Amida, Texina Barivelo, Gillian Burkhardt, Rick Homan, Natalie Kruse, Onja Rahamefy, Leonardine Raharimalala, Zo Raharimanana, Jacqueline Rakotoarisoa, Dimisoa Rakotondramarina, Andriamahenina Ramamonjisoa, Jean René Randriasamimanana, Norbert Ratsimbazafy, Perle Rasanjimanana, Andry Rasamindrakotroka, Marie Clara Chantal Rasoamanarivo, Andry Rasoloarimanana, Solofoson Rakotonandrasana, Benjamin Ravelojaona, Marie Madeleine Razafinoro, Saholinirina Ranivoarimanana Fleur de Line, Cherif Soliman, Richard Steen.
background Sex work is frequently one of the few options women in low-income countries have to generate income for themselves and their families. Treating and preventing sexually transmitted infections (STIs) among sex workers (SWs) is critical to protect the health of the women and their communities; it is also a cost-effective way to slow the spread of HIV. Outside occasional research settings however, SWs in low-income countries rarely have access to effective STI diagnosis.
objectives To develop adequate, affordable, and acceptable STI control strategies for SWs.
methods In collaboration with SWs we evaluated STIs and associated demographic, behavioural, and clinical characteristics in SWs living in two cities in Madagascar. Two months post-treatment and counselling, incident STIs and associated factors were determined. Evidence-based STI management guidelines were developed with SW representatives.
results At baseline, two of 986 SWs were HIV(+); 77.5% of the SWs in Antananarivo and 73.5% in Tamatave had at least one curable STI. Two months post-treatment, 64.9% of 458 SWs in Antananarivo and 57.4% of 481 women in Tamatave had at least one STI. The selected guidelines include speculum exams; syphilis treatment based on serologic screening; presumptive treatment for gonorrhoea, chlamydia, and trichomoniasis during initial visits, and individual risk-based treatment during 3-monthly follow-up visits. SWs were enthusiastic, productive partners.
conclusions A major HIV epidemic can still be averted in Madagascar but effective STI control is needed nationwide. SWs and health professionals valued the participatory research and decision-making process. Similar approaches should be pursued in other resource-poor settings where sex work and STIs are common and appropriate STI diagnostics lacking.
Curable sexually transmitted infections (STIs) are common in Madagascar (Harms et al. 1994; Behets et al. 1996, 1999, 2001b) and sex work is especially visible in the capital and in port and mining areas. To decrease the STI burden and to maintain the still relatively low HIV prevalence, improved primary care of symptomatic STIs has been promoted island-wide, based on research conducted in Antananarivo in 1997 (Behets et al. 1999, 2001b). STI interventions directed specifically at sex workers (SWs) are also needed (Laga et al. 1994) and have been reported as highly cost-effective to prevent HIV in Africa (Creese et al. 2002). However, effective diagnosis of chlamydial and gonococcal infections in female SWs who present for regular follow-up visits requires laboratory diagnosis that is not available in most of Madagascar given its cost, time, personnel, and infrastructure requirements. In addition, the algorithms that were developed in 1997 for syndromic STI treatment cannot be used for routine management of SWs who are most often asymptomatic.
Prior to 2000, the quality of STI care provided to SWs in Madagascar left much to be desired. For instance, clinical exams were not frequently performed and rarely included a pelvic exam. SWs were not screened serologically for syphilis. Vaginal and occasionally cervical secretions were often sub-optimally collected; microscopic exams of the genital secretions, interpretation and subsequent management of the results were frequently inadequate (extracellular Gram negative diplococci recorded as ‘gono (+)’; clue cells not recognized, treatment for gonococcal infection based on positive microscopy, chlamydial infections ignored).
The laboratory tests that were used for follow-up of SWs in Madagascar were of doubtful utility given their inherent failure to diagnose gonococcal and chlamydial infections accurately. Routine screening of SWs for STIs based on vaginal smear microscopy may even be harmful because this practice may create a false sense of security and encourage sexual risk-taking. Local research was, thus, necessary to identify effective and affordable STI screening and treatment strategies for female SWs.
Potential STI management strategies for SWs who present for regular follow-up visits include treatment by aetiology, i.e. based on laboratory diagnosis (Laga et al. 1994); various forms of presumptive treatment (Steen et al. 2000); and treatment based on individual risk assessment (Vuylsteke et al. 1993; Germain et al. 1997; Deceuninck et al. 2000). Improved clinical services for SWs can be expected to have an impact on the women's health and on STI transmission in the community only to the extent that the SWs use those services. In key informant interviews with SWs, it became clear that many women were not satisfied with the care offered in the public clinics. Clinic records showed that only a small proportion of active SWs attended the existing STI care facilities. In Antananarivo, the capital of Madagascar, a total of 1612 SWs had been registered at the Isotry public clinic in October 1999 but only 180 SWs had attended this clinic since the beginning of 1999. At the ‘67 Ha clinic’, a non-governmentalfacility in Antananarivo that provides community-based education and clinical services to STI patients and SWs, 837 consultations by SWs were recorded during the first 9 months of 1999. In Tamatave, the city with the largest port, a total of 505 SWs had been registered at the public clinic in November 1999, while in Diego-Suarez, another port city, 1045 prostitutes were registered through FIVMATA, the local SW's association. SWs in these three cities were encouraged to seek STI care once a month even in the absence of any symptoms, but few did; for instance, at the ‘67 Ha clinic’ only 37% of the consultations by SWs were follow-up visits.
The desire to improve STI preventive and curative services for SWs in Madagascar was shared by a group of researchers, clinicians, and SWs, and we planned to include SWs as active collaborators throughout the research process. Our research was conducted in three cities in Madagascar and included three major components: (i) identification of effective and appropriate STI screening and treatment strategies for SWs in Madagascar; (ii) evaluation of the feasibility and acceptability of improving basic STI services for SWs in partnership with the SWs; (iii) development of national STI treatment guidelines for SWs based on the research data. In this paper, we will only present data that were used directly for the development of the national STI treatment guidelines for SWs.
Female SWs were encouraged to present at the non-governmental ‘67 Ha Clinic’ in Antananarivo, and at the public dispensary in Tamatave, regardless of symptoms, through community-based outreach by peer educators. At enrolment, prevalent STIs, bacterial vaginosis and vulvovaginal candidiasis were assessed as well as risk factors and markers hypothesized to be associated with these conditions. Subsequently, incident STIs and associations with hypothesized risk variables were evaluated.
Active female SWs, 16 years of age or older and not pregnant, who presented to the clinic for newly improved, user-centred STI care (through peer educator referral or spontaneously) were invited to participate in the study. SWs who had given informed consent, who had not been treated at the clinic during the last 3 months, who were willing to comply with treatment regimens and willing and able to return for the scheduled follow-up visit were eligible for enrolment. All consenting, eligible SWs presenting consecutively at the clinics were enrolled.
At the initial visit, the SWs were interviewed by a study clinician using a structured, pre-tested questionnaire in Malagasy. A venous blood sample was collected for syphilis screening (Macro Vue RPR Card Test, Becton Dickinson, Cockeysville, MD), confirmed by Serodia-TPPA (Fujirebio, Tokio, Japan) and HIV using EIA (Genetic Systems Peptide EIA HIV1/2, Sanofi, Pasteur, France), confirmed by Western blot (Bio Rad Novapath TM HIV 1 Immunoblot, CA). The women were asked to give no more than 15 ml of initial stream urine. Urine samples were aliquoted and frozen at −20 °C until shipped on dry ice to the University of North Carolina for evaluation of gonococcal and chlamydial infections using ligase chain reaction (Abbott LCx Probe System, Abbott Laboratories, Abbott Park, IL).
Women were clinically examined, and samples were collected from the posterior fornix using a swab that was subsequently immersed in a culture medium for Trichomonas vaginalis (InPouch, BioMed, San Jose, CA). A second swab from the lateral vaginal walls was rolled onto a slide for Gram staining. Amine odour and pH of the vaginal secretions were assessed. Women were counselled about treatment and prevention and treated presumptively with 1 gm of azithromycin and 500 mg of ciprofloxacine, as cervical infections could not be diagnosed effectively locally. Additional treatment for trichomoniasis, bacterial vaginosis, or reactive syphilis serology was given 1 week later based on local laboratory results.
The follow-up visit, 2 months later, was similar to the initial visit; the questionnaire concentrated on symptoms and risk behaviours during the period since the preceding visit. All clinical and laboratory evaluations performed at the first visit were repeated.
The data from the questionnaires, laboratory and clinical record forms were entered twice into a database and cleaned using EpiInfo 6.04. Data were analysed using SAS V8 (SAS Institute, Cary, NC). The prevalence and incidence of bacterial vaginosis, candidiasis, trichomoniasis, cervical gonococcal and chlamydial infections were determined based on laboratory diagnosis. Characteristics hypothesized to be associated with prevalent and incident conditions, respectively, were examined using Cochran-Mantel-Haenszel chi-square test or Fisher's exact test for categorical variables; t-test or analysis of variance for continuous variables. Depending on the probability distributions of the variables the Wilcoxon rank sum test or the Kruskal–Wallis test were used.
Independent variables hypothesized to be associated with the outcome of interest were entered into a logistic regression model regardless of the bivariate analyses (Sun et al. 1996). Variables were entered into the model in groups: behavioural, socio-demographic, and clinical/laboratory. The sequence of covariate removal from the model was determined by likelihood ratio testing to ensure that the covariate that contributed the least to the fit of the model would be removed first. Variables that did not significantly contribute to the fit of the model were evaluated one at the time to determine if they acted as confounders for any other covariates still in the model. Confounding was determined to exist when removal of the potentially confounding variables changed the odds ratio of a risk factor by at least 10%.
Clinical decision models were developed using the factors found to be associated with disease outcome in multivariate analysis. The sensitivity, specificity, and predictive values of the developed decision models were calculated following standard methodologies (Fletcher et al. 1988). The national guidelines were developed during a 3-day workshop in Madagascar by a group of researchers, local policy makers and clinicians, and SW representatives who had worked as peer educators in Antananarivo, Tamatave, and Diego-Suarez. Local research results were reviewed, international experience presented, and theoretical and practical, including economic elements of decision-making discussed. Subsequently, the participants were divided into three working groups. Two working groups developed independently STI treatment and prevention guidelines from the perspective of the care providers and policy makers, and the third working group formulated guidelines from the perspective of the users. The final national guidelines, based on research results, practical considerations, and local experience and perspectives, were developed by consensus in a plenary session.
The study was approved by the Committee on the Protection of the Rights of Human Subjects of the University of North Carolina and by the ethical review board of the Malagasy Ministry of Health.
A total of 986 female SWs (493 in each city) were evaluated at the first visit. The median age of the SWs was 26 years (range 16–57); the median age at which they had started sex work was 21 years (range 11–43). While these characteristics did not differ by city, other features did and we analysed the data for each city separately.
Only one woman in each city was found to have HIV antibodies at the recruitment visit, but 77.5% of the SWs in Antananarivo and 73.5% of the women in Tamatave had at least one curable STI (Table 1). Follow-up data were available for 458 (92.9%) of the SWs in Antananarivo and 481 (97.6%) in Tamatave. Roughly 2 months after treatment and advice on prevention of STIs, 64.9% of the SWs in Antananarivo and 57.4% of the women in Tamatave had at least one curable STI.
Table 1. STIs detected in FSWs in Madagascar at the initial visit and 2 months after presumptive, directly observed therapy
Antananarivo, proportion (%)
Tamatave, proportion (%)
* Significant difference between Antananarivo and Tamatave, P < 0.05.
† RPR and TPPA reactive; at follow-up visit results are presented only for women with non-reactive RPR at first visit.
Gonococcal (GC) infection
Chlamydial (CT) infection
GC or CT infection
GC, CT or TV*
RPR titre 1:8
Any curable STI
Yeast or pseudohyphea*
Factors associated in logistic regression analysis with gonococcal or chlamydial cervical infections detected 2 months after directly observed presumptive therapy are presented by city in Table 2a and b. Using the beta coefficients of all the significant predictors in these two logistic regression models, receiver operator characteristic (ROC) curves were developed (Figure 1a and b).
Table 2. Factors associated with incident cervical infection detected 2 months after presumptive therapy in FSWs in (a) Antananarivo (n = 308) and (b) Tamatave (n = 461)
OR (95% CI)
* Adjusted odds ratio significantly different from 1, P < 0.05.
† Includes confounders of variables significantly associated with incident cervical infection.
‡ Confounders of variables significantly associated with incident cervical infection. Variables omitted from the final model because of small numbers: recruited client from ports in the past month and prompted report of bleeding between periods.
Age <25 years
1.55 (0.83, 2.90)
At least 10 partners in the past week
4.28 (1.97, 9.29)*
Used condom at last sex act
0.31 (0.11, 0.87)*
Absence of lactobacilli in vaginal flora
2.89 (0.97, 8.62)
5.02 (1.60, 15.68)*
Incident bacterial vaginosis
2.64 (1.33, 5.25)*
Prompted report of abnormal vaginal discharge
2.14 (1.11, 4.11)*
Prompted report of vaginal itching or burning
1.70 (0.77, 3.74)
Prompted report of bleeding between periods
0.52 (0.08, 3.47)
Cervical erosion on exam
0.65 (0.34, 1.25)
Age <25 years
2.35 (1.41, 3.93)*
At least 10 partners in the past week
2.16 (1.05, 4.44)*
Reported condom use at last commercial sex act
4.13 (1.49, 11.44)*
Primary vs. post-primary
2.44 (1.27, 4.71)*
Less than primary vs. post-primary
4.97 (2.39, 10.31)*
Truck driver client in the past month
2.41 (1.27, 4.57)*
Absence of lactobacilli in vaginal flora
7.91 (2.20, 28.40)*
Cervical erosion on exam
2.04 (1.14, 3.65)*
Cervical friability on exam
2.65 (1.47, 4.77)*
Spontaneous complaint of malodorous discharge
0.47 (0.27, 0.83)*
Taxi driver client in the past month
0.73 (0.39, 1.38)
Used condom at last sex act
2.09 (0.37, 11.80)
Prompted report of pain during intercourse
0.38 (0.14, 1.05)
Development of guidelines
The female SWs who participated in the 3-day workshop presented their perspective as representatives of the service users and requested routine serologic screening for syphilis and a speculum exam. Of note, while a consensus existed among the health professionals on the importance of syphilis screening, the value and feasibility of speculum exams had been debated. The SW representatives considered a good medical exam essential to promote patient trust and compliance. The SW representatives reported that the women want to know what condition(s) they are suffering from and they considered good communication between the care provider and the beneficiary indispensable. Presumptive STI therapy was deemed acceptable as long as it would be accompanied by explanations given by the clinician and preferably also by a peer educator. The SW representatives thought that care including treatment should not cost more than 5000 FMG (about 0.77 US$) and that consultations should start at 7:30 a.m. Lastly, more peer educators should be involved in clinic-based STI care and prevention for female SWs. The national guidelines (Table 3) were subsequently developed by consensus based on the following rationale.
Table 3. Summary National Guidelines for STI treatment and prevention in female sex workers in Madagascar
* Treat and counsel for HSV when lesions are recurrent and vesicular.
† Endocervical mucopus or cervical friability or cervical erosion.
1. Serologic screening using RPR; treatment for syphilis if RPR-reactive
2. Speculum exam; treatment for syphilis & chancroid if genital ulcers present*
3. Presumptive treatment of gonococcal and chlamydial infection, and trichomoniasis/bacterial vaginosis (BV)
4. Promotion of condom use, counselling, and invitation for follow-up visit 3 months later
1. Serologic screening using RPR if last result non-reactive or obtained at least 6 months earlier; treatment for syphilis if RPR-reactive
2. Speculum exam; treatment for syphilis and chancroid if genital ulcers present*
3. Risk-based treatment
3.1. Trichomoniasis/BV if malodorous discharge or vaginal pH > 4.5
3.2. Candidiasis if genital itching or burning or woman pregnant
3.3. Gonococcal and chlamydial infection if total risk score at least 3
Risk factor (score)
Age <25 years (1)
At least 10 partners in last week (1)
Vaginal pH > 4.5 (or absence of lactobacilli) (2)
Any sign of cervicitis† or motion tenderness (1)
4. Promotion of condom use, counselling, and invitation for follow-up visit 3 months later
Serologic screening for syphilis using RPR was recommended given the documented high prevalence of syphilis in Madagascar, particularly in SWs; the feasibility of this screening approach if basic resources are made available, and the request of the SW representatives to be screened. Confirmatory treponemal testing was not recommended in this context because of its cost and technical complexity. The service users wanted a speculum exam that allows the clinician to detect genital lesions. Presumptive treatment for gonococcal, chlamydial, and trichomonas infections at this initial visit was preferred over selective, risk-based treatment because of the high prevalence of these STIs. The recommended frequency of the follow-up visits, i.e. every 3 months, was based on the desire of the users, the estimated numbers of SWs in the major cities and the capacity of the existing health care facilities.
The technical working group decided that SWs with non-reactive syphilis screening results should be subsequently screened using RPR every 3 months. SWs with a reactive RPR should be screened again 6 months after treatment. It was judged that most existing health care facilities cannot currently be expected to perform quantitative RPR testing for follow-up. Therefore, and in light of the high prevalence and exposure to syphilis, SWs with a reactive RPR 6 months after adequate treatment should be treated again with two doses of 2.4 MIU benzathine penicillin given at 1-week interval, despite the fact that a sizeable proportion of these women may no longer have infectious syphilis.
The decision to treat a SW for trichomoniasis/bacterial vaginosis in the presence of malodorous vaginal discharge or a vaginal pH greater than 4.5 was based on the overall 95.6% sensitivity (95% CI: 94.1–97.1); the 46.8% specificity (95% CI: 38.9–54.7), and the 88.2% positive predictive value (95% CI: 85.8–90.6) determined during simulation analyses of the empirical research results. The sensitivity of this approach was 91.5% in Antananarivo (95% CI: 87.9–95.1) and 98.0% in Tamatave (95% CI: 98.0–98.0); the specificity was 57.8% in Antananarivo (95% CI: 47.2–68.4) and 33.8% in Tamatave (95% CI: 22.8–44.8); the positive predictive value was 86.1% in Antananarivo (95% CI: 81.8–90.4) and 89.4% in Tamatave (95% CI: 86.5–92.3).
The findings of the multivariate analyses (Table 2) served as the basis to select the algorithm for management of lower genital tract infections due to Neisseria gonorrhoea or Chlamydia trachomatis at follow-up visits during the workshop. During the decision-making process, the stability of the risk factors for cervical infections with regard to place, time, and implementer, the acceptability by stakeholders and environmental constraints as well as the cost and the effectiveness of various options were discussed and made as explicit as possible (Behets et al. 2001a). Age <25 years, at least 10 sex partners during the last week, vaginal pH >4.5 (or absence of lactobacilli in vaginal smear), and clinical signs of cervicitis or motion tenderness were selected as risk factors for the algorithm (Table 3). To avoid the need for systematic microscopic examinations and allow the use of a less costly and less demanding screening test, vaginal pH >4.5 was substituted for absence of lactobacilli.
The ROC curves obtained for the selected algorithm are presented in Figure 1a and b. Using a cut-off value of 3, the sensitivity of the selected algorithm was 83.7% overall (95% CI: 81.2–86.2), 82.6% in Antananarivo (95% CI: 78.7–86.5), and 84.7% in Tamatave (95% CI: 81.4–88.0). The specificity of treating cervical infection in the presence of a risk score of at least 3 was 36.6% overall (95% CI: 33.3–39.9), 30.3% in Antananarivo (95% CI: 25.6–35.0), and 41.2% in Tamatave (95% CI: 36.7–45.7). The positive predictive value of this algorithm was 27.3% overall (95% CI: 24.3–30.3), 26.9% in Antananarivo (95% CI: 22.3–31.5), and 27.6% in Tamatave (95% CI: 23.6–31.6).
The importance of providing adequate STI care and prevention interventions to SWs in Madagascar was demonstrated by the documented high burdens of disease and the high exposures to re-infection. While curable STIs were an important public health problem in these women, only two persons were infected with HIV, suggesting that Madagascar is still in a good position to avoid a disastrous HIV epidemic.
There were some differences between the two cities, illustrating the need to conduct this type of research on STIs in various settings within one country. We found differences in STIs and in associated risk factors between the women evaluated in the capital and the women in the port city. Cervical infection was associated with lower educational levels in Tamatave but not in Antananarivo. Interestingly, reported condom use for the last sex act was associated with a lower risk for cervical infection in the capital and with a higher risk in Tamatave. While this finding could reflect reporting issues, it could also be due to different transmission dynamics as suggested by the fact that sex with a truck driver in the past month was an independent risk factor for incident cervical infection in Tamatave but not in Antananarivo. Most women in this study reported that the majority of their recent sex acts were not protected by condom use, thus condom use at last sex act can be expected to be a poor predictor.
Despite the differences in risk factors for cervical infections by city, a single STI care strategy for SWs in Madagascar was developed that was acceptable to the participating health professionals, and to the SWs who represented the service users. We opted for presumptive treatment of gonococcal and chlamydial infections and trichomoniasis the first time a SW seeks care for STIs while for the 3 monthly follow-up visits these infections are treated based on individual risk assessments. Three monthly repeat visits with treatment based on the results of a clinical examination and on an assessment of individual risk factors were preferred to periodic presumptive treatments to avoid negative psychological reactions when SWs repeatedly receive the same treatment. It was felt that SWs might become uncertain of the quality of the STI services and less open to educational messages when clinicians prescribe the same drugs. Similarly, clinicians may become bored and less efficient in counselling to prevent STIs. Interestingly, economic analyses did not show a convincing benefit of periodic presumptive treatment in terms of reducing the STI prevalence (data not shown).
Clinical algorithms developed for female SWs in other countries do not distinguish first from follow-up visits (Vuylsteke et al. 1993; Germain et al. 1997; Deceuninck et al. 2000). Simple laboratory tests such as the detection of intra-cellular Gram-negative diplococci in cervical smear (Deceuninck et al. 2000) or >10 leucocytes/field in vaginal smears (Germain et al. 1997) have been included elsewhere in algorithms for SWs. The sensitivity of these algorithms ranged from 58% to 71% with specificities from 56% to 80%. We selected an algorithm for cervical infections that does not require laboratory testing but that includes vaginal pH in the risk assessment. This can be measured immediately by the clinician and is inexpensive. In our experience, even simple microscopy can be a strain on human and logistical resources in settings with few resources and the quality is usually highly dependent on location or staff and may vary over time. We selected a few risk factors for clinical management that were judged most practical and stable across geographic locations, health care settings, and over time. As shown by the ROC curves, even the best models obtained by logistic regression performed modestly and did not differ much from the selected algorithm for management of cervical infection. Our algorithm, like all clinical algorithms based on risk assessment, will need periodic re-evaluations since STI prevalence and associated risk factors will evolve, particularly in settings with active prevention programmes.
To develop these national STI management guidelines for female SWs, we followed a decision-making process that was similar to the one we used in 1997 when the national STI treatment algorithms for symptomatic patients were developed. While researchers, public health workers as well as public and private sector clinicians were involved in the decision-making process in 1997, patient representatives were not. The inclusion of peer educators in 2001 offered opportunities to better understand and integrate user perspectives that might ultimately improve acceptability to and use of STI services by SWs.
We believed that participatory action research (Minkler 2000) was necessary to improve STI care and prevention for SWs in Madagascar. Actively involving SWs throughout the research was a learning process for SWs and researchers. Including laypersons with little formal education in decision-making entailed challenges and risks due to limited or inadequate understanding of technical aspects and concepts. During the course of the field research, the SW representatives had acquired basic notions of STI care and prevention. We made efforts to explain in plain language concepts such as antimicrobial resistance, predictive values, and misclassification costs. However, the SW representatives could not be expected to have the same grasp of the technical issues as the health professionals. We debated the pros and cons of including SW representatives in each of the three working groups at the decision-making workshop vs. creating one separate working group constituted of peer educators only. We opted for the latter to enable a more independent voice of the users despite the risk of discrepant and possibly irreconcilable opinions and demands. Fortunately, the conclusions presented by the three working groups at the plenary session showed few and manageable differences and a consensus was reached.
The research project was conceived with an explicit desire to promote empowerment of SWs and their organizations. The active engagement of the peer educators in the research and in the decision-making workshop was a positive experience; the peer educators enthusiastically endorsed the opportunity to contribute actively to decision-making and health professionals had the opportunity to understand more effectively the experiences, outlooks, concerns, and needs of the SWs. However, empowerment of SWs in this project was confined to a predominantly biomedical intervention and can only be considered as one step towards better life conditions and improved economic options for these women.
We are grateful for critical support from Wendy Githens Benazerga.
Authors Dr Jocelyne Andriamiadana, USAID, Antananarivo, Madagascar. E-mail: firstname.lastname@example.org Dr Frieda M.-T. F. Behets, School of Public Health, University of North Carolina at Chapel Hill, 2102A McGavran-Greenberg Hall, CB 7435, Chapel Hill, NC 27599-7400, USA. Tel.: +1 919 966 7440; Fax: +1 919 966 2089; E-mail: email@example.com (corresponding author). Dr Gina Dallabetta, Family Health International, Research Triangle Park, NC, USA. E-mail: firstname.lastname@example.org Kristi McClamroch, Department of Epidemiology, School of Public Health, University of North Carolina at Chapel Hill, 2102A McGavran-Greenberg Hall, CB 7435, Chapel Hill, NC 27599-7400, USA. Dr Adeline Ranaivo, Ministry of Health, Antananarivo, Madagascar. Dr Justin Ranjalahy Rasolofomanana, Institut National de santé dublique et communautaire, 74 Rue du 26 Suin 1960, Antananarivo, Madagascar. E-mail: email@example.com Dr Désiré Rasamilalao, Ministry of Health, Antananarivo, Madagascar. Dr Johannes van Dam, Population Council, Washington, DC, USA. E-mail: firstname.lastname@example.org Dr Kathleen Van Damme, Family Health International, Research Triangle Park, NC, USA. E-mail: email@example.com Georgine Vaovola, FIVMATA, Diego-Suarez, Madagascar. Dr Andry Rasamindkakotroka, E-mail: firstname.lastname@example.org