Objectives Studies examining cervicitis aetiology and prevalence lack comparability due to varying criteria for cervicitis. We aimed to outline cervicitis associations and suggest a best case definition.
Methods A cross-sectional study of 558 women at three sexually transmitted infection clinics in Sydney, Australia, 2006–2010, examined pathogen and behavioural associations of cervicitis using three cervicitis definitions: ‘microscopy’ (>30 pmnl/hpf (polymorphonuclear leucocytes per high-powered field on cervical Gram stain)), ‘cervical discharge’ (yellow and/or mucopurulent cervical discharge) or ‘micro+cervical discharge’ (combined ‘microscopy’ and ‘cervical discharge’).
Results Chlamydia trachomatis (CT), Mycoplasma genitalium (MG), Trichomonas vaginalis (TV) and Neisseria gonorrhoeae (NG) had the strongest associations with cervicitis definitions ‘micro+cervical discharge’: CT adjusted prevalence ratio (APR)=2.13 (95% CI 1.38 to 3.30) p=0.0006, MG APR=2.21 (1.33 to 3.69) p=0.002, TV APR=2.37 (1.44 to 3.90) p=0.0007 NG PR=4.42 (3.79 to 5.15) p<0.0001 and ‘cervical discharge’: CT APR=1.90 (1.25 to 2.89) p=0.003, MG APR=1.93 (1.17 to 3.19) p=0.011, TV APR=2.02 (1.24 to 3.31) p=0.005 NG PR=3.88 (3.36 to 4.48) p<0.0001. Condom use for vaginal sex ‘always/sometimes’ reduced cervicitis risk: (‘micro+cervical discharge’) APR=0.69 (0.51 to 0.93) p=0.016. Combined population attributable risk % (PAR%) of these four pathogens was only 18.0% with a protective PAR% of condoms of 25.7%. Exposures not associated with cervicitis included bacterial vaginosis, Mycoplasma hominis, Ureaplasma urealyticum, herpes simplex virus 1&2, cytomegalovirus, Candida, age, smoking and hormonal contraception.
Conclusions Cervicitis was associated with CT, MG, TV and NG with combined PAR% of these pathogens only 18% in this setting, suggesting other factors are involved. Condoms significantly reduced cervicitis risk. Cervicitis definitions with best clinical utility and pathogen prediction were ‘cervical discharge’ and ‘micro+cervical discharge’.
- EPIDEMIOLOGY (CLINICAL)
- M GENITALIUM
Statistics from Altmetric.com
There is much to clarify about cervicitis, including aetiology, clinical significance, a workable case definition and optimal management. Cervicitis is common, often symptomatic and clinically relevant, particularly through association with pelvic inflammatory disease (PID),1 ,2 preterm birth3 and enhanced risk of HIV transmission.4 Lower genital tract inflammation (cervicitis or leucorrhoea) is thought to be a sensitive indicator of endometritis and subclinical pelvic infection.5 ,6 Chlamydia trachomatis (CT) and Neisseria gonorrhoeae (NG) have been consistently associated with cervicitis,7 ,8 with variable evidence for Mycoplasma genitalium (MG),9–11 Trichomonas vaginalis (TV)7 ,9 ,12 and bacterial vaginosis (BV).9 ,13 ,14 Most cervicitis diagnoses occur in the absence of these pathogens, referred to as non-chlamydial and non-gonococcal cervicitis,7 non-specific cervicitis4 or cervicitis of unknown aetiology.15
Lack of consensus regarding cervicitis case definition used in studies until now limits interpretation and comparability of cervicitis associations and prevalence estimates found.16 Associations also vary depending on testing methods and population studied.7 ,8 The commonly used microscopic Gram stain diagnosis of >30 polymorphonuclear leucocytes per high-powered field on cervical Gram stain (×1000) (>30 pmnl/hpf) lacks practical application in primary care but remains popular in sexually transmitted infection (STI) clinics17 and clinical research. Cervicitis diagnosis based on clinical assessments such as ‘mucopurulent’ or ‘yellow discharge’ may have wider practicality and pathogen prediction.8 ,17–19
This study aimed to examine infectious and non-infectious variables putatively associated with cervicitis8 ,20 using the cervicitis case definition >30 pmnl/hpf on cervical Gram stain (‘microscopy’). These associations were examined using two other cervicitis case definitions: ‘cervical discharge‘(yellow and/or mucopurulent discharge) and a combined microscopy and clinical definition ‘micro+cervical discharge’ (‘microscopy’ plus ‘cervical discharge’). We present complete findings of 558 women recruited to this study and this paper addresses an important issue in the field by attempting to elucidate not only the aetiology but also an optimal case definition for cervicitis.7 ,16–18 From this point, we can more reliably inform appropriate cervicitis management.
Materials and methods
Study protocol and laboratory methods are described previously.12 ,21 In brief, 558 women enrolled in a cross-sectional study in three STI/HIV clinics in Sydney, Australia, between July 2006 and February 2010. These clinics are publicly funded services for ‘priority populations’, defined in the 2006–2009 NSW Sexually Transmissible Infections Strategy.22 The study was designed within the constraints of standard clinic care, collaborating with SEALS Virology Laboratory, Prince of Wales Hospital, Sydney, for development of three in-house multiplex PCRs for mollicute and viral testing.21
A convenience sample was enrolled: consenting women over 18 years requiring vaginal examination based on report of symptoms or requesting examination. Consequently, enrolment was biased towards symptomatic women. During the study, it was clinic policy for clinicians to collect vaginal and cervical swabs, with patient-collected vaginal swabs or urine testing less commonly offered to asymptomatic women. Re-enrolment was permitted if more than 6 months had elapsed since initial enrolment and a new partner was reported. Exclusion criteria included PID, menstruation, pregnancy, no vaginal sex or cervical surgery in the prior 3 months, antibiotic use in the prior month, sexual assault and current use of intrauterine contraceptive device.
All recruiting clinicians (doctors and nurses) underwent standardised on-site training in Gram stain technique and cervicitis criteria, study protocol, definitions and sample collection. Self-reported behavioural variables were collected verbally from women prospectively on purpose-designed record sheets. Variables included age (<25 years/>25 years), ‘current vaginal symptoms’ (report of one or more of vaginal discharge, odour or itch), reported condom use for vaginal sex (‘always/sometimes’ or never) and number of sexual partners (1, >1) in the last 3 months, current commercial sex work, current smoking, current douching, current use of combined oral contraceptive (COC), current use of injectable depot medroxyprogesterone acetate (DMPA)/progestogen-only pill (POP)/Implanon and menstrual phase (follicular or luteal).
Associations with cervicitis using three cervicitis case definitions were considered:
>30 pmnl/hpf (×1000) in three non-adjacent high-powered fields on cervical Gram stain (‘microscopy’)
yellow and/or mucopurulent cervical discharge (‘cervical discharge’)
combination of ‘microscopy’ plus ‘cervical discharge’ (‘micro+cervical discharge’).
Sampling and laboratory methods
The cervix was visualised using a sterile speculum and excess exudate was removed before taking three endocervical sterile swabs (Copan, California, USA). Yellow discharge was defined as yellow colour of the first white cotton swab from the endocervix and mucopurulent discharge as purulent appearance of endocervical secretions. The first endocervical swab was rolled on a glass slide and Gram stained to define women as having cervicitis by the microscopy definition and then cultured for NG on selective media (lysed horse blood agar containing vancomycin, colistin, nystatin and trimethoprim). A second endocervical swab was collected for CT/NG PCR (Amplicor CT/NG (Roche)) according to standard practice at the time. NG positivity was defined as positive NG PCR confirmed by positive culture, with PCR results carrying less reliability in the low prevalence setting. NG culture was routinely performed for case confirmation and antimicrobial resistance surveillance. A third endocervical swab was taken for multiplex PCR tests21 for MG, Mycoplasma hominis (MH), Ureaplasma urealyticum (UU), Ureaplasma parvum (UP), herpes simplex virus type 1 (HSV1), herpes simplex virus type 2 (HSV2), Epstein–Barr virus (EBV), varicella zoster virus (VZV), cytomegalovirus (CMV) and TV. A high vaginal swab was taken for Gram stain, Candida culture and wet preparation for TV if clinically suspicious. TV positivity was defined by PCR alone or positive PCR and positive wet preparation for TV.12 BV was defined by Nugent's criteria on Gram stain. Diagnosis of Candida albicans was by inoculation onto a CHROMagar Candida plate, then incubated for 48 h at room temperature and read at 24 and 48 h.
A laboratory scientist blinded to clinical findings verified clinician Gram stain slides for diagnosis of cervicitis and BV. All discrepant slides were re-examined and consensus diagnoses were reached. All women diagnosed with STIs were treated.
Associations of infectious and non-infectious variables with cervicitis (the ‘disease’) were estimated using prevalence ratios (PRs) and χ2 testing. As this was a cross-sectional study and the outcome of cervicitis was common, PRs were calculated rather than ORs as the best measure of association. PRs were estimated by log-binomial regression. Multivariate analysis involved forward selection of covariates using variables from the univariate analysis with p<0.05. Sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) of the ‘test’ (different cervicitis definitions) for the significant pathogens were calculated using 2×2 tables. The study was powered (80%, p<0.05) to detect PRs of at least 3.0 (assuming cervicitis prevalence of 20% and pathogen prevalence of 5%) or PRs of at least 2.4, assuming pathogen prevalence of 10%. All calculations were performed with SAS V.9.3 (SAS Institute, Cary, North Carolina, USA).
There were 1327 consecutive women approached with 630 (47.5%) eligible. Of the 630 eligible women, 558 (88.6%) were enrolled (including 21 enrolled twice) and 72 (11.4%) declined participation. Most common reasons for ineligibility were as follows: not requiring vaginal examination (36.7%), recent antibiotic use (13.5%) and menstruation or pregnancy (12.8%). Mean and median ages of women recruited were 30.3 and 28.0 years, respectively, with a range of 17.0–66.0 years.
Pathogen prevalence at enrolment was CT 5.8% (95% CI 3.8 to 7.7%), NG 1.1% (95% CI 0.2 to 1.9%), MG 3.8% (95% CI 2.2 to 5.4%) and TV 3.9% (95% CI 2.2 to 5.5%). Cervicitis by the ‘microscopy’ definition was present in 268/558 (48.0%) women, by the ‘cervical discharge’ definition in 144/540 (26.6%) women and by the ‘micro+cervical discharge’ definition in 131/558 (23.5%) women. ‘Current vaginal symptoms’ were reported in (330/553) 59.9% of enrolled women. Cervicitis was strongly associated with vaginal symptoms by all definitions: microscopy (PR 1.63 (95% CI 1.34 to 2.00) p<0.0001), ‘cervical discharge’ (PR=3.20 (95% CI 2.12 to 4.74) p<0.0001) and ‘micro+cervical discharge’ (PR=3.98 (95% CI 2.53 to 6.28) p<0.0001).
The Gram stain cervicitis diagnosis (‘microscopy’) was accurate. Of clinician Gram stain slides, 87.2% were examined by a laboratory scientist with high agreement between clinician and scientist diagnoses of cervicitis (kappa 0.79 (95% CI 0.74 to 0.84)). There was substantial agreement between clinician labels of ‘yellow discharge’ and ‘mucopurulent discharge’ (kappa 0.69 (95% CI 0.61 to 0.76)) and less agreement between the cervicitis case definitions ‘microscopy’ and ‘cervical discharge’ (kappa (0.45 (95% CI 0.39 to 0.52)).
The prevalence of infectious and non-infectious variables and their univariate associations with cervicitis by different cervicitis definitions are shown in table 1. Notably, the PR of cervicitis was higher for all significant infections using the cervicitis definitions ‘cervical discharge’ and ‘micro+cervical discharge’, being highest for ‘micro+cervical discharge’: CT (PR=2.11 (95% CI 1.41 to 3.16) p=0.0003), NG (PR=4.42 (95% CI 3.79 to 5.15) p<0.0001), MG (PR=2.12 (95% CI 1.32 to 3.42) p=0.002), TV (PR=2.10 (95% CI 1.30 to 3.37) p=0.002) and CMV (PR=1.92 (95% CI 1.21 to 3.03) p=0.005).
Also in univariate analysis, the use of condoms for vaginal sex ‘always/sometimes’ in the last 3 months was associated with a significant reduction in cervicitis risk by all cervicitis definitions (PR=0.72, p=0.034 by ‘micro+cervical discharge’ definition) and also a reduction in vaginal symptoms (PR=0.82 (95% CI 0.71 to –0.95) p=0.006). None of the exposures BV, MH, UU, UP, HSV1, HSV2, HSV1 or HSV2, asymptomatic HSV, EBV, VZV, HIV seropositivity, Candida albicans, age <25 years, smoking, COC, DMPA/POP/Implanon or menstrual cycle phase, were associated with cervicitis by any definition.
Multivariate analysis (table 2) found only CT, MG and TV significantly associated with increased cervicitis risk. NG could not be modelled due to small case numbers, but achieved the highest unadjusted PR. The highest adjusted PRs (APRs) for cervicitis were seen for the cervicitis definitions ‘micro+cervical discharge’: CT APR=2.13 (95% CI 1.38 to 3.30) p=0.0006, MG APR =2.21 (95% CI 1.33 to 3.69) p=0.002 TV APR=2.37 (95% CI 1.44 to 3.90) p=0.0007 and for ‘cervical discharge’: CT APR=1.90 (1.25 to 2.89) p=0.003, MG APR=1.93 (1.17 to 3.19) p=0.011 TV APR=2.02(1.24 to 3.31) p=0.005. Using the ‘microscopy’ definition, only CT (APR=1.30 (95% CI 1.01 to 1.68) p=0.045) and MG (APR=1.55 (95% CI 1.24 to 1.95) p=0.0002) were associated with increased cervicitis risk. The use of condoms ‘always/sometimes’ was associated with reduced cervicitis risk by all cervicitis definitions: ‘microscopy’ (APR=0.86 (95% CI 0.74 to 0.99) p=0.047), ‘cervical discharge’ (APR=0.70 (95% CI 0.53 to 0.94) p=0.015) and ‘micro+cervical discharge’ (APR=0.69 (95% CI 0.51 to 0.93), p=0.016).
Population attributable risk% (PAR%) of the four significant pathogens for cervicitis (assuming causative effects) were calculated using APRs from table 2 and unadjusted PR for NG from table 1 for the case definition ‘micro+cervical discharge’. Single PAR% of the four significant pathogens were as follows: CT 6.2%, NG 3.6%, MG 4.4% and TV 5.1% with a combined calculated PAR% of only 18.0%. The PAR% of condom use for vaginal sex ‘always/sometimes’ in the last 3 months was 25.7%, again using the ‘micro+cervical discharge’ definition.
Table 3 outlines the performance of each ‘test’ or cervicitis definition in predicting significant pathogens. The case definition ‘micro+cervical discharge’ had the highest specificities, PPVs and NPVs, but lowest sensitivities.
This is the largest study to date examining the aetiology of cervicitis and the only study in Australian women. We prospectively examined a comprehensive range of infectious and non-infectious variables putatively associated with cervicitis and compared these associations using different cervicitis definitions to suggest a best working case definition.
Infectious exposures significantly associated with increased cervicitis risk in multivariate analysis were CT, MG and TV (and NG by univariate analysis). The association of reported vaginal symptoms with cervicitis by all definitions should direct clinicians to examine symptomatic women for cervicitis. Reported use of condoms for vaginal sex ‘always/sometimes’ in the previous 3 months was associated with reduced cervicitis risk.
The three cervicitis definitions used in this study yielded widely differing cervicitis prevalence, from 48% (‘microscopy’) to 23.5% (‘micro+cervical discharge’). Such divergent results suggest imprecisions with the current criteria used. A consensus definition is required. A similar poor agreement between cervicitis case definitions was also demonstrated by Falk16 who examined definitions, including mucopus and >30 pmnl/hpf, and found only a 40.5% and 35.3% agreement, respectively, in women infected or not infected with either CT or MG.
A case definition for cervicitis with the best clinical utility is one easily made in most clinical settings, with good prediction of pathogens that does not overdiagnose ‘disease’.18 Knowing how reliably the diagnosis of cervicitis predicts an STI might mean the difference between treating presumptively or waiting for test results. The case definition ‘micro+cervical discharge’ had the highest APRs, PPVs and specificities but lowest sensitivities for the significant pathogens CT, NG, MG and TV. The collective PAR% of these pathogens accounted for only 18% of cervicitis, leaving much cervicitis unexplained and not necessarily linked to the presence of a known STI. Balancing practicality, sensitivity and specificity of the ‘test’ or cervicitis definition adopted, the harm of false positives (burdening women and their partners with a disease diagnosis and incumbent risk of antibiotic overuse) probably outweighs the harm of a less sensitive ‘test’ for a condition with unclear infectious aetiology and health ramifications. Thus, we suggest that ‘cervical discharge’ may be the optimal cervicitis case definition to use in primary care and ‘micro+cervical discharge’ definition best for STI services where microscopy is generally available. Marrazzo et al18 also found cervical signs such as mucopus to have better PPV for chlamydia or gonorrhoea than the case definition >30 pmnl/hpf, particularly in women under 25 years.
Individual PPVs for significant pathogens were low and NPVs were high, reflecting the low population STI prevalence, in keeping with other prevalence reports in similar populations.23 These values would increase with higher STI prevalence and more highly triaged situations, but might be lower in community settings.
Little is known about the natural history of cervicitis. It is reasonable to postulate from our results that factors other than known infective agents are involved7 and that cervical inflammation may be part of a normal continuum in some women. The protective effect of condoms, particularly in women without a known pathogen, suggests involvement of other exogenous variables, infectious or otherwise. Inflammatory cytokines have been implicated in cervical inflammation.7 Sharkey et al24 found seminal fluid elicited expression of proinflammatory cytokines and chemokines in the cervix, which may offer some explanation of the protective condom effect. A US study25 of chemokines and cytokines found significantly higher levels of interleukin (IL)-1α, IL-1β, IL-6, IL-8, macrophage inflammatory protein 1 alpha (MIP-1α), RANTES, tumour necrosis factor alpha (TNFα), IL-10, IL-12 and interferon gamma (IFNγ) in women with immature epithelium compared with women with mature epithelium. With the exploration of the vaginal microbiome,14 ,26 new pathogens involved in cervical inflammation may emerge.
There are important negative study findings, of note the lack of associations with previously implicated infections such as BV,8 ,13 ,14 the mollicutes MH and UU and HSV.20 Gaydos et al9 did not find an association between cervicitis and BV (nor with CT, NG or TV) using the definition ‘cervical discharge or mucopurulent discharge or cervical friability’. Both BV and MH have been associated with PID,27 so their association with cervicitis might be expected. A closer look at the microbiome of BV and cervicitis, using quantitative PCR and broad-range 16S rRNA gene PCR with pyrosequencing,14 recently found an association with Mageeibacillus indolicus and possible protective effect of Lactobacillus jensenii.
Of the 20 women in this study positive for either cervical HSV1 or 2, only four were symptomatic (current genital ulceration). Of these symptomatic women, three-fourths had cervicitis (‘microscopy’ definition). When the four women with symptomatic HSV were excluded from the analysis, non-significant associations for all cervicitis definitions were seen with HSV1, HSV2 and HSV1 or 2. Thus qualifying our findings, there was no significant association of asymptomatic cervical HSV infection with cervicitis. Despite strong univariate association, CMV was not associated with cervicitis in multivariate analysis, supporting its role as a bystander rather than cause of cervicitis.20 The most common coinfection with CMV was CT (5/28) (table 4) and CT was the only pathogen potentially confounding the association of CMV. With CT removed from the multivariate model, CMV remained non-significant for all definitions. Additionally, the population from which participants were drawn has approximately 50% seropositivity to CMV28 making any conclusions of an association difficult due to high background rates of CMV latency.
STIs enhance HIV shedding from the cervix.4 We reported an early study finding29 of a significant association between MG and HIV, as well as MG and cervicitis. Manhart et al30 described increased HIV shedding in women with high cervical MG burden; it may be that MG-associated cervicitis in particular is an important HIV transmission driver in populations where both infections are prevalent. Our study did not show an association between HIV positivity and cervicitis.
A curious finding was >1 partner being associated with a reduction in cervicitis by the ‘cervical discharge’ definition (RR=0.71, p=0.038) but this could be partly explained by condom use, which was a confounder in this association.
Study strengths and limitations
As well as clarifying the association of CT, NG, MG and TV with cervicitis and their potential contribution towards cervicitis, we have demonstrated a protective effect of reported condom use. We also present important negative findings concerning less-studied infections such as BV, MH, UU, HSV and CMV. This is the first study to compare variable associations using commonly accepted cervicitis definitions and suggest a best working case definition.
Design was cross-sectional with limited inference on causality, but sample size was large and appropriately powered (80% p<0.05) to detect pathogen PRs of at least 2.5–3.0. Clinical practice and funding constraints directed examination of a convenience sample in a prospective real-life study of symptomatic women and asymptomatic women who chose to have an examination over self-collected samples. Consequently, enrolment rates were high among eligible women but biased towards symptomatic women (60%). This selection bias and the finding that symptoms were strongly associated with cervicitis would have contributed to high cervicitis rates. Recruitment occurred at three STI/HIV clinics using a small group of trained clinicians with advantages for consistency. All cervicitis definitions involved an element of clinician judgement and subjectivity, particularly the non-microscopic case definitions. The findings from this study may be less applicable in lower STI prevalence settings.
This study focused primarily on an extensive range of known infectious variables using in-house published and validated PCR methods with additional prospective behavioural variables relying upon patient history. Exploration of the role of cytokines, utilisation of more novel non-cultivation methods to assess the vaginal microbiome and treatment outcomes were outside the scope of this study.
In summary, we have clarified significant infectious associations of cervicitis and suggest a protective role of condoms. Our findings prompt a re-evaluation of current understanding of cervicitis aetiology, definition and significance and reinforce the need for exploration of other potential infectious and non-infectious factors.
Cervicitis is associated with Chlamydia trachomatis, Neisseria gonorrhoeae, Mycoplasma genitalium and Trichomonas vaginalis. Women with cervicitis should be tested for these pathogens.
The combined population attributable risk% of these pathogens was only 18.0%. Improved knowledge of natural history, vaginal microbiome and cytokines may further define cervicitis aetiology.
The protective effective of condoms, particularly in women without a known pathogen, suggests involvement of other exogenous variables, infectious or otherwise.
The cervicitis definitions ‘cervical discharge’ or ‘micro+cervical discharge’ had better clinical utility and pathogen prediction than the microscopy definition >30 pmnl/hpf (polymorphonuclear leucocytes per high-powered field on cervical Gram stain) alone.
This study was conducted in three publicly-funded (New South Wales Health) STI clinics in Sydney, Australia. Christopher J McIver and Christa McPherson provided expert laboratory input. The staff and patients of the Short Street Centre provided invaluable input.
Handling editor Jackie A Cassell
Contributors MJL is the principal and corresponding author. MJL, PK, RGC and WDR designed the study. MJL and PK coordinated the study and MJL supervised recruitment and managed databases at all sites. MJL was chief investigator. ZWN and WDR developed and carried out the PCR testing. MJL and FLG performed the statistical analysis and interpretation with input from RGC. All authors contributed to and reviewed the final manuscript.
Funding No grants supported this study but MJL received a one off research scholarship(funded by Novartis) in 2006 from the Australian College of Sexual Health Physicians.
Competing interests None declared.
Ethics approval Ethics approval for the three clinical sites was granted by two HRECs: the South Eastern Sydney and Illawarra Area Health Service Human Research Ethics Committee (06/09LUSK) and the Sydney South West Area Health Service Ethics Review Committee (X07-0277).
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.