Vaginal bleeding and diethylstilbestrol exposure during pregnancy: Relationship to genital tract clear cell adenocarcinoma and vaginal adenosis in daughters Gerald B. Sharp, DrPH: and Philip Cole, MD, DrPHb Memphis, Tennessee, and Birmingham, Alabama Comparing 186 cancer cases and 1772 cancer-free controls and reexamining several previously published studies, we found genital tract clear cell adenocarcinoma and vaginal adenosis to be moderately, but nonsignificantly, associated with vaginal gestational bleeding when in utero diethylstilbestrol exposure was statistically controlled. Considering the prevalence of diethylstilbestrol exposure in the general population, the relative risks of vaginal clear cell adenocarcinoma for in utero exposure were 365.6 and 459.0 when vaginal bleeding did and did not occur during the index pregnancy, respectively. The relative risks of vaginal adenosis for such diethylstilbestrol exposure were 15.4 and 92.8, respectively, for these women. The strong associations between in utero diethylstilbestrol exposure and both vaginal adenosis and genital tract clear cell adenocarcinoma cannot be attributed to the occurrence of problem pregnancy. However, among daughters exposed to diethylstilbestrol, maternal vaginal bleeding during the index pregnancy does appear to be associated with reduced risks of vaginal adenosis and vaginal clear cell adenocarcinoma. (AM J OSSTET GVNECOL 1990;162:994-1001.)
Key words: Adenocarcinoma, vaginal neoplasms, diethylstilbestrol, pregnancy complications, epidemiology
The use of diethylstilbestrol by pregnant women increases the risk of vaginal or cervical clear cell adenocarcinoma in female offspring. 1• 3 The two initial casecontrol studies of vaginal clear cell adenocarcinoma and diethylstilbestrol exposure 1• 2 have been criticized for their failure to control for maternal history of problem pregnancy, and critics have continued to dispute the relationship between diethylstilbestrol exposure and vaginal dear cell adenocarcinoma. 4• 8 The speculation is that problem pregnancies, which have been linked to abnormal pregnancy outcomes, might be related to clear cell adenocarcinoma of the genital tract and, thus, that diethylstilbestrol might be related to this cancer only because it was frequently taken by women experiencing problem pregnancies. After an extensive review of the clear cell adenocar-
From the Department of Biostatistics and Epidemiology, College of Graduate Health Sciences, University of Tennessee, Memphis: and the Department of Epidemiology, School of Public Health, University of Alabama at Birmingham.' Supported by an Illinois Cancer Council grant under Grant No. 2 R18 CA20071 from the National Cancer Institute, National Institutes of Health and by Grant No. CA13148-17 from the National Cancer Institute. Received for publication August 18, 1989; revised November 28, 1989; accepted December 29, 1989. Reprint requests: Gerald B. Sharp, DrPH, Department of Biostatistics and Epidemiology, College of Graduate Health Sciences, University of Tennessee, Memphis, 877 Madison Ave., Memphis, TN 38163. 611119142 994
cinoma literature, McFarlane et al,7 concluded that the "existing evidence of the [diethylstilbestrol]lvaginal cancer relation is currently too weak for the causal role of [diethylstilbestrol] to be regarded as established." We address this conclusion by describing the relationships between vaginal and cervical clear cell adenocarcinoma and maternal diethylstilbestrol exposure, while controlling for two of the major indications for maternal diethylstilbestrol therapy. Ours is the first study to compare diethylstilbestrol-positive and -negative clear cell adenocarcinoma cases and controls since the initial case-control studies of vaginal and cervical clear cell adenocarcinoma. 1. 3 Furthermore, estimates of the relative risk of genital tract clear cell adenocarcinoma for maternal diethylstilbestrol exposure have not been previously reported. Vaginal adenosis (the presence of columnar epithelium in the vagina) is strongly associated with both maternal diethylstilbestrol exposure and clear cell adenocarcinoma. g • 10 In several instances vaginal clear cell adenocarcinoma has developed in areas that were previously sites of adenosis. ll • 12 Thus, it seems likely that if a mother's history of problem pregnancy increased her daughter'S risk of vaginal clear cell adenocarcinoma, such a history would also increase her risk of vaginal adenosis. In testing this hypothesis, we provide the first estimates of the relative prevalence of vaginal aden os is for diethylstilbestrol exposure adjusted for vaginal gestational bleeding.
Volume 162 Number 4
Material and methods The noncancer effects of in utero exposure to diethylstilbestrol were studied by the National Cooperative Diethylstilbestrol Adenosis (DES AD) Project, a multicenter research effort organized in the mid-1970s.!3 For the DESAD study, 221,245 obstetric records were reviewed at the Mayo and Gunderson clinics in Minnesota and at physicians' offices and clinics in Boston, Houston, and Los Angeles. Investigators followed and interviewed cohorts of women whose mothers either did or did not take diethylstilbestrol during the pregnancies resulting in their births. One diethylstilbestrol-negative group consisted of sisters of diethylstilbestrol-positive subjects.!3.'5 For our investigation, 1772 cancer-free members of DESAD's record review groups were compared to genital tract clear cell adenocarcinoma cases available as of Feb. 28, 1987 by the Registry for Research on Hormonal Transplacental Carcinogenesis at the University of Chicago (Table I). The Registry has made an effort to include all women with this disease born after 1940, regardless of maternal diethylstilbestrol exposure history. As has been described, physicians, predominantly from the United States but from other countries as well, have submitted information both about such cases and about the mother's index pregnancy history. '7. '8 Registry workers have questioned subjects, their mothers, and other family members about these histories and have used all acquired information to corroborate the diagnosis of clear cell adenocarcinoma, to specify tumor origin site, and to determine the mother's history both of diethylstilbestrol exposure and of conditions associated with this exposure. We limited our study to 186 of the total 522 available cases for whom site of tumor origin was determined, written obstetric records documenting diethylstilbestrol exposure status were available, and the mother's history of bleeding during the index pregnancy could be determined (Table I). The site of tumor origin was vaginal for 117 of these cases, indeterminantly vaginal-cervical for 48, and cervical for 21. Information about maternal diethylstilbestrol exposure and pregnancy bleeding histories for 1772 DESAD subjects was obtained from Tilley et al. '5 All exogenous, nonsteroidal estrogens (including diethylstilbestrol, dienestrol, and hexestrol) were classified as diethylstilbestrol, and the exposure status of both cases and controls was based on written obstetric records. Subjects were classified as positive for bleeding if the mother remembered experiencing bleeding during the index pregnancy or if such bleeding was noted in prenatal records. Otherwise, the classification was negative. Relative risks of cervical and vaginal clear cell adenocarcinoma for maternal history of pregnancy bleeding were calculated after subjects were split into two
DES-related adenocarcinoma and adenosis
995
Table I. Study groups Association
Maternal bleeding and genital tract clear cell adenocarcinoma Diethylstilbestrol exposure and genital tract clear cell adenocarcinoma Diethylstilbestrol exposure and vaginal clear cell adenocarcinoma Diethylstilbestrol exposure and vagi~al adenosls
I
Cases
Controls
186 Registry cases
1772 DESAD participants"
186 Registry cases
Constructed control group*
8 cases'
32 matched controls'
37 neonates and fetuses 9
184 neonates and fetuses 9
*Control group based on prevalence of diethylstilbestrol exposure in general population!6 and prevalence of maternal bleeding in DESAD group.'5
groups based on their diethylstilbestrol exposure status. However, because our controls were selected for the DESAD study on the basis of their diethylstilbestrol exposure, it was not possible to simply exchange cell values to calculate the relative risk of clear cell adenocarcinoma for such exposure when adjusting for maternal bleeding history (Table I). Thus we assumed that we had a group of 1000 controls and that the proportion of diethylstilbestrol-positive controls in this group equaled the 1.5% of women born in the United States at the peak period of diethylstilbestrol usage in this country who are estimated to be diethylstilbestrolpositive.'6 We also assumed that the maternal bleeding status of the 15 controls classified as diethylstilbestrol positive and the 985 controls classified as diethylstilbestrol negative would be proportional to that of our original control group, taking into account exposure status. For example, because 90.5% of diethylstilbestrolnegative DESAD controls were negative for maternal bleeding, we assumed that the same percentage of our 985 hypothetically diethylstilbestrol-negative controls would also be negative. Relative risks of vaginal and cervical clear cell adenocarcinoma for diethylstilbestrol exposure were calculated using this constructed control group while dividing cases and controls into groups positive and negative for maternal bleeding. We were able to calculate a relative risk of vaginal cancer for maternal diethylstilbestrol exposure that was adjusted for pregnancy loss before the index pregnancy (an additional major indication for diethylstilbestrol therapy) by unmatched reanalysis of the initial casecontrol study of vaginal clear cell adenocarcinoma' (Table I). Although that was a matched study, pub-
996 Sharp and Cole
April 1990 Am J Obstet Gynecol
Table II. Maternal vaginal bleeding and diethylstilbestrol exposure status of subjects* Diethylstilbestrol Maternal bleeding
I
70 729
Vaginal casest DESAD controls:j: Relative risk 95% Confidence interval
Diethylstilbestrol negative
No maternal bleeding
Maternal bleeding
34 497 1.4 (0.9-2.1)
2 52
I
No maternal bleeding
II 494 1.7 (0.4-7.9)
*Crude relative risk = 2.0 (95% confidence interval 1.4 to 3.0); summary relative risk = 1.4 (95% confidence interval 0.9 to 2.1). tCases were restricted to those with written maternal obstetric records on file. :j:Data from Tilley et aJ. 15
Table III. Numbers of vaginal cancer cases and hypothetic controls according to maternal bleeding and diethylstilbestrol exposure status
No maternal bleeding, no diethylstilbestrol*
Maternal bleeding, no diethylstilbestrol
II 891 1.0
94 1.7
Casest Hypothetic controls:j: Relative risk
2
No maternal bleeding, diethylstilbestrol
Maternal bleeding, diethylstilbestrol
34 6 459.0
70 9 630.0
*Reference group. tCases restricted to those with tumors of vaginal origin for whom written maternal obstetric records were available. :j:Control group assumes that in a group of 1000 women, 1.5% or 15 would be diethylstilbestrol-positive. 16 Based on the data presented in Table II, 59% of the 15 would be positive and 41 % negative for a maternal history of bleeding during the index pregnancy. Similarly, 9.5% of the remaining 985 diethylstilbestrol-negative controls would be positive and 90.5% would be negative for maternal bleeding.
Table IV. Initial case-control study of vaginal clear cell adenocarcinoma: reanalysis* Diethylstilbestrol positive
Diethylstilbestrol negative
7 0
I 32
Cases Controls Crude relative risk 95% Confidence interval
00
(15.4- 00 )
*Data from Herbst et aJ. 1
lished data were insufficient to allow matched reanalysis. To estimate the relative prevalence of vaginal adenosis for both maternal diethylstilbestrol exposure and vaginal bleeding during pregnancy, we reanalyzed a Boston autopsy study published in 1979 that included 281 female neonates and fetuses conceived during the era when diethylstilbestrol was used for pregnancy support9 (Table I). These investigators reviewed maternal medical records to assess histories of both diethylstilbestrol exposure and bleeding during the relevant pregnancies, and they examined autopsy materials to determine if subjects had vaginal adenosis. Using their published data, we calculated both the relative preva-
lence of vaginal adenosis for diethylstilbestrol exposure, controlling for gestational vaginal bleeding, and the relative prevalence of vaginal adenosis for such bleeding, controlling for maternal diethylstilbestrol exposure. Results
The crude relative risk of vaginal clear cell adenocarcinoma for maternal bleeding was 2.0, with a testbased 95% confidence interval'9 that did not include the null value of 1.0. Controlling for in utero diethylstilbestrol exposure, the Mantel-Haenszel summary estimate of this relative risk20 was 1.4, with a 95% confidence interval (0.9 to 2.1) that did include the null value. Among diethylstilbestrol-positive subjects the relative risk of vaginal clear cell adenocarcinoma for maternal bleeding was 1.4, and among diethylstilbestrol-negative subjects the corresponding relative risk was 1.7, with both 95% confidence intervals including the null value (Table II). Maternal exposure to diethylstilbestrol was strongly associated with vaginal clear cell adenocarcinoma even when maternal history of bleeding during the index pregnancy was statistically controlled. Comparing our constructed control group with our group of vaginal clear cell adenocarcinoma cases, the crude relative risk
DES-related adenocarcinoma and adenosis
Volume 162 Number 4
997
Table V. Relative risk of vaginal cancer for diethylstilbestrol exposure adjusted for mother's history of problem pregnancy* High-risk pregnancyt
I
Diethylstilbestrol positive
Cases Controls Stratified relative risk 95% Confidence interval
7
Low-risk pregnancyt
Diethylstilbestrol negative
Diethylstilbestrol positive
o
o o
o
5
I
Diethylstilbestrol negative 1
27
00
Undefined
(3.4- 00 )
(?-?)
*Data from Herbst et al. 1 tPregnancies were classified as high risk when the mother experienced bleeding during the index pregnancy, pregnancy loss before it, or both; otherwise, pregnancies were classified as low risk.
Table VI. Numbers of fetuses and neonates with vaginal adenosis according to maternal bleeding and diethylstilbestrol exposure status* No maternal bleeding, no diethylstilbestrolt
With vaginal adenosis Without vaginal adenosis Relative prevalence 95% confidence interval
Maternal bleeding, no diethylstilbestrol
6 159 1.0
1 12
2.2 (0.3-19.0)
No maternal bleeding, diethylstilbestrol
21 6 92.8 (39.0-220.6)
Maternal bleeding, diethylstilbestrol
9 7 34.1 ( 13.2-88.2)
*Data from Johnson et al. 9 tReference group.
of vaginal clear cell adenocarcinoma for diethylstilbestrol exposure was 525.3. Stratum-specific relative risk estimates of vaginal clear cell adenocarcinoma for diethylstilbestrol exposure were 365.6 among subjects whose mothers experienced bleeding during the index pregnancy and 459.0 among subjects whose mothers did not (Table III). When subjects whose mothers had both index pregnancy bleeding and diethylstilbestrol exposure were compared with subjects whose mothers were negative for both factors, the relative risk of clear cell adenocarcinoma was 630.0. Because this relative risk would be expected to be 780.3 on the multiplicative scale if no antagonism were involved, vaginal bleeding appears to slightly mitigate the effect of maternal diethylstilbestrol exposure on the incidence of vaginal clear cell adenocarcinoma. However, this antagonism is not strong enough to be defined as such on the additive scale!' Comparing our hypothetic controls with the 21 cervical clear cell adenocarcinoma cases for whom written maternal obstetric records were available, the crude relative risk of clear cell adenocarcinoma for diethylstilbestrol exposure was 59.7. The stratum-specific relative risk of this clear cell adenocarcinoma for diethylstilbestrol exposure among subjects whose mothers experienced pregnancy bleeding was undefined because of a zero cell; it was 13.5 among subjects whose
mothers were negative for such bleeding. Comparing our hypothetic controls with the subjects who had tumors of indeterminant vaginal-cervical origin, relative risks of clear cell adenocarcinoma for diethylstilbestrol exposure were 146.2 among subjects whose mothers were positive for maternal bleeding and 297.0 among subjects whose mothers were negative. This mitigation by maternal bleeding of diethylstilbestrol's effect on the incidence of the tumor was similar to that found when vaginal clear cell adenocarcinoma cases alone were compared. Our reanalysis of the initial case-control study of diethylstilbestrol exposure and vaginal clear cell adenocarcinoma' is described in Tables IV and V. The lower bound of the Fisher-exact 95% confidence interval'2 for the relative risk of vaginal clear cell adenocarcinoma for diethylstilbestrol exposure was 15.4 (Table IV). Maternal diethylstilbestrol exposure was significantly associated with vaginal clear cell adenocarcinoma when analysis was restricted to subjects whose mothers experienced vaginal bleeding during the index pregnancy, loss in a pregnancy preceding it, or both (Table V). Among subjects positive for at least one of these indications for diethylstilbestrol therapy, the lower bound of the Fisher-exact 95% confidence interval for the relative risk of clear cell adenocarcinoma for diethylstilbestrol exposure was 3.4 (Table V).
998
Sharp and Cole
Examining the data of Johnson et ai.,9 we found that for subjects whose mothers took diethylstilbestrol, the relative prevalence of adenosis for maternal bleeding was 0.4. Among subjects whose mothers did not take diethylstilbestrol, this relative prevalence was 2.2 (Table VI); 95% confidence intervals for both relative prevalences included the null value. The relative prevalence of vaginal adenosis for diethylstilbestrol exposure was 15.4 (95% confidence interval 2.1 to 113.4) among offspring of mothers who experienced bleeding during the index pregnancy and 92.8 (95% confidence interval 39.0 to 220.6) among offspring whose mothers did not (Table VI). When subjects positive for both maternal bleeding and diethylstilbestrol exposure were compared with the reference group negative for both factors, the relative prevalence of adenosis declined to 34.1 (95% confidence interval 13.2 to 88.2) (Table VI). Thus our results show that maternal bleeding reduces diethylstilbestrol's effect on vaginal adenosis more than maternal bleeding reduces diethylstilbestrol's effect on clear cell adenocarcinoma. The adenosis-associated antagonism can be classified as such on both additive and multiplicative scales!' Comment
The moderate and generally nonsignificant associations between vaginal gestational bleeding and clear cell adenocarcinoma that we report may be somewhat underestimated. The written records of our DESAD controls came from a small number of physicians who volunteered for the original DESAD study and from the Mayo and Gunderson clinics; these records may be of higher quality than those of Registry subjects, which were submitted by a large number of physicians on request. As a result, records of controls may be more likely than those of cases to include information about vaginal bleeding during pregnancy and, therefore, to be positive for it. This conclusion is supported by the fact that although 9.5% of diethylstilbestrol-negative DESAD subjects had mothers who experienced such bleeding, other studies using retrospective data have reported less than 5% of term pregnancies to be accompanied by such bleeding. 2,.2' Our estimates of the relative risk of clear cell adenocarcinoma for diethylstilbestrol exposure are also conservative. Our assumption that 1.5% of our hypothetic controls would be diethylstilbestrol positive is based on births during the peak period of diethylstilbestrol usage in this country. Because many of our cases were born when diethylstilbestrol use was less prevalent, we have probably overestimated the number of hypothetic controls who would be diethylstilbestrol positive. Our application of unmatched analysis methods to the matched data from the initial study of vaginal clear cell adenocarcinoma, if it biased our relative risk
April 1990 Am J Obstet Gynecol
estimates, would be expected to cause them to be underestimated. 2 ' In all, 7.3% of the diethylstilbestrolnegative subjects in the autopsy study were classified as positive for vaginal gestational bleeding, which is comparable to findings in similar retrospective studies. Because such bleeding is not always reported by obstetric patients, this prevalence is likely underestimated, but because the study investigators were blinded to maternal history, any errors in recording bleeding histories should have occurred randomly. Thus while our reported relative prevalences of adenosis for maternal bleeding mayor may not be underestimated, they should not be overestimated. 25 Our finding of a moderate, but statistically nonsignificant, relationship between maternal history of bleeding during the index pregnancy and vaginal clear cell adenocarcinoma when diethylstilbestrol exposure was statistically controlled does not support the hypothesis suggested by several authors'- 7 that diethylstilbestrol exposure may only appear to be associated with clear cell adenocarcinoma because use of this drug is associated with a history of problem pregnancies. Adjusting for maternal history of bleeding during the index pregnancy (the primary indication for maternal diethylstilbestrol therapy; unpublished observations), the stratum-specific relative risks of vaginal clear cell adenocarcinoma for maternal diethylstilbestrol exposure are 365.6 and 459.0 for daughters whose mothers did and did not experience bleeding, respectively. Even if there were residual confounding, the moderate association of maternal bleeding and vaginal clear cell adenocarcinoma we report could not possibly account for these large, conservatively estimated relative risks of vaginal clear cell adenocarcinoma for diethylstilbestrol, which are among the highest relative risk estimates reported by any large-scale epidemiologic study to date. Our reanalysis of the initial case-control study of vaginal clear cell adenocarcinoma' demonstrates that maternal diethylstilbestrol exposure is significantly associated with vaginal clear cell adenocarcinoma even when analysis is restricted to subjects whose mothers experienced either bleeding during the index pregnancy, pregnancy loss before it, or both. Although we have recently found (unpublished observations) that bleeding in a prior pregnancy is also significantly associated with diethylstilbestrol therapy during pregnancy (and we were not able to control for this factor in this study), it is unlikely that such bleeding could confound the strong relationships of maternal diethylstilbestrol exposure with vaginal adenosis and genital tract clear cell adenocarcinoma. Our findings do not support the conclusion of MacFarlane et ai. 7 that the existing evidence is too weak for the causal role of diethylstilbestrol to be regarded as established. With relative risks of clear cell adenocarcinoma for
Volume 162 Number 4
diethylstilbestrol exposure ranging from 59.7 for cervical cases to 525.3 for vaginal cases and with the relative risk for mixed vaginal and cervical cases falling in between, it is clear that diethylstilbestrol exposure is more strongly associated with vaginal than with cervical clear cell adenocarcinoma. 18 Our crude relative risk of cervical clear cell adenocarcinoma for diethylstilbestrol exposure of 59.7 is considerably higher than the relative risk of 10.2 we calculated using the data of Noller et aI.' Those investigators compared diethylstilbestrol exposure histories of women with cervical clear cell adenocarcinoma who were treated, but not necessarily born, at Mayo Clinic with the diethylstilbestrol exposure histories of women who were born at this clinic. Because 8.9% of the mothers of women who delivered at this clinic from 1946 to 1954 took diethylstilbestrol' compared with 1.5% of women giving birth at a variety of U.S. hospitals when diethylstilbestrol use was most common, the drug appears to have been prescribed more frequently at the Mayo Clinic than at other hospitals during this time. If so, our use of the data of Noller et al. to calculate a relative risk of cervical clear cell adenocarcinoma for diethylstilbestrol exposure would be expected to underestimate the relative risk, and the higher relative risk estimate we report may be more accurate. Our findings show a moderate but consistent tendency for maternal bleeding to be associated with vaginal adenosis and with vaginal clear cell adenocarcinoma when maternal diethylstilbestrol exposure is controlled. These relationships are interesting in light of the possibility of underestimation and of the associations, similar in magnitude, that have been reported between maternal bleeding and congenital malformations in offspring. 26.3o This study depended entirely on the analysis of previously collected and, in some instances, previously published data. Although the causes of genital tract clear cell adenocarcinoma and vaginal adenosis remain of great scientific and historic interest, the rarity of these diseases makes studying their occurrence in young women in any other way difficult. However, although justified, our use of previously collected data imposed limitations in the way we performed our analyses. While subjects in the Registry and DESAD groups do have similar age and racial distributions, we were unable to control these potential confounders. In addition, although two previous comparisons of diethylstilbestrol-positive clear cell adenocarcinoma cases and diethylstilbestrol-positive controls have shown the timing of exposure during pregnancy to be an important risk factor for genital tract clear cell adenocarcinoma,'" 3. we were unable to control for when diethylstilbestrol therapy was begun. Because maternal bleeding during pregnancy is a risk
DES-related adenocarcinoma and adenosis
999
factor for low birth weight offspring,·,··7.29 it is tempting to speculate that hormone fluctuations during pregnancy might explain both the antagonism reported here and the previous finding that women weighing under 5.5 pounds at birth are at increased risk of genital tract clear cell adenocarcinoma.'" Several investigators have reported an association between maternal bleeding and decreased maternal levels of estradiol and progesterone during the first trimester of pregnancy."· 34 Savu and Nunez" suggest that at prescribed doses, diethylstilbestrol levels would have been high enough to disturb the equilibrium between proteinbound and free-circulating estrogens in mothers who took the drug, a situation resulting in increased levels of free estrogens that could cross the placenta. Thus a fetus might be doubly, and perhaps detrimentally, exposed to large amounts of easily transferred drug and to a non regulated excess of maternal estrogens. One might speculate, then, that less fetal damage would occur if maternal hormone levels were reduced in conjunction with maternal bleeding and, thus, that the risk of both adenosis and clear cell adenocarcinoma would be reduced when a diethylstilbestrol-positive mother experienced such bleeding. Because pregnancy bleeding and history of spontaneous abortion were primary but competing indications for diethylstilbestrol therapy, women taking diethylstilbestrol who experienced one condition would have been less likely to have experienced the other. Thus women who took diethylstilbestrol because of prior spontaneous abortions would have been more likely to have experienced higher first-trimester hormone levels than women who took the drug because of pregnancy bleeding. If the hormone hypothesis were true, a comparison of diethylstilbestrol-positive clear cell adenocarcinoma cases and controls should reveal a maternal history of spontaneous abortion to be a risk factor for clear cell adenocarcinoma, as has been reported.'· For similar reasons, this possible association between hormone levels during pregnancy and risk of vaginal adenosis and clear cell adenocarcinoma in offspring might partially account for the previous finding that a fall birth is a significant risk factor for genital tract clear cell adenocarcinoma when diethylstilbestrol-positive subjects are compared. 3• Female hormone levels have been reported to vary seasonally,36 and female fetuses undergoing in utero development when maternal hormone levels are at a seasonal peak may be at greater risk of adenosis and clear cell adenocarcinoma when diethylstilbestrol is taken than are fetuses undergoing intrauterine development at other times of the year. However, because the risk of both vaginal adenosis and genital tract clear cell adenocarcinoma are increased with early exposure to diethylstilbestrol during pregnancy,9. 31. 32 the most likely and least speculative
1000 Sharp and Cole
explanation for the antagonism we report is that an association exists between indication for diethylstilbestrol therapy and the time during pregnancy when therapy was begun. Women who experienced bleeding during pregnancy and were placed on diethylstilbestrol therapy may have started taking the drug later in the index pregnancy than did women who took diethylstilbestrol based on their prior pregnancy history. Although the role of maternal bleeding in causing vaginal adenosis and cervical and vaginal clear cell adenocarcinoma remains unclear, we believe we have conclusively shown that no scientific basis exists for the speculation that diethylstilbestrol is spuriously associated with the tumor because it was used to treat problem pregnancies. The present study demonstrates that when maternal bleeding is controlled, the relative risks of vaginal clear cell adenocarcinoma for in utero diethylstilbestrol exposure are among the highest ever reported in epidemiologic research. We thank Dr. Arthur L. Herbst at the University of Chicago for generously providing access to Registry data. REFERENCES 1. Herbst AL, Ulfelder H, Poskanzer DC. Adenocarcinoma of the vagina: association of maternal stilbestrol therapy with tumor appearance in young women. N Engl] Med 1971 ;284:878-81. 2. Greenwald P, Barlow JJ, Nasca PC, Burnett WS. Vaginal cancer after maternal treatment with synthetic estrogens. N Engl] Med 1971;285:390-2. 3. Noller KL, Decker DG, Lanier AP, Kurland LT. Clearcell adenocarcinoma of the cervix after maternal treatment with synthetic estrogens. Mayo Clin Proc 1972; 47:629-30. 4. Feinstein AR, Horwitz Rl. An algebraic analysis of biases due to exclusion, susceptibility, and protopathic prescription in case-control research.] Chronic Dis 1981 ;34:393403. 5. Mantel N. Some problems with investigations and epidemiologicstudies relating to pregnancy. Progr Clin Bioi Res 1985;163:45-8. 6. Horwitz RI, McFarlane M], Brennan TA, Feinstein AR. The role of susceptibility bias in epidemiologic research. Arch Intern Med 1985;145:909-12. 7. McFarlane M], Feinstein AR, Horwitz Rl. Diethylstilbestrol and clear cell vaginal carcinoma: reappraisal of the epidemiologic evidence. Am] Med 1986;81 :855-63. 8. Feinstein AR. Epidemiologic investigation. Science 1989;243: 1256. 9. Johnson LD, Driscoll SG, Hertig AT, Cole PT, Nickerson R]. Vaginal adenosis in stillborns and neonates exposed to diethylstilbestrol and steroidal estrogens and progestins. Obstet Gynecol 1979;53:671-9. 10. Scully RE, Welch WR. Pathology of the female genital tract after prenatal exposure to diethylstilbestrol. In: Herbst AL, Bern HA, eds. Developmental effects of diethylstilbestrol (DES) in pregnancy. New York: ThiemeStratton, 1981 :26-45. 11. Monaghan ]M, Sirisena LAW. Stilboestrol and vaginal clear-cell adenocarcinoma syndrome. Br Med] 1978; 1: 1588-90. 12. Anderson B, Watring WG, Edinger DD ]r, Small EC, Netland AT, Safaii H. Development of DES-associated clear-
April 1990 Am J Obstet Gynecol
13.
14. 15.
16.
17. 18.
19. 20. 21. 22. 23. 24. 25. 26. 27. 28.
29. 30. 31.
32.
33.
cell carcinoma: the importance of regular screening. Obstet Gynecol 1979;53:293-9. Labarthe D, Adam E, Noller KL, et al. Design and preliminary observations of National Cooperative Diethylstilbestrol Adenosis (DESAD) project. Obstet Gynecol 1978;51 :453-8. Nash S, Tilley BC, Kurland LT, et al. Identifying and tracing a population at risk: the DESAD project experience. Am] .Public Health 1983;73:253-9. Tilley BC, Barnes AB, Bergstralh E, et al. A comparison of pregnancy history recall and medical records: implications for retrospective studies. Am] Epidemiol 1985; 121 :269-81. Melnick S, Cole P, Anderson D, Herbst A. Rates and risks of diethylstilbestrol-related clear-cell adenocarcinoma of the vagina and cervix: an update. N Engl ] Med 1987;316:514-6. Herbst AL, Kurman R], Scully RE, Poskanzer DC. Clearcell adenocarcinoma of the genital tract in young females: registry report. N Engl] Med 1972;287: 1259-64. Herbst AL, Robboy S], Scully RE, Poskanzer DC. Clearcell adneocarcinoma of the vagina and cervix in girls: analysis of 170 Registry cases. AM] OBSTET GYNECOL 1974;119:713-24. Miettinen OS. Estimability and estimation in case-referent studies. Am] Epidemiol 1976;130:226-35. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease.] Nat! Cancer Inst 1959;22:719-48. Rothman K]. Modern epidemiology. Boston: Little, Brown, 1986:237-83,311-26. Gart]. The comparison of proportions: a review of significance tests, confidence intervals and adjustments for stratification. Rev Int Stat Inst 1971;39:148-69. Fedrick], Anderson ABM. Factors associated with spontaneous pre-term birth. Br ] Obstet Gynaecol 1976;83: 342-50. Funderburk S], Guthrie D, Meldrum D. Outcome of pregnancies complicated by early vaginal bleeding. Br] Obstet Gynaecol 1980;87:100-5. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. Belmont, Calif: Lifetime Learning Publications, 1982:220-41. Peckham CH. Uterine bleeding during pregnancy. I. When not followed by immediate termination of pregnancy. Obstet Gynecol 1970;35:937-41. South], Naldrett]. The effect of vaginal bleeding in early pregnancy on the infant born after the 28th week of pregnancy.] Obstet Gynaecol Br Commonw 1973;80:236-41. Ornoy A, Benady S, Kohen-Raz R, Russell A. Association between maternal bleeding during gestation and congenital anomalies in the offspring. AM] OBSTET GYNECOL 1976; 124:474-8. Berkowitz GS, Harlap S, Beck G], Freeman DH, Baras M. Early gestational bleeding and pregnancy outcome: a multivariable analysis. Int] EpidemioI1983;12:165-73. Strobino B, Pantel-Silverman]. Gestational vaginal bleeding and pregnancy outcome. Am ] Epidemiol 1989; 129:806-15. Herbst AL, Cole P, Norusis M], Welch WR, Scully RE. Epidemiologic aspects and factors related to survival in 384 cases of clear cell adenocarcinoma of the vagina and cervix. AM] OBSTET GYNECOL 1979; 135:876-83. Herbst AL, Anderson S, Hubby MM, Haenszel WM, Kaufman RH, Noller KL. Risk factors for the development of diethylstilbestrol-associated clear cell adenocarcinoma: a case-control study. AM ] OBSTET GYNECOL 1986; 154:814-22. Tulchinsky D, Hobel C]. Plasma human chorionic gonadotropin, estrone, estradiol, estriol, progesterone, and 17 alpha-hydroxy-progesterone in human pregnancy. IV. Early normal pregnancy. AM ] OBSTET GYNECOL 1973; 117:884-93.
DES-related adenocarcinoma and adenosis
Volume 162 Number 4
34. Nygren K, Johansson EDB, Wide L. Evaluation of the prognosis of threatened abortion from the peripheral plasma levels of progesterone, estradiol, and human chorionic gonadotropin. AM J OBSTET GYNECOL 1973; 116:916-22. 35. Savu L, Nunez EA. Diethylstilbestrol inhibits the estrogen-
binding activity of pregnancy plasma: possible role in DES-associated pathology. Clin Chern 1985;31:1409-10. 36. Hughes A, Jacobson HI, Wagner RK, Jungblut pw. Ovarian-independent fluctuations of estradiol receptor levels in mammalian tissues. Mol Cell Endocrinol 1976; 5:379-88.
Are choroid plexus cysts an indication for second-trimester amniocentesis? Beryl R. Benacerraf, MD, Bernard Harlow, PhD, and Fredric D. Frigoletto, Jr., MD Boston, Massachusetts Previous series that described fetuses with choroid plexus cysts have been too small to determine whether there is an association with trisomy 18 sufficiently high to warrant amniocentesis. To address this issue, we studied the incidence of choroid plexus cysts and other ultrasonographic abnormalities in 26 consecutive fetuses (13.5 to 36 weeks' gestation) with trisomy 18. Twenty of these 26 fetuses had major sonographic anomalies suggestive of aneuploidy. Seventeen of these 26 fetuses were 15 to 20 weeks and 5 of 17 (30%) had choroid plexus cysts. Six of our total 26 affected fetuses had no sonographic anomalies and therefore, on the basis of our data, 30% of these (1.8 fetuses) with trisomy 18 would have choroid plexus cysts without other findings. The incidence of choroid plexus cysts in all second-trimester fetuses (including normal fetuses and those with trisomy 18) is reportedly 1%. Given the known incidence of trisomy 18 (3/10,000), we calculated a total presumptive sample of 86,667 patients to yield our 26 fetuses wiht trisomy 18. Our hypothetical sample has 86,641 (86,667 - 26) fetuses without trisomy 18, 858 of which would have choroid plexus cysts. Thus there would be one fetus with trisomy 18 for every 477 normal fetuses with choroid plexus cysts with no other defect seen. If amniocentesis were done to seek trisomy 18 in all second-trimester fetuses with choroid plexus cysts, two normal fetuses would be lost for every one with trisomy 18 identified. (AM J OSSTET GVNECOL 1990;162:1001-6.)
Key words: Choroid plexus cysts, trisomy 18, ultrasound, cytogenetics Chudleigh et aLI first described the benign nature of choroid plexus cysts observed ultrasonographically in the second-trimester fetus. Subsequently many others have agreed that second-trimester choroid plexus cysts are usually without sequelae: most choroid plexus cysts regress by 24 weeks and the neonates are normal!-7 On occasion these cysts have been associated with aneuploidy, particularly trisomy 18. Because of this association, some investigators have recommended cytogenetic studies for second-trimester fetuses with choroid plexus cysts. "-12 The observations by Fitzsimmons et aI., 13 who studied the brains of 14 fetuses with trisomy 18 at autopsy and found a strong correlation between this abnormal karyotype and choroid plexus cysts, have From the Departments of Obstetrics and Gynecology and Radiology, Brigham and Women's Hospital, Harvard Medical School. Receivedforpublicationjune 29,1989; revised September 15,1989; accepted September 20, 1989. Reprint requests: Beryl R. Benacerraf, MD, Diagnostic Ultrasound Associates, 333 Longwood Ave., Boston, MA 02115.
611116823
fueled this controversy. To critically assess this debate, the incidence of choroid plexus cysts in secondtrimester fetuses with and without trisomy 18 as well as associated anomalies must be determined. Review of several series demonstrates considerable variation in the general incidence of second-trimester choroid plexus cysts from series to series (0.18%, 0.65%, 0.80%, and 2.5%).2.4· 5.H However, if the average of these four numbers is taken, then approximately 1% of all secondtrimester fetuses would be expected to have choroid plexus cysts, an estimate in agreement with the experience in our laboratory. Furthermore even though choroid plexus cysts are more commonly present in fetuses with trisomy 18, this is a rare chromosomal abnormality that has a known incidence of three in every 10,000 births, an important consideration in the decision to recommend amniocentesis. II Clearly, the multitude of small series describing two to 41 cases of choroid plexus cysts each is insufficient to determine whether amniocentesis is really indicated when these cysts are discovered sonographically.I-12 In
1001
Key words: Adenocarcinoma, vaginal neoplasms, diethylstilbestrol, pregnancy complications, epidemiology
The use of diethylstilbestrol by pregnant women increases the risk of vaginal or cervical clear cell adenocarcinoma in female offspring. 1• 3 The two initial casecontrol studies of vaginal clear cell adenocarcinoma and diethylstilbestrol exposure 1• 2 have been criticized for their failure to control for maternal history of problem pregnancy, and critics have continued to dispute the relationship between diethylstilbestrol exposure and vaginal dear cell adenocarcinoma. 4• 8 The speculation is that problem pregnancies, which have been linked to abnormal pregnancy outcomes, might be related to clear cell adenocarcinoma of the genital tract and, thus, that diethylstilbestrol might be related to this cancer only because it was frequently taken by women experiencing problem pregnancies. After an extensive review of the clear cell adenocar-
From the Department of Biostatistics and Epidemiology, College of Graduate Health Sciences, University of Tennessee, Memphis: and the Department of Epidemiology, School of Public Health, University of Alabama at Birmingham.' Supported by an Illinois Cancer Council grant under Grant No. 2 R18 CA20071 from the National Cancer Institute, National Institutes of Health and by Grant No. CA13148-17 from the National Cancer Institute. Received for publication August 18, 1989; revised November 28, 1989; accepted December 29, 1989. Reprint requests: Gerald B. Sharp, DrPH, Department of Biostatistics and Epidemiology, College of Graduate Health Sciences, University of Tennessee, Memphis, 877 Madison Ave., Memphis, TN 38163. 611119142 994
cinoma literature, McFarlane et al,7 concluded that the "existing evidence of the [diethylstilbestrol]lvaginal cancer relation is currently too weak for the causal role of [diethylstilbestrol] to be regarded as established." We address this conclusion by describing the relationships between vaginal and cervical clear cell adenocarcinoma and maternal diethylstilbestrol exposure, while controlling for two of the major indications for maternal diethylstilbestrol therapy. Ours is the first study to compare diethylstilbestrol-positive and -negative clear cell adenocarcinoma cases and controls since the initial case-control studies of vaginal and cervical clear cell adenocarcinoma. 1. 3 Furthermore, estimates of the relative risk of genital tract clear cell adenocarcinoma for maternal diethylstilbestrol exposure have not been previously reported. Vaginal adenosis (the presence of columnar epithelium in the vagina) is strongly associated with both maternal diethylstilbestrol exposure and clear cell adenocarcinoma. g • 10 In several instances vaginal clear cell adenocarcinoma has developed in areas that were previously sites of adenosis. ll • 12 Thus, it seems likely that if a mother's history of problem pregnancy increased her daughter'S risk of vaginal clear cell adenocarcinoma, such a history would also increase her risk of vaginal adenosis. In testing this hypothesis, we provide the first estimates of the relative prevalence of vaginal aden os is for diethylstilbestrol exposure adjusted for vaginal gestational bleeding.
Volume 162 Number 4
Material and methods The noncancer effects of in utero exposure to diethylstilbestrol were studied by the National Cooperative Diethylstilbestrol Adenosis (DES AD) Project, a multicenter research effort organized in the mid-1970s.!3 For the DESAD study, 221,245 obstetric records were reviewed at the Mayo and Gunderson clinics in Minnesota and at physicians' offices and clinics in Boston, Houston, and Los Angeles. Investigators followed and interviewed cohorts of women whose mothers either did or did not take diethylstilbestrol during the pregnancies resulting in their births. One diethylstilbestrol-negative group consisted of sisters of diethylstilbestrol-positive subjects.!3.'5 For our investigation, 1772 cancer-free members of DESAD's record review groups were compared to genital tract clear cell adenocarcinoma cases available as of Feb. 28, 1987 by the Registry for Research on Hormonal Transplacental Carcinogenesis at the University of Chicago (Table I). The Registry has made an effort to include all women with this disease born after 1940, regardless of maternal diethylstilbestrol exposure history. As has been described, physicians, predominantly from the United States but from other countries as well, have submitted information both about such cases and about the mother's index pregnancy history. '7. '8 Registry workers have questioned subjects, their mothers, and other family members about these histories and have used all acquired information to corroborate the diagnosis of clear cell adenocarcinoma, to specify tumor origin site, and to determine the mother's history both of diethylstilbestrol exposure and of conditions associated with this exposure. We limited our study to 186 of the total 522 available cases for whom site of tumor origin was determined, written obstetric records documenting diethylstilbestrol exposure status were available, and the mother's history of bleeding during the index pregnancy could be determined (Table I). The site of tumor origin was vaginal for 117 of these cases, indeterminantly vaginal-cervical for 48, and cervical for 21. Information about maternal diethylstilbestrol exposure and pregnancy bleeding histories for 1772 DESAD subjects was obtained from Tilley et al. '5 All exogenous, nonsteroidal estrogens (including diethylstilbestrol, dienestrol, and hexestrol) were classified as diethylstilbestrol, and the exposure status of both cases and controls was based on written obstetric records. Subjects were classified as positive for bleeding if the mother remembered experiencing bleeding during the index pregnancy or if such bleeding was noted in prenatal records. Otherwise, the classification was negative. Relative risks of cervical and vaginal clear cell adenocarcinoma for maternal history of pregnancy bleeding were calculated after subjects were split into two
DES-related adenocarcinoma and adenosis
995
Table I. Study groups Association
Maternal bleeding and genital tract clear cell adenocarcinoma Diethylstilbestrol exposure and genital tract clear cell adenocarcinoma Diethylstilbestrol exposure and vaginal clear cell adenocarcinoma Diethylstilbestrol exposure and vagi~al adenosls
I
Cases
Controls
186 Registry cases
1772 DESAD participants"
186 Registry cases
Constructed control group*
8 cases'
32 matched controls'
37 neonates and fetuses 9
184 neonates and fetuses 9
*Control group based on prevalence of diethylstilbestrol exposure in general population!6 and prevalence of maternal bleeding in DESAD group.'5
groups based on their diethylstilbestrol exposure status. However, because our controls were selected for the DESAD study on the basis of their diethylstilbestrol exposure, it was not possible to simply exchange cell values to calculate the relative risk of clear cell adenocarcinoma for such exposure when adjusting for maternal bleeding history (Table I). Thus we assumed that we had a group of 1000 controls and that the proportion of diethylstilbestrol-positive controls in this group equaled the 1.5% of women born in the United States at the peak period of diethylstilbestrol usage in this country who are estimated to be diethylstilbestrolpositive.'6 We also assumed that the maternal bleeding status of the 15 controls classified as diethylstilbestrol positive and the 985 controls classified as diethylstilbestrol negative would be proportional to that of our original control group, taking into account exposure status. For example, because 90.5% of diethylstilbestrolnegative DESAD controls were negative for maternal bleeding, we assumed that the same percentage of our 985 hypothetically diethylstilbestrol-negative controls would also be negative. Relative risks of vaginal and cervical clear cell adenocarcinoma for diethylstilbestrol exposure were calculated using this constructed control group while dividing cases and controls into groups positive and negative for maternal bleeding. We were able to calculate a relative risk of vaginal cancer for maternal diethylstilbestrol exposure that was adjusted for pregnancy loss before the index pregnancy (an additional major indication for diethylstilbestrol therapy) by unmatched reanalysis of the initial casecontrol study of vaginal clear cell adenocarcinoma' (Table I). Although that was a matched study, pub-
996 Sharp and Cole
April 1990 Am J Obstet Gynecol
Table II. Maternal vaginal bleeding and diethylstilbestrol exposure status of subjects* Diethylstilbestrol Maternal bleeding
I
70 729
Vaginal casest DESAD controls:j: Relative risk 95% Confidence interval
Diethylstilbestrol negative
No maternal bleeding
Maternal bleeding
34 497 1.4 (0.9-2.1)
2 52
I
No maternal bleeding
II 494 1.7 (0.4-7.9)
*Crude relative risk = 2.0 (95% confidence interval 1.4 to 3.0); summary relative risk = 1.4 (95% confidence interval 0.9 to 2.1). tCases were restricted to those with written maternal obstetric records on file. :j:Data from Tilley et aJ. 15
Table III. Numbers of vaginal cancer cases and hypothetic controls according to maternal bleeding and diethylstilbestrol exposure status
No maternal bleeding, no diethylstilbestrol*
Maternal bleeding, no diethylstilbestrol
II 891 1.0
94 1.7
Casest Hypothetic controls:j: Relative risk
2
No maternal bleeding, diethylstilbestrol
Maternal bleeding, diethylstilbestrol
34 6 459.0
70 9 630.0
*Reference group. tCases restricted to those with tumors of vaginal origin for whom written maternal obstetric records were available. :j:Control group assumes that in a group of 1000 women, 1.5% or 15 would be diethylstilbestrol-positive. 16 Based on the data presented in Table II, 59% of the 15 would be positive and 41 % negative for a maternal history of bleeding during the index pregnancy. Similarly, 9.5% of the remaining 985 diethylstilbestrol-negative controls would be positive and 90.5% would be negative for maternal bleeding.
Table IV. Initial case-control study of vaginal clear cell adenocarcinoma: reanalysis* Diethylstilbestrol positive
Diethylstilbestrol negative
7 0
I 32
Cases Controls Crude relative risk 95% Confidence interval
00
(15.4- 00 )
*Data from Herbst et aJ. 1
lished data were insufficient to allow matched reanalysis. To estimate the relative prevalence of vaginal adenosis for both maternal diethylstilbestrol exposure and vaginal bleeding during pregnancy, we reanalyzed a Boston autopsy study published in 1979 that included 281 female neonates and fetuses conceived during the era when diethylstilbestrol was used for pregnancy support9 (Table I). These investigators reviewed maternal medical records to assess histories of both diethylstilbestrol exposure and bleeding during the relevant pregnancies, and they examined autopsy materials to determine if subjects had vaginal adenosis. Using their published data, we calculated both the relative preva-
lence of vaginal adenosis for diethylstilbestrol exposure, controlling for gestational vaginal bleeding, and the relative prevalence of vaginal adenosis for such bleeding, controlling for maternal diethylstilbestrol exposure. Results
The crude relative risk of vaginal clear cell adenocarcinoma for maternal bleeding was 2.0, with a testbased 95% confidence interval'9 that did not include the null value of 1.0. Controlling for in utero diethylstilbestrol exposure, the Mantel-Haenszel summary estimate of this relative risk20 was 1.4, with a 95% confidence interval (0.9 to 2.1) that did include the null value. Among diethylstilbestrol-positive subjects the relative risk of vaginal clear cell adenocarcinoma for maternal bleeding was 1.4, and among diethylstilbestrol-negative subjects the corresponding relative risk was 1.7, with both 95% confidence intervals including the null value (Table II). Maternal exposure to diethylstilbestrol was strongly associated with vaginal clear cell adenocarcinoma even when maternal history of bleeding during the index pregnancy was statistically controlled. Comparing our constructed control group with our group of vaginal clear cell adenocarcinoma cases, the crude relative risk
DES-related adenocarcinoma and adenosis
Volume 162 Number 4
997
Table V. Relative risk of vaginal cancer for diethylstilbestrol exposure adjusted for mother's history of problem pregnancy* High-risk pregnancyt
I
Diethylstilbestrol positive
Cases Controls Stratified relative risk 95% Confidence interval
7
Low-risk pregnancyt
Diethylstilbestrol negative
Diethylstilbestrol positive
o
o o
o
5
I
Diethylstilbestrol negative 1
27
00
Undefined
(3.4- 00 )
(?-?)
*Data from Herbst et al. 1 tPregnancies were classified as high risk when the mother experienced bleeding during the index pregnancy, pregnancy loss before it, or both; otherwise, pregnancies were classified as low risk.
Table VI. Numbers of fetuses and neonates with vaginal adenosis according to maternal bleeding and diethylstilbestrol exposure status* No maternal bleeding, no diethylstilbestrolt
With vaginal adenosis Without vaginal adenosis Relative prevalence 95% confidence interval
Maternal bleeding, no diethylstilbestrol
6 159 1.0
1 12
2.2 (0.3-19.0)
No maternal bleeding, diethylstilbestrol
21 6 92.8 (39.0-220.6)
Maternal bleeding, diethylstilbestrol
9 7 34.1 ( 13.2-88.2)
*Data from Johnson et al. 9 tReference group.
of vaginal clear cell adenocarcinoma for diethylstilbestrol exposure was 525.3. Stratum-specific relative risk estimates of vaginal clear cell adenocarcinoma for diethylstilbestrol exposure were 365.6 among subjects whose mothers experienced bleeding during the index pregnancy and 459.0 among subjects whose mothers did not (Table III). When subjects whose mothers had both index pregnancy bleeding and diethylstilbestrol exposure were compared with subjects whose mothers were negative for both factors, the relative risk of clear cell adenocarcinoma was 630.0. Because this relative risk would be expected to be 780.3 on the multiplicative scale if no antagonism were involved, vaginal bleeding appears to slightly mitigate the effect of maternal diethylstilbestrol exposure on the incidence of vaginal clear cell adenocarcinoma. However, this antagonism is not strong enough to be defined as such on the additive scale!' Comparing our hypothetic controls with the 21 cervical clear cell adenocarcinoma cases for whom written maternal obstetric records were available, the crude relative risk of clear cell adenocarcinoma for diethylstilbestrol exposure was 59.7. The stratum-specific relative risk of this clear cell adenocarcinoma for diethylstilbestrol exposure among subjects whose mothers experienced pregnancy bleeding was undefined because of a zero cell; it was 13.5 among subjects whose
mothers were negative for such bleeding. Comparing our hypothetic controls with the subjects who had tumors of indeterminant vaginal-cervical origin, relative risks of clear cell adenocarcinoma for diethylstilbestrol exposure were 146.2 among subjects whose mothers were positive for maternal bleeding and 297.0 among subjects whose mothers were negative. This mitigation by maternal bleeding of diethylstilbestrol's effect on the incidence of the tumor was similar to that found when vaginal clear cell adenocarcinoma cases alone were compared. Our reanalysis of the initial case-control study of diethylstilbestrol exposure and vaginal clear cell adenocarcinoma' is described in Tables IV and V. The lower bound of the Fisher-exact 95% confidence interval'2 for the relative risk of vaginal clear cell adenocarcinoma for diethylstilbestrol exposure was 15.4 (Table IV). Maternal diethylstilbestrol exposure was significantly associated with vaginal clear cell adenocarcinoma when analysis was restricted to subjects whose mothers experienced vaginal bleeding during the index pregnancy, loss in a pregnancy preceding it, or both (Table V). Among subjects positive for at least one of these indications for diethylstilbestrol therapy, the lower bound of the Fisher-exact 95% confidence interval for the relative risk of clear cell adenocarcinoma for diethylstilbestrol exposure was 3.4 (Table V).
998
Sharp and Cole
Examining the data of Johnson et ai.,9 we found that for subjects whose mothers took diethylstilbestrol, the relative prevalence of adenosis for maternal bleeding was 0.4. Among subjects whose mothers did not take diethylstilbestrol, this relative prevalence was 2.2 (Table VI); 95% confidence intervals for both relative prevalences included the null value. The relative prevalence of vaginal adenosis for diethylstilbestrol exposure was 15.4 (95% confidence interval 2.1 to 113.4) among offspring of mothers who experienced bleeding during the index pregnancy and 92.8 (95% confidence interval 39.0 to 220.6) among offspring whose mothers did not (Table VI). When subjects positive for both maternal bleeding and diethylstilbestrol exposure were compared with the reference group negative for both factors, the relative prevalence of adenosis declined to 34.1 (95% confidence interval 13.2 to 88.2) (Table VI). Thus our results show that maternal bleeding reduces diethylstilbestrol's effect on vaginal adenosis more than maternal bleeding reduces diethylstilbestrol's effect on clear cell adenocarcinoma. The adenosis-associated antagonism can be classified as such on both additive and multiplicative scales!' Comment
The moderate and generally nonsignificant associations between vaginal gestational bleeding and clear cell adenocarcinoma that we report may be somewhat underestimated. The written records of our DESAD controls came from a small number of physicians who volunteered for the original DESAD study and from the Mayo and Gunderson clinics; these records may be of higher quality than those of Registry subjects, which were submitted by a large number of physicians on request. As a result, records of controls may be more likely than those of cases to include information about vaginal bleeding during pregnancy and, therefore, to be positive for it. This conclusion is supported by the fact that although 9.5% of diethylstilbestrol-negative DESAD subjects had mothers who experienced such bleeding, other studies using retrospective data have reported less than 5% of term pregnancies to be accompanied by such bleeding. 2,.2' Our estimates of the relative risk of clear cell adenocarcinoma for diethylstilbestrol exposure are also conservative. Our assumption that 1.5% of our hypothetic controls would be diethylstilbestrol positive is based on births during the peak period of diethylstilbestrol usage in this country. Because many of our cases were born when diethylstilbestrol use was less prevalent, we have probably overestimated the number of hypothetic controls who would be diethylstilbestrol positive. Our application of unmatched analysis methods to the matched data from the initial study of vaginal clear cell adenocarcinoma, if it biased our relative risk
April 1990 Am J Obstet Gynecol
estimates, would be expected to cause them to be underestimated. 2 ' In all, 7.3% of the diethylstilbestrolnegative subjects in the autopsy study were classified as positive for vaginal gestational bleeding, which is comparable to findings in similar retrospective studies. Because such bleeding is not always reported by obstetric patients, this prevalence is likely underestimated, but because the study investigators were blinded to maternal history, any errors in recording bleeding histories should have occurred randomly. Thus while our reported relative prevalences of adenosis for maternal bleeding mayor may not be underestimated, they should not be overestimated. 25 Our finding of a moderate, but statistically nonsignificant, relationship between maternal history of bleeding during the index pregnancy and vaginal clear cell adenocarcinoma when diethylstilbestrol exposure was statistically controlled does not support the hypothesis suggested by several authors'- 7 that diethylstilbestrol exposure may only appear to be associated with clear cell adenocarcinoma because use of this drug is associated with a history of problem pregnancies. Adjusting for maternal history of bleeding during the index pregnancy (the primary indication for maternal diethylstilbestrol therapy; unpublished observations), the stratum-specific relative risks of vaginal clear cell adenocarcinoma for maternal diethylstilbestrol exposure are 365.6 and 459.0 for daughters whose mothers did and did not experience bleeding, respectively. Even if there were residual confounding, the moderate association of maternal bleeding and vaginal clear cell adenocarcinoma we report could not possibly account for these large, conservatively estimated relative risks of vaginal clear cell adenocarcinoma for diethylstilbestrol, which are among the highest relative risk estimates reported by any large-scale epidemiologic study to date. Our reanalysis of the initial case-control study of vaginal clear cell adenocarcinoma' demonstrates that maternal diethylstilbestrol exposure is significantly associated with vaginal clear cell adenocarcinoma even when analysis is restricted to subjects whose mothers experienced either bleeding during the index pregnancy, pregnancy loss before it, or both. Although we have recently found (unpublished observations) that bleeding in a prior pregnancy is also significantly associated with diethylstilbestrol therapy during pregnancy (and we were not able to control for this factor in this study), it is unlikely that such bleeding could confound the strong relationships of maternal diethylstilbestrol exposure with vaginal adenosis and genital tract clear cell adenocarcinoma. Our findings do not support the conclusion of MacFarlane et ai. 7 that the existing evidence is too weak for the causal role of diethylstilbestrol to be regarded as established. With relative risks of clear cell adenocarcinoma for
Volume 162 Number 4
diethylstilbestrol exposure ranging from 59.7 for cervical cases to 525.3 for vaginal cases and with the relative risk for mixed vaginal and cervical cases falling in between, it is clear that diethylstilbestrol exposure is more strongly associated with vaginal than with cervical clear cell adenocarcinoma. 18 Our crude relative risk of cervical clear cell adenocarcinoma for diethylstilbestrol exposure of 59.7 is considerably higher than the relative risk of 10.2 we calculated using the data of Noller et aI.' Those investigators compared diethylstilbestrol exposure histories of women with cervical clear cell adenocarcinoma who were treated, but not necessarily born, at Mayo Clinic with the diethylstilbestrol exposure histories of women who were born at this clinic. Because 8.9% of the mothers of women who delivered at this clinic from 1946 to 1954 took diethylstilbestrol' compared with 1.5% of women giving birth at a variety of U.S. hospitals when diethylstilbestrol use was most common, the drug appears to have been prescribed more frequently at the Mayo Clinic than at other hospitals during this time. If so, our use of the data of Noller et al. to calculate a relative risk of cervical clear cell adenocarcinoma for diethylstilbestrol exposure would be expected to underestimate the relative risk, and the higher relative risk estimate we report may be more accurate. Our findings show a moderate but consistent tendency for maternal bleeding to be associated with vaginal adenosis and with vaginal clear cell adenocarcinoma when maternal diethylstilbestrol exposure is controlled. These relationships are interesting in light of the possibility of underestimation and of the associations, similar in magnitude, that have been reported between maternal bleeding and congenital malformations in offspring. 26.3o This study depended entirely on the analysis of previously collected and, in some instances, previously published data. Although the causes of genital tract clear cell adenocarcinoma and vaginal adenosis remain of great scientific and historic interest, the rarity of these diseases makes studying their occurrence in young women in any other way difficult. However, although justified, our use of previously collected data imposed limitations in the way we performed our analyses. While subjects in the Registry and DESAD groups do have similar age and racial distributions, we were unable to control these potential confounders. In addition, although two previous comparisons of diethylstilbestrol-positive clear cell adenocarcinoma cases and diethylstilbestrol-positive controls have shown the timing of exposure during pregnancy to be an important risk factor for genital tract clear cell adenocarcinoma,'" 3. we were unable to control for when diethylstilbestrol therapy was begun. Because maternal bleeding during pregnancy is a risk
DES-related adenocarcinoma and adenosis
999
factor for low birth weight offspring,·,··7.29 it is tempting to speculate that hormone fluctuations during pregnancy might explain both the antagonism reported here and the previous finding that women weighing under 5.5 pounds at birth are at increased risk of genital tract clear cell adenocarcinoma.'" Several investigators have reported an association between maternal bleeding and decreased maternal levels of estradiol and progesterone during the first trimester of pregnancy."· 34 Savu and Nunez" suggest that at prescribed doses, diethylstilbestrol levels would have been high enough to disturb the equilibrium between proteinbound and free-circulating estrogens in mothers who took the drug, a situation resulting in increased levels of free estrogens that could cross the placenta. Thus a fetus might be doubly, and perhaps detrimentally, exposed to large amounts of easily transferred drug and to a non regulated excess of maternal estrogens. One might speculate, then, that less fetal damage would occur if maternal hormone levels were reduced in conjunction with maternal bleeding and, thus, that the risk of both adenosis and clear cell adenocarcinoma would be reduced when a diethylstilbestrol-positive mother experienced such bleeding. Because pregnancy bleeding and history of spontaneous abortion were primary but competing indications for diethylstilbestrol therapy, women taking diethylstilbestrol who experienced one condition would have been less likely to have experienced the other. Thus women who took diethylstilbestrol because of prior spontaneous abortions would have been more likely to have experienced higher first-trimester hormone levels than women who took the drug because of pregnancy bleeding. If the hormone hypothesis were true, a comparison of diethylstilbestrol-positive clear cell adenocarcinoma cases and controls should reveal a maternal history of spontaneous abortion to be a risk factor for clear cell adenocarcinoma, as has been reported.'· For similar reasons, this possible association between hormone levels during pregnancy and risk of vaginal adenosis and clear cell adenocarcinoma in offspring might partially account for the previous finding that a fall birth is a significant risk factor for genital tract clear cell adenocarcinoma when diethylstilbestrol-positive subjects are compared. 3• Female hormone levels have been reported to vary seasonally,36 and female fetuses undergoing in utero development when maternal hormone levels are at a seasonal peak may be at greater risk of adenosis and clear cell adenocarcinoma when diethylstilbestrol is taken than are fetuses undergoing intrauterine development at other times of the year. However, because the risk of both vaginal adenosis and genital tract clear cell adenocarcinoma are increased with early exposure to diethylstilbestrol during pregnancy,9. 31. 32 the most likely and least speculative
1000 Sharp and Cole
explanation for the antagonism we report is that an association exists between indication for diethylstilbestrol therapy and the time during pregnancy when therapy was begun. Women who experienced bleeding during pregnancy and were placed on diethylstilbestrol therapy may have started taking the drug later in the index pregnancy than did women who took diethylstilbestrol based on their prior pregnancy history. Although the role of maternal bleeding in causing vaginal adenosis and cervical and vaginal clear cell adenocarcinoma remains unclear, we believe we have conclusively shown that no scientific basis exists for the speculation that diethylstilbestrol is spuriously associated with the tumor because it was used to treat problem pregnancies. The present study demonstrates that when maternal bleeding is controlled, the relative risks of vaginal clear cell adenocarcinoma for in utero diethylstilbestrol exposure are among the highest ever reported in epidemiologic research. We thank Dr. Arthur L. Herbst at the University of Chicago for generously providing access to Registry data. REFERENCES 1. Herbst AL, Ulfelder H, Poskanzer DC. Adenocarcinoma of the vagina: association of maternal stilbestrol therapy with tumor appearance in young women. N Engl] Med 1971 ;284:878-81. 2. Greenwald P, Barlow JJ, Nasca PC, Burnett WS. Vaginal cancer after maternal treatment with synthetic estrogens. N Engl] Med 1971;285:390-2. 3. Noller KL, Decker DG, Lanier AP, Kurland LT. Clearcell adenocarcinoma of the cervix after maternal treatment with synthetic estrogens. Mayo Clin Proc 1972; 47:629-30. 4. Feinstein AR, Horwitz Rl. An algebraic analysis of biases due to exclusion, susceptibility, and protopathic prescription in case-control research.] Chronic Dis 1981 ;34:393403. 5. Mantel N. Some problems with investigations and epidemiologicstudies relating to pregnancy. Progr Clin Bioi Res 1985;163:45-8. 6. Horwitz RI, McFarlane M], Brennan TA, Feinstein AR. The role of susceptibility bias in epidemiologic research. Arch Intern Med 1985;145:909-12. 7. McFarlane M], Feinstein AR, Horwitz Rl. Diethylstilbestrol and clear cell vaginal carcinoma: reappraisal of the epidemiologic evidence. Am] Med 1986;81 :855-63. 8. Feinstein AR. Epidemiologic investigation. Science 1989;243: 1256. 9. Johnson LD, Driscoll SG, Hertig AT, Cole PT, Nickerson R]. Vaginal adenosis in stillborns and neonates exposed to diethylstilbestrol and steroidal estrogens and progestins. Obstet Gynecol 1979;53:671-9. 10. Scully RE, Welch WR. Pathology of the female genital tract after prenatal exposure to diethylstilbestrol. In: Herbst AL, Bern HA, eds. Developmental effects of diethylstilbestrol (DES) in pregnancy. New York: ThiemeStratton, 1981 :26-45. 11. Monaghan ]M, Sirisena LAW. Stilboestrol and vaginal clear-cell adenocarcinoma syndrome. Br Med] 1978; 1: 1588-90. 12. Anderson B, Watring WG, Edinger DD ]r, Small EC, Netland AT, Safaii H. Development of DES-associated clear-
April 1990 Am J Obstet Gynecol
13.
14. 15.
16.
17. 18.
19. 20. 21. 22. 23. 24. 25. 26. 27. 28.
29. 30. 31.
32.
33.
cell carcinoma: the importance of regular screening. Obstet Gynecol 1979;53:293-9. Labarthe D, Adam E, Noller KL, et al. Design and preliminary observations of National Cooperative Diethylstilbestrol Adenosis (DESAD) project. Obstet Gynecol 1978;51 :453-8. Nash S, Tilley BC, Kurland LT, et al. Identifying and tracing a population at risk: the DESAD project experience. Am] .Public Health 1983;73:253-9. Tilley BC, Barnes AB, Bergstralh E, et al. A comparison of pregnancy history recall and medical records: implications for retrospective studies. Am] Epidemiol 1985; 121 :269-81. Melnick S, Cole P, Anderson D, Herbst A. Rates and risks of diethylstilbestrol-related clear-cell adenocarcinoma of the vagina and cervix: an update. N Engl ] Med 1987;316:514-6. Herbst AL, Kurman R], Scully RE, Poskanzer DC. Clearcell adenocarcinoma of the genital tract in young females: registry report. N Engl] Med 1972;287: 1259-64. Herbst AL, Robboy S], Scully RE, Poskanzer DC. Clearcell adneocarcinoma of the vagina and cervix in girls: analysis of 170 Registry cases. AM] OBSTET GYNECOL 1974;119:713-24. Miettinen OS. Estimability and estimation in case-referent studies. Am] Epidemiol 1976;130:226-35. Mantel N, Haenszel W. Statistical aspects of the analysis of data from retrospective studies of disease.] Nat! Cancer Inst 1959;22:719-48. Rothman K]. Modern epidemiology. Boston: Little, Brown, 1986:237-83,311-26. Gart]. The comparison of proportions: a review of significance tests, confidence intervals and adjustments for stratification. Rev Int Stat Inst 1971;39:148-69. Fedrick], Anderson ABM. Factors associated with spontaneous pre-term birth. Br ] Obstet Gynaecol 1976;83: 342-50. Funderburk S], Guthrie D, Meldrum D. Outcome of pregnancies complicated by early vaginal bleeding. Br] Obstet Gynaecol 1980;87:100-5. Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. Belmont, Calif: Lifetime Learning Publications, 1982:220-41. Peckham CH. Uterine bleeding during pregnancy. I. When not followed by immediate termination of pregnancy. Obstet Gynecol 1970;35:937-41. South], Naldrett]. The effect of vaginal bleeding in early pregnancy on the infant born after the 28th week of pregnancy.] Obstet Gynaecol Br Commonw 1973;80:236-41. Ornoy A, Benady S, Kohen-Raz R, Russell A. Association between maternal bleeding during gestation and congenital anomalies in the offspring. AM] OBSTET GYNECOL 1976; 124:474-8. Berkowitz GS, Harlap S, Beck G], Freeman DH, Baras M. Early gestational bleeding and pregnancy outcome: a multivariable analysis. Int] EpidemioI1983;12:165-73. Strobino B, Pantel-Silverman]. Gestational vaginal bleeding and pregnancy outcome. Am ] Epidemiol 1989; 129:806-15. Herbst AL, Cole P, Norusis M], Welch WR, Scully RE. Epidemiologic aspects and factors related to survival in 384 cases of clear cell adenocarcinoma of the vagina and cervix. AM] OBSTET GYNECOL 1979; 135:876-83. Herbst AL, Anderson S, Hubby MM, Haenszel WM, Kaufman RH, Noller KL. Risk factors for the development of diethylstilbestrol-associated clear cell adenocarcinoma: a case-control study. AM ] OBSTET GYNECOL 1986; 154:814-22. Tulchinsky D, Hobel C]. Plasma human chorionic gonadotropin, estrone, estradiol, estriol, progesterone, and 17 alpha-hydroxy-progesterone in human pregnancy. IV. Early normal pregnancy. AM ] OBSTET GYNECOL 1973; 117:884-93.
DES-related adenocarcinoma and adenosis
Volume 162 Number 4
34. Nygren K, Johansson EDB, Wide L. Evaluation of the prognosis of threatened abortion from the peripheral plasma levels of progesterone, estradiol, and human chorionic gonadotropin. AM J OBSTET GYNECOL 1973; 116:916-22. 35. Savu L, Nunez EA. Diethylstilbestrol inhibits the estrogen-
binding activity of pregnancy plasma: possible role in DES-associated pathology. Clin Chern 1985;31:1409-10. 36. Hughes A, Jacobson HI, Wagner RK, Jungblut pw. Ovarian-independent fluctuations of estradiol receptor levels in mammalian tissues. Mol Cell Endocrinol 1976; 5:379-88.
Are choroid plexus cysts an indication for second-trimester amniocentesis? Beryl R. Benacerraf, MD, Bernard Harlow, PhD, and Fredric D. Frigoletto, Jr., MD Boston, Massachusetts Previous series that described fetuses with choroid plexus cysts have been too small to determine whether there is an association with trisomy 18 sufficiently high to warrant amniocentesis. To address this issue, we studied the incidence of choroid plexus cysts and other ultrasonographic abnormalities in 26 consecutive fetuses (13.5 to 36 weeks' gestation) with trisomy 18. Twenty of these 26 fetuses had major sonographic anomalies suggestive of aneuploidy. Seventeen of these 26 fetuses were 15 to 20 weeks and 5 of 17 (30%) had choroid plexus cysts. Six of our total 26 affected fetuses had no sonographic anomalies and therefore, on the basis of our data, 30% of these (1.8 fetuses) with trisomy 18 would have choroid plexus cysts without other findings. The incidence of choroid plexus cysts in all second-trimester fetuses (including normal fetuses and those with trisomy 18) is reportedly 1%. Given the known incidence of trisomy 18 (3/10,000), we calculated a total presumptive sample of 86,667 patients to yield our 26 fetuses wiht trisomy 18. Our hypothetical sample has 86,641 (86,667 - 26) fetuses without trisomy 18, 858 of which would have choroid plexus cysts. Thus there would be one fetus with trisomy 18 for every 477 normal fetuses with choroid plexus cysts with no other defect seen. If amniocentesis were done to seek trisomy 18 in all second-trimester fetuses with choroid plexus cysts, two normal fetuses would be lost for every one with trisomy 18 identified. (AM J OSSTET GVNECOL 1990;162:1001-6.)
Key words: Choroid plexus cysts, trisomy 18, ultrasound, cytogenetics Chudleigh et aLI first described the benign nature of choroid plexus cysts observed ultrasonographically in the second-trimester fetus. Subsequently many others have agreed that second-trimester choroid plexus cysts are usually without sequelae: most choroid plexus cysts regress by 24 weeks and the neonates are normal!-7 On occasion these cysts have been associated with aneuploidy, particularly trisomy 18. Because of this association, some investigators have recommended cytogenetic studies for second-trimester fetuses with choroid plexus cysts. "-12 The observations by Fitzsimmons et aI., 13 who studied the brains of 14 fetuses with trisomy 18 at autopsy and found a strong correlation between this abnormal karyotype and choroid plexus cysts, have From the Departments of Obstetrics and Gynecology and Radiology, Brigham and Women's Hospital, Harvard Medical School. Receivedforpublicationjune 29,1989; revised September 15,1989; accepted September 20, 1989. Reprint requests: Beryl R. Benacerraf, MD, Diagnostic Ultrasound Associates, 333 Longwood Ave., Boston, MA 02115.
611116823
fueled this controversy. To critically assess this debate, the incidence of choroid plexus cysts in secondtrimester fetuses with and without trisomy 18 as well as associated anomalies must be determined. Review of several series demonstrates considerable variation in the general incidence of second-trimester choroid plexus cysts from series to series (0.18%, 0.65%, 0.80%, and 2.5%).2.4· 5.H However, if the average of these four numbers is taken, then approximately 1% of all secondtrimester fetuses would be expected to have choroid plexus cysts, an estimate in agreement with the experience in our laboratory. Furthermore even though choroid plexus cysts are more commonly present in fetuses with trisomy 18, this is a rare chromosomal abnormality that has a known incidence of three in every 10,000 births, an important consideration in the decision to recommend amniocentesis. II Clearly, the multitude of small series describing two to 41 cases of choroid plexus cysts each is insufficient to determine whether amniocentesis is really indicated when these cysts are discovered sonographically.I-12 In
1001
