Cancer Incidence in a Cohort of Infertile Women

American Journal of Epidemiology Copyright © 1998 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved

Vol. 147, No. 11 Printed in U.S.A.

A BRIEF ORIGINAL CONTRIBUTION

Cancer Incidence in a Cohort of Infertile Women

Baruch Modan, 1 - 2 Elaine Ron,3 Liat Lerner-Geva,1 Tzvia Blumstein,1 Joseph Menczer,4 Jaron Rabinovici,5 Gabriel Oelsner,5 Laurence Freedman,1 Shlomo Mashiach,5 and Bruno Lunenfeld6

breast neoplasms; endometrial neoplasms; infertility; ovarian neoplasms; ovulation induction

Over the last three decades, several series of cases and numerous case reports of ovarian cancer's occurring in women treated with fertility drugs have been published (1-9). In addition, few epidemiologic studies observed a possible relation between ovulationinducing drugs and cancer incidence (10-15). In 1987, our group published a report (16) based on a follow-up of 2,575 infertile women. An excess of endometrial but not of ovarian cancer was noted. The current report presents an updated follow-up, with emphasis on the hormonal status of the patients and the treatment they received.

Received for publication July 14, 1997, and accepted for publication December 2, 1997. Abbreviations: Cl, confidence interval; SIR, standardized incidence ratio. 1 Department of Clinical Epidemiology, Chaim Sheba Medical Center, Tel Hashomer, Israel. 2 The Stanley Steyer Institute for Cancer Epidemiology and Research, Tel Aviv University, Tel Aviv, Israel. 3 Radiation Epidemiology Branch, National Cancer Institute, Bethesda, MD. 4 Department of Gynecology, Edith Wolfson Medical Center, Holon, Israel. 5 Department of Obstetrics and Gynecology, Chaim Sheba Medical Center, Tel Hashomer, Israel. 6 Bar Han University, Ramat Gan, Israel.

MATERIALS AND METHODS Study cohort

The study cohort consists of women who were diagnosed for infertility between January 1964 and December 1974 at the Chaim Sheba Medical Center, Tel Hashomer, Israel. Patients were identified by reviewing the medical records of the general gynecologic and specific infertility clinics. Included in the study cohort were patients presenting with primary or secondary infertility who visited the clinics more than once. To define the relation between hormonal status and cancer incidence, the study cohort was divided into four categories (16). Based on clinical criteria, patients were classified as follows: 1) women with unopposed estrogen, that is, adequate endogenous estrogen with progesterone deficiency, 2) women with both estrogen and progesterone deficiencies, 3) women with adequate endogenous estrogen and progesterone and normal ovulation with mechanical infertility or infertility due to the male partner (nonhormonal infertility), and 4) women with unknown hormonal status. Patients were considered to have had ovulation induction treatment if they received at least one cycle of human menopausal gonadotropins or clomiphene citrate or both.

1038

Downloaded from aje.oxfordjournals.org by guest on July 16, 2011

Among 2,496 infertile Israeli women treated between 1964 and 1974, 143 cancer cases were observed as compared with 116.1 expected (standardized incidence ratio (SIR) = 1.2,95% confidence interval (Cl) 1.0-1.5) through 1991. Site-specific analysis revealed 12 ovarian cancers versus 7.2 expected (SIR = 1.6, 95% Cl 0.8-2.9), 21 endometrial cancers versus 4.3 expected (SIR = 4.85, 95% Cl 3.0-7.4), and 59 breast cancers versus 46.6 expected (SIR = 1.3,95% Cl 0.96-1.6). Sensitivity analysis revealed that confounding was unlikely to explain the raised risk of endometrial cancer, but nulliparity might explain the increased risk of ovarian cancer. The excess of endometrial cancer was prominent among patients with normal estrogen production but progesterone deficiency (SIR = 9.4,95% Cl 5.0-16.0). The risk for ovarian cancer was similar among the total groups of treated and untreated patients (SIR = 1.7 vs. 1.6). The standardized incidence ratio for endometrial cancer was higher among the treated group than the untreated group, although not significantly. Treatment with ovulation-inducing drugs does not appear to increase the risk for ovarian cancer, but its role cannot be completely excluded. Am J Epidemiol 1998;147:1038-42.

Cancer and Infertility

Cancer identification

The study cohort file was linked to the Israel Cancer Registry to identify cancer cases through December 1991. The completeness of cancer ascertainment by the Israel Cancer Registry ranges from 90 to 95 percent, depending on the cancer site (17). Record linkage was accomplished by computer matching the study file with the Cancer Registry data file using patients' national identity numbers, names, and demographic data as scored matching variables. Cancer diagnoses were verified by reviewing the original histopathologic report for each case. Eight cases of carcinoma in situ were included (one larynx, one endometrium, two cervix, and four breast), since the Cancer Registry file included in situ cases in the population reference group as well. Statistical analysis

RESULTS

The 2,496 women in the study cohort contributed 54,413 women-years to the follow-up, with a mean of 21.4 (standard deviation, 3.9) years. The mean age at entry was 28.7 (standard deviation, 5.7) years, and the mean age at the end of follow-up was 50.0 (standard deviation, 7.0) years. During the follow-up period, 143 cases of cancer were identified (116.1 expected; standardized incidence ratio (SIR) =1.2, 95 percent confidence interval (CI) 1.0-1.5) (table 1). Endometrial cancer was markedly in excess, with 21 cases observed versus 4.3 expected (SIR = 4.8, 95 percent CI 3.0-7.4), while the risks for ovarian cancer (SIR = 1.6, 95 percent CI 0.8-2.9) and breast cancer (SIR = 1.3, 95 percent CI 0.96-1.6) were somewhat increased. Am J Epidemiol

Vol. 147, No. 11, 1998

TABLE 1. Observed and expected cancer among 2,496 women evaluated for infertility, Chaim Sheba Medical Center, Tel Hashomer, Israel, 1964-1991

All sites Ovary Endometrium Breast Melanoma Thyroid Otherf

Observed (no.)

Expected (no.)

143 12

116.1 7.2 4.3 46.6 7.0

21 59 8 8 35

5.4

41.1

SIR«

95% C I '

1.2 1.6

1.0-1.5 0.8-2.9 3.0-7.4 0.96-1.6 0.5-2.2 0.6-2.9 0.6-1.2

4.8 1.3 1.1 1.5 0.9

* SIR, standardized incidence ratio; CI, confidence interval. f Lip, one; tongue, one; mouth, one; nasopharynx, one; stomach, two; colon, one; rectum, six; pancreas, one; larynx, two; lung, one; bone, one; connective tissue, one; cervix, three; kidney, one; brain, three; adrenal, one; lymphosarcoma, two; Hodgkin's lymphoma, two; non-Hodgkin's lymphoma, two; leukemia, one; and unspecified, one.

Women treated for infertility differ from the general population by a number of cancer-related parameters. Therefore, evaluation of a possible confounding effect that might explain the observed elevated risks was performed using sensitivity analyses (table 2). Confounding by nulliparity, obesity, and contraceptive use or hysterectomy cannot fully explain the increased risk ratio (SIR = 4.8) for endometrial cancer. On the other hand, the elevated risk observed for ovarian cancer (SIR = 1.6) could be explained by a much higher prevalence of nulliparity or family history among the study cohort compared with the reference population. The former is clearly a possible explanation for the increase; a nulliparity rate of 0.5 in the study group versus 0.12 in the reference population would be sufficient. We evaluated the association between cancer incidence and hormonal status (table 3). In the group of 967 patients with unopposed estrogen, the observed number of endometrial cancers was 9.4 times the expectation (95 percent CI 5.0-16.0), while the risk for the other sites was not significantly elevated. In the group of women with nonhormonal infertility, only the ovarian cancer excess was significant. In an attempt to assess whether any increased risk could be linked to ovulation induction treatment, we calculated the observed and expected cancer incidence for treated and untreated patients (table 4). The risk for ovarian cancer was similar among the total groups of treated and untreated patients (SIR = 1 . 7 vs. 1.6, respectively). Six cases of ovarian cancer were observed in the group of patients treated with clomiphene citrate alone compared with 2.3 expected (SIR = 2.7, 95 percent CI 0.97-5.8), but this standardized incidence ratio was not significantly different from that of the untreated group (SIR = 1.6; chi-square = 0.34 on 1 df). The standardized incidence ratio for endometrial


Standardized incidence ratios were computed as a ratio of observed to expected cancers. Confidence intervals were estimated using the procedure described in the publication by Rothman and Boice (18). Expected numbers of cancer were computed by applying age, country of origin, and year-specific national cancer incidence rates to the relevant categories of personyears. Treated and untreated groups were compared using the chi-square test, and the results were adjusted for hormonal status using stratification methods. Methods of sensitivity analyses were used to control for confounders whose prevalence was unknown in the study cohort (19, 20). A sensitivity analysis will determine how much confounding would need to be present if certain confounders were to explain the differing outcomes (20). For each possible confounder, an agreeable reported relative risk was chosen from the literature.

1039

1040

Modan et al. TABLE 2.

Sensitivity analyses for the controlling of possible confounders

Cancer site

SIR*

Endometrium

Ovary

4.8

1.6

Reported relative riskt

Risk (actor

Nulliparity Obesity Oral contraceptives Hysterectomy

3 4 0.5 0.3

Nulliparity Family history Oral contraceptives Tubal ligation

3

5 0.4 0.6

Population prevalence^

Assumed study prevalence§

Expected

SIRfl

0.12 0.10 0.33 0.06

1 1 0 0

2.4 3.0 1.2

0.12 0.005 0.33 0.04

1 1

2.4

1.04

4.9 1.24 1.02

0 0

* SIR, standardized incidence ratio. t References 28-34. $ The population prevalence among Israeli controls (matched for age), according to the World Health Organization Collaborative Study on Neoplasia and Steroid Contraceptives (35). § Assumed as extreme to demonstrate the maximum effect that confounding on this factor would have on the SIR. H The SIR that one would expect to observe if there were no raised risks in the cohort except that due to the stated prevalence of the risk factor. Expected SIR (for relative risk >1)

study prevalence + 1/(reported relative risk - 1) population prevalence + 1/(reported relative risk - 1) =

study prevalence - 1/(1 - reported relative risk) population prevalence -1/(1 - reported relative risk)

TABLE 3. Observed and expected cancers by levels of estrogen and progesterone, Chaim Sheba Medical Center, Tel Hashomer, Israel, 1964-1991 Hormonal status

Cancer site

Observed (no.)

Expected (no.)

Estrogen (+)/progesterone (-) (n = 967)

All Ovary Endometrium Breast Other

58 2 13 23 20

39.9 2.5 1.4 16.3


2 0 0 0 2

7.2 0.5

Estrogen (+)/progesterone (+) (n=694)


Unknown (n = 706)


Estrogen (-)/progesterone (-) (n=129)

19.7

SIR*

95%CI»

1.4

1.1-1.9 0.1-2.9 5.0-16.0 0.9-2.1 0.6-1.6

0.8 9.4 1.4 1.0 0.3

0.03-1.0

0.3 3.1 3.3

0.6

0.1-2.2

36 6 4 16 10

35.5 2.2 1.4 14.0 17.9

1.0 2.7 2.8 1.1 0.6

0.7-1.4 1.0-6.0 0.8-7.2 0.7-1.9 0.3-1.0

47 4 4

33.5 2.0 1.2 13.5 12.6

1.4 1.9 3.2 1.5 1.5

1.0-1.9 0.5-4.8 0.9-8.3 0.9-2.3 0.9-2.4

20 19

* SIR, standardized incidence ratio; Cl, confidence interval.

cancer was twice as high among the group treated with ovulation induction than among the untreated (SIR = 6.8 vs. 3.3; ratio = 2.1), although not significantly different (chi-square =1.94 on 1 df). Moreover, some

of this difference is mostly explained by the patients' hormonal status as shown by stratified analysis (adjusted SIR ratio =1.4 for the proportion of women with unopposed estrogen in the treated vs. untreated group). Am J Epidemiol

Vol. 147, No. 11, 1998


Expected SIR (for relative risk