A reanalysis of the Cu-7 intrauterine contraceptive device clinical trial and the incidence of pelvic inflammatory disease: A paradigm for assessing intrauterine contraceptive device safety

Roy and Azen

later. The point about using the femur length is interesting and may be of value. However, the ultrasonographer must know what formula is being used in these projections that are taken off the machine and plotted on a curve to obtain a fetal weight estimation. Again, this goes back to the original point that it may be better to choose a single parameter and use that as a predictor of weight rather than using a formula reduction. To address Dr. Parer's questions, as in any clinical practice, this was a mixed population where some

June 1994 Am J Obstet Gynecol

babies were growing normally and others were growing abnormally. However, I suppose a baby that was dropping from the 30th to the 20th to the 10th percentile and then was delivered at the 10th percentile may represent an error. It might be reasonable to take these growth-retarded babies out and observe than on three occasions. Then you would be more likely to get the correct result compared with a single observation. However, even doing that may not put us back in business in terms of repetitive scanning.

A reanalysis of the Cu-7 intrauterine contraceptive device clinical trial and the incidence of pelvic inflammatory disease: A paradigm for assessing intrauterine contraceptive device safety Subir Roy, MD, and Colleen Azen, MS Los Angeles, California OBJECTIVE: We calculated and compared the incidence of pelvic inflammatory disease in a 10% random sample of the Cu-7 intrauterine contraceptive device (G.D. Searle & Co., Skokie, III.) clinical trial with the rates reported to the Food and Drug Administration and those in subsequent trials published in the world literature. STUDY DESIGN: A 10% random sample of the Cu-7 clinical trial was examined because calculations had demonstrated this random sample to be sufficient in size (n = 1614) to detect a difference in rates of pelvic inflammatory disease from those reported to the Food and Drug Administration. An audit of a subset of the patient files, compared with the original files in Skokie, Illinois, confirmed that the files available for analysiS were complete. Standard definitions were used to identify cases of pelvic inflammatory disease and to calculate rates of pelvic inflammatory disease. The world literature on Cu-7 clinical trials was reviewed. RESULTS: The calculated crude and Pearl index rates of pelvic inflammatory disease were consistent with those rates previously reported to the Food and Drug Administration and published in the medical literature. Life-table pelvic inflammatory disease rates were not different between nulliparous and parous women and pelvic inflammatory disease did not differ from basal annual rates in fecund women. CONCLUSION: On the basis of the analysis of this 10% sample, the pelvic inflammatory disease patient rates reported to the Food and Drug Administration for the entire Cu-7 clinical trial are accurate and are similar to those published in the world literature. (AM J OBSTET GYNECOL 1994;170:1606-16.)

Key words: Pelvic inflammatory disease, Cu-7, intrauterine contraceptive device, clinical trial

From the Departments of Obstetrics and Gynecology and Biometry, University of Southern California School of Medicine. Presented by invitation at the Sixtieth Annual Meeting of the Pacific Coast Obstetrical and Gynecological Society, Blaine, Washington, September 7-12, 1993. Reprint requests: Subir Roy, MD, Department of Obstetrics and Gynecology, LAC/USC Women's Hospital, 1240 North Mission Road, Los Angeles, CA 90033. Copyright © 1994 by Mosby-Year Book, Inc. 0002-9378/94 $3.00 + 0 6/6/54573

1606

When contraceptive choice in the world today is analyzed, the intrauterine contraceptive device (IUD) is probably the second most commo~y used, reliable, reversible method of preventing pregnancy; only oral hormonal contraceptives are used more frequently. 1 It is unfortunate that the litigation climate in the United States in the 1980s has significantly curtailed both research and availability of contraceptive choices for

Roy and Azen

Volume 170, Number 6 Am J Obstet Gynecol

American men and women today.2 An example of this phenomenon is the discontinuance of the Cu-7 IUD by G.D. Searle & Co. (Skokie,Ill.) notwithstanding continued Food and Drug Administration approval and worldwide endorsement by the medical community." Reports in the medical literature since 1986 have continued to demonstrate both the reasonable safety and efficacy of the Cu-7 and copper IUDs. 4 -6 Among the litigation-prompted claims has been the suggestion by attorneys for plaintiffs that in the early 1970s, when the Cu-7 was approved, the Food and Drug Administration may not have had accurate information concerning Searle's worldwide study of the Cu-7 and the rate of pelvic. inflammatory disease occurring within the study population. At the time of the initial filing of its New Drug Application (NDA 17-708) in August 1971, Searle reported a pelvic inflammatory disease rate of 2.02 per 100 users per year. Mter 20 months the rate was l.84 per 100 users per year, which was reported to the Food and Drug Administration in December 1972. In a published article after completion of the study, O'Brien et al. 7 reported a pelvic inflammatory disease rate of 18 per 100 woman years (or l.8/100 woman years) on the U.S. portion of Searle's clinical trial. In 1984, as part of the litigation against Searle, attorneys representing plaintiffs commissioned Professor Miriam Orleans of the Department of Preventive Medicine and Biometrics, University of Colorado Medical School, Boulder, Colorado, to obtain a 10% random sample of > 16,000 subjects who were included in the Searle clinical trial. A 10% sample would be adequate to estimate pelvic inflammatory disease rates of 2%, the estimated annual rate for women of reproductive age, 8 within an accuracy of 0.69% with 95% confidence (e.g., 95% confidence interval l.31 to 2.69). TestifYing as an expert epidemiologist for plaintiffs, Dr. Orleans detailed the protocol that she followed to obtain a statistically random 10% sample of the Searle clinical trial. Initially, Dr. Orleans expressed an opinion that her analysis of the sample yielded a crude pelvic inflammatory disease rate in excess of 6%. However, the need for reanalysis was later conceded by Orleans when she learned that her nonmedical staff has mistakenly interpreted recorded data on the patient follow-up forms. 9 The investigation reported herein was undertaken at the request of G.D. Searle. We were asked to perform an independent evaluation of the rates of pelvic inflammatory disease in a 10% sample of Cu-7 users in the Searle clinical trial. This analysis would also permit a comparison with the clinical trials published in the world literature. Finally, a judgment could be made as to the safety of this widely used method of contraception and whether American women should have the option of IUDs as a viable contraceptive method.

1607

The G.D. Searle Cu-7 clinical trial was performed throughout the world by > 200 clinically experienced investigators treating patients from all socioeconomic groups. It was designed so that subjects underwent a pelvic· examination on each clinic visit, enabling any change from prior examinations to be noted. The Orleans 10% sample was the basis for the data analyzed in this article because it was a random sample of sufficient size to test the hypothesis of whether pelvic inflammatory disease rates reported to the Food and Drug Administration were accurate, and the sampling techniques cannot be criticized as having been generated by Searle. A total of 1614 patient files have been reviewed, with data extracted and analyzed by standard statistical methods to determine the incidence of pelvic inflammatory disease occurring within the Searle clinical trial of the Cu-7. Material and methods

Sample validation. All of the information contained in the Cu-7 New Drug Application, Dr. Orleans' testimony,9 and copies of the patient files were provided by representatives ofG.D. Searle. The protocol followed by Dr. Orleans for generating a 10% random sample of the Searle clinical trial was as follows: The original patient files in the Searle clinical trial were provided to Dr. Orleans, who assigned a number to each of the > 16,000 patient files. A computer-generated 10% random sample of these numbers was obtained by Dr. Orleans and matched with the previously numbered patient files. 9 This analysis is of those patient files selected by Dr. Orleans in accordance with the random number list. Dr. Orleans ultimately included 1628 patient files in her analysis of the study sample. One of the patients never had the Cu-7 inserted so that patient has been excluded here. Six patient files included in her analysis did not match the random number list and have therefore been excluded from this analysis. None of the patients had pelvic inflammatory disease. The remaining seven patient files randomly selected by Dr. Orleans and included in her data analysis could not be identified because of the manner in which she recorded her data. None of the seven had pelvic inflammatory disease according to Dr. Orleans' analysis, and in any event their inclusion would not have had any statistical effect on this analysis, except perhaps to have slightly lowered the rate of pelvic inflammatory disease in the sample. These 1614 files, placed in ascending order by the Orleans number, were consecutively numbered from 1 to 1614. As a check on the randomness of the sample, an analysis was also undertaken to determine whether the major investigators ir. the New Drug Application (investigators with more than 100 subjects, as provided by G.D. Searle) had similar percentages of subjects in

1608 Roy and Azen

both the sample group and the Searle clinical population. As a check on the completeness of the records provided by G.D. Searle for analysis, we tested the assumption of 99% completeness of the records (i.e., that there was page-for-page agreement with information in the original files in 99% of our files; any number of missing pages would qualify as an incomplete file). To detect a 3% deviation from this rate of 99%, with 90% power at IX = 0.05, required a check of a random sample of 250 of the 1614 Orleans sample files. A random sample of 250 of 1614 numbers was generated with the Statistical Analysis System (SAS) procedure PLAN.IO These numbers were matched to the Orleans numbers, and the corresponding subject files were identified and selected. The date of each visit and total numbers of pages in each file were recorded. At G.D. Searle's warehouse in Skokie, in our presence, Searle personnel identified and provided the original files corresponding to the random number list of Orleans' sample, which was provided to them on site. We carried out an audit of these files. The original files were compared to the records in our files for numbers of pages and dates of each clinic record. Data analysis. A coding sheet was developed for extracting information from each patient file. Factors that were coded included the Orleans sample number, the Searle clinical study number, age, parity, insertion date, event date, and event code. If subjects were reinserted with a Cu-7, similar information was compiled for all subsequent insertions. This report is based on the first insertion only. A diagnosis of pelvic inflammatory disease was made if data on the investigator report form mentioned upper-genital-tract infection (endometritis, pelvic infection, salpingitis, oophoritis, adnexitis, or pelvic inflammatory disease) regardless of whether the IUD was removed or remained in situ. Additionally, because Searle required that investigators note "changes in pelvic findings" on follow-up forms that were completed with each patient visit, cases that logically could have been excluded because they were not specifically diagnosed as such were nonetheless included as having pelvic inflammatory disease if upper-genital-tract infection could conceivably have been present on the basis of the recorded information. In other words, the method of recording information provided by the forms permitted an overinclusive and consistently applied definition of the diagnosis of pelvic inflammatory disease. Months of use were tabulated from insertion until study termination, regardless of the causes for terminating the study. Subjects who became pregnant had 2 weeks added to the date of onset of their last menstrual period and were dropped as of that date. Expulsions were recorded on the date that they occurred. Removals

June 1994 Am J Obstet Gynecol

for bleeding or pain, or both, were recorded on the date of removal. Other removals, whether for planned pregnancy or no further need, were recorded on the date of removal. For subjects who were considered lost to follow-up or who moved away with the Cu-7 in situ, who had not previously experienced pelvic inflammatory disease, and for whom there was no further information, we backtracked to their last visit for which there was a recorded physical examination documenting that the device was in place, 3 months was added to the date of the last visit, and the patient was dropped from the study. We included as having pelvic inflammatory disease all women who experienced pelvic inflammatory disease whether or not the IUD was removed when pelvic inflammatory disease occurred and no matter what the eventual" reason for removal from the Searle study was or whether the woman was later declared lost to follow-up in the Searle study. The data were entered into an SAS computer database, printed, and checked against the original handgenerated listing. Demographic data (age, parity, and months of use) were tabulated by parity and by diagnosis of pelvic inflammatory disease. The crude rate of pelvic inflammatory disease for selected periods of use was calculated as the cumulative incidence of pelvic inflammatory disease cases to that time divided by the number of original insertions. The Pearl index for pelvic inflammatory disease at selected periods of use was calculated as follows: 100 x Cumulative evidence of pelvic inflammatory disease cases x 12 .;Cumulative months of use of Cu-7 to that time. The usual techniques based on the binomial distribution were used to generate 95% confidence intervals around the crude and Pearl index pelvic inflammatory disease rates. 11 The BMDP Statistical Program lL was used to generate life-table analyses of pelvic inflammatory disease by parity (0, > 0, and the entire population). 12 A singledecrement life-table analysis was used. This method of analysis includes terminations for reasons other than the one in question in a statistically valid manner. To calculate woman months of use, the program computes the exact number of days from insertion date to event date, divided by 30.4375, the average number of days per month, including leap years (any given 4-year cycle will have 1461 days). The total woman months of use, along with the numbers of events by time periods from the life-table analysis, permitted the calculation of Pearl index percentages for pelvic inflammatory disease. The BMDP program permitted the statistical comparison of the life-table curves by the use of the generalized Wilcoxon test. The SAS program was used to make summary tabulations and to compare means between pelvic inflammatory disease versus non-pelvic inflammatory disease groups.

Roy and Azen 1609

Volume 170, Number 6 Am J Obstet Gynecol

Table I. Demographic information related to parity Nulliparous (n = 466)

I

Mean

SEM

Mean

0.2

27.3 2.2 19.4

23.4

Age (yr)

o

Parity Duration or use (rno)

Parous (n = 1148)

o

16.3

0.6

I

SEM

Significance

0.2 0.003 0.5

p p

= 0.0000 = 0.0000

p=

0.0001

Table II. Demographic information related to pelvic inflammatory disease Pelvic inflammatory disease (n Mean Age (yr)

Parity Duration or use (rno)

25.5 1.5 14.5

1

= 48)

No pelvic inflammatory disease (n

SEM

Mean

0.8 0.2 2.0

26.2 1.6 18.6

Results

The major investigators (i.e., those having > 100 subjects in the clinical trial) reported in the New Drug Application were checked to determine whether approximately 10% of their subjects were represented in the Orleans sample. The X2 goodness-of-fit calculation with 52 degrees of freedom was 47.8, p = 0.63, signifying that the major investigators were appropriately represented in the 10% Orleans sample. The check on completeness of the 250 randomly selected files resulted in finding two admission forms in the original files that were not in the files in our possession. OtheIWise, all files in our possession were complete and identical to the original files. With our predetermined definition for completeness, the number of incomplete files was 2 of 250 (0.8%), thus confirming our assumption of 99% completeness (or 1% incomplete files) in the sample of 1614. Additionally, this check on completeness added no substantive information not already in our files. The life-table analysis of this sample confirmed that after 4 years of use, there were few nulliparous (n = 48) and parous (n = 173) users. Because the intended period of use of the Cu-7 is 3 years, the focus of the life-table results and discussion is for rates up to this time. O'Brien et al.' reported on 4 years of use; therefore we include analysis up to this date as well. However, in the interest of completeness, pelvic inflammatory disease events and rates are provided for the entire experience, which is up to 60 months of use. The 466 nulliparous and 1148 parous subjects differed significantly for age, parity, and months of use (fable I). Forty-eight of the 1614 patients had pelvic findings that came within the overinclusive definition of upper-genital-tract infections as imposed for this analysis during the period following the first insertion of the

I

= 1566)

SEM

Significance

0.1 0.04 0.4

p = 0.035

P= P=

0.0.78 0.06

Cu-7. Investigators noted a diagnosis of "pelvic inflammatory disease" in 20 subjects, "endometritis" in 13 subjects, "adnexitis" in 3 subjects, and "endometritis and adnexitis" in 1 subject. Eleven additional subjects were included as having "pelvic inflammatory disease" because of a constellation of findings, including one with "gonococci," 3 with .~ tender uterus on pelvic examination and receiving antibiotic therapy, 2 treated with antibiotics for pelvic infection (not otheIWise specified), 1 with a diagnosis of infection (location unspecified), 1 with a tender mass (location unspecified), 1 with pain in the pelvis and a temperature of 105 F, 1 with yellowish white discharge and symptoms of upper-genital-tract infection and treated with antibiotics, and 1 with an abscess (location unspecified). In the overall sample, these 48 subjects with the diagnosis of pelvic inflammatory disease did not differ significantly from the 1566 subjects who did not have pelvic inflammatory disease by age, parity, or months of use of the Cu-7 (Table II). The crude pelvic inflammatory disease rate (number of pelvic inflammatory disease cases per number of initial Cu-7 insertions) and 95% confidence intervals by time of use are shown in Table III. The Pearl index for pelvic inflammatory disease (number of cases of pelvic inflammatory disease x 12 x 100/Cumulative woman months of use) with corresponding 95% confidence intervals is shown in Table IV. The life-table pelvic inflammatory disease rates for all subjects are shown in Table V. Comparison of pelvic inflammatory disease life-table rates between nulliparous and parous subjects demonstrated no differences (generalized Wilcoxon X2 = 1.13, P = 0.29). An analysis including all insertions did not reveal statistically significantly different rates from those presented and has not been included. 0

1610 Roy and Azen

June 1994 Am J Obstet Gynecol

Table III. Crude pelvic inflammatory disease rate (percent) of 1614 users Duration of use (mo)

No.

%

SEM

12 20 24 36 48 60

28 36 38 43 46 48

1.73 2.23 2.35 2.66 2.85 2.97

0.32 0.37 0.38 0.40 0.41 0.42

95% Confidence interval 1.08-2.38 1.49-2.97 1.59-3.11 1.86-3.46 2.03-3.67 2.13-3.81

Table IV. Pearl index pelvic inflammatory disease rates

Duration of use (mo)

Cumulative No. of pelvic inflammatory disease cases

Cumulative woman months

Pearl index

SEM

12 20 24 36 48 60

28 36 38 43 46 48

14,643 20,977 23,380 27,535 29,205 29,794

2.29 2.06 1.95 1.87 1.89 1.93

0.43 0.34 0.33 0.28 0.28 0.28

95% Confidence interval 1.43-3.15 1.38-2.74 1.29-2.61 1.31-2.43 1.33-2.45 1.37-2.49

Table V. Life-table pelvic inflammatory disease rates for all subjects

Interval of use (mo)

Cumulative pelvic inflammatory disease cases

Cumulative pelvic inflammatory disease rate per 100 woman years

Pelvic inflammatory disease rate (95% confidence interval)

0-12 12-20 20-24 24-36 36-48 48-60

28 36 38 43 46 48

2.03 2.24 3.23 3.55 4.83 6.69

1.23-2.83 1.38-3.09 2.14-4.32 2.37-4.73 3.20-6.46 4.03-9.02

Comment This analysis and its conclusions differ from those of Orleans in that a clinician, not an epidemiologist, reviewed and coded the recorded data for the clinical diagnosis of pelvic inflammatory disease, applying standard statistical tests to the data and making a comparison with the world literature. The importance of having a clinician review the recorded data is demonstrated by the observation that Professor Orleans' nonmedical staff mistakenly included patients with vaginitis or condyloma acuminatum or other lower genital tract conditions as having pelvic inflammatory disease without their having any evidence of upper-genital-tract infection, thereby inappropriately elevating the pelvic inflammatory disease rate. A 10% random sample of such a large clinical trial (> 16,000 subjects) permits confidence that conclusions arrived at here are an accurate reflection of the experience of the overall Searle clinical trial. If there had been an incidence of pelvic inflammatory disease in the Searle clinical trial that went unrecognized, this 10% sample would have exposed its existence. The rate of pelvic inflammatory disease in this 10% sample is not

statistically different from the rate reported by Searle, demonstrating what the medical community involved with contraceptive issues has known - that reports of pelvic inflammatory disease with the Cu-7 in place are uncommon (see Pearl rates, below)!·6. 8. 13·35 The case report files provided to us for analysis by G.D. Searle fulfilled our criteria of completeness on the basis of our audit of 250 of the Orleans sample of 1614 cases. Thus we are confident of the accuracy and completeness of the data that were analyzed. The pelvic inflammatory disease rates in the sample, calculated at 20 months both as a crude rate (2.23 [1.49 to 2.97]) and by use of the Pearl index (2.06 [1.38 to 2.74]), do not differ statistically from the pelvic inflammatory disease rate of 1.84 reported by Searle to the Food and Drug Administration in the New Drug Application filed in December 1972 because both 95% confidence intervals include the reported rate. The Pearl index rate and 95% confidence intervals at 48 months were 1.89 and 1.33 to 2.45. This rate is essentially identical to the report of O'Brien et al. 7 for the entire American subject experience of 18 per 1000 woman years of 1.8 per 100 woman years (365 pelvic

Volume 170, Number 6 Am J Obstet Gynecol

inflammatory disease cases with 242,169 woman months of use). By comparison, the overall Pearl index pelvic inflammatory disease rate was < 1% in a summary of 27 independently reported clinical trials performed around the world and published from 1972 to 1989 (Cu-7 users, 11,811; cumulative woman months of use, 214,937; pelvic inflammatory disease cases, 134).4.6, s, 13·35 The life-table pelvic inflammatory disease rates, a different but more statistically rigorous way of looking at these data, are consistent and similar to those reported for the use of the Cu-7 by numerous clinical trials, s, IS, 22, 26, 30, 31, 33-35 Of interest, some literature has suggested an increased incidence of pelvic inflammatory disease among nulliparous subjects using IUDs 36 ; however, this sample supports the data reported by O'Brien et aV that Searle's clinical trial found a slightly higher incidence in parous subjects as opposed to nulliparous ones, although the analysis of this sample did not reach statistical significance. The annualized rates of pelvic inflammatory disease, between 1% and 2%, are similar to the pelvic inflammatory disease rates reported for fecund, sexually active women. s The number of subjects continuing in the study beyond 4 years was sparse, and therefore, as has been suggested by Jain and Sivin,37 these data should not be considered reliable. They have been provided here simply for completeness sake. The manner in which loss to follow-up was handled in this study is consistent with the approach that has been used by O'Brien et al. 7 in the analysis of the Searle data, with the recommendations of Orleans,9 and with the practical implications of the Jain and Sivin report.'7 In this analysis of the study sample, 3 months was added to the last visit for patients who were known to be using the Cu-7 when they moved out of the area or were later determined to be lost to follow-up, regardless of what the investigator had recorded as the date when the patient became lost to follow-up. The effect of this algorithm was to eliminate the possibility of including subjects who were in fact lost to follow-up, thus ensuring that the incidence of pelvic inflammatory disease was not artificially depressed by woman months of use for which there was no information regarding patient experience. However, it does represent the realistic observation that all of these individuals did not stop use on the date of their last recorded visit, which would represent the most conservative but unrealistic manner to treat this situation. Eliminating the 3 months negligibly affects the pelvic inflammatory disease rates. In any event, in a single-decrement life-table analysis as used here, loss to follow-up is recorded as a termination category and, as the analysis shows, no deleterious impact on pelvic inflammatory disease rates is noted. Additionally, on the basis of this study size, including the pattern of follow-up, which reflects all termination categories including loss to follow-up through 60

Roy and Azen

1611

months of use, the data are consistent with pelvic inflammatory disease rates that fall between 1.4% and 2.5% per year with 95% confidence.'s Finally, there is no credible way to know whether subjects who are lost to follow-up experience a better, the same, or a worse outcome compared with those who remain in the study. There have been several recent reports that are relevant to the findings of our analysis. The first was by the World Health Organization, which has recently reported on its experience of pelvic inflammatory disease associated with IUD use from prospective efficacy studies.'9 Its analysis differed from ours in several important aspects. They used more stringent criteria for the diagnosis pelvic inflammatory disease, focusing on the identification of potential risk factors and comparing rates according to parity. We used a more liberal definition of pelvic inflammatory disease and were more interested in the overall global pelvic inflammatory disease rates, which accounts for the differences between these studies. Our daily life-table pelvic inflammatory disease rates did not demonstrate the "break" at 20 days after insertion; rather the pelvic inflammatory disease rates observed in this study were fairly evenly distributed with time of use. Our report supports the large body of evidence accumulated from prospective clinical trials over the past 20 years, attesting to the safety and efficacy of the Cu-7, as well as other "modern" IUDs. Any suggestion that prospective clinical trials are not able to detect the occurrence of low-incidence events such as pelvic inflammatory disease with IUD use is not borne out by our analysis of these data. Certainly the prospective clinical trials of "modern IUDs," coupled with our 10% sample of the 16,000 patients in the Searle clinical trial, demonstrate a consistency not found in case-control studies, which are often performed to assess these associations although they cannot be used to demonstrate causality. A review of the available pertinent case-control literature demonstrates this point. Two National Institutes of Health-supported investigations basically found no association of copper IUD use with infertility (used as a surrogate for "silent pelvic inflammatory disease"):o, 41 but one of the groups has subsequently reported a correlation of ectopic pregnancy with prior or current IUD use, including copper IUDs,, 2, 43 As referenced in their discussions, these studies are inconsistent with other reports by the same and other groups. Thus the case-control reports make it difficult to gain an understanding of this issue because of their inconsistent conclusions, which are in contrast with the large body of consistent evidence accumulated from prospective clinical trials. In the last study, a critical reappraisal of IUDs, Bromham44 concluded that " ... the IUD is a highly effective and safe form of reversible contraception, It

1612 Roy and Azen

does not cause pelvic inflammatory disease, ectopic pregnancy or infertility; it could be used by most nulligravidae; ... it acts by preventing fertilisation not implantation. These statements contrast sharply with many of our long-standing prejudices about the method." Thus the second most widely used method of reversible contraception worldwide, the IUD in its various forms, should continue to be available to or approved for use by suitably selected American women. In conclusion, analysis of the 10% study sample of the Searle Cu-7 clinical trial demonstrated it to be a random sample. The audit of the patient files confirmed that the files available for analysis were complete. The calculated crude and Pearl index rates of pelvic inflammatory disease were consistent with those rates previously reported to the Food and Drug Administration and published in the medical literature. Life-table pelvic inflammatory disease rates were not different between nulliparous and parous women and indicated pelvic inflammatory disease rates no different from basal annual rates in fecund women. On the basis of analysis of this 10% sample the pelvic inflammatory disease rates reported to the Food and Drug Administration for the entire Searle Cu-7 clinical trial are accurate. Clinical trials, especially of this size, are appropriate, indeed are required by the Food and Drug Administration, for the study of safety and efficacy of devices and drugs.· One cannot perform case-control studies until the devices and drugs are in general use, and as noted, the results of such studies are not necessarily consistent, thus demonstrating the importance of prospective clinical trials. REFERENCES 1. Fertility and family planning surveys: an update. Popul Rep M 1985;8:289-308. 2. Mastroianni Ljr, Donaldson Pj, Kane IT. Developing new contraceptives: obstacles and opportunities. Washington: National Academy Press, 1990. 3. After contraception: dispelling rumors about later child bearing. Popul Rep j 1984;28:109-712. 4. Bisset AM, Dingwall-Fordyce I, Hamilton MK. A study of event rates of intrauterine devices over a two year period. Br j Fam Plann 1987;13:55-61. 5. Manuilova AI, Dubnitskaya LV. A comparative study ofthe acceptability of the inert and copper-containing IUD's. Akush Ginekol (Mosk) 1988;11:47-50. 6. Apelo RA, Ramos RM, Bernardo E, Champion CB. A 3-year evaluation of the TCu 380Ag and the Cu-7. Int j Gynecol Obstet 1989;28:269-73. 7. O'Brien FB, Stewart WC, Sturtevant FM. Incidence of pelvic inflammatory disease in clinical trials with Cu-7 (intrauterine copper contraceptive): a statistical analysis. Contraception 1983;27: 111-22. 8. Hasson HM. Clinical experience with intrauterine devices in a private practice. Adv Contracept 1985;1:51-61. 9. Marder v Searle, 630 F Supp 1087 (D Md 1986). 10. SAS user's guide: statistics, version 5 edition. Cary, North Carolina: SAS Institute, 1985. 11. Fleiss jL. In: Statistical methods for rates and proportions. New York: john Wiley, 1973.

June 1994

Am J Obstet Gynecol

12. Dixon Wj, ed. Volume 2: BMDP statistical software manual: to accompany the 1990 software release. Berkeley, California: University of California Press, 1990. 13. Bernstein G, Israel R, Seward P, et al. Clinical experience with the Cu-7 intrauterine device. Contraception 1972;6: 99-107. 14. Chen Y-P. Clinical evaluation of copper-seven, an intrauterine contraceptive device (IUD). j Obstet Gynecol Rep China 1973;12:153-63. 15. Levin HM, Columbi Dj, Bare WW. The Cu-7: a metallic copper intrauterine contraceptive device. j Reprod Med 1974;12:165-6. 16. Macourt D. Evaluation of the Gravigard intrauterine copper contraceptive device. Medj Aust 1974;1:717-8. 17. Newtonj, Eliasj, McEwanj, Mann G. Intrauterine contraception with the Copper 7: evaluation after two years. BMj 1974;3:447-50. 18. Tejuja S, Choudhury SD, Malhetra U, Socana NC. Clinical trial with the Copper 7 intrauterine device-one year experience in India. Contraception 1974;10:351-61. 19. Matsumoto S. Clinical studies on Cu-7: copper-IUD study group on Cu-7. jap j Fertil Steril 1976;21:70-82. 20. Simcock BW. A three-year trial of the Gravigard (Cu-7) intrauterine device. Med j Aust 1976;2:297-9. 21. Thorborg P, Copper T-Kabi and Gravigard: report of one year's experience with two copper IUDs in a maternity welfare clinic. Ugeskr Laeger 1976;138:342-3. 22. Zipper j, Medel M, Dabancens A, et al. Studies on a new seven device (small seven) containing copper and copperzinc. Amsterdam: Excerpta Medica, 1976:201-6; International congress series 370. 23. Nebel W, Currie JL, Lassiter RE. Clinical experience with the Copper 7 intrauterine device. Fertil Steril 1978;30: 516-8. 24. Heng GT. Use-effectiveness of the Copper-7 intrauterine device in a Malaysian Family Planning Clinic. Med J Malaysia 1979;33:352-5. 25. Muhlemann M, Obolensky W. Intrauterine contraception with the Cu-7: evaluation of 503 observations. Med Hyg 1979;37:4438-45. 26. Bouchardj, Albert R, Cerskus I. Use ofthe Cu-7 device in a family planning clinic. Can Fam Physician Med Fam Can 1980;26:1507-14. 27. Gillett PC, Lee NH, Yuzpe AA, Cerskus I. A comparison of the efficacy and acceptability of the Copper-7 intrauterine device following immediate or delayed insertion after first-trimester therapeutic abortion. Fertil Steril 1980;34: 121-4. 28. KuligjW, RauhJL, Burket RL, Cabot HM, Brookman RR. Experience with the Copper 7 intrauterine device (Cu-7) in an adolescent population. j Pediatr 1980;96:746-50. 29. Larrson B, Hagstrom B, Viberg L, Hamberger L. Long term clinical experience with the Cu-7-IUD: evaluation of the prospective study. Contraception 1981;23:387-97. 30. Sellors JW. The Copper-7 intrauterine contraceptive device: five-year evaluation. Can Med Assoc J 1981;125: 717-20. 31. World Health Organization. Interval IUD insertion in parous women: A randomized multicenter comparative trial of the Lippes Loop D, the T-Cu220C and the Copper 7. Contraception 1982;26: 1-22. 32. Goh TH, Sinnathuray TA, Sivanessaratnam V, Sen DK. A randomized comparative evaluation of the Copper-7, Multiload Copper-250 and T-Copper-220C IUDs. Contraception 1983;27:75-83. 33. World Health Organization. IUD insertion following spontaneous abortion: a clinical trial of the TCu 220C, Lippes Loop D, and Copper. 7. Stud Fam Plann 1983;14:109-14. 34. World Health Organization. IUD insertion following termination of pregnancy: a clinical trial of the TCu 220C, Lippes Loop D and Copper 7. Stud Fam Plann 1983;14: 99-108.

Volume 170, Number 6

Am J Obstet Gynecol

35. Sandmire HF, Cavanaugh RA. Long-term use ofintrauterine contraceptive devices in a private practice. AM] OBSTrr GYNECOL 1985;152:169-75. 36. Ory HW. A review of the association between intrauterine devices and acute pelvic inflammatory disease. ] Reprod Med 1978;20:200-4. 37. Jain AK, Sivin I. Life-table analysis ofIUDs: problems and recommendations. Study Fam Plann 1977;8:25-47. 38. Sprott DA. In: Statistical inference with small samples. Waterloo, Canada: University of Waterloo, 1972. 39. Farley TMM, Rosenberg M], Rowe P], Chen], Merik O. Intrauterine devices and pelvic inflammatory disease: an international perspective. Lancet 1992;339:785-8. 40. Daling ]R, Weiss NS, Metch B], et al. Primary tubal infertility in relation to the use of an intrauterine device. N Engl] Med 1985;312:937-41. 41. Cramer DW, Schiff I, Schoenbaum SC, et al. Tubal infertility and the intrauterine device. N Engl] Med 1985;312: 942-7. 42. Rossing MA, Daling ]R, Weiss NS, et al. Past use of an intrauterine device and risk of tubal pregnancy. Epidemiology 1993;4:245-51. 43. Rossing MA, Daling ]R, Voight LF, Stergachis AS, Weiss NS. Current use of an intrauterine device and risk oftubal pregnancy. Epidemiology 1993;4:252-8. 44. Bromham DR. Intrauterine contraceptive devices-a reappraisal. Br Med Bull 1993;49:100-21.

Editors' note: This manuscript was revised after these discussions were presented. Discussion DR. RICHARD M. SODERSTROM, Seattle, Washington. In early 1974 the AH. Robins Company, the maker of the Dalkon Shield, temporarily removed their IUD from the marketplace pending a review by the Food and Drug Administration of isolated cases of septic death associated with women who became pregnant while wearing the Dalkon Shield. Through the summer months, a cadre of experts reviewed the current data and published articles. By the late fall, a panel of experts agreed that the Dalkon Shield could be sold once again, provided its multifilament tail be replaced by a monofilament tail. In that short time the number of product liability suits against the AH. Robins Company had grown to such proportions that they chose to terminate its production and sales. Prompted by a landslide of allegations about this "defective device," Congress passed the Health Device Act in May 1976, giving the Food and Drug Administration a new power of surveillance and control over medical devices that was consistent with their control over drugs. Advisory panels in all specialties - including gynecology-were formed and "high-risk" devices were chosen for evaluation and classification, with an eye toward setting safety and efficacy standards. By January 1977, the gynecology panel, chaired by Elizabeth Connell, set forth new device regulations for future IUDs. As a member of "Connell's Commandos," I was asked to review the literature on uterine perforation and pelvic inflammatory disease associated with IUD use. U sing the methods of literature search available at the time (before the use of "key words"), no article was found dealing with pelvic inflammatory disease and

Roy and Azen

1613

IUD use until 1972, a time after the studies on the Cu-7 had been submitted. One article from the Centers for Disease Control, entitled "Pelvic inflammatory disease in an indigent population," demonstrated a possible increased risk of pelvic inflammatory disease in women wearing IUDs, but the authors failed to comment on this in their statistical review or in their conclusions. 1 Despite this, many suits against the AH. Robins Company were filed, and the suits were awarded to plaintiffs over the "duty to inform" about an increased risk of pelvic inflammatory disease with the use of the IUD. In 1982, a Dalkon Shield lawsuit in Seattle ended in a hung jury for the second time! The patient admitted to having 26 sexual partners over the period of IUD usage, six of whom she could remember only by their first name. Each side had presented more than 20 experts, each speaking to any association of the Dalkon shield in a patient with severe pelvic inflammatory disease that devloped while she was wearing the Dalkon Shield. Mter this second trial, the judge was asked why such a case would end that way. He stated that it was "because all the experts for each side were credible and with outstanding credentials." Eleven years later the battle continues, as exposed in a commentary following Letters to the Editor by Paul G. McDonough, MD, an editor of Fertility and Sterility, the commentary being entitled, "Flak Jackets for Editors.'" He expounded on the editorial staffs surprise at the motional content of the letters written as a retort to a guest editorial by David Eschenbach, MD, reviewing the subject of pelvic inflammatory diseases and pelvic inflammatory disease. S What is amazing is the fervor that persists because all of the IUDs produced before 1976 (except for the Progestasert) have been voluntarily withdrawn from the marketplace by the manufacturer for > 10 years. Although the field of epidemiology has progressed by leaps and bounds, there is little unanimity of opinion about the causal association of IUDs and pelvic inflammatory disease. I was schooled to be a statistician, but I flounder in the interpretation of contemporary articles that address risk factors and their associations, such as Dr. Roy's learned treatise. What is a practicing gynecologist to do or think while this battle persists? Several basic approaches may help. The Achilles' heel of any study, if one exists, will be some form of bias. There are three major kinds of bias. The first is called selection bias. For instance, hospitalized patients may have different characteristics from patients with the same disease treated in an outpatient facility. Selecting controls within the hospital can be a bigger problem. The second is information bias. The mother of a baby with a birth defect may have better recall of drug exposures during her pregnancy than a mother with a normal child. This third bias, confounding bias, is an extraneous factor that causes a distortion of the association between an exposure and a disease and can create spurious conclusions in either direction. It

1614

Roy and Azen

must satisfY both of two conditions: (1) It is a risk factor for the study disease; (2) it is associated with the study exposure but is not a consequence of exposure. A confounder's association with disease must occur in the absence of exposure. The number of sexual partners per case is a good example of a confounding bias in a study exploring the association of IUD use and pelvic inflammatory disease. The more information collected on confounding variables, the easier it is to remove the bias in the analysis. As one reads controlled studies, a critical review of the control selection is mandatory. Cases and controls should have had an equal chance for the disease to develop. Controls and cases should have had an equal chance to be exposed and examined for the disease. The same information, obtained by similar methods of collection, should be gathered from controls and cases. Aids to assist recall of exposure may be used provided the chance of exposure to these aids in another context is unlikely. For example, picutres of different IUDs that are used to evaluate a relationship between IUDs and pelvic inflammatory disease may prompt spurious recall if the patient has been exposed to such pictures during a news media discussion of the same subject. In closing, I will discuss causation. The term cause embraces an elusive concept that has been elaborated on from antiquity. A causal association requires that factors thought to be causative must precede those events that are regarded as their effects. A second condition that supports a causal interpretation of an association is the repeated observation of the association under different conditions of study. On the other hand, similar studies that yield diverse results weaken a causal interpretation. The larger the relative risk, the less likely the association is to be spurious. The tradition of scientific explanation requires that a claim of proof of cause and effect must specifY the mechanism by which the effect is produced. In general, this requires an experimental study. Thus most observation studies end with an opinion or judgment about causality, not a claim of proof. Having said all that, I will not join a side but will add Dr. Roy's data to my collection of articles on the subject; perhaps in retirement I might be satisfied that all of the criteria for a conclusion have been met. Several facts do seem clear. Although the numbers are small compared to the Cu-7 experience of the 1970s, our current experience with the copper-bearing ParaGard T380A (Gynopharma Inc., Somerville, N.].) shows that the outcome statistics are good, paralleling the experience of our European colleagues, who have many IUDs to choose from and who do so freely. Have we developed the proper guidelines for choosing the right candidate to use an IUD, or are we more responsive to patients who report IUD-associated complaints? Is the design of contemporary IUDs safer? I applaud Dr. Roy and colleagues for their attempt to set the record straight. Only you, the reader (audience), can decide whether their goal satisfies your curiosity.

June 1994 Am J Obstet Gynecol

There are many who still have doubts, but when in doubt, make it sound convincing. REFERENCES 1. Wright NH, Laemmle P. Acute pelvic inflammatory disease in an indigent population. AM J OBSTET GYNECOL 1968;101: 979-90. 2. McDonough PG. Flak jackets for editors [Editorial comment]. Fertil Steril 1993;60:191. 3. Eschenbach DA. Earth, motherhood, and the intrauterine device. Fertil Steril 1992;57: 1177-9.

DR. JAMES CAILLOUETIE, Pasadena, California. I would like to know whether Dr. Roy and Dr. Soderstrom have accessed the Food and Drug Administration Adverse Drug Reaction database through the Freedom of Information Act? The reason I bring this up is that I was confounded by the difference in number of ovarian cysts and multiphasic oral contraceptives. My database did not compare to that provided to me by the Food and Drug Administration and the three companies manufacturing the multiphasic pills. I discovered the Freedom of Information Act and that they have a complete database of the Food and Drug Administration Adverse Drug Reaction reports. Those who do these studies need to access the database of the Food and Drug Administration through the Freedom of Information Act. DR. SIMON HENDERSON, San Francisco, California. I am wondering whether your indicator, pelvic inflammatory disease, was really strong enough to warrant your excellent statistical analysis. What about the rest of the "iceberg" (i.e., subclinical infection)? The reason I bring this up is that in the early 1980s, when chlamydia could first be cultured and identified by serum titers, I found that in my infertility-based practice, chlamydia cultures were negative in 100 patients so I did not do any more culturing for chlamydia. But I did continue to do titers and it was really illuminating. About 60% of the patients had positive chlamydia titers and none of these patients had any history of pelvic inflammatory disease. The patients who had the positive titers were then divided into two groups, those who had IUDs and those who did not. The IUD group had 74% positive titers and the non-IUD group, about 40%. When we operated on these patients we found that there was a good correlation between chronic inflammatory tubal damage and positive chlamydia titers, and this has since been shown in several articles. Looking pelvic inflammatory disease and subclinical pelvic inflammatory disease with the end point being infertility, one is concerned that the whole subject of subclinical tubal-ovarian infections has been largely ignored by these original IUD articles. One can imagine that the IUD probably works as a contraceptive or contragestational agent because it elicits a chronic inflammatory reaction inside the uterus. The IUD certainly does produce chronic endometritis with an inflammatory exudate. If reported this finding years and years ago when I was performing endometrial washings. If a patient contracts gonorrhea

Volume 170, Number 6 Am J Obstet Gynecol

or chlamydia in the presence of an IUD, the already compromised defense mechanism in the uterus might make it more likely that these organisms will ascend to the tube and perhaps, in the case of chlamydia, in a subclinical manner. Many investigators now consider that chlamydia can often produce an acute salpingitis from a subclinical base, and it is possible that the presence of an IUD augments this inflammatory reaction. DR. GEORGE LEE, San Francisco, California. When you presented your data, you estimated roughly a 2% pelvic inflammatory disease rate. Is that rate above and beyond that of a control group in the original report? In other words, is that 2% related to an iiIcreased incidence of pelvic inflammatory disease in the IUD group compared with a control group without IUDs? If we are dealing with fertile women who wish to preserve their fertility, a 2% increase, if solely related to the IUD, is pretty significant. DR. CHARLES MARCH, Los Angeles, California. You mentioned that this is an appropriate method of contraception for those women who are selected appropriately. On the basis of not only a review of the literature but your review of these data, who are appropriate candidates and what preinsertion and postinsertion studies ought to be done? DR. MARy MARTIN, San Francisco, California. Those of us who had the impression that IUDs are associated with infection also thought that the risk was associated with insertion. A recent resident survey suggested that about 50% of residents were graduating without ever having placed an IUD. The question are as follows: Is it true that the insertion is a high-risk event? Is experience a factor? Should inexperienced individuals prescribe antibiotics? DR. JOSEPH IiANss, Phoenix, Arizona. I will add a fourth category of bias, and that is a physician bias. I am older than many of you and I recall that in 1960 Dr. Clyde Randall, when he presented IUDs to the student body at the University of Buffalo, talked about the wishbone pessary. He stated that it was made of platinum, was worn only by prostitutes, was inserted only by chiropractors, and every obstetrician had a moral obligation to remove them and under no circumstance to reinsert them. The following year, in either 1961 or 1962, Dr. Jack Lippes talked to the same group of students and introduced the Lippes Loop. DR. Roy (Closing). Dr. Caillouette, I am aware that the Food and Drug Administration does have files of drug-device events records, but this particular analysis did not avail itself of that as an additional source. We were looking at the incidence of pelvic inflammatory disease within the framework of the studv population. In response to Dr. Henderson, I disagree with the basic premise that the IUD produces "chronic endometritis with an inflammatory exudate." To the contrary, the presence of an IUD produces a sterile inflammatory response that is not evidence of a "compromised host." Indeed, Moyer l has reported "In the Macaca mulatta, the presence of an IUD in the uterine

Roy and Azen

1615

cavity intensifies the inflammatory response to bacteria and sperm eliminating each more quickly than would normally occur if an intrauterine foreign body would not be present." With. respect to the diagnosis of pelvic inflammatory di!\ease, this particular study was designed so that these patients, who were known to the clinicians, were undergoing pelvic examinations on a regular, routine basis at 1, 2, 3, 6, 12, and 18 months after insertion, as well as interim visits if necessary. The regularity Of the pelvic examinations, coupled with the request that the clinicians note "changes in pelvic findings," reduces any likelihood that the clinicians missed a low-grade infection. In the literature sexually transmitted chlamydia has been associated with lowgrade infection ("su~clinical infection"), but there are no consistent data associating IUD usage and ectopic gestation or infertility with chlamydial titers. The extensive number of clinical trials, which seem to be fairly consistent in their conclusion that the IUD does not cause pelvic inflammatory disease, stand in contrast to the case-control studies, even by the same authors, which differ in their conclusions over time. To answer Dr. Lee, the 2% annual pelvic inflammatory disease rate quoted in this study is not 2% above the basal rate. The 2% figure appears to be the annual rate of pelvic inflammatory disease in fecund women, which is approximately the same as what was shown by our analysis. Dr. March asked about appr~priate patient selection, which obviously is a critical issue, and I think it reflects what we have learned over the last 25 to 30 years. The IUD does not protect against the development .of upper-genital-tract infection. For those women who are planning future childbearing and who are not in a mutually monogamous relationship, it is not a contraceptive method that should be considered first. But for those women, even nulliparous ones, for whom steroidal methods of contraception or coitally related methods are not suitable, the IUD may be selected appropriately. If patients have evidence of cervical or vaginal discharge or if they have any evidence of tenderness of the upper genital tract, we know that IUDs should not be inserted. It is probably not necessary to perform white blood cell counts or treat with antibiotics. On the other hand, Dr. David Grimes and others haye done studies where they have shown benefits from using prophylactic antibiotics at the time of IUD insertion if the findings at pelvic examination are entirely normal. To further clarify this question an ongoing National Institutes of Health-funded study currently underway by the Los Angeles Regional Family Planning groups is examining this issue. Dr. Martin mentioned the insertion procedure and the lack of experience of our residents in the insertion technique. There is an insertion-associated risk of upper genital tract infection that is probably not related to the experience of the inserter. If a standard protocol is followed, where your insertion is not done in any

Cameron, Gown, and Tamimi

individuals who have any evidence of infection, only those who do not, and the cervix is prepared before insertion, then the act of insertion is probably not related to postinsertion infection. For most types of IUDs insertion is performed with the withdrawal technique, in which the holding tube is placed into the uterus and then the device is released by withdrawal of the outer sheath, and the IUD is left in situ, much the same way as levonorgestral implants are inserted. I agree that there is physician bias; each of us is biased. We also have volition, and I think that, as Dr.

June 1994 Am J Obstet Gynecol

Soderstrom so carefully noted, there is a diversity of opinion regarding this topic. It is a matter of making a judgment as to what we believe is the truth and to provide this information to our patients so that, with our help, they can choose the method of contraception suitable to them. REFERENCE 1. Moyer DL, Shaw ST Jr, Darwish N, et aI. Investigations of intrauterine devices in Macaca mulatta monkeys. Acta En-

docrinol 1972;71:381-6.

Expression of c-erb B-2 oncogene product in persistent gestational trophoblastic disease Brian Cameron, MD: Allen M. Gown, MD,b and Hisham K. Tamimi, MD-

Seattle, Washington OBJECTIVE: Much debate exists on the initiation of chemotherapy for women at risk for persistent gestational trophoblastic disease. This is a result of a lack of earty predictors for the development of persistent gestational trophoblastic disease after evacuation of a complete hydatidiform mole, because the only current reliable method of detection and diagnosis lies in persistent or rising postmolar ~-human chorionic gonadotropin values. We used immunocytochemical techniques to retrospectively study the expression of the c-erb B-2 oncogene product in formalin-fixed, paraffin-embedded trophoblastic tissues as a potential indicator of the· development of persistent gestational trophoblastic disease. STUDY DESIGN: In this retrospective study 56 trophoblastic tumors were examined by means of immunocytochemical techniques to stain for the oncogene product for evidence of c-erb B-2 expression. Our 56 cases included original tissue from 20 cases of complete mole that progressed to persistent gestational trophoblastic disease, seven cases of choriocarcinoma after term pregnancy or abortion, and 29 cases of hydatidiform mole representing postevacuation, spontaneously regressing disease (including one partial mole). We also studied 11 cases of first-trimester trophoblast and 15 cases of term placenta as additional controls. RESULTS: Our results showed positive immunostaining for c-erb B-2 gene product in one case of persistent gestational trophoblastic disease, with negative staining in all other cases in the study groups and controls. CONCLUSION: Analysis for the significance of c-erb B-2 expression in persistent gestational trophoblastic disease showed that this correlation between c-erb B-2 expression and persistent gestational trophoblastic disease is not significant, suggesting that future efforts should be directed at the involvement of different oncoproteins. (AM J OasTET GYNECOL 1994; 170: 1616-22.)

Key words: Trophoblastic disease, oncogene, c-erb B-2 oncogene

From the Departments of Obstetrics and Gynecology" and Pathology, b University of Washington. Presented at the Sixtieth Annual Meeting of the Pacific Coast Obstetrical and Gynecological Society, Blaine, Washington, September 7-12, 1993. Reprint requests: Hisham K. Tamimi, MD, University of Washington, Department of Obstetrics and Gynecology (RH-20), Seattle, WA 98195. Copyright © 1994 by Mosby-Year Book, Inc. 0002-9378/94 $3.00 + 0 6/6/54574

1616

Recent advances demonstrated various clinicopathologic factors associated with a. definite increased risk for persistent gestational trophoblastic disease after evacuation of molar pregnancy. To date, none of the following factors have allowed unequivocal detection of a subgroup of patients who will predictably have persistent gestational trophoblastic disease. These factors include clinical presentation, and conventional histologic, immuno-