Pelvic inflammatory disease: current concepts in pathogenesis, diagnosis and treatment.

Current recommendations for antimicrobial treatment regimens in PID were published in 2010 ( Table 4 ), and are scheduled for update in 2014. [ 67 ] A guiding principle for selection of antimicrobial therapy for PID is that the regimen should cover N. gonorrhoeae and C. trachomatis , regardless of results of diagnostic testing for these pathogens. Therapy of gonorrhea, and therefore PID, shifted away from fluoroquinolone-based regimens between the 2006 and 2010 iterations of the CDC Treatment Guidelines, given the rapid emergence of fluoroquinolone resistance. [ 68 ] In 2012, after reports of increasing prevalence of cefixime-resistant gonorrhea, the guidelines were changed again to drop cefixime as one of the first line outpatient treatment options for cervicitis. [ 69 ] With additional cephalosporin resistance to gonorrhea reported as discussed in the article by Barbee and Dombrowski elsewhere in this issue, the potential for development of resistance that could compromise treatment of gonorrhea-associated PID is of great concern.

Ultrasonography can also be used to aid in diagnosis of PID and direct treatment. A finding of thickened, fluid-filled tubes have an 85% sensitivity and 100% specificity for endometritis among women with clinically diagnosed PID. [ 55 ] Timor-Tritsch detailed the various transvaginal sonographic markers of acute tubal inflammatory disease, including dilated tubal shape, abnormal wall structure, increased wall thickness (≥5mm), and presence of pelvic peritoneal fluid (free fluid or inclusion cyst). [ 56 ] In a study comparing 30 patients with clinical PID confirmed with laparoscopy and 20 normal women, power Doppler demonstrating tubal hyperemia was 100% sensitive and 80% specific for PID; in addition, altered tubal shape, structure, and wall thickness were seen in an overwhelming majority of patients with pyosalpinx. [ 57 ] Magnetic resonance imaging (MRI), with its highly sensitive and specific ability to identify thickened, fluid-filled tubes, pyosalpinx, pelvic free fluid, and tubo-ovarian abscesses, has also been proposed as a diagnostic modality for PID; however, it is very costly and not easily accessible or applicable to women seeking outpatient evaluation for possible PID. [ 58 ]

In the 1970’s the Dalkon Shield intrauterine device (IUD) was associated with increased rates of PID, and let to significant concerns about the safety of IUDs in women at risk for sexually transmitted infections (STI). [ 91 ] Modern IUDs, including the levonorgestrel IUD (Mirena) and the copper IUD (Paraguard), have not been associated with an increased risk of PID over the long term. [ 92 ] There does appear to be a slightly increased rate of PID in the 20 days post insertion: in one study the rate of PID during this time was 9.66/1000 women, while after that it was 1.38/1000 women. [ 93 ] A review of studies assessing PID after IUD insertion in the presence of gonoccoal or chlamydial cervicitis showed an increased, but overall quite low risk (0–5%). [ 94 ] Recent studies suggest that screening for GC/CT at the time of insertion, as opposed to requiring a negative test prior to the procedure, does not significantly increase adverse sequelae. [ 95 ] There does not appear to be any difference in risk of PID with hormone-containing compared with copper IUDs. [ 96 ] The presence of an intrauterine device (IUD) at time of diagnosis of acute PID does not alter the management, and empiric removal of the IUD is not indicated. [ 67 , 97 ]

PID is associated with significant reproductive morbidity, which appears to be reduced with prompt, proactive treatment of cervicitis and lower genital tract infections. It is a clinical diagnosis, and providers should maintain a high index of suspicion when presented with a reproductive age women complaining of abdominal and pelvic pain. Sexually transmitted infections are commonly associated with PID, but vaginal anaerobes also appear to be involved, and antibiotic coverage for these pathogens should be considered when treating women with severe symptoms or pelvic abscesses.

Practically, in the clinic or emergency department, when a sexually active woman presents with lower abdominal or pelvic pain, PID must be considered in the differential diagnosis, which also includes appendicitis, ectopic pregnancy, ovarian torsion, intrapelvic bleeding, rupture of an adnexal mass, endometriosis, and gastroenteritis. [ 45 ] Key components of the physical exam include: abdominal exam, including palpation of the right upper quadrant, vaginal speculum exam, including inspection of the cervix for friability and mucopurulent cervical discharge, bimanual exam, assessing for cervical motion, uterine, or adnexal tenderness, as well as pelvic masses. Microscopic evaluation of a sample of cervicovaginal discharge to assess for T. vaginalis , bacterial vaginosis, and/or leukorrhea. The clinical presentation of PID is quite variable ( Table 1 ), thus a high index of suspicion is necessary. Symptoms may differ depending on the pathogens responsible. In the PID Evaluation and Clinical Health (PEACH) trial, women with PID associated with C. trachomatis or M. genitalium took almost one week longer to present to care than women with gonorrhea-associated PID, suggesting milder symptoms. [ 19 ] Women with gonococcal infection are more likely to have fever, adnexal tenderness, mucopurulent cervicitis and an elevated peripheral white blood count (WBC). [ 46 ]

In early studies of PID, N. gonorrhoeae was the most commonly isolated pathogen, and is still more likely to cause severe symptoms than other pathogens. [ 13 , 17 – 19 ] However, as the prevalence of gonorrhea has decreased, its importance as a causal agent for PID has diminished. [ 20 , 21 ] Chlamydia trachomatis remains a significant pathogen associated with PID, detected in up to 60% of women with confirmed salpingitis or endometritis. [ 22 – 24 ] Mycoplasma genitalium has been independently associated with PID, though its prevalence is low in most populations that have been studied. [ 25 , 26 ] The proportion of cases of PID that involve non-gonococcal, non-chlamydial etiologyranges between 9–23% in women with confirmed salpingitis or endometritis, even as diagnostic testing for gonorrhea and chlamydia become more sensitive. [ 7 , 22 , 24 , 27 , 28 ] In these cases, the microbial community is often diverse and includes anaerobes like Peptostreptococcus spp . and Prevotella spp . [ 23 , 27 ] Even in women with gonorrhea or chlamydia, detection of anaerobes in the upper genital tract is frequent and is associated with more severe disease. [ 16 , 22 ] In a study of Kenyan women with laparoscopically confirmed salpingitis, polymerase chain reaction (PCR) assay of tubal samples for the bacterial 16S rRNA gene identified multiple species, including several associated with BV such as Atopobium vaginae, Leptotrichia spp., Peptostreptococcus spp , and Prevotella spp . [ 29 ]

Women with PID have an increased risk of ectopic pregnancy, infertility and chronic pelvic pain due to tubal scarring and damage from inflammation. In the PEACH trial 36% of participants reported chronic pelvic pain; women with 2 or more episodes of PID were at highest risk. [ 98 ] In a cohort study of women with laparoscopically confirmed salpingitis in Sweden, followed for a mean of 94 months, the infertility rate was 16%, 67% of which was attributable to tubal factor infertility compared to an infertility rate of 2.7% in women without salpingitis. Of women who became pregnant, 9% of women with salpingitis had an ectopic compared to 1.9% of control women. [ 21 ] The risk of infertility increased with severity of salpingitis and number of episodes of PID. Chlamydial cervicitis also increases the risk of ectopic with repeat infections; women with 3 or more episodes had 4.5 times increased odds of PID. [ 21 ] In the PEACH trial, upper genital tract detection of gonorrhea, chlamydia, or endometritis was sufficient to confirm the diagnosis of PID. However, there were no differences in reproductive health outcomes between women with and without endometritis or upper genital tract infection. [ 42 ] A more recent study of women with lower genital tract infection but no PID by clinical criteria used a permissive definition of endometritis (1 plasma cell per hpf) and showed a 40% decrease in pregnancy rates among women with endometritis (or subclinical PID). [ 44 ] The differences in these analyses may be due to a slightly higher rate of C. trachomatis infection in the latter study.

There is little clarity on the need for empiric coverage for anaerobic bacteria when PID is diagnosed, in part because there is a lack of clear understanding of the contribution of anaerobes to pathogenesis in PID. Several studies have shown that BV is associated with PID, with BV-associated anaerobic bacteria present in endometritis, but that BV may not actually cause acute PID. [ 16 , 61 , 74 ] However, other data have not shown any long-term reproductive sequelae of histologically-diagnosed anaerobic endometritis, even when treatment with a cephalosporin-based regimen with poor anaerobic coverage was provided. [ 42 ] Few studies have specifically examined microbiologic cure rates of antimicrobial treatment regimens targeting anaerobic bacteria. Some BV-associated microbes may form a biofilm on the endometrial surface, which could limit the ability of antibiotics to eliminate colonization. [ 75 ] The CDC currently recommends consideration of treatment regimens with anaerobic coverage until data suggest equivalent prevention of reproductive sequelae in treatment regimens lacking anaerobic coverage. [ 67 ] Anaerobic coverage should be included in women with a tubo-ovarian abscess and with BV, regardless of the latter’s potential etiologic role in the development of acute PID. [ 67 , 76 ]

The therapeutic goal for the treatment of PID is two-fold: short-term microbiologic and clinical cure and long-term prevention of sequelae, namely tubal infertility, ectopic pregnancy, and chronic pelvic pain. Since the 1980s, PID therapy has shifted from the inpatient setting to the outpatient setting, with a 68% decline in hospitalization [ 4 ] due in part to several studies showing equivalent short-term outcomes with outpatient versus inpatient therapy for mild-moderate PID. [ 59 , 60 ] Between 1995 and 2001, 89% of all PID visits occurred in the ambulatory setting. [ 4 ] Current criteria for inpatient hospitalization are summarized in Box 2 . The PEACH trial compared inpatient administration of parenteral cefoxitin and doxycycline (parenteral/oral) versus outpatient administration of intramuscular cefoxitin and oral doxycycline and found no short-term (30-day) differences in microbiologic or clinical cure [ 61 ] or long term differences in reproductive health outcomes ( Table 3 ). [ 62 ] A secondary analysis among participants of the PEACH trial also saw no long-term outcome differences by treatment group among those with clinically confirmed endometritis or upper genital tract gonorrheal or chlamydial infection. [ 61 , 62 ] Interestingly, a representative subpopulation from PEACH revealed a 70% mean treatment adherence rate, with only 17% of participants taking doxycycline exactly as prescribed. [ 63 ] Similar rates of poor adherence to doxycycline or tetracycline prescribed for outpatient therapy of STI have been seen, particularly in the setting of gastrointestinal side effects. [ 64 , 65 ] This may explain the relatively high rates of ongoing disease and long-term sequelae in the PEACH cohort. While many of the original efficacy studies mandated inpatient IV treatment for 48–96 hours before switching to oral therapy, [ 61 , 66 ] current practice is to treat with IV medications until there is clinical improvement for 24 hours or more. However, since IV doxycycline can cause significant phlebitis, and its oral bioavailability is comparable to parenteral, an earlier switch to oral doxycycline can be made if a patient is tolerating oral medications. [ 61 ]

Among patients who qualified for outpatient therapy, re-evaluation of clinical status should occur within 72 hours, or sooner if indicated. If no meaningful clinical response is detected, patients with PID may require inpatient hospitalization, transition to parenteral antibiotics, further diagnostic tests, including additional laboratory studies and imaging to evaluate for possible tubo-ovarian abscess (TOA), and possible surgical intervention. Empiric treatment for gonorrhea and chlamydia is recommended for all male sexual partners within the past 60 days, or the most recent sexual partner if >60 days ago, regardless of symptoms or the result of gonorrhea and chlamydia testing in the female patient with PID. [ 67 ] Women diagnosed with PID should be offered an HIV test at the time of diagnosis. Repeat testing for gonorrhea or chlamydia in 3 to 6 months is recommended if initial testing was positive for either infection. [ 67 ]

Because PID is a clinical diagnosis, laboratory data or imaging studies are not usually necessary, but they can be helpful in establishing the diagnosis or in defining its severity. [ 51 ] In the PEACH trial, which enrolled women with abdominal pain, pelvic tenderness and evidence of lower genital tract inflammation, an elevated leukocyte count (≥ 10,000 cells/mL) had 41% sensitivity and 76% specificity for the presence of endometritis. [ 52 ] The presence of ≥1 neutrophil per 1000x field saline wet mount of vaginal discharge had 91% sensitivity and 26.3% specificity for endometritis. [ 53 ] In another cohort study, an elevated ESR (> 15mm/hr) had 70% sensitivity and 52% specificity for endometritis or salpingitis. Elevated WBC had 57% sensitivity and 88% specificity, while presence of increased numbers of vaginal neutrophils (≥3/HPF) had 78% sensitivity and 39% specificity. [ 54 ] In a cohort of women at high risk for pelvic infections, absence of vaginal while blood cells had excellent negative predictive value (95%). [ 53 ] These data suggest that if an evaluation of a saline microscopy of vaginal fluid reveals no white blood cells (leucorrhea), an alternative diagnosis to PID should be considered.

As gonorrhea and chlamydia contribute over half to three-quarters of PID, screening for and treating these infections should decrease the incidence and sequelae. Four randomized trials have examined whether this strategy is effective. The earliest was conducted between 1990 and 1992 in Seattle, Washington. Over 1000 women in a managed care organization were randomized to receive an invitation for chlamydia screening, then were followed for a year and compared to approximately 1600 women receiving standard care. Although only 64% of the intervention group were screened, during the 12 month follow-up there were 9 PID cases in the screening group and 33 in the control group (RR 0.44 (0.2, 0.9)). [ 99 ] A second study used cluster randomization to randomize students at 17 high schools to receive the offer of chlamydia screening and then followed them for PID over 12 months. At one year the PID incidence was 2.1% in the screening group (of whom 48% were screened) and 4.2% in the control group. [ 100 ] Most recently, the Prevention Of Pelvic Infections (POPI) trial in England enrolled sexually active women under age 27 and randomized them to early vs. delayed screening for chlamydia. The early screening group had a chlamydia prevalence of 5.4%, and 15/1191 (1.3%) developed PID over the course of the study. In the delayed group, 5.9% had chlamydia detected on their enrollment swab when it was tested a year later. During that time, 23/1186 (1.9%) in that group developed PID. The relative risk for PID in those with early screening was not significant (0.65 (0.34, 1.22), [ 32 ] but the study was underpowered given the low rate of PID.

Although not part of CDC’s recommendations, newer data suggest that parenteral followed by oral azithromycin, either as monotherapy or in combination with doxycycline and metronidazole, produces clinical cure rates of 97–98% at 2 weeks post-treatment and microbiologic cure rates of 90–94% at 6 weeks post-treatment. [ 70 ] The azithromycin-based regimens were compared to third-generation cephalosporin-based regimens or parenteral amoxicillin-based regimens and showed no statistically significant difference in clinical or microbiologic cure rates, although the study had a high dropout rate and low proportion of anaerobic bacteria isolated from endocervix and endometrium. [ 70 ] Another trial compared intramuscular ceftriaxone plus oral azithromycin versus the standard of oral doxycycline and found higher rates of clinical and histologic cure with azithromycin, although rates of cure in both arms were less than 80%. [ 71 ] While M. genitalium is neither tested for or considered when choosing therapy, newer evidence suggests a greater than 4-fold higher risk of treatment failure with a cefoxitin/doxycycline regimen when M. genitalium is present, although there were no differences in reproductive sequelae or recurrent PID, as discussed in the article by Manhart elsewhere in this issue. [ 72 ] Doxycycline has poor efficacy against M. genitalium (cure rates from 17–94%), and while azithromycin is more effective (67–100% cure), moxifloxacin seems to be the most effective treatment. [ 73 ] In cases of persistent PID not responsive to standard therapy, testing for and treatment of M. genitalium should be considered, and presumptive therapy with moxifloxacin may be warranted.

The clinical diagnosis of PID is based on recommendations from the Centers for Disease Control and Prevention (CDC). Minimum diagnostic criteria (see Box 1 ) have been set with a high sensitivity and low specificity, in order to detect as many cases of clinical disease as possible, thus potentially avoiding the long-term reproductive sequelae and economic costs associated with delayed diagnosis and lack of treatment. In a cohort of patients with suspected PID who underwent laparoscopyin Lund, Sweden,, PID was considered when a patient presented with lower abdominal pain and at least twp of the following: abnormal vaginal discharge, fever, vomiting, menstrual irregularities, urinary symptoms, proctitis symptoms, marked tenderness of the pelvic organs on bimanual, palpable adnexal mass or ESR > 15mm/hr. Only 65% of women suspected to have PID using these criteria actually had salpingitis. [ 47 ] A 2003 re-analysis of data from this cohort demonstrated that the combination of fever > 38.3°C, elevated ESR, and adnexal tenderness achieved the highest combination of sensitivity and specificity, 65% and 66%, respectively, for acute salpingitis. [ 48 ] In other words, these criteria would have a 35% false negative rate for predicting laparoscopically determined PID. It is difficult to calculate the exact sensitivity and specificity of the CDC diagnostic criteria, as there at least two potential “gold standards” for a true positive diagnosis of PID: salpingitis at laparoscopy or endometritis on endometrial biopsy. Since laparoscopy is expensive, invasive and not part of a standard evaluation of PID, many studies use endometritis as a marker of upper genital tract infection and inflammation. Endometritis and salpingitis are correlated; histologic endometritis has a sensitivity of 89–92% and specificity of 63–87% for laparoscopically-diagnosed acute salpingitis, with only 7–22% of patients with clinically suspected PID having salpingitis without endometritis. [ 28 , 41 , 49 , 50 ] However, while presence and severity of salpingitis is correlated with risk of ectopic pregnancy and infertility, [ 21 ] endometritis is not as consistently associated with these outcomes. [ 42 ] This may be due to the fact that not all women with endometritis have salpingitis ( Table 2 ), thus diluting the association.

In the United States in 2000, there were an estimated 1.2 million medical visits for PID, [ 1 ] a number that has been decreasing since 1985. [ 2 – 4 ] This decrease is attributed in part to widespread adoption of screening for Chlamydia trachomatis , the goal of which is to identify and treat asymptomatic cases of cervicitis before they can progress to PID. [ 5 ] Estimated direct medical costs associated with PID and its sequelae (ectopic pregnancy, chronic pelvic pain and tubal infertility) were as high as 1.88 billion USD in 1998, even though the majority of women receive care as outpatients. [ 6 ] Risk factors for PID are the same as those for acquisition of sexually transmitted diseases: multiple sexual partners, young age, smoking, and illicit drug use. [ 6 – 9 ] Douching has been implicated in some studies, and has been observed to double a woman’s risk of upper genital tract infection. [ 8 , 10 , 11 ] Oral contraceptive use has been associated with lower rates of clinical PID, though it is not clear whether this is due to fewer infections or fewer symptoms, and thus under diagnosis. [ 12 – 14 ] Bacterial vaginosis (BV) has also been associated with PID, though primarily in cross-sectional studies that were unable to determine causality. [ 15 ] In the prospective Gyn Infections Follow-through study (GIFT), women with BV at enrollment did not have higher risk for PID over 4 years of follow-up, though women with Neisseria gonorrhoeae and C. trachomatis did. [ 16 ]

Mathematical modeling based on epidemiologic and microbiologic studies suggests that 8–10% of women with C. trachomatis infection will develop PID if not treated, [ 30 ] although in studies that followed women with chlamydial endocervical infection without treatment, the rate was even lower. [ 31 , 32 ] When both the lower and upper genital tract are sampled, there is a clear gradient of infections, with a higher proportion of women testing positive at the vagina and/or cervix, fewer in the endometrium, and less frequently in the fallopian tubes. [ 23 , 24 , 27 ] One component of protection from bacterial ascent is the physical barrier of the cervix and its mucus barrier. Endometrial detection of gonorrhea or chlamydia is more frequent in the proliferative phase of the menstrual cycle [ 18 ] when cervical mucus is thinner, [ 33 ] and the peristaltic contractions of the uterus move fluid cephalad. [ 34 ] There is also likely an immunologic component to the cervical barrier; genetic polymorphisms in toll-like receptor (TLR) genes appear to increase the risk of upper genital tract infection [ 24 ] as do certain HLA class II alleles, suggesting that individual differences in immune function may increase the risk of developing PID in the setting of cervical infection. Tubal damage is best described in the context of chlamydial infection and appears to be related both to an innate immune inflammatory response initiated by the epithelial cells infected by C. trachomatis , [ 35 ] and to an adaptive T-cell response. [ 36 , 37 ] Although antibody titers to chlamydial antigens are increased in severe disease, [ 38 , 39 ] higher titers have not been associated with worse reproductive outcomes. [ 40 ] In human studies, evaluation of tubal inflammation is difficult without surgical intervention; thus, many studies use endometritis as a marker for tubal inflammation. Kiviat et al correlated the presence of both neutrophils and plasma cells in endometrial biopsies with visible salpingitis. [ 41 ] In a cohort of women with mild-moderate PID who were treated with broad-spectrum antibiotics, the presence of either neutrophils or plasma cells in an endometrial biopsy was not associated with decreased fertility. [ 42 ] Plasma cells alone were found in 33% of endometrial samples of low risk women [ 43 ] and were not associated with laparoscopic abnormalities, but in women at high risk of sexually transmitted infections, plasma cell endometritis appears to be associated with decreased fertility. [ 44 ] The heterogeneity of these findings suggests that there is a range of individual immune response to upper genital tract infection, and that not all women have the same likelihood of reproductive sequelae from PID.

Although the presenting signs and symptoms of a TOA are not often distinct from those with salpingitis/endometritis, there are often more objective signs of infection and inflammation. A large series of patients with ultrasound- or surgically-confirmed TOA found that 60% had a temperature >37.8°C, 68% had a leukocytosis >10,000 cells/mL, 26% had nausea, and 19% had chronic abdominopelvic pain. [ 77 ] In women with PID, palpation of an adnexal mass on physical exam, significant pain limiting proper evaluation of the adnexa, severe illness, or lack of clinical response to antimicrobial therapy should prompt imaging studies. In addition, imaging can be helpful to evaluate for alternative diagnoses such as appendicitis, ovarian torsion or cyst rupture. Inpatient observation is recommended for at least 24 hours among hemodynamically stable women with a tubo-ovarian abscess, with the aim of observing for early signs of sepsis or potential abscess rupture. Surgical exploration on initial evaluation is indicated in the setting of an acute abdomen and signs of sepsis or hemodynamic instability, particularly if a ruptured TOA is suspected. Antimicrobial therapy should be parenteral to start, and should include clindamycin or metronidazole to cover anaerobes. [ 67 ] Antimicrobial therapy alone, with appropriate anaerobic coverage and the ability to penetrate and function in abscess cavities, is effective in 70–84% of women. [ 77 , 78 ] In one cohort of women admitted with TOA, 60% of those with an abscess larger than 10cm needed surgical management compared to 20% of those with 4–6cm abscesses. [ 78 ] When no clinical improvement is noted within 72 hours of antibiotic initiation, minimally-invasive drainage of the abscess or surgical management can be pursued; however, significant clinical deterioration at any time point usually indicates the need for surgical exploration. [ 77 ] A study of empiric transvaginal ultrasound-guided aspiration of TOAsat the time of diagnosis, in concert with antimicrobial therapy, revealed that the procedure is safe, well tolerated, and averted surgical management in 93% of cases. [ 79 ]