Chlamydia trachomatis antibody titers and hysterosalpingography in predicting tubal disease in infertility patients

Transcription

1 FERTILITY AND STERILITY Copyright c 1994 The American Fertility Society Printed on acid-free paper in U. S. A. Chlamydia trachomatis antibody titers and hysterosalpingography in predicting tubal disease in infertility patients Susan F. Meikle, M.D.t Xiaozheng Zhang, M.D.t William M. Marine, M.D.t B. Ned Calonge, M.D.t:\: Richard F. Hamman, M.D., Dr.P.H.t George Betz, M.D., Ph.D. University of Colorado Health Sciences Center and the Colorado Permanente Medical Group, Denver, Colorado Objective: To determine if the number of diagnostic laparoscopies done on women without tubal adhesive disease could be reduced by testing for tubal disease with Chlamydia trachomatis antibody titers and hysterosalpingography (HSG), either singly or together. Design: Historical prospective chart review Setting: The Colorado Kaiser Permanente Reproductive Endocrinology Clinic. Patients: All 703 infertility patients who had C. trachomatis antibody titers done from March 2, 1988 to April 30, The final study group was comprised of 218 patients who had antibody titers, HSG, and laparoscopy. Interventions: None. Main Outcome Measure: Sensitivity, negative predictive value, and false-positive rate were the test characteristics of interest. Tubal disease was identified by laparoscopy. Results: For HSG testing, the sensitivity was 78% and the negative predictive value was 85%. For C. trachomatis titers, the sensitivity was also 78% and the negative predictive value was 82%. Ninety-five percent confidence intervals for sensitivity and negative predictive value overlapped, indicating that there was no significant difference. However, false-negative rates were the same for the two tests, but false-positive rates were lowest for HSG and series testing.. Conclusions: To minimize false-positive tests and thus, to minimize unnecessary laparoscopies, HSG testing either alone or combined with the C. trachomatis antibody titer as series tests yielded a significantly lower false-positive rate. In our study group, if both tests were negative, tubal disease was identified on laparoscopy in only 5% of cases. Choice of most cost-effective test sequence will depend on who bears the cost. Further studies of cost-benefit using well-defined testing sequences are needed to determine if C. trachomatis antibody titers in series with HSG would be more cost effective than HSG alone in detecting tubal disease. Fertil Steril1994;62: Key Words: C. trachomatis antibody titers, predictive value, false-positive rate, infertility, tubal disease Infertility, defined as 1 year of unprotected intercourse without conception, is an important public health problem (1-3). Infertility due to tubal disease affects 10% to 20% of North American couples (4). The most common disease that alters the ana- Received September 2, 1993; revised and accepted March 24, * Supported a grant from the Colorado Kaiser Permanente Research and Development Committee, Denver and training grant 5033AH18002 from the University of Colorado Depart- ment of Preventive Medicine, Denver, Colorado. t Department of Preventive Medicine and Biometrics, University of Colorado. :j: Department of Preventive Medicine, Colorado Permanente Medical Group. Department of Reproductive Endocrinology, Colorado Permanente Medical Group. II No reprints available. Corresponding author and present address: Susan F. Meikle, M.D., Department of Obstetrics and Gynecology, Denver General Hospital, 777 Bannock Street, Denver, Colorado (FAX: ). Meikle et ai. Predicting tubal disease 305

2 r tomic integrity ofthe tube is salpingitis (5, 6). Epidemiologic studies over the last two decades have demonstrated rising rates of sexually transmitted diseases associated with a secondary epidemic of pelvic inflammatory disease, resulting in a tertiary epidemic of tubal occlusion, infertility, and ectopic pregnancy (EP) (4,7). Neisseria gonorrhea and Chlamydia trachomatis are the two organisms most frequently related to upper genital tract infection in the United States (8,9). Genital chlamydial infections are estimated to occur at a rate two times that of N. gonorrhea infections (8). A definite role for C. trachomatis as an etiologic agent for salpingitis has been demonstrated in the United States and Canada and specific serotypes have been identified (8, 10-12). The sequence of sexually transmitted disease (STD) to salpingitis and then to infertility appears common but not predictable. Asymptomatic pelvic infection has been reported, especially with C. trachomatis. Evidence for this is the presence of tubal disease on laparoscopy in women with no history of an acute episode of pelvic inflammatory disease (PID) (6, 9, 11). Consequently, the diagnosis of tubal disease cannot rely solely on presence or absence of a history of PID. Commonly used methods of diagnostic testing for tubal disease are hysterosalpingography (HSG) and laparoscopy. Hysterosalpingography has been used in the evaluation of infertility to demonstrate tubal patency, myomata, and uterine anomalies for >70 years (13). Hysterosalpingography has been routine in many infertility centers as a preliminary investigation tool because it is less costly and less invasive than laparoscopy. In comparing HSG with laparoscopy, however, laparoscopy is superior in detecting peritoneal adhesions and endometriosis (13, 14). Although laparoscopy is considered the gold standard for evaluation of tubal disease, it is invasive and costly. Noninvasive methods that yield comparable diagnostic results would be useful in a cost contained system. Laboratory tests such as enzyme-linked immunosorbant assays (ELISA) and indirect fluorescent antibody (IF A) testing have been proposed as alternative methods of testing for tubal disease. Microimmunofluorescence (MIF) has been considered the most reliable test for detecting past exposure to C. trachomatis but is not commercially available (9). Enzyme-linked immunosorbant assays are widely available and compare favorably with MIF testing (12, 16). Among women with tubal disease infertility, the prevalence of positive C. trachomatis 306 Meikle et al. Predicting tubal disease antibody titers is higher than in pregnant controls without tubal disease in several settings (2, 4, 10, 11,15). Indirect fluorescent antibody testing is also widely available but requires more technical training to perform than ELISA. This technique involves use of cell culture and closely resembles MIF. False positives are minimized with IF A. Knowledge of test or operating characteristics, which include sensitivity, specificity, and predictive values, can help in interpreting results and in selection of the best test to be performed. We targeted a negative predictive value, the likelihood that a negative test result indicates the absence of disease; sensitivity, the proportion of people with a disease who have a positive test; and the false-positive rate, the probability that a test will be positive when the disease is absent, as the test characteristics most useful in evaluating their ability to detect the absence of tubal disease. With two different tests available, combination testing became an option. Combination testing can increase the performance of certain test characteristics. Two types of combination testing are generally used: series and parallel. Series testing requires both tests to be positive to make the diagnosis while parallel testing requires either test to be positive for diagnosis. Series testing increases specificity and parallel testing increases sensitivity (17). As medical costs in this country rise, a new attention to quality and cost of care has become apparent. The goal of this research was to compare the diagnostic utility of two less expensive, less invasive tests, C. trachomatis antibody titer assay and HSG, either alone or in combination, to laparoscopy for tubal disease. Our hypothesis was that C. trachomatis antibody testing either alone or in combination with HSG would predict tubal disease absence so that laparoscopic testing would not be required in patients without tubal disease. Such patients could then proceed with a medically managed treatment plan. MATERIALS AND METHODS A historical prospective study was carried out on patients who sought treatment for infertility at Kaiser Permanente in Denver, Colorado from March 2, 1988 to April 30, Chlamydia trachomatis titers were first obtained in 1988 by one ofthe two physicians in this infertility practice and by 1990, titers were part of routine protocol. Patients were included in the study if a C. trachomatis anti- Fertility and Sterility

3 body titer was recorded in their chart. Routine history and physical exams were carried out on all patients. In general, most patients underwent a period of medical management to determine ovulatory function before laparoscopy. Hysterosalpingography was considered part of the initial evaluation and results were available for 80% of the population. Laparoscopy was scheduled without regard to C. trachomatis antibody titer. The C. trachomatis titer test used during this time period was an IF A for immunoglobulin (Ig)G antibodies against C. trachomatis using serotype L2 cells as antigen, distributed as the Pharmacia VIRGO test kit by Schiapparelli Biosystems, Inc. (Columbia, MD). This assay compares well with the MIF technique of the same laboratory (Vivian Abadom, personal communication). The same test has been used at Kaiser Permanente for the entire time period ofthe study. Charts were abstracted by one of two authors (S.F. Meikle or X. Zhang) on a standardized data collection form. Data collected included past medical history, infertility history, surgical history, treatment methods, HSG results, and laparoscopy results. Hysterosalpingography films were reviewed by one of two reproductive endocrinologists whether the film was obtained outside or within the Kaiser Permanente system. Laparoscopy reports were coded on a standardized form by one author (S.F. Meikle) blinded to the C. trachomatis antibody titer. Data entry was 100% verified and 5% of charts and laparoscopy reports were recoded for agreement. There was 98% agreement on the charts that were reabstracted. Tubal disease was determined by review of the laparoscopic operative report and included any adhesions involving the tube, clubbing of the tube, obstruction to dye, or hydrosalpinx. Adhesions limited to the ovary were not included in this definition. Tubal disease due to endometriosis was not separately categorized but did not appear to be a confounding factor. Only 16% of patients with tubal disease on laparoscopy had evidence of endometriosis versus 30% of those patients classified as free of tubal disease. Tubal disease seen on HSG was defined as obstruction to dye or abnormal dye patterns as indicated by the film reviewer. When. laparoscopic and HSG findings did not agree, laparoscopic findings were considered final. A history of PID, diethylstilbesterol (DES) exposure, or intrauterine device (IUD) use was per patient report. Statistical analysis was performed using SAS software version 6.03 (SAS Institute, Cary, NC). The Wilcoxon rank-sum test was used to compare titer distributions. Confidence intervals around operating characteristics were calculated using the normal theory approximation to the binomial distribution (18). Differences in proportions and means were calculated using the x 2 test and analysis of variance, respectively. Study Population Laparoscopy and HSG results were not available on all patients. Of the original 703 patients with titers, 280 underwent laparoscopy either at Kaiser Permanente or before enrolling in the clinic, leaving 423 patients without laparoscopy results. Because of the period of medical management, there was a several month delay before performing laparoscopy, during which time 186 patients became pregnant. The remaining 237 patients that did not undergo laparoscopy were either currently undergoing infertility therapy, had been referred for IVF services elsewhere, had lost insurance coverage, or had withdrawn from clinic services. Of the 280 patients who underwent laparoscopy, we excluded those with previous tubal and/or ovarian surgeries (n = 33) such as salpingoophorectomies or tubal reversals because we could not determine whether tubal disease was due to pelvic infection or surgery. If there was more than one laparoscopy report available, we chose the report of the date closest to the titer or before tubal or ovarian surgery. Of the remaining 247 patients, 29 either did not undergo HSG or the films were not available for review, leaving the study group of 218 with both laparoscopy and HSG results. Patients with previous appendectomies were not excluded. Although the proportion of patients with previous appendectomies was not significantly different in the study group (6%) compared with the entire population (5.6%), the number of appendectomies was higher in the study group with tubal disease (n = 9) compared with the study group without tubal disease (n = 4; P = 0.04). However, laparoscopy reports did not indicate involvement of the appendiceal stump for any of these patients. Fully 72% of patients had undergone infertility evaluations elsewhere before the evaluation at the Kaiser Permanente center. Thus, C. trachomatis titers were not always drawn before laparoscopy. Antichlamydial antibody titers have been reported to remain elevated on the average for >5 years (6). Eighty-five percent of laparoscopies were performed within 1 year of the titer. Overall, 79% of Meikle et ai. Predicting tubal disease 307

4 r laparoscopies and 56% of HSGs were performed in the same Kaiser Permanente clinic and subsequent to the C. trachomatis titer. When the data were analyzed to compare titer distributions in the group with antibody titers done before laparoscopy to the group with antibody titers done after laparoscopy, there was no significant difference in distribution (P = 0.2). Thus, we pooled these patients into the final study group of 218. Patients RESULTS There were 703 subjects with C. trachomatis titers. Ofthese, 218 had both HSG and laparoscopy reports available and will be the subject of most analyses. A distribution of the entire population is shown in Figure 1. The distributions of the C. trachomatis antibody titers in the group of patients who underwent laparoscopy and HSG (n = 218 total) by presence or absence of disease is shown in Figure 2. Women with tubal disease had significantly higher antibody titers than women without tubal disease (P < ). We were interested in the antibody titer cutoff point that would yield the highest negative predictive value and lowest false-positive rate, becuaw false-positive test results would lead to unnecessary laparoscopic testing. Negative predictive value is driven by prevalence of disease and by test sensitivity, and thus, we were also interested in a cutoff with high sensitivity. A receiver operator characteristic (ROC) curve is presented in Figure 3, which shows the true positive rate, the false-positive rate, and the specificity of the C. trachomatis antibody assay at each titer level in the study group. An ROC curve can be used to decide the optimum cutoff point depending on the purpose of the test (17). The <1:8 1:8 1:16 1:32 1:64 1:128 1:256>1:256 Chlamydia Antibody Titer Figure 2 Chlamydia trachomatis antibody titer distributions., with tubal disease (n = 87); D, without tubal disease (n = 131, P = ). highest sensitivity and negative predictive value with the lowest false-positive rate for the C. trachomatis antibody titer assay were obtained when the cutoff for a positive result was placed at >1:16. Table 1 displays the descriptive characteristics of three patient subgroups: the study group with all True Positive Rete SpecifiCity L-.L.-L...I---L-l..-L.-L-...I.-J ~1:512 O~~~~~~-r~~ o Fal.. Polllive Rete t Figure 1 Study population. Figure 3 Receiver operator characteristic curve for Chlamydia trachomatis antibody titers. *Sensitivity; tloo% - specificity. 308 Meikle et al. Predicting tubal disease Fertility and Sterility

5 Table 1 Descriptive Characteristics of Selected Patient Subgroups Age (y)* Married Single Gravidityt None One or more Living childrent None One or more Years attempted pregnancy* Previous IUD uset DES exposuret History of PlOt Pregnant Study group controls Not pregnant (n = 218) (n = 186) (n = 237) 32 ± (95) 10 (5) 121 (56) 97 (44) 178 (82) 40 (17) 4.1 ± (14) 4 (2) 17 (8) * Values are means ± SD. t Values in parentheses are percents. *P < ± (95) 9 (5) 67 (36) 119 (64) 139 (75) 47 (25) 2.5 ± (16) 3 (1.6) 11 (6) 32 ± 5.5t 214 (90*) 23 (10) 118 (50*) 119 (50) 184 (78) 53 (12) 4.2 ± 3.5* 46 (19) 1 (0.4) 19 (8) tests available, subjects who became pregnant before laparoscopy, and women lost to follow-up. Marital status was significantly different among the three groups. Patients in the lost to follow-up group tended to be single. As expected, a significantly higher proportion of patients were nulligravida in the study group and nonpregnant group compared with the pregnant controls. The period of medical management used in the clinic effectively removed patients from the study group with prior successful pregnancies who were experiencing abnormalities of ovulatory function and timing. These subjects became the pregnant controls. There was no difference in PID, IUD, or DES exposure, which are known risk factors for tubal disease, among the three groups. For all three groups combined, the prevalence of tubal disease was 23%, comparable to the expected 10% to 20% reported in the literature (4). The numbers of subjects categorized by their diagnostic test results and compared with laparoscopy are presented in Table 2. These data were used to calculate the operating characteristics for C. trachomatis antibody titers and HSG alone as well as for parallel and series testing. Table 3 shows are the operating characteristics, 95% confidence intervals, and the false-positive and -negative rates for each type of testing compared with laparoscopy. Chlamydia trachomatis antibody titers and HSG testing yielded equal sensitiv- ities of 78% and negative predictive values of 82% and 85%, respectively. Focusing on the test characteristics of importance for our study, the sensitivity and negative predictive value of C. trachomatis titers and HSG are not significantly different as indicated by the overlapping confidence intervals. However, the significantly lower false-positive rate for HSG testing makes it the superior test for our purposes. Combining the tests in series increased the specificity while decreasing the negative predictive value and false-positive rate. However, confidence intervals for sensitivity and negative predictive value overlapped between series testing and HSG, indicating that although a larger sample size may clearly differentiate between the two types of testing, this sample did not. The lowest false-positive rate was obtained with series testing although this confidence interval came very close to overlapping with the HSG result. Parallel testing yielded the highest sensitivity and negative predictive value at the expense ofthe highest false-positive rate. Thus, for our study goal of reducing unnecessary laparoscopies, HSG and series testing provided the highest sensitivities and negative predictive values while minimizing the false-positive rates. We also examined different definitions for the gold standard to see if it made substantial differences in results for C. trachoma tis testing. When adhesions were grouped into unilateral or bilateral disease on laparoscopy for the study group, the 95% confidence intervals for the sensitivity, negative predictive value, and false-positive rate for the C. trachomatis assay overlapped with the 95% confidence intervals for the standard definition of tubal disease. In another approach, we considered an intrauterine pregnancy to be absence of tubal disease and an EP to be presence of tubal disease among subjects who became pregnant. When these women were added to the study group, the operating characteristics for the C. trachomatis titer also over- Table 2 Diagnostic Test Results Tubal Disease on HSG/chlamydia Laparoscopy titer results + Total +HSG/+Titer HSG/-Titer HSG/+Titer HSG/-Titer Total Meikle et at. Predicting tubal disease 309

6 r Table 3 Operating Characteristics and False-Positive and -Negative Rates for Tubal Disease Using Laparoscopy as the Gold Standard (n = 218)* Operating characteristics Chlamydia trachomatis titer HSG Seriest Parallel:j: Sensitivity 78 (69, 87) Specificity 64 (56, 72) Positive predictive value 59 (50,68) Negative predictive value 82 (75,89) False-positive rate 36 (28,44) False-negative rate 22 (13, 31) * Values in parentheses are 95% confidence limits. t Series testing means both tests are positive. 78 (69, 87) 61 (51, 71) 95 (90, 100) 84 (78,90) 95 (91, 100) 53 (44,62) 76 (67,85) 88 (80, 96) 58 (48, 68) 85 (79,91) 78 (72, 84) 95 (90, 100) 16 (10, 22) 5 (1,9) 47 (38, 56) 22 (13, 31) 39 (29, 49) 5 (0, 10) :j: Parallel testing means either test is positive. lapped those of the standard definition. Thus, the C. trachomatis antibody titer performed similarly with either definition of the gold standard. We were interested to learn if the diagnosis of tubal disease had any prognostic significance for pregnancy outcome so we examined the pregnancy outcome data for the study group. In the study group with tubal disease, the intrauterine pregnancy rate for the first recorded pregnancy was 22% (n = 19) versus 42% (n = 55; P < 0.01) for the study group without tubal disease. Numbers for EPs were too small to analyze, with only one EP in the tubal disease group compared with five in the group without tubal disease. DISCUSSION Based on sensitivity, the negative predictive value, and the false-positive rate as the most important test characteristics for reducing unnecessary laparoscopies, we determined that HSG and C. trachomatis antibody titers by themselves yield comparable results for sensitivity and negative predictive value, but HSG yields a significantly lower falsepositive rate. Consequently, if choosing the single test to provide the highest sensitivity and negative predictive value with the lowest false-positive rate for this health maintenance organization (HMO) population, HSG would be the choice. For combination testing, series testing performed as well as HSG and perhaps better when considering the false-positive rate. A larger sample size may more clearly delineate between the two types of tests. Thus, our original hypothesis that C. trachomatis antibody testing would aid in reducing unnecessary laparoscopies was only partially supported. Choosing the appropriate test or tests to diagnose tubal disease will depend on the clinical context and financial capacity of the individual or system. For example, a patient may decide that HSG is too uncomfortable to obtain and may choose to have an antibody titer performed in lieu ofhsg. The health care system may choose to continue with HSG testing alone because the number of false positives (16%) is minimized compared with either C. trachomatis titers alone (36%) or parallel testing (47%), thus decreasing the number of patients undergoing expensive and invasive laparoscopy. However, a cost analysis would have to be done to weigh the cost oflaparoscopy against the cost of medical management' which can also be expensive over time. It would also be necessary to measure the benefits of laparoscopy, such as increased fertility rates or shortened time from attempted pregnancy to conception, in this population. This study has several limitations. Although the population subgroups have been clearly defined, the element of time of enrollment in the clinic has not been evaluated here. Longer enrollment for patients will result in a higher laparoscopy rate as well as fluctuating pregnancy rates. Other limitations of the study include the IF A test used, which may cross-react significantly with other chlamydial organisms, limiting the specificity of the test. Additionally, there is tubal disease not associated with C. trachomatis or gonorrhea in up to 25% to 50% of cases, which limits the use of the C. trachomatis antibody assay in detecting ali previous cases of PID (19). We did not exclude patients with previous appendectomies and this may have misclassified a small amount of tubal disease in our study group. Our definition of tubal disease, any amount of adhesions involving the tube, may have diminished the difference in tubal disease and C. trachomatis titers between the study group and pregnant controls. It is likely that small, filmy adhesions do not preclude pregnancy and may be present in the 310 Meikle et al. Predicting tubal disease Fertility and Sterility

7 women who became pregnant before laparoscopy. Also, without direct visualization of the tubal mucosa, both HSG and laparoscopy will miss microscopic disease that may play an important role in tubal dysfunction. Comparison to other studies in the literature is limited by the use of multiple types of assays and lack of information about operating characteristics of each type of test compared with a standard such as cell culture or MIF. However, Minassian et al. (20) reported the predictive values of C. trachomatis antibody titers as measured by an ELISA test as a means to select either HSG or laparoscopy for detection of tubal disease for patients pooled from several private practices in Pennsylvania. These authors reported on 258 infertility patients who had C. trachomatis antibody titers drawn as part of an initial infertility evaluation. Fifty-eight patients had C. trachomatis antibody assays, HSG, and laparoscopy, a smaller group than presented here. Results showed that the absence of C. trachomatis antibody correctly predicted the absence of tubal adhesions in 86.9% of patients who underwent laparoscopy. With the addition of a normal HSG, the predictive value for no tubal disease was raised to 95.2% although no measure of variance was included. Our study is similar in design but differs in population, prevalence of tubal disease on laparoscopy, and type of antibody assay. The prevalence of tubal disease was twice as high in the patients who underwent laparoscopy in Minassian's study (87%) compared with ours (43 %) despite apparently simi- 1ar definitions of tubal disease. Generalization of Minassian's results and our results to other patient populations should be done carefully, taking into account that the prevalence of tubal disease will change the predictive values. The assay used in our study may have a higher negative predictive value when used in a clinical setting with a lower prevalence of tubal disease, making it a more useful test. Cost to the patient and cost to the health care system need to be considered when selecting the testing regimen. In this clinical setting, false-positive tests increase cost to the medical care system by increasing the number of laparoscopies. False negative tests increase cost to the medical care system through ineffective medical management. In summary, C. trachomatis antibody titers by the IF A method described here did not perform as well as HSG when targeting the false-positive rate in this HM 0 setting. Series testing, however, was comparable to HSG testing alone for detecting tubal disease when compared with laparoscopy and a larger sample size may show it to be superior. In a subpopulation of our study group, if both tests were negative, tubal disease was identified in only 5% of cases. When considering all test results, however, the performance of series testing does not appear significantly different enough from the performance of HSG testing alone to justify the added cost. In this HMO infertility clinic setting, HSG alone provides optimal test characteristic performance in detecting tubal disease while minimizing the laparoscopy rate. REFERENCES 1. Speroff L, Glass RH, Kase NG. Clinical Gynecologic Endocrinology and Infertility. 4th ed. Baltimore: Williams & Wilkins, 1989: Reniers J, Collet M, Frost, Leclerc A, Ivanoff B, Meheus A. Chlamydial antibodies and tubal infertility. Int J Epidemiol 1989;18: Cates W, Wasserheit IN. Genital chlamydia infections: epidemiology and reproductive sequelae. Am J Obstet Gynecol 1991;164: Kosseim M, Brunham RC. Fallopian tube obstruction as a sequela to Chlamydia trachomatis infection. Eur J Clin Microbiol 1986;5: Mattingly RF, Thompson JD. Te Linde's operative gynecology. 6th ed. Philadelphia: J.B. Lippincott Company, 1985: Ruijs GJ, Kauer FM, Jager S, Schroder FP, Schirm J, Kremer J. Further details on sequelae at the cervical and tubal level of Chlamydia trachoma tis infection. in infertile women. Fertil Steril 1991;56: Sweet RL, Gibbs RS. Infectious diseases of the female genital tract, Part 1. Baltimore: Williams & Wilkins, 1985: Cates W, Rolfs RT, Aral SO. Sexually transmitted diseases, pelvic inflammatory disease, and infertility: an epidemiologic update. Epidemiol Rev 1990;12: Barnes RC. Laboratory diagnosis of human Chlamydial infections. Clin Microbiol Rev 1989;2: Conway D, Caul EO, Hull MG, Glazener CM, Hodgson J, Clarke SKR, et al. Clamydial serology in fertile and infertile women. Lancet 1984;1: Walters MD, Eddy CA, Gibbs RS, Schachter J, Holden AE. Antibodies to Chlamydia trachomatis and risk for tubal pregnancy. Am J Obstet Gynecol 1988;159: Finn MP, Ohlin A, Schachter J. Ep.zyme-linked immunosorbant assay for immunoglobulin G and M antibodies to Chlamydia trachomatis in human sera. J Clin Microbiol 1983;17: Siegler AM. Hysterosalpingography. Fertil Steril 1983; 40: Corson SL. The role of laparoscopy in the infertility workup. J Reprod Med 1977;18: Wessels PH, Viljoen GJ, Marais NF, de Beer JAA, Smith M, Gericke A. The prevalence, risks, and management of Chlamydia trachomatis infections in fertile and infertile patients from the high socioeconomic bracket of the South African population. Fertil Steril 1991;56: Meikle et a1. Predicting tubal disease 311

8 r 16. Evans RT, Taylor-Robinson D. Developemnt and evaluation of an enzyme-linked immunosorbant assay (ELISA), using chlamydial group antigen, to detect antibodies to Chlamydia trachomatis. J Clin Pathol 1982;35: Griner PF, Mayewski RJ, Mushlin AI, Greenland P. Selection and interpretation of diagnostic tests and procedures. Principles and applications. Ann Intern Med 1981;94: Rosner B. Fundamentals of biostatistics. 3rd ed. Boston: PWS-Kent Publishing Company, 1990:17l. 19. Rice PA, Schachter J. Pathogenesis of pelvic inflammatory disease: what are the questions? JAMA 1991;266: Minassian SS, Wu CH, Jungkind D, Gocial B, Filer R, Glassner M. Chlamydia antibody, as determined with an enzyme-linked immunosorbent assay, in tubal factor infertility. J Reprod Med 1990;35: Meikle et al. Predicting tubal disease Fertility and Sterility