Correspondence
Screening for Chlamydia to Prevent Pelvic Inflammatory Disease
N Engl J Med 1996; 335:1531-1533November 14, 1996
- Article
To the Editor:
In the study by Scholes et al. (May 23 issue),1 which examined whether screening for and treatment of chlamydia prevented pelvic inflammatory disease, 1009 women were randomly assigned to screening and treatment, but only 645 were actually screened. Of these 645 women, 44 had positive tests and received treatment, with 7 cases of pelvic inflammatory disease subsequently identified. However, the authors did not state whether any of the cases occurred in the 44 women who tested positive for cervical chlamydial infection. The remaining 364 women were not screened, received no intervention at all, and hence do not appear to differ from the controls.
In this group of 364 women, two cases of pelvic inflammatory disease were reported. Assuming a mean follow-up period of 11.42 months (since the length of follow-up in this group is not reported), the incidence of pelvic inflammatory disease is 4.81 per 10,000 woman-months, much lower than that in the intervention or control group. Therefore, inclusion of the 364 women in the screening-and-treatment group represents a “misclassification of exposure (intervention)” at base line and results in a biased estimate of the risk.2 A more logical approach would be to add the 364 unscreened women to the usual-care group (which was also not screened) or exclude them from the analysis.
If the 364 participants are included in the usual-care group, the incidence of pelvic inflammatory disease is 9.47 and 15.6 per 10,000 woman-months in the screening-and-treatment group and the usual-care group, respectively. The relative risk of pelvic inflammatory disease in the screening-and-treatment group is 0.60 (95 percent confidence interval, 0.22 to 1.3; P = 0.22). Alternatively, if the 364 are excluded from the analysis, the respective rates change to 9.47 and 18.0 per 10,000 woman-months (relative risk, 0.52; 95 percent confidence interval, 0.19 to 1.2; P = 0.11). Thus, in either case the results would not achieve statistical significance. In addition, we believe that Cox regression models3 would be a better choice than logistic regression for adjusting for the confounding factors in this study.
3 ReferencesElfatih I.M. Abter, M.D.
Mohammad A. Mahmud, M.D., Ph.D.
Beth Israel Medical Center, New York, NY 100031
Scholes D ,Stergachis A ,Heidrich FE ,Andrilla H ,Holmes KK ,Stamm WE . Prevention of pelvic inflammatory disease by screening for cervical chlamydial infection. N Engl J Med 1996;334:1362-1366
Full Text | Web of Science | Medline2
Information bias. In: Kleinbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. London: Lifetime Learning, 1982:220-41.
3
Cox DR . Regression models and life-tables. J R Stat Soc [B] 1972;34:187-220
To the Editor:
We applaud Scholes and her colleagues for their efforts to show the effectiveness of a population-based approach to the selective screening of women at risk for chlamydia. Although they chose the most powerful study design — the randomized, controlled trial — they committed a methodologic mistake that makes the results hard to interpret. Premature randomization of subjects, before the intervention, may affect the chance that a subject will eventually participate in a trial, leading to a potential mismatch between groups and possible bias.1
In this case, all 36,547 female members of a health maintenance organization who were 18 to 34 years of age were randomly allocated to the two study groups, and then recruitment began. Each woman in the two cohorts should have had the same probability of receiving one or the other treatment (standard care or the offer of screening for chlamydia). However, recruitment was pursued much more avidly, apparently with telephone calls (from blinded callers?), in the intervention group than in the control group, resulting in a shortfall of approximately 20 percent, or 400 women, in the control group. In their haste to recruit the intervention group they may have biased the composition of the two groups; it is now impossible to tell for certain. The purpose of randomization is to attempt to achieve two groups that are similar with respect to known as well as unknown risk factors. The validity of the statistical testing of the results of a trial is based on this.
This example of putting the cart before the horse in a randomized trial should serve as a reminder that random allocation should be delayed until the last possible moment before the intervention. In this case it would have been better to randomize the women after they had been determined to fulfill the eligibility criteria.
1 ReferencesJohn Sellors, M.D.
McMaster University, Hamilton, ON L8V 1C3, CanadaJorma Paavonen, M.D.
Helsinki University Hospital, 00290 Helsinki, FinlandTo the Editor:
Scholes and colleagues suggest that screening for cervical chlamydial infection can prevent pelvic inflammatory disease. Any intervention that could reduce the incidence of pelvic inflammatory disease would make a major contribution to female reproductive health. The diagnosis of pelvic inflammatory disease can be difficult, as the authors correctly state. In the context of such a study, it would have been unrealistic to expect laparoscopic confirmation of pelvic inflammatory disease. The clinical diagnosis of pelvic inflammatory disease has a relatively low specificity, and laparoscopic evaluation of clinically diagnosed cases has confirmed the diagnosis only in 15 percent1 to 75 percent2 of cases. Scholes and colleagues did not measure the number of episodes of pelvic inflammatory disease but rather the number of contacts between women with clinical symptoms or findings suggestive of pelvic inflammatory disease and health care providers.
Important differences in the screening and control groups could explain a difference in the use of services by women with symptoms of pelvic inflammatory disease. The intensive amount of contact (the subjects were contacted by letter; contacted to confirm their appointments; seen at a clinical appointment, which included vaginal examination; informed of the results of the tests; and treated in some cases) between health care providers and women in the screening but not the control group could have been associated in the subjects' minds with health education. This could have changed the sexual behavior (such as the use of condoms) of women in the screening group. Furthermore, after screening, the women were reassured. One could argue that health education, rather than screening and subsequent treatment, was responsible for the differences observed.
2 ReferencesRüdiger Pittrof, M.R.C.O.G.
London School of Hygiene and Tropical Medicine, London WC1E 7HT, United Kingdom1
Stacey CM, Munday PE. Abdominal pain in women attending a genitourinary medicine clinic: who has PID? Int J STD-AIDS 1994;5:338-42.
2
Morcos R ,Frost N ,Hnat M ,Petrunak A ,Caldito G . Laparoscopic versus clinical diagnosis of acute pelvic inflammatory disease. J Reprod Med 1993;38:53-56
Web of Science | Medline
To the Editor:
We wish to report a case of pelvic inflammatory disease due to Coccidioides immitis. This fungus is endemic in the southwestern United States. Coccidioidomycosis primarily affects the lungs but can spread to the meninges, skin, and bone. Although extremely rare, pelvic inflammatory disease due to coccidioides has been reported.1
A 56-year-old woman presented with a one-month history of fever and pelvic pain. On presentation, the patient was febrile. Physical examination revealed cervical-motion tenderness and bilateral inguinal adenopathy. The cervical mucosa was erythematous and friable. A Papanicolaou smear of the cervix revealed spherules of C. immitis organisms on staining. Histologic analysis and culture of inguinal lymph nodes revealed C. immitis organisms. The patient then underwent total abdominal hysterectomy with bilateral salpingo-oophorectomy. No other evidence of systemic coccidioidomycosis was present. Treatment with ketoconazole was initiated for a short period, with a satisfactory outcome.
Curiously, the patient's husband had previously had testicular coccidioidomycosis and had undergone orchiectomy, which revealed chronic orchitis and epididymitis. Examination of the testicular mass revealed granulomata with spherules of C. immitis. A culture of prostatic secretions also grew the organism. This patient had evidence of pneumonia on a chest radiograph, presumably the source of dissemination. He was treated with ketoconazole, with a good outcome.
It is tempting to speculate that the case in our patient represents one of sexually transmitted coccidioidomycosis. The possibility of sexual transmission of this disease has been investigated but never confirmed.1,2 The rarity of genitourinary coccidioidomycosis in either sex, the absence of systemic disease in the woman, and the presence of C. immitis organisms in the prostatic fluid of her husband support the possibility of sexual transmission of this disease, thus adding another organism to the list of those causing pelvic inflammatory disease through sexual transmission.
2 ReferencesJose A. Perez, Jr., M.D.
Edward L. Arsura, M.D.
Jennifer J. Abraham, M.D.
Kern Medical Center, Bakersfield, CA 933051
Saw EC ,Smale LE ,Einstein H ,Huntington RW Jr . Female genital coccidioidomycosis. Obstet Gynecol 1975;45:199-202
Web of Science | Medline2
Hart WR ,Prins RP ,Tsai JC . Isolated coccidioidomycosis of the uterus. Hum Pathol 1976;7:235-239
CrossRef | Web of Science | Medline
Author/Editor Response
The authors reply:
To the Editor: Drs. Abter and Mahmud suggest that it might have been more appropriate to place the women in the intervention group who were not screened in the control group or to exclude them from the analysis. However, the intention-to-treat analysis (in this instance, intention-to-screen), in which the original group assignment is maintained, regardless of compliance, is the analytic strategy that best preserves the strengths of the random group assignment.1,2 In addition, this analytic strategy provides a more useful evaluation of screening activities as they occur in actual practice than would an analysis of compliers alone.
We agree with Drs. Sellors and Paavonen that randomization after participants have been proved eligible and successfully recruited is the optimal approach, although other strategies have been proposed.3 Our study design, population-based approach, and health maintenance organization setting offered many methodologic advantages. Nonetheless, the demands of contacting this large volume of potential participants, identifying eligible women, and conducting screening with sufficient time and resources for follow-up necessitated group assignment earlier in the recruitment process. In the article we discussed the possibility that selection bias may have resulted. Although the possibility of bias cannot be ruled out, we disagree that it is not possible to evaluate whether such a bias may have occurred. Most important, we found the final study groups to be very similar with respect to important prognostic and other demographic variables (as shown in Table 1 of our article), and the relative risk remained strong and stable even after adjustment for potentially confounding variables (as shown in Table 2). These factors, in conjunction with a standardized recruitment strategy and the respondents' lack of knowledge about their group assignment, all provide evidence that selection bias, if present, did not have a major effect on the results.
Mr. Pittrof raises the possibility that the study results may reflect the effects of health education rather than screening. Our screening protocol contained no health-education components, and behavioral research unfortunately indicates that the amount of contact we had with our participants is unlikely to induce the adoption of safer sex practices, such as condom use.4,5
The report by Perez et al. of a case of pelvic inflammatory disease resulting from C. immitis infection is a useful reminder that pelvic inflammatory disease is a polymicrobial condition. Our study tested for the most important pathogen in pelvic inflammatory disease in our study population, but a proportion of cases, which will vary in different populations, will certainly not be found by screening for chlamydia.
5 ReferencesDelia Scholes, Ph.D.
Group Health Cooperative of Puget Sound, Seattle, WA 98101King K. Holmes, M.D., Ph.D.
Andy Stergachis, Ph.D.
Walter E. Stamm, M.D.
University of Washington, Seattle, WA 981951
Meinert CL. Clinical trials: design, conduct, and analysis. Vol. 8 of Monographs in epidemiology and biostatistics. New York: Oxford University Press, 1986.
2
Hulley SB, Cummings SR. Designing clinical research: an epidemiologic approach. Baltimore: Williams & Wilkins, 1988.
3
Zelen M . A new design for randomized clinical trials. N Engl J Med 1979;300:1242-1245
Full Text | Web of Science | Medline4
Aral SO ,Peterman TA . Defining behavioral methods to prevent sexually transmitted diseases through intervention research. Infect Dis Clin North Am 1993;7:861-873
Web of Science | Medline5
Kamb ML, Douglas JM, Rhodes F, et al. A multi-center randomized controlled trial evaluating HIV prevention counseling (Project Respect): preliminary results. Presented at the 11th International Conference on AIDS, Vancouver, B.C., July 7–12, 1996. abstract.
- Citing Articles (1)
Citing Articles
1
LARS ØSTERGAARD. (1999) Diagnosis of urogenital Chlamydia trachomatis infection by use of DNA amplification. APMIS 107:S89, 5-36
CrossRef
