Correspondence
BRCA1 Mutations and Survival in Women with Ovarian Cancer
N Engl J Med 1997; 336:1254-1257April 24, 1997
- Article
To the Editor:
Rubin and colleagues (Nov. 7 issue)1 conclude that advanced epithelial ovarian cancers in patients with germ-line BRCA1 mutations have a better prognosis than their sporadic counterparts. Although this may eventually prove to be correct, the methods used in this study raise several questions.
Important prognostic information about the patients with BRCA1 mutations, such as the status of residual disease and the type of chemotherapy used, was said to be unavailable from the patients' charts, although these data are critical in choosing a proper control group. Did the authors attempt to check the original operative reports; to make direct contact with the surgeons, the medical oncologists, or both; and, when possible, to ask the patients? Since many patients and family members know what kind of chemotherapy was administered, I am surprised that this basic information was not available.
Histology and disease stage are important prognostic factors in ovarian tumors, yet in only 14 cases were slides reviewed to confirm the diagnosis. Why were slides from the remaining 39 patients unavailable? Is it possible that the distribution of histologic subtypes and grades shown in Table 1 of the article would have differed if the authors had undertaken a centralized review of all available slides?
The authors make no direct mention of how rigorously they assessed survival status. Was this done by reviewing the charts, calling patients directly, or examining death certificates? Readers should be given enough information for them to have confidence in the quality of the data presented.
In view of these uncertainties in the BRCA1-mutation group, I am at a loss to know how the authors can confidently choose a representative control group for comparison. It is not enough to select the control group on the basis of age and stage alone, because there are many other influences on outcome that could easily bias the results. The median survival of 29 months in the control group suggests that those patients represent a group at particularly high risk, with suboptimally debulked disease,2 and this may therefore artifactually magnify any potential survival advantage associated with the presence of a BRCA1 mutation. These concerns lead me to believe that the authors have drawn conclusions about the influence of BRCA1 mutations on prognosis that are not entirely justified by their data.
2 ReferencesStephen A. Cannistra, M.D.
Dana–Farber Cancer Institute, Boston, MA 021151
Rubin SC ,Benjamin I ,Behbakht K , et al. Clinical and pathological features of ovarian cancer in women with germ-line mutations of BRCA1. N Engl J Med 1996;335:1413-1416
Full Text | Web of Science | Medline2
McGuire WP ,Hoskins WJ ,Brady MF , et al. Cyclophosphamide and cisplatin compared with paclitaxel and cisplatin in patients with stage III and stage IV ovarian cancer. N Engl J Med 1996;334:1-6
Full Text | Web of Science | Medline
To the Editor:
Rubin et al. conclude that “as compared with sporadic ovarian cancers, cancers associated with BRCA1 mutations appear to have a significantly more favorable clinical course.” This is a hazardous conclusion, because of the dissimilarities between the patients (who had germ-line BRCA1 mutations) and the controls (who had ovarian cancer but no BRCA1 mutations).
First, all the controls were “treated relatively recently by gynecologic oncologists at an academic medical center,” whereas many of the patients “were treated less recently in a community setting.” Without more data on treatment and cause of death, it may be speculated that the difference in survival between the two groups may have been due to more aggressive chemotherapy in the controls, which could have increased morbidity and mortality among these women. Second, it is not stated what proportion of the controls was seen at the academic medical center because first-line chemotherapy in the community had failed, thereby biasing the selection of this group toward women with a poorer prognosis. Although the median survival in the control group was similar to that recorded by others,1 those investigators studied patients with either stage IV disease or incompletely resected stage III disease who were also treated in a university setting.
Finally, although the patients studied by Rubin et al. were matched to the controls on the basis of tumor stage, it is likely that unlike the controls, some of the patients had family members with a history of ovarian cancer. Heightened awareness among these patients may have led to diagnosis and treatment earlier in the course of stage III disease (Rubin et al. give mortality data for 38 patients with stage III disease without specifying the substage [A, B, or C] and for 5 patients with stage IV disease). This possible earlier diagnosis may have resulted in a higher proportion of complete tumor resections among the patients than among the controls.
1 ReferencesS. Elizabeth Whitmore, M.D.
Johns Hopkins University, Baltimore, MD 212051
McGuire WP ,Hoskins WJ ,Brady MF , et al. Cyclophosphamide and cisplatin compared with paclitaxel and cisplatin in patients with stage III and stage IV ovarian cancer. N Engl J Med 1996;334:1-6
Full Text | Web of Science | Medline
To the Editor:
Rubin et al. suggest that ovarian cancer associated with BRCA1 mutations has a more favorable clinical course than sporadic ovarian cancer. Although we would all like to believe that there might be some good news for BRCA1 carriers, I think this interpretation is premature, given the potential for lead-time bias in the diagnoses of the women with BRCA1 mutations.1 Since most of those women probably had family histories of breast cancer, ovarian cancer, or both, they would probably have been participating in screening programs for ovarian cancer. Thus, their ovarian cancers could have been detected earlier than usual. Even in the case of stage III and IV disease, the presymptomatic state could have been present for years in some women.
1 ReferencesRobert D. Burk, M.D.
Albert Einstein College of Medicine, Bronx, NY 10461-16021
Walter SD ,Day NE . Estimation of the duration of a pre-clinical disease state using screening data. Am J Epidemiol 1983;118:865-886
Web of Science | Medline
To the Editor:
The data presented by Rubin et al. cannot be used as evidence of better survival among women with germ-line mutations of BRCA1. A bias in their study arises from a potential discrepancy in the course of disease between the patients and the selected controls. Since BRCA1 mutations were detected in the families of patients with ovarian cancer only in 1994 and the overall study population included patients identified much earlier (as the median survival of 77 months indicates), it is only natural that there were more long-term survivors among the patients, who would have had their tests performed years after their diagnoses. In other words, those patients would have had to survive for a number of years in order to be tested, whereas patients with shorter survival whose disease was diagnosed during the same years would not have reached this stage because of their early deaths.1 Also, most of the patients included in the case group who were given their diagnoses in 1995 and 1996 would probably still be alive at the time of the analysis.
As the article stated, tissue samples from some of the patients were tested posthumously, but it is unlikely that either this factor or matching according to disease stage would eliminate this bias. It would be helpful if Rubin et al. gave the distribution of the years of diagnosis and of the determination of the BRCA1 status in the two groups, as well as the distribution of mutation-positive and mutation-negative patients, to support the analysis as presented.
Incidentally, the authors state in the last paragraph of their article that “germ-line BRCA1 mutations occur in only a small proportion of all ovarian cancers.” This statement is incorrect in the light of several recent observations.2,3
3 ReferencesBaruch Modan, M.D.
Chaim Sheba Medical Center, Tel Hashomer 52621, Israel1
Modan B . Computing the length of survival in long-term disease. JAMA 1965;192:609-612
Web of Science | Medline2
Muto MG ,Cramer DW ,Tangir J ,Berkowitz R ,Mok S . Frequency of the BRCA1 185delAG mutation among Jewish women with ovarian cancer and matched population controls. Cancer Res 1996;56:1250-1252
Web of Science | Medline3
Modan B ,Gak E ,Sade-Bruchim RB , et al. High frequency of BRCA1 185delAG mutation in ovarian cancer in Israel. JAMA 1996;276:1823-1825
CrossRef | Web of Science | Medline
To the Editor:
Rubin et al. suggest that carriers of BRCA1 mutations with ovarian carcinoma survive longer than members of a comparison group matched for age and disease stage. But the groups they studied differed in location and time of diagnosis and treatment.
In an ongoing population-based study of carriers of BRCA1 mutations in southern Sweden (unpublished data), we compared the survival of 38 patients with BRCA1 germ-line mutations and ovarian cancer diagnosed between 1958 and 1995 with that of a group of 97 women matched with the carriers for disease stage, age, calendar year of diagnosis, and histopathological features. All the women were treated at the same institution. We found no significant difference between the groups in survival (P = 0.45) (Figure 1Figure 1
Overall Survival of 38 Patients with Ovarian Cancer and BRCA1 Mutations and 97 Controls Matched for Age and Disease Stage.). When the patients with BRCA1 mutations were compared with the matched controls by the McNemar test, the relative risk of mortality was 1.3 (95 percent confidence interval, 0.5 to 3.1). Although we did not exclude potential carriers of BRCA1 germ-line mutations from our control group, we believe that the survival of women carrying such mutations is the same as that of other patients with ovarian cancer, if not worse.Oskar Johannsson, M.D.
Jonas Ranstam, Ph.D., C.Stat.
Åke Borg, Ph.D.
Håkan Olsson, M.D.
University Hospital Lund, S-221 85 Lund, SwedenTo the Editor:
Rubin et al. estimate the 5-year survival rate among 43 women with advanced ovarian cancer and BRCA1 mutations to be 60 percent, and the 10-year survival rate to be 42 percent. These rates are much higher than those in a comparison group of women with nonhereditary cancer. This remarkable outcome of stage III and IV ovarian cancers prompts close scrutiny.
It is difficult to estimate survival rates from historical data because of the possibility that living patients were selectively included in the study. This is of particular concern in molecular epidemiologic studies, because it is often easier to determine the mutation status of living patients than that of deceased ones (the success rate is higher when DNA typing is performed with blood samples than with stored tissue). A family with cancer comes to the physician's attention when a family member seeks counseling, and a mutation analysis is attempted if a living patient is available for study. For example, if there was ovarian cancer in four deceased members of a family but not in any living member, mutation analysis is generally not offered. Furthermore, to estimate survival rates one needs data on the year of birth, the year of the cancer diagnosis, and the year of death. Patients for whom data are missing are excluded. For living patients only data on the first two variables are required, but for deceased patients it is necessary to have data on all three. Deceased patients are preferentially excluded if information on the age at death is not collected for the entire sample.
We have performed a survival analysis similar to that of Rubin et al. in a group of 44 Canadian patients with ovarian cancer and BRCA1 mutations, but in our study there were several differences. First, we excluded the index patients (the proband), because including them would bias the study population toward surviving patients. Second, we included all patients with ovarian cancer, regardless of grade and stage. Finally, we included all family members with invasive ovarian cancer, without restricting our study population to subjects whose carrier status was confirmed by molecular studies. In our experience, over 90 percent of patients with invasive ovarian cancer in families with identified BRCA1 mutations are carriers.1 The median actuarial survival in our cohort was 2.6 years, as compared with 6.4 years in the study of Rubin et al. The 5-year survival rate in our study was 32.6 percent, and the 10-year survival rate 14.9 percent — rates similar to those in the control group of Rubin et al.
We believe that a possible explanation for the survival differences between our two studies is selection bias, and that it will be important to address this issue by conducting prospective studies of women with ovarian cancer, with and without BRCA1 mutations.
1 ReferencesJean-Sebastien Brunet, M.Sc.
Steven A. Narod, M.D.
Women's College Hospital, Toronto, ON M5G 1N8, CanadaPatricia Tonin, Ph.D.
William D. Foulkes, M.B., Ph.D.
McGill University, Montreal, QC H3G 1A4, Canada1
Narod S ,Tonin P ,Lynch H ,Watson P ,Feunteun J ,Lenoir G . Histology of BRCA1-associated ovarian tumours. Lancet 1994;343:236-236
CrossRef | Web of Science | Medline
Author/Editor Response
Dr. Rubin replies:
To the Editor: At each participating center, my colleagues and I included all BRCA1-related cases for which the basic clinical information was available. There was no requirement for the patients to be treated or even seen at that institution, and in many instances they were identified as BRCA1-mutation carriers only after death, through an analysis of archival tissues performed at the request of a relative. No patients were identified through screening or surveillance programs. Because the BRCA1-related patients were treated at dozens of hospitals, often many years ago, we could not obtain some important clinical data, such as information on the substage of disease and the amount of residual tumor. Accordingly, we selected the patients in our control group without knowing this information.
Cannistra raises the question of conducting a centralized review of histologic material. Although we acknowledge the importance of such a review, histologic subtypes are not important prognostic factors in ovarian cancer, except for tumors of the clear-cell variety, which constitute about 5 percent of epithelial ovarian cancers and were not represented among either our patients or our controls.1
With regard to Whitmore's comments, the members of the control group all received their primary treatment at the University of Pennsylvania Medical Center; none were referred after treatment elsewhere had failed. The suggestion that patients treated by subspecialists in an academic medical center would have a poorer survival than patients treated in a community setting is counterintuitive and contrary to published evidence.2
The data of Johannsson et al. and Brunet et al. are interesting, but without knowing more about their methods we hesitate to speculate about why their conclusions differ from ours. Brunet et al. suggest that the estimation of survival rates requires fewer data points for living patients than for dead ones, and that this could be a potential source of bias. For the widely accepted Kaplan–Meier method3 we used, the same three data points are needed for each patient, living or dead: the date of diagnosis, the date of the last follow-up, and the status at that follow-up (alive or dead).
We wish to emphasize that nothing in our data suggests that patients with BRCA1-related ovarian cancer should be treated differently from patients with sporadic ovarian cancer, and we agree completely that a prospective study is needed to clarify these issues.
3 ReferencesStephen C. Rubin, M.D.
University of Pennsylvania Medical Center, Philadelphia, PA 19104-42831
Ozols RF, Rubin SC, Thomas G, Robboy S. Epithelial ovarian cancer. In: Hoskins WJ, Perez CA, Young RC, eds. Principles and practice of gynecologic oncology. 2nd ed. Philadelphia: Lippincott–Raven, 1997:919-86.
2
Eisenkop SM ,Spirtos NM ,Montag TW ,Nalick RH ,Wang HJ . The impact of subspecialty training on the management of advanced ovarian cancer. Gynecol Oncol 1992;47:203-209
CrossRef | Web of Science | Medline3
Kaplan EL ,Meier P . Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53:457-481
CrossRef | Web of Science
- Citing Articles (4)
Citing Articles
1
Nancie Petrucelli, Mary B. Daly, Gerald L. Feldman. (2010) Hereditary breast and ovarian cancer due to mutations in BRCA1 and BRCA2. Genetics in Medicine 12:5, 245-259
CrossRef2
Henry T. Lynch, Murray Joseph Casey, Carrie L. Snyder, Chhanda Bewtra, Jane F. Lynch, Matthew Butts, Andrew K. Godwin. (2009) Hereditary ovarian carcinoma: Heterogeneity, molecular genetics, pathology, and management. Molecular Oncology 3:2, 97-137
CrossRef3
Christina S. Chu, Mark A. Morgan, Thomas C. Randall, Christina A. Bandera, Stephen C. Rubin. (2001) Survival of BRCA1 Negative Ovarian Cancer Patients Based on Family History. Gynecologic Oncology 83:1, 109-114
CrossRef4
G. Bruce Mann, Patrick I. Borgen. (1998) Breast cancer genes and the surgeon. Journal of Surgical Oncology 67:4, 267-274
CrossRef
