing JP, et al. Survival after breast cancer in Ashkenazi Jewish BRCA1 and BRCA2 mutation carriers. J Natl Cancer Inst 1999;91: Frank TS, Manley SA, Oloade OI, et al. Sequence analysis BRCA1 and BRCA2: correlation mutations with family history and ovarian cancer risk. J Clin Oncol 1998;16: Stratton JF, Gayther SA, Russell P, et al. Contribution BRCA1 mutations to ovarian cancer. N Engl J Med 1997;336: Kalan E, Meier P. Nonarametric estimation from incomlete observations. J Am Stat Assoc 1958;53: Kelsey J, Whittemore AS, Evans AS, et al. Methods in observational eidemiology. New York, NY: Oxford University Press, Rothman KJ, Boice JD Jr. Eidemiologic analysis with a rogrammable calculator. Washington, DC: US Government Printing Office, (NIH ublication no ). 17. Brown BW Jr, Hollander N. Statistics. A biomedical introduction. New York, NY: John Wiley & Sons, Wong C, DiCioccio RA, Allen HJ, et al. Mutations in BRCA1 from fixed araffin-embedded tissue can be artifacts reservation. Cancer Genet Cytogenet 1999;107: Scully R, Chen J, Plug A, et al. Association BRCA1 with Rad51 in mitotic and meiotic cells. Cell 1997;88: