Abstract

BACKGROUND: Gestational trophoblastic disease refers to a spectrum of rare benign and malignant gynecologic disorders whose pathogenesis is not well understood. Recent studies from China and the United States have raised the hypothesis that long-term use of oral contraceptives before conception may increase the risk of gestational trophoblastic tumors. A multicenter case-control study of gestational trophoblastic tumors was undertaken to test this hypothesis. METHODS: Telephone interviews were conducted with 235 case patients, including 50 with gestational choriocarcinoma, and 413 control subjects matched on recentness of pregnancy, age at pregnancy, and area of residence. Relative risks (odds ratios) were computed by conditional logistic regression. Reported P values are two-sided. RESULTS: The relative risk estimate for ever having used oral contraceptives before the index pregnancy was 1.9 (95% confidence interval [CI] = 1.2-3.0), and the risk increased with duration of use ( P for trend = .05). The estimate was highest for women who used oral contraceptives during the cycle in which they became pregnant (relative risk = 4.0; 95% CI = 1.6-10), but there was no consistent pattern according to the time interval since last use. Separate analyses of choriocarcinoma and persistent mole yielded similar results, i.e., the relative risk estimates for oral contraceptive use were 2.2 (95% CI = 0.8-6.4) and 1.8 (95% CI = 1.0-3.0), respectively. Control for the number of sexual partners, which was independently associated with risk ( P for trend = .05), did not materially change the results. CONCLUSIONS: This study, the largest to date, indicates that long duration of oral contraceptive use before conception increases the risk of gestational trophoblastic tumors. These findings may provide clues to the pathogenesis of this rare disease. Changes in use of oral contraceptives are not warranted, however, because the incidence attributable to oral contraceptive use is very low.

Gestational trophoblastic diseases are uncommon disorders arising from gestational products. They comprise the following: hydatidiform mole, complete or partial; invasive mole; choriocarcinoma developing from molar pregnancy or, less often, normal pregnancy; and placental site trophoblastic tumor, the rarest. Hydatidiform moles, the commonest of these, are characterized by chromosomal abnormalities: the complete mole, although diploid, is androgenetic, bearing a homologous set of chromosomes of paternal origin; the partial mole is typically triploid. The risk of molar pregnancy appears to be increased if conception occurs at a relatively advanced or a very young age ( 1 - 5 ). The incidence appears to be highest in Asian countries, but contributory factors have not been identified ( 1 - 5 ).

In most cases, molar pregnancy is a self-limited process, completely resolved following evacuation of the hydatidiform mole. Occasionally, however, persistent growth of molar tissue or development of choriocarcinoma or placental site trophoblastic tumor ensues, necessitating additional therapy. The latter tumors may also arise during nonmolar pregnancies, following a similar course. The pathogenesis of gestational trophoblastic tumor is not understood. Findings from a study of 75 cases of gestational choriocarcinoma suggest that lower than normal estrogen levels may be associated with risk of choriocarcinoma, perhaps due to a disruption of normal ovulation ( 6 ).

A few studies ( 6 - 9 ) have raised the hypothesis that oral contraceptive use, before pregnancy, may increase the risk of gestational trophoblastic tumors. We investigated this hypothesis in a multicenter case-control study, the largest study of gestational trophoblastic tumors to date.

S ubjects and M ethods

Case Patients

Most women with gestational trophoblastic tumors are referred to specialized centers for treatment. Case patients in the present study were identified at eight such centers located in the United States: Brigham and Women's Hospital (Harvard University), Duke University, Brewer Trophoblastic Disease Center (Northwestern University), University of Minnesota, Yale University, The University of Texas, M. D. Anderson Cancer Center University of California at Irvine, and Long Beach Memorial Hospital. Possible case patients with gestational trophoblastic tumor were identified by personnel in the Division of Gynecologic Oncology at each center. Final selection for inclusion of case patients was made by the study pathologist (S. G. Driscoll) and the principal investigator (J. R. Palmer) after review of histologic slides, pathology reports, discharge summaries, and case notes. The research was approved by the Institutional Review Boards of each participating institution, and subjects gave informed consent for review of their medical records and histologic specimens.

Women who received chemotherapy following evacuation of hydatidiform moles were classified under persistent mole (postmolar gestational trophoblastic tumor). Criteria for chemotherapy following a mole were consistent across the participating centers, with only minor variations: following molar evacuation, patients were so treated if their serum human chorionic gonadotropin levels plateaued over a period of at least 3 consecutive weeks, if they became re-elevated (>20% increase) in the absence of a new pregnancy, or if there was evidence of metastatic disease. The diagnostic classification of a case by the local center was accepted unless the slides of the original tissue specimen showed no evidence of trophoblastic disease. Slide review was used to classify the original mole as complete, partial, or undetermined. Patients were classified under choriocarcinoma, rather than persistent mole, if the prior pregnancy was not molar or if the histologic tissue slides strongly suggested choriocarcinoma: that is, if there was histologic evidence of hyperplastic and dysplastic trophoblast unaccompanied by villi.

Case ascertainment and enrollment data are shown in Table 1. Of 359 potential case patients identified, 11% could not be located. Thirty-four potential case patients did not meet eligibility criteria of being at least 18 years of age, speaking English, and having a telephone. Six refused to participate and four had died.

We attempted to obtain histologic slides, pathology reports, and discharge summaries or other physician notes that gave a history of presenting illness for all 274 case patients interviewed. Medical records were obtained for all patients except 12 and slides were obtained for all but 52. Slides were reviewed by the study pathologist (S. G. Driscoll) without knowledge of exposure status. Of the 262 patients whose medical records were reviewed (including 210 with slide review), 16 (6%) did not meet case criteria: no persistent disease following molar evacuation (eight patients) or no histologic evidence of trophoblastic disease (hydropic abortus, four patients; ectopic pregnancy, four patients). Because 94% of patients met case criteria when records and slides were reviewed, we elected to include the 12 patients without records or slides, for a total of 258 case patients.

Twenty-three potential case patients, 17 aged 39 years or older and six who had been enrolled close to the end of the study, were later excluded because suitable control subjects were not identified. The excluded case patients were compared with included case patients of the same ages and were found to be similar with respect to oral contraceptive use, number of sexual partners, and years of education. Thus, there were 235 case patients in the final study group: 182 with persistent mole, 50 with choriocarcinoma, and three with placental site trophoblastic tumors.

Control Subjects

Control subjects matched to case patients on interval since last pregnancy, age at the time of that pregnancy, and area of residence were selected by random-digit dialing. We sought up to two control subjects per case patient. The matching criteria specified that the control subject had been pregnant within 5 years of the calendar year of the case patient index pregnancy (the molar pregnancy or, if there was no molar pregnancy, the pregnancy immediately preceding diagnosis of choriocarcinoma) and that her age at the end of that pregnancy be within 2 years of the age of the matched case patient at the end of her index pregnancy. To match on area of residence, a grid of 999 telephone numbers was created by using the area code and first four digits of the case's home telephone number and all permutations of 0-9 for the last three digits. The numbers were called in random order and household respondents were queried until a suitable match was found.

A total of 41 734 telephone numbers were called, with up to 12 attempts to reach each number. Of these numbers, 21 506 were unsuitable: 7989 represented business phones; 8234 were disconnected numbers; 1411 had been changed to a new number; 2959 had no answer after 12 attempts; in 905, the respondent did not speak English; and in eight, the respondent was hearing impaired. Of the 20 228 households contacted, 1102 (5%) refused to give information. From the remainder, 532 potential control subjects were identified. Thirty-five (6.6%) control subjects chose not to participate and 22 (4.1%) could not be recontacted for interview. Of the 475 potential control subjects interviewed, 62 were excluded from the final analyses, either because the case patients to which they had been matched were excluded or because they constituted third control subjects for case patients. Thus, the final control series comprised 413 women.

Data Collection

Data collection was carried out from January 1991 through June 1996, with telephone interviews conducted by nurse interviewers. The interview included questions on demographic factors, height and weight, cigarette smoking, usual diet, reproductive history including use of oral contraceptives and other hormonal medications, and sexual history. A detailed reproductive history was obtained by asking about each pregnancy and each episode of contraceptive use, with the aid of a special calendar. The interviewer prompted more complete recall by asking specifically about those intervals of time for which neither pregnancy nor contraceptive use was reported. Participants were asked their age at first sexual intercourse, number of sexual partners, and whether they had ever been diagnosed with any of a list of sexually transmitted infections (gonorrhea, syphilis, venereal warts, genital herpes, chlamydia, or trichomonas). Medical records were used only for confirmation of diagnosis and not for information on exposures. Medical records of the control subjects were not reviewed.

We counted as exposures only those events that occurred before the beginning of the index pregnancy. For case patients, the index pregnancy was defined as the molar pregnancy or, if a choriocarcinoma was not preceded by a molar gestation, the most recent pregnancy prior to diagnosis. For control subjects, the index pregnancy was the matched pregnancy; this was usually the control's most recent pregnancy. This definition of exposure was particularly important for oral contraceptive use, because most women are advised to use oral contraceptives to prevent pregnancy for at least a year after molar evacuation. Oral contraceptive use after the molar pregnancy was not counted as exposure.

Statistical Analysis

The matching was retained in all analyses, with relative risks (RRs) (odds ratios) and two-sided 95% confidence intervals (CIs) estimated by conditional logistic regression ( 10 ). Matched RRs were estimated from analyses that took into account all matching factors—age at end of index pregnancy, calendar year of index pregnancy, and region of residence. Multivariate RRs were estimated from conditional logistic models that controlled for race/ethnicity, duration of oral contraceptive use, body mass index (weight in kg divided by height in m 2 ), age at menarche, and lifetime number of sexual partners. Years of education, parents' education, religion at birth, and use of other hormonal medications were evaluated as potential confounders but were not included in the final models. Unadjusted and multivariate results were very similar, as shown in Tables 2 and 3 . Estimates given in the text are from the multivariate analyses. Two-sided P values for tests of trend were estimated by conditional logistic regression; for oral contraceptive use, the trend tests were carried out among users only. Analyses were carried out in the entire study group and separately for choriocarcinomas and persistent moles.

R esults

The median age at the end of the index pregnancy was 29 years for both the case patients and the control subjects. Seven percent of the case patients and six percent of the control subjects were under 20 years of age; 20% of the case patients and 15% of the control subjects were age 35 years and older. Case patients and control subjects were similar with respect to level of education and household income: 40% of the case patients and 39% of the control subjects were college graduates, while 7% and 5%, respectively, had not finished high school; and 46% of the case patients and 40% of the control subjects had a family income of at least $45 000, while 20% and 15%, respectively, had incomes of less than $15 000. Seventy-four percent of the case patients and 84% of the control subjects were white.

Oral contraceptive use before the index pregnancy was associated with an increased risk of gestational trophoblastic tumor (Table 2). The overall RR estimate was 1.9 (95% CI = 1.2-3.0), and the highest estimates were observed for the longest durations of use: for 3-6 years, 7-9 years, and 10 or more years of use, RR estimates were 1.9 (95% CI = 1.1-3.3), 2.3 (95% CI = 1.1-4.6), and 2.5 (95% CI = 1.2-5.3), respectively. A test of trend among users yielded a P value of .05. The association was present in each age category—less than 25, 25-34, and 35 or more years (data not shown). Control for factors related to sexual activity (number of sexual partners, age at first intercourse, and history of sexually transmitted disease) did not materially change the estimates, and the association was present among women who reported having had only one sexual partner (Table 2 ). The RR estimate was greatest for women who were using oral contraceptives during the cycle in which they conceived, but there was no consistent pattern according to number of months since last use among those women who had stopped use before conception (Table 2 ). Use of nonhormonal contraceptives during the cycle in which conception occurred was not associated with an increased risk (RR = 1.1 [95% CI = 0.7-1.9]).

Separate analyses of the 50 choriocarcinoma case patients and 182 cases of persistent mole gave similar results. In the choriocarcinoma analysis, the RR estimates were 2.2 (95% CI = 0.8-6.4) for any oral contraceptive use, 2.8 (95% CI = 0.7-10) for greater than or equal to 7 years of use, and 4.6 (95% CI = 1.0-22) for use during the cycle in which conception occurred. For persistent mole, the RR estimates were 1.8 (95% CI = 1.0-3.0) for any use, 2.6 (95% CI = 1.3-5.6) for greater than or equal to 7 years of use, and 3.2 (95% CI = 1.2-8.6) for use during the cycle in which conception occurred.

Factors related to sexual activity are considered in Table 3. Participants chose the range that best represented their total number of sexual partners before the index pregnancy: 1, 2-4, 5-9, 10-19, or 20 or more. The RR estimate for 10-19 sexual partners was 2.0 (95% CI = 1.1-3.6), with no further elevation for 20 or more partners. A similar elevation for 10 or more sexual partners was observed among the subgroup of women who had never used oral contraceptives before the index pregnancy. Early age at first intercourse and history of sexually transmitted disease (STD) (the latter analyzed as at least one, at least two, and each STD separately) were not associated with an increased risk. The point estimates were similar for choriocarcinoma and persistent mole: for choriocarcinoma the RR estimate for greater than or equal to 10 sexual partners was 1.7 (95% CI = 0.5-6.3), and for persistent mole, the comparable estimate was 2.1 (95% CI = 1.1-4.1).

D iscussion

This study provides strong evidence of a positive association of oral contraceptive use with risk of developing a gestational trophoblastic tumor. There was a statistically significant trend with increasing duration of use, and 10 or more years of use was associated with more than a doubling in risk. The results were consistent across age groups. Risk appeared to be highest for women who were taking oral contraceptives during the cycle in which they conceived. Although long-term users of oral contraceptives tended to have had more sexual partners, the association of oral contraceptive use with risk of trophoblastic tumor was independent of several measures of sexual activity. Whether oral contraceptive users are at a greater risk of hydatidiform mole or only of malignant trophoblastic disease cannot be determined from this study.

Having had 10 or more sexual partners before the index pregnancy was associated with a doubling in risk. The association was independent of oral contraceptive use. Early age at first intercourse and history of sexually transmitted diseases were not associated with risk.

Several previous studies ( 6-9,11-14 ) have assessed the risk of gestational trophoblastic disease in relation to antecedent oral contraceptive use. In a study from China, which included 142 patients with invasive moles and 189 whose moles were not invasive, the RR estimate for 4 or more years of oral contraceptive use was 2.6 (95% CI = 0.9-6.9) ( P for trend = .03) ( 9 ). Four other studies concerning the risk of noninvasive hydatidiform mole yielded inconsistent results: two found an increased risk ( 11 , 13 ) and two found no association ( 12 , 14 ).

By contrast, studies of malignant trophoblastic disease have indicated an increased risk associated with long-duration oral contraceptive use prior to conception ( 6 - 9 ). In a case-control study of 28 patients with choriocarcinoma identified in the Philippines, the RR estimate for ever use of oral contraceptives was 6.4 (95% CI = 2.2-18.8) ( 7 ). These findings were viewed with caution, however, because of small sample size, low participation rate, and the possibility of differential recall between case patients and control subjects. A similar association of oral contraceptive use and choriocarcinoma was observed in data from a large multipurpose case-control cancer study ( 8 ). A detailed analysis, which compared case patients with other hospital patients who had been pregnant at least once, yielded RR estimates for greater than or equal to 5 years of oral contraceptive use of 6.0 (95% CI = 1.3-28) for choriocarcinoma (23 cases) and 2.8 (95% CI = 0.9-8.8) for invasive mole (34 cases). A case-control study of 75 patients with choriocarcinoma reported an odds ratio of 2.00 (95% CI = 0.88-4.93) for ever use of oral contraceptives before the index pregnancy ( 6 ). The odds ratio increased to 3.0 for greater than or equal to 4 years of use, with a P for trend of .07 (Buckley J: personal communication, 1988).

A satisfactory biologic explanation for an association between oral contraceptives and the development of gestational trophoblastic tumor has not been established. Genetic studies of complete hydatidiform mole, the type most likely to result in persistent disease, indicate that the mole originates through androgenesis, the egg nucleus being either absent or inactivated ( 15 ). The descriptive epidemiology of trophoblastic disease suggests that ovarian dysfunction may be involved in the etiology: the incidence is highest for very young women, who have experienced menarche in the recent past and may not yet be having regular cycles, and for older women, who may also be experiencing ovarian dysfunction as they approach the menopause. Perhaps the common thread among factors that influence risk is interference with ovulation, either physiologically caused or induced by exogenous agents such as oral contraceptives. One possibility is that long-duration oral contraceptive use may damage the ova or interfere with meiosis and thus yield ova with absent or inactivated nuclei.

The present results also raise the possibility that a sexually transmitted infection plays an independent role in the cause of the disease. Although history of STD was not associated with risk, having had 10 or more sexual partners was associated with a doubling in risk. The STD information was based on self-reports about the most common STDs and it is likely that there was undiagnosed, underreported, and misreported disease. While clinical and laboratory studies have not pointed to an infectious cause for trophoblastic disease, the present findings suggest that investigations along these lines may be worthwhile.

Selection bias is unlikely to explain the current results because participation rates were high for both case patients and control subjects. Eighty-nine percent of the potential case patients were located; of those who met eligibility criteria, 96% participated. For control subjects, 95% of the households contacted provided sufficient information to determine whether an eligible control subject resided in that household, and 89% of eligible control subjects participated in the study. As with any study that used random-digit dialing for control selection, the true participation rate is unknown because some unknown proportion of households contacted may have been untruthful about whether there was an eligible resident.

For both case patients and control subjects, questionnaire items were asked with regard to the period before the index pregnancy. This was important because women who have a molar pregnancy are usually advised to use oral contraceptives for the next 12 months to prevent pregnancy ( 16 ). Thus, inclusion of data on oral contraceptive use following the index pregnancy would have biased the results in the direction of a positive association. Several previous studies ( 17 - 21 ) have indicated that oral contraceptive use following evacuation of a mole does not increase the risk of developing a gestational trophoblastic tumor from that gestation. So that control subjects would have approximately the same period of opportunity for oral contraceptive use, each control must have had a recent pregnancy at approximately the same age and calendar year as the matched case patient's index pregnancy. In fact, 31% of case patients and control subjects had their index pregnancies at the same year of age; of the remainder, approximately half of the control subjects had the index pregnancy at an age younger than the matched case patient and half at older ages.

The results are based on self-reported data. If case patients had more complete recall of their oral contraceptive use than control subjects, this would have biased the results in the direction of a positive association. However, a high proportion of control subjects (76%) reported use of oral contraceptives, and this figure is in line with the expected prevalence for this birth cohort of U.S. women ( 22 ). In addition, studies of self-reported oral contraceptive use have indicated that women report the duration of oral contraceptive use accurately ( 23 , 24 ). To record the number of sexual partners, we asked respondents to pick out of five choices the range of numbers that best reflected their experience. This approach was intended to desensitize the question. In addition, all interviews were conducted by telephone by a female nurse who was not known to the respondent, which may have allowed respondents to feel more comfortable answering that question.

The most likely potential confounding factor would have been the subject's age, which is strongly related to risk of trophoblastic disease. This factor was controlled by matching on the subject's age at index pregnancy (within 2 years) and by carrying out matched analyses. In addition, we controlled for three different measures of exposure to sexually transmitted infection and found the RR estimates for oral contraceptive use to be virtually unchanged. Similarly, oral contraceptive use did not appear to confound the relation between number of sexual partners and risk of trophoblastic tumor.

In conclusion, the present findings suggest that oral contraceptive use before pregnancy increases the risk of gestational trophoblastic tumors and that the risk is highest for long durations of use and use in the month in which conception occurs. Future laboratory research regarding the effects of oral contraceptives on the ovary may provide new insights as to the pathogenesis of gestational trophoblastic tumors. The results concerning number of sexual partners, although preliminary, raise the hypothesis of an infectious component in the cause of gestational trophoblastic disease.

Since the incidence of trophoblastic tumors is low in the United States where oral contraceptive use is quite common and higher in some developing countries where use is less common, there must be other predisposing factors. Even if there were a causal relation, the number of cases due to oral contraceptive use would be quite low. Based on an estimated incidence in the United States of one in 5000 pregnancies and a prevalence of use of 75%, the attributable risk of gestational trophoblastic tumors among oral contraceptive users is about one in 10 000 pregnancies. Thus, changes in clinical usage of oral contraceptives are not warranted.

Table 1.

Enrollment of women with gestational trophoblastic tumors


 
No. of patients
 
Potential case subjects 359 
 Unable to locate 41 
Potential case subjects contacted 318 
 Ineligible due to language 27 
 Ineligible due to age 
 Ineligible due to no telephone 
Eligible case subjects 284 
 Refused interview 
 Deceased 
Eligible case subjects interviewed 274 
 Excluded based on record or   slide review 16 
Final case series 258 
Final case series after exclusion   of 23 cases for whom no    control was identified 235 
 Persistent mole 182 
  Complete mole 96 
  Partial mole 51 
  Undetermined type 35 
 Choriocarcinoma 50 
  Following birth 24 
  Following complete mole 
  Following partial mole 
  Following mole, not otherwise    specified 14 
 Placental site trophoblastic tumor 

 
No. of patients
 
Potential case subjects 359 
 Unable to locate 41 
Potential case subjects contacted 318 
 Ineligible due to language 27 
 Ineligible due to age 
 Ineligible due to no telephone 
Eligible case subjects 284 
 Refused interview 
 Deceased 
Eligible case subjects interviewed 274 
 Excluded based on record or   slide review 16 
Final case series 258 
Final case series after exclusion   of 23 cases for whom no    control was identified 235 
 Persistent mole 182 
  Complete mole 96 
  Partial mole 51 
  Undetermined type 35 
 Choriocarcinoma 50 
  Following birth 24 
  Following complete mole 
  Following partial mole 
  Following mole, not otherwise    specified 14 
 Placental site trophoblastic tumor 

Table 2.

Risk of gestational trophoblastic tumor in relation to oral contraceptive (OC) use, according to duration of use and interval since last use


 
No. of case subjects
 
No. of control subjects
 
Matched relative risk estimate * (95% confidence interval)
 
Multivariate relative risk estimate (95% confidence interval)
 
Duration of OC use before index pregnancy 
 No use 36 98  1.0   1.0  
 Any use 199 315 1.8 (1.2-2.8) 1.9 (1.2-3.0) 
 <1 y 24 47 1.5 (0.8-2.8) 1.5 (0.7-2.9) 
 1-2 y 48 85 1.6 (0.9-2.7) 1.6 (0.9-2.9) 
 3-6 y 71 111 1.9 (1.1-3.1) 1.9 (1.1-3.3) 
 7-9 y 32 42 2.4 (1.3-4.7) 2.3 (1.1-4.6) 
 ⩾10 y 24 29 2.5 (1.2-5.1) 2.5 (1.2-5.3) 
 Unknown — — 
   P for trend = .03 § P for trend = .05 § 
Duration of OC use before index pregnancy among women reporting one sexual partner 
 No use 13 46  1.0   1.0  
 <1 y 17 1.7 (0.6-4.9) 1.7 (0.6-5.2) 
 1-2 y 14 28 1.9 (0.8-4.7) 1.9 (0.7-5.0) 
 3-6 y 18 23 2.4 (1.0-5.9) 2.6 (1.0-6.6) 
 ⩾7 y 10 3.7 (1.1-12.2) 3.2 (0.9-11.5) 
 Unknown — — 
   P for trend = .04 § P for trend = .04 § 
Interval from last use of OCs to index pregnancy 
 No use 36 98  1.0   1.0  
 Use during cycle in which conception occurred 16 11 3.8 (1.6-9.1) 4.0 (1.6-10.0) 
 <1 y 62 110 1.5 (0.9-2.6) 1.7 (1.0-2.9) 
 1-4 y 62 80 2.2 (1.3-3.7) 2.3 (1.3-4.1) 
 5-9 y 28 66 1.2 (0.7-2.3) 1.3 (0.7-2.5) 
 ⩾10 y 31 48 2.0 (1.0-3.9) 1.8 (0.9-3.7) 
   P for trend = .95 § P for trend =.83 § 

 
No. of case subjects
 
No. of control subjects
 
Matched relative risk estimate * (95% confidence interval)
 
Multivariate relative risk estimate (95% confidence interval)
 
Duration of OC use before index pregnancy 
 No use 36 98  1.0   1.0  
 Any use 199 315 1.8 (1.2-2.8) 1.9 (1.2-3.0) 
 <1 y 24 47 1.5 (0.8-2.8) 1.5 (0.7-2.9) 
 1-2 y 48 85 1.6 (0.9-2.7) 1.6 (0.9-2.9) 
 3-6 y 71 111 1.9 (1.1-3.1) 1.9 (1.1-3.3) 
 7-9 y 32 42 2.4 (1.3-4.7) 2.3 (1.1-4.6) 
 ⩾10 y 24 29 2.5 (1.2-5.1) 2.5 (1.2-5.3) 
 Unknown — — 
   P for trend = .03 § P for trend = .05 § 
Duration of OC use before index pregnancy among women reporting one sexual partner 
 No use 13 46  1.0   1.0  
 <1 y 17 1.7 (0.6-4.9) 1.7 (0.6-5.2) 
 1-2 y 14 28 1.9 (0.8-4.7) 1.9 (0.7-5.0) 
 3-6 y 18 23 2.4 (1.0-5.9) 2.6 (1.0-6.6) 
 ⩾7 y 10 3.7 (1.1-12.2) 3.2 (0.9-11.5) 
 Unknown — — 
   P for trend = .04 § P for trend = .04 § 
Interval from last use of OCs to index pregnancy 
 No use 36 98  1.0   1.0  
 Use during cycle in which conception occurred 16 11 3.8 (1.6-9.1) 4.0 (1.6-10.0) 
 <1 y 62 110 1.5 (0.9-2.6) 1.7 (1.0-2.9) 
 1-4 y 62 80 2.2 (1.3-3.7) 2.3 (1.3-4.1) 
 5-9 y 28 66 1.2 (0.7-2.3) 1.3 (0.7-2.5) 
 ⩾10 y 31 48 2.0 (1.0-3.9) 1.8 (0.9-3.7) 
   P for trend = .95 § P for trend =.83 § 
*

Relative risk estimates derived from unadjusted conditional logistic regression.

The following factors were controlled in the multivariate model by conditional logistic regression: age at menarche, body mass index (weight in kg divided by height in m 2 ), race, and number of sexual partners.

Reference category.

§

P values are two-sided.

Table 3.

Risk of gestational trophoblastic tumor in relation to sexual activity


 
No. of case subjects
 
No. of control subjects
 
Matched relative risk estimate * (95% confidence interval)
 
Multivariate relative risk estimate (95% confidence interval)
 
No. of sexual partners 
 1 62 124  1.0   1.0  
 2-4 80 140 1.1 (0.7-1.7) 1.1 (0.7-1.7) 
 5-9 44 87 1.1 (0.6-1.7) 1.0 (0.6-1.7) 
 10-19 37 43 2.0 (1.1-3.4) 2.0 (1.1-3.6) 
 ⩾20 11 17 1.6 (0.7-3.9) 2.0 (0.8-5.2) 
 Unknown — — 
   P for trend = .05 § P for trend = .05 § 
No. of sexual partners among women who had never used oral contraceptives before index pregnancy 
 1 13 46  1.0   1.0  
 2-4 12 26 1.7 (0.7-4.4) 1.6 (0.6-4.3) 
 5-9 13 1.3 (0.4-4.9) 1.6 (0.4-6.4) 
 ⩾10 13 1.9 (0.6-6.1) 1.9 (0.6-6.6) 
 Unknown — — 
   P for trend = .05 § P for trend = .03 § 
Age at first intercourse, y 
 <16 25 41  1.0   1.0  
 16, 17 84 127 1.1 (0.6-2.0) 1.5 (0.8-2.7) 
 18, 19 67 126 0.8 (0.5-1.5) 1.2 (0.6-2.2) 
 20-24 50 102 0.8 (0.4-1.5) 1.2 (0.6-2.5) 
 ⩾25 16 0.7 (0.2-1.9) 1.4 (0.4-4.5) 
 Unknown — — 
   P for trend = .85 § P for trend = .83 § 
History of sexually transmitted disease 
 No 182 344  1.0   1.0  
 Yes 53 69 1.5 (1.0-2.3) 1.3 (0.8-2.0) 

 
No. of case subjects
 
No. of control subjects
 
Matched relative risk estimate * (95% confidence interval)
 
Multivariate relative risk estimate (95% confidence interval)
 
No. of sexual partners 
 1 62 124  1.0   1.0  
 2-4 80 140 1.1 (0.7-1.7) 1.1 (0.7-1.7) 
 5-9 44 87 1.1 (0.6-1.7) 1.0 (0.6-1.7) 
 10-19 37 43 2.0 (1.1-3.4) 2.0 (1.1-3.6) 
 ⩾20 11 17 1.6 (0.7-3.9) 2.0 (0.8-5.2) 
 Unknown — — 
   P for trend = .05 § P for trend = .05 § 
No. of sexual partners among women who had never used oral contraceptives before index pregnancy 
 1 13 46  1.0   1.0  
 2-4 12 26 1.7 (0.7-4.4) 1.6 (0.6-4.3) 
 5-9 13 1.3 (0.4-4.9) 1.6 (0.4-6.4) 
 ⩾10 13 1.9 (0.6-6.1) 1.9 (0.6-6.6) 
 Unknown — — 
   P for trend = .05 § P for trend = .03 § 
Age at first intercourse, y 
 <16 25 41  1.0   1.0  
 16, 17 84 127 1.1 (0.6-2.0) 1.5 (0.8-2.7) 
 18, 19 67 126 0.8 (0.5-1.5) 1.2 (0.6-2.2) 
 20-24 50 102 0.8 (0.4-1.5) 1.2 (0.6-2.5) 
 ⩾25 16 0.7 (0.2-1.9) 1.4 (0.4-4.5) 
 Unknown — — 
   P for trend = .85 § P for trend = .83 § 
History of sexually transmitted disease 
 No 182 344  1.0   1.0  
 Yes 53 69 1.5 (1.0-2.3) 1.3 (0.8-2.0) 
*

Relative risk estimates derived from unadjusted conditional logistic regression.

The following factors were controlled in the multivariate model by conditional logistic regression: age at menarche, body mass index (weight in kg divided by height in m 2 ), race, and duration of oral contraceptive use.

Reference category.

§

P values are two-sided.

Supported by Public Health Service grant R01CA52223 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services.

L. Rosenberg currently receives financial support from Ortho-McNeil, a manufacturer of oral contraceptives.

References

(1)
Grimes DA. Epidemiology of gestational trophoblastic disease.
Am J Obstet Gynecol
 
1984
;
150
:
309
-18.
(2)
Bracken MB, Brinton LA, Hayashi K. Epidemiology of hydatidiform mole and choriocarcinoma.
Epidemiol Rev
 
1984
;
6
:
52
-75.
(3)
Bracken MB. Incidence and aetiology of hydatidiform mole: an epidemiologic review.
Br J Obstet Gynaecol
 
1987
;
94
:
1123
-35.
(4)
Hayashi K, Bracken MB, Freeman DH Jr, Hellenbrand K. Hydatidiform mole in the United States (1970-1977): a statistical and theoretical analysis.
Am J Epidemiol
 
1982
;
115
:
67
-77.
(5)
Palmer JR. Advances in the epidemiology of gestational trophoblastic disease.
J Reprod Med
 
1994
;
39
:
155
-62.
(6)
Buckley JD, Henderson BE, Morrow CP, Hammond CB, Kohorn EI, Austin DF. Casecontrol study of gestational choriocarcinoma.
Cancer Res
 
1988
;
48
:
1004
-10.
(7)
Baltazar JC. Epidemiological features of choriocarcinoma.
Bull World Health Organ
 
1976
;
54
:
523
-32.
(8)
Rosenberg L, Palmer JR, Shapiro S. Gestational trophoblastic disease and use of contraceptives [letter].
Am J Obstet Gynecol
 
1989
;
161
:
1087
-8.
(9)
Brinton LA, Wu BZ, Wang W, Ershow AG, Song HZ, Li JY, et al. Gestational trophoblastic disease: a case-control study from the People's Republic of China.
Am J Obstet Gynecol
 
1989
;
161
:
121
-7.
(10)
Breslow NE, Day NE, Halvorsen KT, Prentice RL, Sabai C. Estimation of multiple relative risk functions in matched case-control studies.
Am J Epidemiol
 
1978
;
108
:
299
-307.
(11)
Berkowitz RS, Bernstein MR, Harlow BL, Rice LW, Lage JM, Goldstein DP, et al. Case-control study of risk factors for partial molar pregnancy.
Am J Obstet Gynecol
 
1995
;
173
:
788
-94.
(12)
Messerli ML, Lilienfeld AM, Parmley T, Woodruff JD, Rosenshein NB. Risk factors for gestational trophoblastic neoplasia.
Am J Obstet Gynecol
 
1985
;
153
;
294
-300.
(13)
Berkowitz RS, Cramer DW, Bernstein MR, Cassells S, Driscoll SG, Goldstein DP. Risk factors for complete molar pregnancy from a case-control study.
Am J Obstet Gynecol
 
1985
:
152
:
1016
-20.
(14)
La Vecchia C, Franceschi S, Parazzini F, Fasoli M, DeCarli A, Gallus G, et al. Risk factors for gestational trophoblastic disease in Italy.
Am J Epidemiol
 
1985
;
121
:
457
-64.
(15)
Jacobs PA, Wilson CM, Sprenkle JA, Rosenshein NB, Migson BR. Mechanisms of origin of complete hydatidiform moles.
Nature
 
1980
;
286
:
714
-6.
(16)
Smith DB, O'Reilly SM, Newlands ES. Current approaches to diagnosis and treatment of gestational trophoblastic disease.
Curr Opin Obstet Gynecol
 
1993
;
5
:
84
-91.
(17)
Curry SL, Schlaerth JB, Kohorn EI, Boyce JB, Gore H, Twiggs LB, et al. Hormonal contraception and trophoblastic sequelae after hydatidiform mole (a Gynecology Oncology Group Study).
Am J Obstet Gynecol
 
1989
;
160
:
805
-11.
(18)
Berkowitz RS, Goldstein DP, Marean AR, Bernstein M. Oral contraceptives and post molar trophoblastic disease.
Obstet Gynecol
 
1981
;
58
:
474
-7.
(19)
Ho Yuen B, Burch P. Relationship of oral contraceptives and the intrauterine contraceptive devices to the regression of concentrations of the beta subunit of human chorionic gonadotropin and invasive complications after molar pregnancy.
Am J Obstet Gynecol
 
1983
;
145
:
214
-17.
(20)
Morrow P, Nakamura R, Schlaerth J, Gaddis O Jr, Eddy G. The influence of oral contraceptives on the postmolar human chorionic gonadotropin regression curve.
Am J Obstet Gynecol
 
1985
;
151
:
906
-14.
(21)
Deicas RE, Miller DS, Rademaker AW, Lurain JR. The role of contraception in the development of postmolar gestational trophoblastic tumor.
Obstet Gynecol
 
1991
;
78
:
221
-6.
(22)
Dawson DA. Trends in use of oral contraceptives—data from the 1987 National Health Interview Survey.
Fam Plann Perspect
 
1990
;
22
:
169
-72.
(23)
van Leeuwen FE, van Duijn CM, Camps MH, Kempers BA, Mentjens MF, Mulder HB, et al. Agreement between oral contraceptive users and prescribers: implications for case-control studies.
Contraception
 
1992
;
45
:
399
-408.
(24)
Nischan P, Ebeling K, Thomas DB, Hirsch U. Comparison of recalled and validated oral contraceptive histories.
Am J Epidemiol
 
1993
;
138
:
697
-703.