METHODS

The study protocol was reviewed and ethically approved by the national ethics committees of Bénin, Burkina Faso, Ghana, Guinée, and Mali. Women who gave consent were enrolled in 11 health facilities in the cities of Cotonou (Centre de Santé de la Circonscription Urbaine, Centre de Santé Béthesda, Centre Médical St-Jean, Maternité Lagune), Ouagadougou (Centre Médical de Wemtenga), Accra (Adabraka polyclinic), Kumasi (Suntreso polyclinic), Conakry (Centre de Santé de Madina, CHU de Donka), and Bamako (Centre de Santé d’Asacobafa and Commune-6) if they presented with vaginal discharge as their main complaint. Exclusion criteria were being a SW, having abdominal pain, or pregnancy. A standardised questionnaire was completed which gathered demographic and behavioural information as well as a review of current complaints. A pelvic examination was performed, the appearance of the discharge was noted, the cervix was visualised with a speculum and appropriate specimens were obtained. All women were treated with a single dose of 2 g of metronidazole, as well as with clotrimazole vaginal cream for 3 days. They were asked not to have intercourse during the next week, but no partner notification was carried out. Participants were randomised to one of two groups: (i) in group A, participants were asked to come back 7 days later only if there was no improvement in the discharge; (ii) in group B, all participants were asked to come back after 7 days, regardless of response to treatment. On day 7, a combination of ciprofloxacin (500 mg single dose)/doxycycline (100 mg twice a day for a week) was given to all women in group A, and to women in group B with no response or only partial response to treatment. Participants were asked to come back on day 14 if there was no improvement, and treatment was then left to the discretion of the local medical staff.

A cervical swab was obtained from all participants on day 0 for detection of pathogens by polymerase chain reaction (PCR). The presence of Neisseria gonorrhoeae and Chlamydia trachomatis was documented using a commercial PCR assay (Amplicor, Roche Diagnostics, Branchburg, NJ, USA) while an in-house PCR was used for detection of Trichomonas vaginalis, as described elsewhere.⁹ The cervical swabs were taken with the Amplicor collection kit according to recommended procedures, kept at 4–8°C for a variable duration (from a few days up to 6 months) and sent to a central laboratory in Canada. Data were entered, verified, and analysed using Epi-Info 6.04. 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.

RESULTS

A total of 770 women were enrolled between September 2000 and July 2002. Twenty four were excluded because they either had abdominal pain (20), were pregnant (one), or both (three), leaving 746 women for analysis. For 15 women, the sample was lost and never reached the central laboratory, and in five participants inhibitors were present in the specimen and it was not possible to carry out the PCR, leaving 726 participants for whom results of all PCR tests were available. Table 1 shows the prevalence of N gonorrhoeae and/or C trachomatis according to various characteristics. Overall, gonoccocal and chlamydial infections were found in only 1.9% (14) and 3.2% (23) of participants respectively. The prevalence of gonococcal infection was significantly lower in Ghana, Bénin, and Burkina Faso than in Guinée and Mali. Neither age, marital status, the presence of a new sexual partner, the presence of urethral discharge in the partner, nor reported condom use by the partner were associated with the presence of cervical infections. Chlamydial and gonococcal infections were absent in women who denied any sexual activity in the past 3 months. Gonococcal or chlamydial infections were more frequent in women without any formal education. Gonococcal infection was marginally more frequent among women who complained of dysuria and in those with pelvic pain on examination, while chlamydial infection was more common in women with dyspareunia and with pus on the cervical swab, but the majority of women with these symptoms or signs did not have any cervical infection.

The distribution of some characteristics, including risk factors, clinical findings, and the prevalence of various pathogens according to the two groups determined by randomisation is shown in table 2. The two groups were well balanced for demographic and behavioural characteristics, and did not differ either for marital status, education, condom use by the partner, or the presence of other symptoms such as dysuria, dyspareunia, vulvar pruritus and fever (data not shown). Women in group B (to come back regardless of symptoms) were more often found to have an inflammatory cervix, bleeding after sampling, or a genital ulcer than participants in group A.

Thirty eight per cent (140/371) of women in group A came back on day 7 compared to 62% (231/375) of women in group B (p<0.001). Table 3 shows the frequency of a visit on day 7 in each group. Within each treatment group, there was no significant difference in the frequency of a return visit whether or not a cervical pathogen had been found on day 0. In group B, women with cervical infections tended to be less likely to come back on day 7 than those without a cervical infection.

Table 4 summarises the subjective response to treatment according to the presence of various pathogens on day 0. Surprisingly, the frequency of a complete or partial response was the same in both groups. Within each group, the presence of a cervical infection on day 0 was not predictive of a worse symptomatic response. The subjective response reported by the patient and the presence or absence of a discharge seen by the healthcare provider were strongly correlated (data not shown). As also shown in table 4, the presence of NG/CT on admission was not associated with discharge seen on day 7. In such circumstances, strategy A (come back if not improved, treat for NG/CT all those who come back) had, for the proper treatment of cervical infections, a sensitivity of 60% (6/10), a specificity of 64% (224/350), and a positive predictive value of only 5% (6/132). Strategy B (ask all to come back, treat for NG/CT those with no response or only partial response) had a sensitivity of 23% (5/22), a specificity of 66% (226/344), and a positive predictive value of 4% (5/123).

DISCUSSION

The prevalence of cervical infections with NG (1.9%), CT (3.2%) and NG/CT (4.4%) that we documented among women with vaginal discharge in five west African countries was somewhat lower than that observed in Bénin a few years ago (NG: 5.7%, CT: 2.1%, NG/CT: 7.8%)¹⁰ and much lower than that reported in east or southern Africa, where NG was found in 7–17%, CT in 5–11%, and NG/CT in 11–20% of women with vaginal discharge.^11–14 This lower prevalence of gonococcal and chlamydial infections in west African women with vaginal discharge was mirrored at the population level in the four cities study in which the prevalence of cervical infections in women aged 15–49 was lower in Cotonou (NG: 0.9%, CT: 1.3%) than in Yaoundé (2.7%, 9.4%), Kisumu (0.9%, 4.5%), and Ndola (2.3%, 2.9%).¹⁵ A similar low level of cervical infections was also noted in surveys of women of reproductive age in the Gambia (NG: 0%; CT 0.8–1.2%).^16,17 In contrast, the prevalence of these agents in large community surveys of women aged 15–39 years in Masaka and Rakai, Uganda and Mwanza, Tanzania, adjusted for selection biases and diagnostic errors, was 1.8–2.3% for NG and 1.6–13.0% for CT.¹⁸

The higher prevalence of NG/CT among women with vaginal discharge in east and southern Africa might be to some extent explained by the presence of SW in these study populations and their deliberate exclusion from the present study. A sample of women consulting in a health centre for vaginal discharge might include a large fraction of non-SW with a low prevalence of cervical infections, and a small fraction of SW with a high prevalence of cervical infections, resulting in a measure of prevalence which is intermediate between these two extremes. We also excluded women who spontaneously complained of abdominal pain, which might have corresponded in some of them to pelvic inflammatory disease caused by NG and/or CT.

Although we did not obtain simultaneously, in the same cities, cervical samples of sexually active women who did not have vaginal discharge, there is little evidence from our data that in west Africa NG and/or CT infections are associated with the occurrence of vaginal discharge. If such an association exists, it must be very weak indeed. The prevalence of NG and CT in women with vaginal discharge in Cotonou that we measured (1.2%, 1.2%) was nearly identical to what was documented at the population level in the same city 3 years earlier (0.9%, 1.3%), the latter measure using urine which is slightly less sensitive in women than a cervical specimen.¹⁵ Such findings are comparable to those reported from Bangladesh, where the prevalence of cervical infections was similar in women complaining of vaginal discharge and in women attending antenatal clinics, family planning clinics, or consulting for pregnancy interruption.¹⁹

Given this low prevalence, it will be impossible to locally identify, in west Africa, risk factors that are predictive of cervical infections. Among our sample of 726 participants with vaginal discharge, only 2/31 (6%) women with NG/CT reported that their partner had a urethral discharge. Neither age, being unmarried, nor having a new partner in the past 3 months was associated with a higher risk of cervical infection. Among the risk factors that used to be recommended by WHO, the only one which was associated with NG/CT in our study population was the number of sexual partners in the past 3 months. When taken together, the previously WHO recommended risk factors for the presumptive treatment of cervical infections among women presenting with vaginal discharge, still used in several west African countries (partner symptomatic or any two of: age<21 years, single, more than one partner, or new partner in past 3 months) had a sensitivity of 37.5%, a specificity of 74.8%, and a dismal positive predictive value of 6.4% which was merely 1.5-fold higher than the prevalence of cervical infections. Other studies of the same risk factors for the identification of cervical infections among women with vaginal discharge in Africa have documented a positive predictive value which was only 1.3–1.7 times higher than the prevalence.^10–13

It should then come as no surprise that women with a cervical infection were not any more likely to show up for a return visit on day 7 or, if they came, to report that there was no improvement in their symptoms than women who did not have a cervical infection. Since gonococcal and chlamydial infections are not associated with the symptom of vaginal discharge, they cannot influence the likelihood of non-response to a first line treatment targeting only the microbial agents of vaginitis. We believe it is futile to look for other risk factors predictive of the presence of NG/CT in women with vaginal discharge in west Africa. The current recommendation that risk factors for cervicitis should be identified locally⁵ seems like an invitation to use, in public health programmes, two or three characteristics which happened to have been associated with agents of cervicitis by chance. Strategies for the control of gonorrhoea and chlamydia, and the reduction of their potentially disastrous consequences, need to be redesigned. Our data indicate a more fruitful approach. Within the five countries that we surveyed, the prevalence of NG/CT in women was significantly lower in Ghana and Bénin (5/280, 1.8%), where large scale interventions for sex workers have been ongoing for several years,^20,21 compared to Burkina Faso, Mali, and Guinée (27/446, 6.1%) (p = 0.01), where such interventions did not exist, at least not on a city wide scale. Although causality can not be inferred and other cultural and behavioural might impact on the prevalence of NG/CT, this suggests that interventions aiming to increase condom use and to improve the management of STI among SW eventually lead to a reduction in the prevalence and incidence of gonococcal and chlamydial infection among the general adult female population. The core group of SW and their clients offer a relatively untapped opportunity for the control of gonococcal and chlamydial infections and, of course, for the control of HIV. Rather than waiting endlessly for a rapid test which would be cheap, easy to perform, highly sensitive and highly specific, resources need to be oriented immediately towards making available preventive and curative services for SW, not only in pilot projects in capital cities, but in each urban centre where at least 20 such women sell their services, aiming at a nationwide coverage.

In west Africa, women who are not SW who present with vaginal discharge without abdominal pain should be treated only for agents of vaginitis. Less than 5% of them have a cervical infection, which is probably not higher than the population prevalence among women in childbearing years. The few who have a cervical infection probably do so incidentally, and their discharge might be caused by concomitant vaginitis. The syndromic management of vaginal discharge in west Africa should focus on the relief of symptoms through an effective treatment of common causes of vaginitis, preferably with a 7 day course of metronidazole (which is superior to single dose metronidazole in women with bacterial vaginosis²²) combined with either topical clotrimazole or generic fluconazole orally.⁶ Rather than trying to address both categories of patients in the same flow chart,⁵ we think it might be more fruitful to recommend an algorithm for the management of vaginal discharge in non-SW women, focusing on vaginitis, and a second algorithm for the management of vaginal discharge in SW, the truly high prevalence high risk population in west Africa, for which treatment of gonococcal and chlamydial infection should be considered as well.