As heterosexual transmission of human immunodeficiency virus (HIV) has increased in the United States,1 pelvic inflammatory disease (PID) among HIV-infected women has become increasingly common. From 6% to 22% of American women diagnosed with PID are HIV-infected,2,3 percentages up to ten-fold greater than those among sexually active women without PID.4,5 Human immunodeficiency virus-related immunodeficiency might facilitate ascent of vaginal flora to the upper genital tract, induce upper-tract disease from lower-tract flora considered nonpathogenic in immunocompetent women, increase severity, or impair therapeutic response. Retrospective studies of the influence of HIV on PID have lacked standard diagnostic criteria, standard treatment, or important clinical and microbiologic endpoints,6 so we conducted the first prospective study in American women.
Materials and Methods
From 1992 to 1994, research nurses in the emergency and outpatient departments of Jackson Memorial Hospital and Mt. Sinai Medical Center recruited participants by using informed consent procedures consistent with the Declaration of Helsinki. Recruited women had known HIV infections or no past positive HIV tests and at least two of four symptomatic criteria for PID (lower abdominal pain, abnormal vaginal discharge, abnormal vaginal bleeding, and fever [by self-report or temperature of at least 38C at registration]); negative pregnancy tests; and preliminary diagnoses of PID (including “rule out PID”). Women were ineligible if they were outside the 15–44-year age range; were incarcerated; were mentally incapable of informed consent; had pregnancy outcomes, pelvic or abdominal surgeries, laparoscopy, or endometrial curettage in the past 6 weeks; had hysterectomy or a bilateral salpingectomy; used an intrauterine device in the past 12 months; had penicillin allergy (among women treated as outpatients); or had immunodeficiency conditions unrelated to HIV (among women not known to be HIV-infected at enrollment). Gynecologists enrolled eligible women with signs associated with laparoscopically confirmed salpingitis: tenderness at two or more sites (lower abdomen, cervix, uterus, or adnexa) plus one or more of mucopurulent cervicitis, temperature of at least 38C, or an adnexal or pelvic mass diagnosed by transabdominal sonogram or palpation.7
Nurses questioned women about demographic characteristics and medical histories. Gynecologists used standard scales to rate vaginal and endocervical discharge (zero = absent to three = copious), and tenderness of abdomen (direct and rebound in each quadrant scored separately), cervix, uterus, and each adnexa (zero = absent to three = severe). For women who had sonograms within 24 hours of enrollment, sonographers scored the largest diameters of masses considered infectious (ie, complex echogenicity, associated free peritoneal fluid) as zero = no mass, one = up to 2 cm, two = 2–5 cm; three = more than 5 cm. Masses judged noninfectious (eg, functional ovarian cysts) were scored as zero. When women declined sonograms or could not be scheduled, gynecologists scored masses with infectious characteristics using the largest palpated diameter. A clinical severity score was calculated by summing tenderness scores (zero–36-point range) and mass scores (zero–six-point range), yielding zero–42 possible points.8 Gynecologists collected vaginal and endocervical swabs and transcervical endometrial biopsies with self-sheathing suction curettes, using methods to avoid contamination with vaginal flora. We also collected endometrial biopsies from 14 HIV-infected women who had cervical neoplasia but no clinical signs of PID or bacterial vaginal infections to compare with microbiology of HIV-infected women with PID.
Hospital admission was recommended if women were aged younger than 18 years; nulligravid; had a PID diagnosis within 6 months; had evidence of peritonitis, sepsis, or pelvic mass; had purulent culdocentesis aspirate, had a temperature of at least 38C; had no improvement with oral PID therapy before enrollment; or had questionable adherence to outpatient therapy. Known or suspected HIV infection was not an admission criterion. Inpatients were treated with intravenous (IV) gentamicin (loading dose 2 mg/kg body weight; maintenance dose 1.5 mg/kg body weight every 8 hours) and clindamycin, 900 mg, every 8 hours, according to 1993 CDC guidelines.9 At discharge, women were prescribed oral doxycycline, 100 mg, twice daily for 14 − X days (if no abscess was diagnosed) or 21 − X days (if an abscess was diagnosed), where X was the number of days antibiotics were received in the hospital. Outpatients were given one 250-mg intramuscular (IM) dose of ceftriaxone and prescribed oral doxycycline, 100 mg, twice daily for 14 days if no abscess was suspected or for 21 days if an abscess was suspected. (Some women with abscesses declined hospitalization.) Doxycycline-allergic women were prescribed oral erythromycin, 500 mg, four times daily for 14 days. To enhance adherence and follow-up, nurses offered free medication or vouchers, transportation, and child care, and phoned women periodically. Women were offered analgesics after examinations. Three methadone-dependent women received their usual doses during the study. Three of 11 HIV-infected women taking antiretroviral therapy (zidovudine monotherapy for eight) before enrollment had documented use during the study.
A second evaluation was scheduled 2 to 4 days after enrollment. All HIV-infected women were scheduled for third evaluations 5 to 9 days after their 14–21-day antibiotic courses (19–30 days after enrollment). A randomly selected sample of HIV-noninfected women who returned for second evaluations was scheduled for third evaluations because sample size necessitated only a subset. Gynecologists evaluated inpatients daily who worsened between scheduled evaluations and evaluated outpatients who sought care outside scheduled visits. To assess pain more accurately at follow-up examinations, analgesics were usually withheld 6 hours before, and outpatients were asked to withhold analgesics 6 hours before examinations. Increased abdominal pain, new or worsening peritonitis, a new or enlarging pelvic mass, or fever 48 hours after enrollment were indications to prescribe a third antibiotic (IV ampicillin or pipercillin), consider surgery, continue hospitalization, or admit outpatients. Outpatients who were later hospitalized were prescribed standard inpatient antibiotics and repeat courses of doxycycline after discharge.
Standard methods were used to test specimens for HIV antibodies, leukocyte counts, lymphocyte phenotyping, erythrocyte sedimentation rate, syphilis serology, and illicit drug toxicology. Microbiologists unaware of patient serostatus tested vaginal secretions for Trichomonas vaginalis (wet mount and culture)10 and for organisms associated with bacterial vaginosis (Gram stain).11 They tested endocervical secretions and endometrial biopsy tissue for Neisseria gonorrhoeae (Gram stain and culture)10 and Chlamydia trachomatis (culture and polymerase chain reaction [PCR] [Roche Diagnostic Systems, Branchburg, NJ]).10,12 Biopsies were tested for genital mycoplasmas (isolation for Mycoplasma hominis by culture for New York specimens; isolation and speciation for M hominis, M fermentans, and M genitalium by culture and PCR for Miami specimens),13,14Ureaplasma urealyticum (culture),13T vaginalis (culture),10 and facultative and anaerobic bacteria (culture).10 Endocervical and biopsy specimens collected at the third visit were tested for N gonorrhoeae or C trachomatis if these organisms had been isolated at enrollment. Gonorrhea isolates were tested for susceptibility to prescribed antibiotics. Resistance criteria to clindamycin and gentamicin are not published, so we applied aerobic bacteria criteria.15,16
A gynecologic pathologist unaware of patient serostatus examined hematoxylin-eosin–stained biopsy tissue. Acute endometritis was diagnosed when intra-glandular neutrophil collections or glandular or stromal microabcesses not associated with menstrual endometria were seen. Plasma cell endometritis was diagnosed when at least one plasma cell per 120 x field was seen within endometrial stroma.17
We compared HIV-seronegative and HIV-seropositive women (combining women with known and unknown HIV serostatus at enrollment) by using uncorrected χ2, Fisher exact, Student t, and Wilcoxon rank-sum tests.18 Two-tailed P < .05 was considered significant. Many women did not return as scheduled, so we analyzed outcome variables by using generalized estimating equations. Those assumed gamma distribution errors and log-linear models that were quadratic functions of time since enrollment.19
Among 852 women who met symptomatic criteria for PID, 539 were eligible and 456 agreed to participate. Among 216 women who met case definition, nine HIV-noninfected women were excluded from analysis because PID was later excluded. Of the remaining 207 women, HIV serostatus was unknown at enrollment for 178 (86.0%; 163 HIV-noninfected women, 15 HIV-infected women) and 29 were known to be HIV-infected. Among women with initially unknown serostatus, 8.4% were HIV-infected.
Demographic characteristics, medical history, drug use, PID risk factors, and history of bacterial sexually transmitted diseases (STDs) did not differ significantly by serostatus with few exceptions (Table 1). Human immunodeficiency virus–infected and HIV-noninfected women reported an average of one prior episode of PID (treated in all but one HIV-noninfected woman), and similar proportions (13.6% versus 6.8%) reported PID in the past month. More HIV-infected women recently had used antiretroviral drugs and antibiotics (Table 1). Antibiotic use was much more common in HIV-infected women whose serostatus was known than unknown at enrollment (58.6% versus 6.7%, P < .05). Many women recently had used pain-altering drugs, including illicit analgesics (75.7%), marijuana (24.8%), and cocaine (20.9%). Only methadone use was significantly more common among HIV-infected women.
The duration and character of abnormal vaginal bleeding, discharge, and pain before enrollment did not differ by serostatus. Although more HIV-infected women reported fevers before enrollment, the presence of fever and most other signs at enrollment did not differ significantly by serostatus (Table 2). More human immunodeficiency virus-infected women had infectious adnexal masses diagnosed by sonogram (but not by palpation), yielding higher severity scores, but differences were not significant. Human immunodeficiency virus-infected women had significantly lower leukocyte and CD4 T-lymphocyte counts, and significantly higher sedimentation rates (Table 2).
The 24 HIV-infected women with CD4 T-lymphocyte percentages of less than 21 had significantly lower leukocyte counts than the 20 less immunodeficient HIV-infected women. Other signs did not differ significantly by CD4 T-lymphocyte percentage, including fever (58.3% versus 35.0%), adnexal masses (25.0 versus 40.0%), and scores of abdominal tenderness (8 versus 8 points), pelvic tenderness (9.5 versus 9 points), and severity (17 versus 16 points).
About half of HIV-infected women whose serostatus was unknown at enrollment (53.5%) and of HIV-noninfected women (50.3%) were hospitalized initially, compared with 79.3% of women known to be HIV-infected at enrollment (Table 2). Fewer HIV-infected women with known serostatus were admitted for severe signs (82.6%) than HIV-infected women with unknown serostatus (100%) or HIV-noninfected women (85.4%). Human immunodeficiency virus-infected women with known serostatus were more likely to be admitted because adherence to outpatient therapy was questioned (21.7%) than HIV-infected women with unknown serostatus (12.5%) or HIV-noninfected women (14.6%). Known or suspected HIV infection was not a recommended admission criterion; however, 47.8% of HIV-infected women with known serostatus were admitted partly for that reason, compared with 0% of HIV-infected women with unknown serostatus and 4.9% of HIV-noninfected women.
Compared with women who had not used antibiotics in the 6 weeks before enrollment, recent antibiotic users were significantly less likely to report profuse vaginal discharge (29 of 62 [46.8%] versus 97 of 144 [67.4%]) and uterine tenderness (50 of 62 [80.7%] versus 132 of 144 [91.7%]). They also had significantly lower median abdominal tenderness scores (4.5 versus 8 points), severity scores (13 versus 18 points), and leukocyte counts (8000 versus 13,000 cells/mm3). Those differences by prior antibiotic use were seen regardless of serostatus. The 26 HIV-infected women and the 118 HIV-noninfected women who had not used antibiotics before enrollment reported similar pain and discharge symptoms and analgesic use when they sought care. In that subgroup, as among all women, HIV-infected women had higher severity scores (19.5 versus 16 points; P = .05). However, abdominal tenderness scores (8 versus 7 points; P = .09) and diagnosis of masses by sonogram or palpation (34.6% versus 22.0%) did not differ significantly by HIV serostatus.
All but one outpatient (HIV-infected) and seven inpatients (6 HIV-noninfected) initially received the antibiotics prescribed by protocol. Adherence to outpatient therapy measured by patient reports and pill counts did not differ by serostatus. At least 73% of women who returned after completing therapy reported taking at least a 10-day course of antibiotics. Many women took drugs that might affect pain or antibiotic absorption, but only methadone and other oral narcotic use before the second visit was significantly more commonly documented in HIV-infected women (13.2% versus 1.5%).
At first follow-up visits, examining physicians knew the serostatus of only 3.0% of women whose serostatus was unknown at enrollment. The proportions of women evaluated 2–4 days after enrollment did not vary by serostatus (86.4% versus 86.5%) (Table 3). Twenty-eight HIV-infected women returned for evaluation 19–30 days after enrollment. Fifty-nine HIV-noninfected women who returned for second visits were randomly sampled for continued enrollment, and 37 returned for evaluation 19–30 days after enrollment.
There were no important differences in clinical outcomes by HIV serostatus in categoric analysis (Table 3) or generalized estimation equation analysis (Figure 1), including proportions with subjective or objective clinical improvement, fever, masses, abdominal pain, and discharge or declines in severity score and mass size. Among women treated initially as outpatients, only two (15.4%) HIV-infected women and nine (11.1%) HIV-noninfected women were later hospitalized because they did not improve. Among initially hospitalized women, HIV-infected women had significantly longer median hospital stays (7 versus 4 days) overall and in subanalyses that excluded women with conditions unrelated to PID that might have delayed discharge (eg, pneumonia, tuberculosis). We found no important differences in clinical outcomes by serostatus in subanalyses of inpatients, outpatients, and women who received antibiotics strictly by protocol, and among HIV-infected women, by initial CD4 T-lymphocyte percentage (less than 21% versus at least 21%).
Seven (15.9%) HIV-infected women and six (3.7%) HIV-noninfected women had invasive procedures to more definitively diagnose or treat PID within a few weeks of enrollment, ie, hysterectomy, salpingectomy, lysis of adhesions, diagnostic laparoscopy or laparotomy, culdocentesis, or endometrial curettage. Some women (five HIV-infected, three HIV-noninfected) had procedures because they did not improve with medical therapy. The remaining women had laparoscopies or culdocenteses to exclude other causes of abdominal pain or laparotomies to treat uterine myomas. Gynecologists tended to report more gross adnexal abnormalities in HIV-infected women during surgery, but differences by serostatus were not significant.
Possibly pathogenic organisms were isolated from endocervical or endometrial specimens of 82.5% of women. Mycoplasmas and streptococci were isolated more commonly from HIV-infected women. Isolation rates of other organisms did not differ by serostatus (Table 4). Among 21 HIV-infected and 67 HIV-noninfected women in whom neither N gonorrhoeae nor C trachomatis was isolated from endometrial or endocervical specimens, isolation of streptococci (38.1% versus 17.9%; P = .07) and mycoplasmas (50.0% versus 13.7%; P < .05) was more common among biopsies of HIV-infected women. However, all but one woman (HIV-infected) who had mycoplasmas but not N gonorrhoeae and C trachomatis isolated had other potentially pathogenic facultative bacteria isolated.
Among 26 HIV-infected and 118 HIV-noninfected women who had not taken antibiotics before enrollment, significantly more HIV-infected women had mycoplasmas (57.1% versus 24.7%) and streptococci (36.4% versus 16.0%) isolated from biopsies. Human immunodeficiency virus–infected women with CD4 T-lymphocyte percentages less than 21 had significantly more facultative bacteria (86.4% versus 56.3%) and streptococci (50.0% versus 12.5%) isolated from biopsies than HIV-infected women with higher CD4 T-lymphocyte percentages. Isolation of mycoplasmas (58.8% versus 36.4%), Gardnerella vaginalis (14% versus 6%), staphylococci (36% versus 25%), N gonorrhoeae (21.7% versus 42.1%), and C trachomatis (13.6% versus 33.3%) did not differ significantly by CD4 T-lymphocyte percentage.
None of the women who had N gonorrhoeae, C trachomatis, anaerobes, or mycoplasmas isolated from endocervical or endometrial specimens at enrollment and had biopsies taken at least 72 hours after therapy started had evidence of persistent infection.
Of 14 HIV-infected women without PID who had biopsies, six had facultative bacteria isolated. None had streptococci or mycoplasmas isolated.
From 43–51% of N gonorrhoeae isolates from 65 endocervical and 45 endometrial tested specimens were resistant to penicillin, tetracycline, or clindamycin. Resistance did not differ by serostatus. Gentamicin-resistant endocervical isolates were more common among HIV-infected women (100% versus 63%, P = .07). Gentamicin resistance among endometrial isolates did not differ by serostatus (61% versus 62%).
Human immunodeficiency virus–infected women were significantly less likely to have acute endometritis but were significantly more likely to have plasma cell endometritis (Table 4). Among women without pre-enrollment antibiotic use, plasma-cell endometritis was significantly more common in HIV-noninfected women (ten of 13 [77.0%] versus 15 of 53 [28.3%]). Of the eight HIV-infected and seven HIV-noninfected women who had diagnostically adequate biopsies taken 8–19 days after enrollment, only one HIV-infected woman was newly diagnosed with acute endometritis. Five HIV-infected women and four HIV-noninfected women had plasma cell endometritis, of whom four HIV-infected women and two HIV-noninfected women had the diagnosis at enrollment.
Compared with HIV-noninfected women, HIV-infected women had more sonographically diagnosed adnexal masses but otherwise presented with similar signs and symptoms. Nevertheless, HIV-infected women generally responded as well to the antibiotics now recommended by CDC.20 We used a standard case definition (that was highly predictive of isolation of endometrial pathogens) and standard treatment that enhanced the validity of diagnostic and clinical outcome measures, respectively. Diagnostic and admission biases related to the knowledge of serostatus at enrollment6 were minimized because clinicians and sonographers were aware of the serostatus of only 14.0% and 4.6% of women, respectively.
Our study confirmed others that showed that HIV-infected women with PID tended to have lower leukocyte counts or more adnexal masses.4,5,21–23 We also found that HIV-infected women lacked the endometrial inflammation characteristic of acute endometritis but were more likely to have plasma cell endometritis, a diagnosis associated with laparoscopically diagnosed acute salpingitis23,24 and chronic endometritis.17 It is unlikely that diagnostic criteria or endometrial sampling were applied differently by serostatus. Our results, therefore, suggested that HIV infection might impair acute inflammatory responses to bacterial infections in the endometrium. That might limit containment of infection in the endometrium, resulting in more severe tubal inflammation, as other studies suggested.23 Alternatively, HIV-infected women might harbor chronic, partially treated endometrial or tubal infections resulting from long-term antibiotic use or immunodeficiency.
Although the power to detect the minor differences in clinical status among HIV-infected women by initial CD4 T-lymphocyte percentage was less than 50%, our study suggested that signs and therapeutic responses of women with CD4 T-lymphocyte percentages less than 21 are not substantially worse than those of more immunocompetent women. Two African studies found that severely immunodeficient HIV-infected women (CD4 T-lymphocyte percentages less than 14) had significantly higher severity scores and longer hospitalizations than more immunocompetent women.23 They also tended to have more abscesses, but differences were not statistically significant.22,23
The small numbers of HIV-infected women forced us to combine in analysis HIV-infected women with known serostatus and HIV-infected women whose serostatuses were unknown at enrollment. Compared with HIV-infected women with unknown serostatus, more HIV-infected women with known serostatus had taken antibiotics before enrollment and had milder disease at enrollment, despite lower median CD4 T-lymphocyte percentages. Combining all HIV-infected women in analysis probably underestimated the underlying differences by serostatus in clinical severity and microbiology because pre-enrollment antibiotic use was associated with milder disease and less bacterial isolation, regardless of serostatus. Among women who had not used antibiotics before enrollment, HIV-infected women still had significantly higher clinical severity scores. Analgesic use and duration of symptoms before seeking care did not differ significantly by serostatus. The slightly more severe signs of HIV-infected women, albeit lessened by the confounding effects of pre-enrollment antibiotic use, are best explained by differences in clinical response to pelvic infection or bacterial etiology. This finding supports the CDC surveillance definition for AIDS that classifies PID as a disease that might be complicated by HIV.25
According to a MEDLINE search from 1984 to 1999 using the search terms adnexitis, HIV, and endometritis, this study was the first to report the influence of HIV on endometrial pathogens other than N gonorrhoeae and C trachomatis, microbiologic cure, antibiotic susceptibility, and histopathology in American women. Isolation of most bacteria did not differ significantly by serostatus, as other research showed2,4,5,22,23; however, isolation of genital mycoplasmas and streptococci from biopsies was associated with HIV infection among all women, the subgroup lacking N gonorrhoeae and C trachomatis isolation, and the subgroup without pre-enrollment antibiotic use in whom microbiologic data are most meaningful. Although we cannot exclude the possibility that organisms isolated from biopsies were vaginal contaminants, it is not likely that sampling adequacy or contamination differed by serostatus.
Our findings suggest that HIV-infected women might be predisposed to endometrial colonization by mycoplasmas and streptococci. Those organisms commonly colonize the vaginas of sexually active women26 and might be more likely to colonize HIV-infected womenFriedmann W, Schafer A, Schwartlander B. Opportunistic diseases in HIV-infected women [abstract]. Proceedings of IX International AIDS Conference; Berlin, 1993;1:60). Greater use of antibiotics by HIV-infected women (as we observed) might also selectively reduce colonization by organisms other than mycoplasmas.27 If HIV impairs the mucosal or systemic immunity that normally prevents ascent of vaginal organisms to the endometrium, mycoplasmas and streptococci might affect PID pathogenesis. Two findings support that hypothesis: Those organisms were not isolated from the biopsies of HIV-infected women without PID, and HIV-infected women were significantly more likely to have plasma cell endometritis, a condition associated with streptococci and mycoplasma infection.28,29 Although those organisms might act as primary pathogens, the finding that nearly all women with mycoplasmas had other endometrial pathogens isolated suggests that mycoplasmas might modulate pathogenesis initiated by other bacteria. If HIV-infected women are more susceptible to mycoplasma and streptococci infections, optimal therapy might require antibiotics that are highly active against them. Reports of tetracycline-resistant mycoplasmas27 and streptococci30 and the high prevalence of gentamicin- and clindamycin-resistant N gonorrhoeae in our study highlight the need to monitor antibiotic resistance and periodically reevaluate PID treatment guidelines.
Despite the potential concerns about antibiotic coverage, our study of CDC-recommended antibiotics20 and studies of other regimens5,21–23 found that HIV-infected women generally responded as well as HIV-noninfected women. That should reassure clinicians who treat without knowledge of serostatus. All women infected initially with N gonorrhoeae, C trachomatis, anaerobes, and mycoplasmas cleared their infections after antibiotic therapy, even though more HIV-infected women had gentamicin-resistant endocervical N gonorrhoeae. That was consistent with other research.23 Our findings refute early theoretic concerns that HIV infection might cause fulminant, therapeutically unresponsive PID.31 Our study also showed that appropriate antibiotics reduce treatment failure. Appropriate antibiotics also reduce cervicovaginal HIV shedding,32 drug toxicities and resistance, and cost. Interventions to improve clinician and patient adherence to appropriate treatment are needed when adherence is poor.33,34
Human immunodeficiency virus-infected women in our study were more likely to be hospitalized, but hospitalization imperfectly classified women with severe disease because more HIV-infected women were hospitalized to ensure adherence. However, HIV-infected women were hospitalized longer, partly to resolve abscesses, as other studies showed.5,23 Human immunodeficiency virus–infected women generally responded well to outpatient therapy, as other studies showed22 (Bukusi E, Stevens C, Cohen C. Impact of HIV on acute pelvic inflammatory disease in a Nairobi outpatient clinic [abstract]. Proceedings of XI International AIDS Conference; Vancouver; 1996;1:161). Past recommendations to hospitalize all HIV-infected women with PID9 might not be warranted without standard indications such as abscesses.
Our therapeutic response data might be limited by several factors. Differences in pain-related outcomes might have been diminished because more HIV-infected women took narcotics after enrollment. However, other factors that might have influenced response (illicit drug use, antibiotic adherence, resistance to most antibiotics) did not differ significantly by serostatus. Only 13 HIV-infected women were treated initially as outpatients, yielding a power of less than 50% to detect the minor differences in therapeutic response by serostatus. Few HIV-infected women were taking anti-retroviral therapy, mostly zidovudine monotherapy. Our results are therefore most generalizable to uninsured HIV-infected women who are least likely to take potent combination therapy.35 We only evaluated short-term therapeutic responses, the primary concern of clinicians in emergency and outpatient departments.
The high prevalence of HIV infection among PID patients in this and other studies2–5,21–23 underscores the need to offer routinely HIV counseling and testing. Despite longstanding national recommendations to offer counseling and testing to all PID patients,36 many hospitals cannot because of competing needs to counsel pregnant women about HIV (personal communication, C. Barbosa, R. Sperling, 1995). Our study highlights that when counseling resources are limited, women with PID—who are likely to be HIV-infected and to have engaged recently in unprotected sex—merit special attention.
1. Wortley PM, Fleming PL. AIDS in women in the United States: Recent trends. JAMA 1997;278:911–6.
2. Safrin S, Dattel BJ, Hauer L, Sweet RL. Seroprevalence and epidemiologic correlates of human immunodeficiency virus infection in women with acute pelvic inflammatory disease. Obstet Gynecol 1990;75:666–70.
3. Shannon J, Benrubi GI. Epidemiology of pelvic inflammatory disease at University Medical Center, Jacksonville. J Fla Med Assoc 1991;78:158–61.
4. Hoegsberg B, Abulafia O, Sedlis A, Feldman J, DesJalais D, Landesman S, et al. Sexually transmitted diseases and human immunodeficiency virus infection among women with pelvic inflammatory disease. Am J Obstet Gynecol 1990;163:1135–9.
5. Barbosa C, Macasaet M, Brockmann S, Sierra MF, Xia Z, Duerr A. Pelvic inflammatory disease and human immunodeficiency virus infection. Obstet Gynecol 1997;89:65–70.
6. Irwin KL, Rice RJ, Sperling RS, O'Sullivan MJ, Brodman M. Potential for bias in studies of the influence of human immunodeficiency virus infection on the recognition, incidence, clinical course, and microbiology of pelvic inflammatory disease. Obstet Gynecol 1994;84:463–9.
7. Hager WD, Eschenbach DA, Spence MR, Sweet RL. Criteria for diagnosis and grading of salpingitis. Obstet Gynecol 1983;61:113–4.
8. Thompson SE III, Hager WD, Wong KH, Lopez B, Ramsey C, Allen SD, et al. The microbiology and therapy of acute pelvic inflammatory disease in hospitalized patients. Am J Obstet Gynecol 1980; 136:179–86.
9. Centers for Disease Control and Prevention. 1993 sexually transmitted diseases treatment guidelines. MMWR Morb Mortal Wkly Rep 1993;42:75–81.
10. Balows A, Hausler WJ, Herrmann KL, Isenberg HD, Shadomy HJ, eds. Manual of clinical microbiology. 5th ed. Washington DC: American Society for Microbiology, 1991.
11. Nugent RP, Krohn MA, Hillier SL. Reliability of diagnosing bacterial vaginosis is improved by a standardized method of Gram stain interpretation. J Clin Microbiol 1991;29:297–301.
12. Loeffelholz MJ, Lewinski CA, Silver SR, Purohit AP, Herman SA, Buonagurio DA, et al. Detection of Chlamydia trachomatis in endocervical specimens by polymerase chain reaction. J Clin Microbiol 1992;30:2847–51.
13. Cassell GH, Blanchard A, Duffy L, Crabb D, Waites KB. Mycoplasmas. In: Howard BJ, Keiser JF, Smith TE, Weissfeld AS, Tilton RC, eds. Clinical and pathogenic microbiology. St Louis: Mosby-Year Book, 1994:491–502.
14. Blanchard A, Yanez A, Dybvig K, Watson HL, Griffiths G, Cassell GH. Evaluation of intraspecies genetic variation within the 16S rRNA gene of Mycoplasma hominis and detection by polymerase chain reaction. J Clin Microbiol 1993;31:1358–61.
15. National Committee for Clinical Laboratory Standards. Methods for dilution antimicrobial susceptibility tests for bacteria that grow aerobically. Standard M7-A3. Villanova, PA: National Committee for Clinical Laboratory Standards, 1993.
16. National Committee for Clinical Laboratory Standards. Performance standards for antimicrobial disk susceptibility tests. 5th ed. Standard M2-A5. Villanova, PA: National Committee for Clinical Laboratory Standards, 1993.
17. Kurman RJ, Mazur MT. Benign diseases of the endometrium. In: Kurman RT, ed. Blaustein's pathology of the female genital tract. New York: Springer Verlag, 1994:367–410.
18. SAS Institute. SAS/STAT user's guide: Version 6. 4th ed. Cary, NC: SAS Institute, 1990.
19. Zeger SL, Liang KY. Longitudinal data analysis for discrete and continuous outcomes. Biometrics 1986;42:121–30.
20. Centers for Disease Control and Prevention. 1998 guidelines for treatment of sexually transmitted diseases. MMWR Morb Mortal Wkly Rep 1998;47:79–86.
21. Korn AP, Landers DV, Green JR, Sweet RL. Pelvic inflammatory disease in human immunodeficiency virus–infected women. Obstet Gynecol 1993;82:765–8.
22. Kamenga MC, De Cock KM, St Louis ME, Toure CK, Zakaria S, N'gbichi JM, et al. The impact of human immunodeficiency virus infection on pelvic inflammatory disease: A case-control study in Abidjan, Ivory Coast. Am J Obstet Gynecol 1995;172:919–25.
23. Cohen CR, Sinei S, Reilly M, Bukusi E, Eschenbach D, Holmes KK, et al. Effect of human immunodeficiency virus type 1 infection upon acute salpingitis: A laparoscopic study. J Infect Dis 1998;178:1352–8.
24. Kiviat NB, Wolner-Hanssen P, Eschenbach DA, Wasserheit JN, Paavonen JA, Bell TA, et al. Endometrial histopathology in patients with culture-proved upper genital tract infection and laparoscopically diagnosed acute salpingitis. Am J Surg Pathol 1990;14:167–75.
25. Centers for Disease Control and Prevention. 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Morb Mortal Wkly Rep 1992;41:1–18.
26. McCormack WM, Almeida PC, Bailey PE, Grady EM, Lee YH. Sexual activity and vaginal colonization with genital mycoplasmas. JAMA 1972;221:1375–7.
27. Westrom L, Mardh PA. The effect of antibiotic therapy on mycoplasma in the female genital tract. In vitro and in vivo studies on the sensitivity of Mycoplasma hominis and T-Mycoplasmas to tetrocyclines and other antibiotics. Acta Obstet Gynecol Scand 1971;50:25–31.
28. Paavonen J, Kiviat N, Brunham RC, Stevens CE, Kuo CC, Stamm WE, et al. Prevalence and manifestations of endometritis among women with cervicitis. Am J Obstet Gynecol 1985;152:280–6.
29. Hillier SL, Kiviat NB, Hawes SE, Hasselquist MB, Hanssen PW, Eschenbach DA, et al. Role of bacterial vaginosis-associated micro-organism in endometritis. Am J Obstet Gynecol 1996;175:435–41.
30. Hasselquist MB, Hillier S. Susceptibility of upper-genital tract isolates from women with pelvic inflammatory disease to ampicillin, cefpodoxime, metronidazole, and doxycycline. Sex Transm Dis 1991;18:146–9.
31. Peterson HB, Walker CK, Kahn JG, Washington AE, Eschenbach DA, Faro S. Pelvic inflammatory disease. Key treatment issues and options. JAMA 1991;266:2605–11.
32. Ghys PD, Fransen K, Diallo MO, Ettiegne-Traore V, Coulibaly IM, Yeboue KM, et al. The associations between cervicovaginal HIV shedding, sexually transmitted diseases and immunosuppression in female sex workers in Abidjan, Cote d'Ivoire. AIDS 1997;11:F85–93.
33. Hessol NA, Priddy FH, Bolan G, Baumrind N, Vittinghoff E, Reingold AL, et al. Management of pelvic inflammatory disease by primary care physicians. A comparison with Centers for Disease Control and Prevention Guidelines. Sex Transm Dis 1996;23:157–63.
34. Brookoff D. Compliance with doxycycline therapy for outpatient treatment of pelvic inflammatory disease. South Med J 1994;87:1088–91.
35. Shapiro MF, Morton SC, McCaffrey DF, Senterfitt JW, Fleishman JA, Perlman JF, et al. Variations in the care of HIV-infected adults in the United States: Results from the HIV Cost and Services Utilization Study. JAMA 1999;281:2305–15.
36. Centers for Disease Control. 1989 sexually transmitted disease treatment guidelines. MMWR Morb Mortal Wkly Rep 1989;38:2.