During the past decade, Mycoplasma genitalium has been identified as a relatively common cause of non-gonococcal urethritis, both in industrialised countries and in the developing world.^1–4 Ongoing investigations to define its pathogenicity in women have yielded somewhat contradictory findings. M genitalium was found more frequently in Japanese, American, and Swedish women with cervicitis than among controls, but no such association was found in France.^5–8 In Kenyan women, M genitalium infection was associated with acute endometritis⁹ but its prevalence was low among women with salpingitis.¹⁰M genitalium was recovered more frequently in samples from British women with pelvic inflammatory disease than in controls.¹¹ In Guinea-Bissau, M genitalium infection was not associated with adverse outcomes of pregnancy.¹² To identify the contribution of M genitalium to the aetiology of cervicitis among female sex workers (FSW) in sub-Saharan Africa and its relative importance in the overall burden of sexually transmitted infections (STI), we looked for M genitalium in cervical samples of women during cross sectional studies of FSW in Ghana and Bénin.

METHODS

FSW were recruited at FSW/STI clinics in Accra (Ghana), Cotonou and Porto Novo (Bénin) between March 2001 and April 2002, during the initial visit of a longitudinal study whose primary goal was to evaluate the effectiveness of monthly antibiotics on the prevalence of gonococcal and chlamydial infections, the results of which will be reported elsewhere.¹³ The population of FSW thus represented a cross sectional sample, and most did not complain of symptoms of genital infection. The study protocol was approved by the national ethical committees of Bénin and Ghana. After informed consent was obtained, a questionnaire gathering demographic and behavioural information was administered. A pelvic examination was immediately carried out by a medical doctor or midwife; the cervix was visualised with a speculum. Clinical signs potentially indicating cervicitis were cervical discharge, pus on the cervical swab (“swab test”), bleeding after cervical sampling, inflammatory cervix (oedema and erythema). The presence of cervical motion tenderness was also recorded. A cervical swab was taken with the Amplicor collection kit according to recommended procedures, kept at 4–8°C, and sent to a central laboratory in Canada. During this initial visit, antibiotics were given when clinically indicated.

The presence of Neisseria gonorrhoeae and Chlamydia trachomatis was documented using a commercial polymerase chain reaction (PCR) assay (Amplicor, Roche Diagnostics, Branchburg, NJ, USA). All repeatedly N gonorrhoeae positive samples were retested using an in-house 16S rRNA PCR assay¹⁴; samples were considered positive when confirmed by the 16S rRNA PCR assay. In-house PCR were used for detection of M genitalium and Trichomonas vaginalis, as described elsewhere.^3,4 The laboratory personnel performing these assays was blinded to the previously recorded clinical findings.

Data were entered and verified using Epi-Info 6.04, and analysis was performed with Stata 6.0. Proportions were compared with the χ² test using Yates’s correction or, when an expected cell value was less than 5, with Fisher’s exact test. Logistic regression was used for multivariate analysis.

RESULTS

A total of 382 different FSW were recruited in Ghana and 444 in Bénin. In Ghana, they were all home based; in Bénin, 192 were brothel based, 223 were home based, and 29 usually picked up clients on the street. The prevalence of M genitalium (MG) infection was 26.3% (217/826), compared to 16.0% (132/826) for N gonorrhoeae (NG), 3.4% (28/826) for C trachomatis (CT) and 23.1% (191/826) for T vaginalis (TV). Table 1 shows the frequency of co-infections among these women. The majority of women in whose cervical sample MG was found did not harbour any other co-pathogen.

Table 2 displays the association between sexually transmitted pathogens, sociodemographic, and behavioural characteristics among the FSW, in univariate analyses. All four pathogens were more prevalent in FSW in Bénin than in Ghana (within Bénin, they were more prevalent in Porto Novo than in Cotonou, except for TV), but this was significantly so only for NG. While the prevalence of CT decreased with age, this was not seen for the other pathogens. MG and TV were significantly less prevalent in women with post-primary education than in their less educated counterparts, and more prevalent in women who were divorced or widowed. NG was less common in FSW who charged more than the average price for each intercourse. The presence of MG was not associated with the average number of clients per week, but was more common among FSW who had five or more clients the previous day, and in women who did not use a condom with all of their clients of the preceding week. NG and TV infections were also more common in women who did not use a condom with all clients of the preceding week. MG or NG infections were two times more frequent among HIV infected FSW than among their seronegative colleagues. Having had sex with a “boyfriend” in the previous week was not associated with a higher prevalence of MG, NG, CT, or TV (data not shown). In a logistic regression model, the independent correlates for infection with MG were HIV infection (adjusted odds ratio (AOR) 2.9; 95% confidence interval (CI) 2.0 to 4.3; p<0.001), having only primary school (AOR: 2.4, 95% CI: 1.2 to 4.7) or no formal education at all (AOR 2.5; 95% CI 1.2 to 4.9; p = 0.01), experience of sex work shorter than 1 year (AOR 1.8; 95% CI 1.2 to 2.8; p = 0.005) and working in Porto Novo (AOR 1.7; 95% CI 1.0 to 2.7; p = 0.04) or Cotonou (AOR 1.6; 95% CI 1.1 to 2.4; p = 0.01) rather than in Accra.

Table 3 compares the frequency of clinical signs among women with or without various pathogens. The presence of vaginal discharge was significantly but weakly associated with each of these four pathogens. Vulvar inflammation was uncommon and not associated significantly with any pathogen (data not shown). Each of the four signs potentially indicating cervicitis (cervical discharge, pus on the cervical swab, bleeding after sampling, inflammatory cervix) as well as cervical motion tenderness were associated with each of the four pathogens. By and large, the unadjusted odds ratios for signs of cervicitis were higher for NG and CT, and lower for MG and TV. The presence of a genital ulcer on clinical examination was not associated with a positive PCR for MG, NG, CT, or TV (data not shown). However, as shown in table 4, the associations between signs of cervicitis or cervical motion tenderness and the presence of MG were stronger among women in whom no vaginal discharge was present; for three signs, there was a significant interaction.

Logistic regression analyses were carried out for each of the four signs of cervicitis and for cervical motion tenderness as the outcomes, in which the odds ratios for each pathogen were adjusted for the presence of the other three pathogens and, in three of the models, for age (in models for cervical discharge and pus on swab, age did not improve the fit and was not kept). The presence of vaginal discharge did not improve the fit of any of the models and was thus not further considered. When adjusted for the presence of NG, CT, and MG, the presence of TV was no longer significantly associated with any of the five clinical signs and was also removed from the final models. As shown in table 5, NG was strongly and independently associated with each of the four signs of cervicitis as well as with cervical motion tenderness. The adjusted odds ratios for CT were lower than those of NG, and either similar or somewhat higher than those of MG. However, owing to its higher prevalence, the associations between MG and signs of cervicitis were all statistically significant; there was no significant association with cervical motion tenderness.

Table 6 shows the sensitivity, specificity, positive predictive value, and negative predictive values of two clinical scores to identify among FSW the presence of agents of cervicitis, without any laboratory back-up. The score with the highest PPV/prevalence ratio was the one incorporating the five signs and defining as positive any woman in whom at least two of these signs were found. This score had a sensitivity of 43% and a positive predictive value of 36% in identifying FSW with gonococcal infection. When taking into consideration all three pathogens, the positive predictive value jumped to 61% but the sensitivity dropped to 29%.

DISCUSSION

Among 826 FSW seen during a cross sectional survey, associations were found between M genitalium and the presence of four signs generally thought to be indicative of cervicitis. As this was a post hoc initiative exploiting data accumulated during the course of another study,¹³ we did not obtain Gram stains of cervical secretions which might have given us a more objective definition of cervicitis.⁶ However, this point is debatable as the cut off for the presence of mucopurulent cervicitis used by other researchers varies from ⩾10, ⩾20, and ⩾30 polymorphonuclear leucocytes per oil immersion field,¹⁵ underlining that there is no clear cut cytological definition of cervicitis.

Is the association between M genitalium and signs of cervicitis causal? Some of the Bradford Hill¹⁶ criteria have been met by studies addressing this issue, including ours: consistency (an association between M genitalium and cervicitis has been found in Japan, the United States, Sweden, and now west Africa^5–7), biological plausibility (experimental infections in primates¹⁷), analogy (M genitalium is clearly associated with urethritis in men,^1–4 agents of urethritis such as N gonorrhoeae and C trachomatis cause cervicitis in women, thus it would make sense for M genitalium to cause cervicitis as well), and coherence (many women with cervicitis have no known pathogen found¹⁵). Temporality and dose response have not been addressed by the literature so far, and the strength of the associations (that is, the odds ratio) is rather weak. The only study that did not find an association between M genitalium and cervicitis included only women with a vaginal discharge⁸: our findings suggest that, perhaps as a consequence of its modest pathogenicity, the signs of M genitalium induced cervicitis are difficult to elicit in women who also have vaginitis. Ultimately, experimental data will provide the definitive answer: does the treatment of women with cervicitis, among whom M genitalium but not N gonorrhoeae nor C trachomatis has been recovered, with a drug active against M genitalium lead to the disappearance of cervicitis in a timely manner?

In this population of west African FSW, among whom there was a high frequency of condom use during transactional sex, M genitalium was eight times more prevalent than C trachomatis and 1.6 times more prevalent than N gonorrhoeae. It might be that the natural history of M genitalium cervical infection is characterised by a longer duration of infection, and this could be further compounded by the frequent co-infection with HIV. Alternatively, as M genitalium is less strongly associated with signs of cervicitis than are the gonococcus and C trachomatis, M genitalium infected sex workers might seek care and receive antibiotics less often than when they are infected with the other two pathogens.

C trachomatis was remarkably uncommon (3.4%) among this population of FSW, even among those aged 30 years or less (5.7%). Low prevalence of C trachomatis among west African FSW has been noted in recent surveys in Cotonou (5.1%),¹⁸ in a multicentre study which included sites in Côte d’Ivoire and Bénin (1.9% in Abidjan and 5.9% in Cotonou)¹⁹ (B Mâsse, personal communication), and in the four cities study which documented a prevalence of 4% among FSW in Cotonou, compared to 8% in Kisumu (Kenya), 9% in Ndola (Zambia), and 18% in Yaoundé (Cameroon)²⁰ (all of which used nucleic acid amplification tests (NAAT)). This was mirrored at the population level, with a prevalence of C trachomatis as low as 1.3% among a random sample of adult women in Cotonou, compared to 4.5% in Kisumu, 2.9% in Ndola, and 9.4% in Yaoundé,²¹ perhaps reflecting to some extent the long standing presence in Bénin, Ghana, and Côte d’Ivoire of interventions targeting FSW, with relatively high rates of condom use and reasonable access to specialised clinics where treatment against C trachomatis and N gonorrhoeae is provided when cervicitis is suspected. The prevalence of gonorrhoea was not appreciably lower in sex workers of Cotonou than in the other three cities²⁰: C trachomatis being less transmissible than N gonorrhoeae,²² its prevalence might be more readily lowered by a high rate of condom use during transactional sex. Recent studies of FSW from east Africa, also using NAAT, have documented somewhat higher C trachomatis prevalences: 7% in Kenya, 12% in Tanzania, 13% in South Africa, and 16–17% in Madagascar.^23–26 Whether populations vary in their intrinsic, biological susceptibility to sexually transmitted pathogens or simply differ in the frequency of condom use during transactional sex deserves further study.

The population attributable fraction of a given pathogen with regard to a given syndrome depends on both the prevalence of the pathogen, and the strength (odds ratio) of the association. In this case, taking as the measure of association the means of the adjusted odds ratios of the four signs of cervicitis shown in table 5 (4.9 for N gonorrhoeae, 2.6 for C trachomatis, 1.7 for M genitalium) and the prevalences that we measured, it is possible to estimate that 38% of cases of clinical cervicitis are caused by N gonorrhoeae, 16% by M genitalium, and 5% by C trachomatis. The gonococcus thus remains the predominant agent of cervicitis in this population of FSW, but M genitalium contributes more than C trachomatis.

Several independent correlates of M genitalium infection among FSW were identified. It was more frequent in women with little or no formal education, which might be related to health seeking behaviour or condom use. M genitalium was also more common in FSW who had been involved for less than 1 year in the trade, which might relate to the same factors, or to a lack of acquired immunity to the pathogen. Work in Porto Novo or Cotonou rather than in Accra was a risk factor, which we believe is related to a lower rate of condom use in Bénin as documented in surveys of clients.^27–29 In the current study, adjustment for condom use (as a marker of the number of unprotected intercourses) reported by the FSW did not enhance the fit of the model and was thus not kept in the final model, emphasising that self reports by FSW might be unreliable, leading to the presence of residual confounding. In this context, the rather strong independent association between M genitalium and HIV infections, also described in other studies^12,30 should be interpreted cautiously. HIV infection was also associated with the presence of gonococcal cervicitis, which could be interpreted as an additional argument suggesting that there was some residual confounding. Other potential mechanisms, which could not be sorted out because of the cross sectional design, are either that M genitalium might facilitate the transmission of HIV from clients to FSW, that HIV might prolong the duration of untreated M genitalium infection, and that HIV might reduce the effectiveness of antibiotics, given for other reasons, against M genitalium. Only longitudinal studies could answer these questions.

Where does that lead us with regard to the syndromic management of STI among FSW in developing countries with no access to laboratory back up? For FSW, the administration of a treatment against agents of cervicitis seems warranted when they are found to have two or more of the following signs: cervical discharge, pus on swab, bleeding after sampling, inflammatory cervix, cervical motion tenderness. One could also drop cervical discharge, and consider as having cervicitis any FSW with at least one of the four other signs. Either strategy would lead to the proper treatment of almost half of the cases of gonococcal infections and, if one considers the presence of N gonorrhoeae, C trachomatis, and M genitalium, in the overtreatment of 39–43% of women found to fulfil these criteria. The higher prevalence of sexually transmitted pathogens among sex workers result in a positive predictive value much higher than among women not involved in transactional sex, and these recommendations cannot be extrapolated to other populations. Furthermore, the identification of signs of cervicitis by polyvalent health care providers who carry out speculum examinations infrequently might be less reliable.³¹ This argues in favour of organising the care of FSW around specialised nurses or medical officers, working within primary healthcare institutions. The proper control of these infections among the core groups of FSW might be the key to reducing their prevalence in the general adult population.³²