Risk factors for treatment denial and loss to follow-up in an antiretroviral treatment cohort in Kenya
Corresponding Author Heiko Karcher, Institute of Tropical Medicine and International Health, Spandauer Damm 130, 14050 Berlin, Germany. Tel.: +49 30 30116720; Fax: +49 30 30116710; E-mail: firstname.lastname@example.org
Objectives: To evaluate risk factors for treatment denial and loss to follow-up in an antiretroviral treatment (ART) cohort in a rural African setting in western Kenya.
Method: Sociodemographic and clinical data of patients enrolled in an ART cohort were collected within 18 months of an observational longitudinal study and analysed by logistic and Cox regression models.
Results: Of 159 patients with treatment indication 35 (22%) never started ART. Pregnancy [adjusted odds ratio (AOR) 3.60, 95% confidence interval (CI) 1.10–11.8; P = 0.035] and lower level of education (AOR 3.80, 95% CI 1.14–12.7; P = 0.03) were independently associated with treatment denial. The incidence of total loss of patients under therapy was 43.2 per 100 person years (pys) (mortality rate 19.2 per 100 pys plus drop out rate 24 per 100 pys). Older age [adjusted hazard ratio (AHR) 1.06, 95% CI 1.01–1.12; P = 0.04], AIDS before starting treatment (AHR 5.83, 95% CI 1.15–29.5; P = 0.03) and incomplete adherence to treatment (AHR 1.05, 95% CI 1.03–1.07; P < 0.001) were independent risk factors for death. Incomplete adherence also independently predicted drop out because of other reasons (AHR 1.06, 95% CI 1.04–1.09; P < 0.001).
Conclusion: Pregnancy and lower level of education, higher age, advanced AIDS stage and impaired compliance to ART were identified as risk factors for treatment denial and death, respectively. Adequate counselling strategies for patients with these characteristics could help to improve adherence and outcome of treatment programmes in resource-limited settings.
Objectifs Evaluer les facteurs de risque associés au refus du traitement et à la perte au suivi dans une cohorte de traitement antirétroviral (ARV) dans une région rurale africaine dans l'ouest du Ke nya.
Méthode Des données sociodémographiques et cliniques de patients enrôlés dans une cohorte de traitement ARV ont été collectées au cours de 18 mois d'une étude d'observation longitudinale et analysées par le modèle de la régression logistique et celui de Cox.
Résultats Sur 159 patients avec indication pour traitement ARV, 35 (22%) n'ont jamais commencé le traitement. La grossesse (rapport de cotes ajusté AOR = 3,60; IC95%: 1,10–11,8; p = 0,035) et un niveau d’éducation plus faible (AOR = 3,80; IC95%: 1,14–12,7; p = 0,03) étaient indépendamment associés au refus du traitement. L'incidence totale de la perte de patients sous traitement était de 43,2 par 100 personnes/année (taux de mortalité: 19,2 par 100 personnes/année et taux des abandons: 24 par 100 personnes/année). Un âge plus élevé (rapport du hasard ajusté: AHR = 1,06; IC95%: 1,01–1,12; p = 0,04), développement du SIDA avant commencement du traitement (AHR = 5,83; IC95%: 1,15–29,5; p = 0,03) et adhérence incomplète au traitement (AHR = 1,05; IC95%: 1,03–1,07; p < 0,001) étaient des facteurs de risque indépendants pour la mort. L'adhérence incomplète prévoyait aussi indépendamment l'abandon du traitement dûà d'autres raisons (AHR = 1,06; IC95%: 1,04–1,09; p < 0,001).
Conclusion La grossesse et un niveau d’éducation plus faible, un âge plus élevé, l’état avancé du SIDA et la compliance altérée au traitement ARV ont été identifiés comme des facteurs de risque pour le refus du traitement et la mort. Des stratégies de conseil pour les patients présentant ces caractéristiques pourraient aider à améliorer l'adhérence et les résultats des programmes de traitement dans les régions à ressource s limitées.
Objetivos Evaluar los factores de riesgo para la no aceptación de tratamiento y la pérdida de seguimiento en una cohorte con tratamiento antirretroviral (TAR) en un área rural Africana, al oeste de Kenia.
Método Se recolectaron datos sociodemográficos y clínicos de pacientes que participaban en una cohorte de TAR durante los 18 meses de un estudio longitudinal observacional, y fueron analizados mediante métodos de regresión logística y de Cox.
Resultados De 159 pacientes con indicación de tratamiento, 35 (22%) nunca comenzaron TAR. El embarazo (Razón de posibilidades ajustada, RPA, 3.60, 95% IC 1.10–11.8; p = 0.035) y un menor nivel de educación (RPA 3.80, 95% IC 1.14–12.7; p = 0.03) estaban asociados de forma independiente con el rechazo del tratamiento. La incidencia de la pérdida total de pacientes bajo terapia fue de 43.2 por 100 persona años (tasa de mortalidad 19.2 por 100 persona años más una tasa de abandono de 24 por 100 persona años). Una mayor edad (Razón de riesgo ajustada, RRA, 1.06, 95% IC 1.01–1.12; p = 0.04), el tener SIDA antes de comenzar el tratamiento (RRA 5.83, 95% IC 1.15–29.5; p = 0.03) y una adherencia incompleta al tratamiento (RRA 1.05, 95% IC 1.03–1.07; p < 0.001) fueron factores de riesgo independientes para mortalidad. Una adherencia incompleta también predecía de forma independiente el abandono del estudio por otras razones (RRA 1.06, 95% IC 1.04–1.09; p < 0.001).
Conclusión El embarazo y un menor nivel de educación, una mayor edad, un estado avanzado del SIDA y una no adherencia al TAR fueron identificados como factores de riesgo para rehusar el tratamiento y la muerte, respectivamente. Una estrategia adecuada de aconsejamiento para pacientes con estas características podría ayudar a mejorar la adherencia y los resultados de programas de tratamiento en lugares con recursos limitados.
Antiretroviral therapy (ART) is becoming increasingly implemented in countries of sub-Saharan Africa. Several reports demonstrate generally good adherence and treatment outcome comparable with studies from developed countries (Orrell et al. 2003; Akileswaran et al. 2005; Ivers et al. 2005; Laurent et al. 2005; Wools-Kaloustian et al. 2006). Most of the published data, however, arises from supervised cohorts in urban centers, whereas such evidence from treatment programmes integrated into routine healthcare in rural areas is still scarce (Stevens et al. 2004; Ivers et al. 2005; Van Oosterhout et al. 2005). Most studies report mainly clinical, immunological and virological data, but little is known about factors influencing loss to follow-up. As it is crucial for the success of treatment that patients stay on therapy it is important to analyse factors increasing or impeding adherence.
To this end, we analysed sociodemographic and clinical variables with regard to their possible influence on treatment denial and loss to follow-up in an ART programme which was an integral part of the routine health service of a District Hospital in Kenya.
Study design, study population and study site
The study was conducted at Migori District Hospital, Nyanza Province, Kenya between April 2004 and September 2005. Patients received ART in the outpatient clinic of the hospital. ART was a component of a PMTCT (prevention of mother-to-child transmission of HIV) Programme which was started 2 years earlier in April 2002 in Migori District as a cooperation between the Kenyan Ministry of Health and GTZ (German Agency for Technical Co-operation and Development). Eligible for ART were women who participated in the PMTCT Programme, their partners and children. Both antiretroviral and supportive drugs were free of charge as was laboratory monitoring.
Clinical procedures and adherence measurement
The standard regimen for antiretroviral long-term therapy was a generic fixed dose combination of stavudine, lamivudine and nevirapine (TriomuneTM, CIPLA, Mumbai, India). Treatment indication followed international and national guidelines (Ministry of Health 2002; WHO 2005a) and was considered if patients were in CDC (Centers for Disease Control and Prevention) stage C or had a CD4 cell count twice below 350 cells/μl. Pregnant women were screened as early as possible during pregnancy and, in case of a treatment indication, long-term therapy was started before delivery in order to minimize the risk of vertical HIV transmission. Patients were counselled thoroughly regarding benefits of therapy, possible side effects and the importance of adherence. All patients who met the criteria for treatment indication and who had received counselling were enrolled into the treatment programme. Appointments for clinical and laboratory investigations were scheduled at baseline and at months 0.5, 1, 2, 4, 6, 9, 12, 15 and 18 after start of therapy. Clinical examination comprised staging of HIV infection and screening for concurrent diseases. Laboratory investigations consisted of complete blood counts and alanine aminotransferase; CD4 counts were evaluated at baseline and every six months thereafter.
Adherence to ART was assessed by patient self-reports, visual analogue scale (VAS) and pill counts at every visit. A missed visit was defined as 0% adherence. Patients were defined as lost to follow-up if they had not shown up within 4 months after a scheduled appointment. These patients were then traced once in order to assess the underlying cause. Accordingly, the classification loss to follow-up was divided into death and drop out because of other reasons. Treatment denial was defined as patients being enrolled into the treatment programme but never starting antiretroviral therapy.
Statistical analysis was performed using SPSS version 12.0 (SPSS Inc., Chicago, IL, USA). Medians were compared using the Mann-Whitney-U-test for independent samples. The Chi-squared test was used to compare categorical variables. Incidence for drop out as well as mortality rates were calculated. Kaplan-Meier estimates were used to analyse the cumulative probability of surviving and the probability of stopping treatment. Patients who dropped out were censored on their last documented visit and patients who died on the date of death.
The following sociodemographic and clinical data were collected at baseline and used for regression analysis: age, sex, occupation, level of education, ethnic group, confession and travel distance to the hospital; CD4 cell count, HIV stage, and pregnancy status. Adherence was calculated for all three methods at each visit and averaged over time (cumulative mean adherence). Adherence was calculated using the intention-to-treat analysis: patients who dropped out or died were included in the analysis and adherence was defined as 0% after the date of loss to follow-up. Missing data at one clinical visit were calculated by using the mean value of both the preceding and subsequent visits. As most of the deaths and drop outs occurred during the first 6 months of treatment, the mean cumulative adherence at the beginning of therapy (after 2 months) was chosen as a covariate for the regression analysis.
The association of baseline variables with treatment denial was evaluated in univariate analysis; variables significant at the level of P < 0.1 were included in an analysis of logistic regression. Cox proportional hazard regression was used to analyse unadjusted and adjusted hazard ratios for death and drop out; variables significant in univariate analysis (P < 0.1) were included in the multivariate model. The multivariate analysis was run by forward and backward selection. For risk factors identified in multivariate analysis, the attributable risk (AR) and the population attributable risk (PAR) of the study population were indicated.
This study was conducted according to the principles of the Declaration of Helsinki and approved as part of the evaluation protocol of the ART programme by the national and regional health authorities in Kenya.
Thirty-five (22%) of 159 patients failed to begin ART after programme enrollment. Fifty-three of these 159 patients were pregnant women, 20 (38%) of whom denied therapy. Four patients (11%) had died and one patient had a severe psychiatric disorder resulting in treatment denial. Death was because of cryptococcal meningitis in two cases and Pneumocystis jiroveci pneumonia and bacterial pneumonia in one case each.
The association of baseline variables with treatment denial is shown in Table 1. In univariate analysis, variables associated with treatment denial were age <30 years (P = 0.005), lower level of education (P = 0.007), no AIDS diagnosis prior to therapy (P = 0.006) and pregnancy (P = 0.001). In multivariate analysis, variables remaining independently associated with treatment denial were lower level of education (P = 0.03) and pregnancy (P = 0.035). The AR for lower education was 0.73 and the PAR was 0.60. The AR and PAR for pregnancy were 0.72 and 0.54, respectively.
Table 1. Logistic regression for factors associated with treatment denial
|Age||150|| || || || || || || |
| ≤30 years||83||29||3.50||1.40–8.70||0.005*||1.10||0.27–4.50||0.9|
| >30 years||67||10||1.00|| || ||1.00|| || |
|Sex||158|| || || || || || || |
| Female||116||24||1.90||0.73–5.00||0.18|| || || |
| Male||42||14||1.00|| || || || || |
|Occupation||154|| || || || || || || |
| Yes||78||21||0.97||0.44–2.10||0.93|| || || |
| No||76||21||1.00|| || || || || |
|Education||152|| || || || || || || |
| Secondary||72||11||1.00|| || ||1.00|| || |
|Ethnic group||131|| || || || || || || |
| Other||111||30||1.64||0.57–4.73||0.36|| || || |
| Luo||20||21||1.00|| || || || || |
|Confession||152|| || || || || || || |
| Protestant||119||21||1.19||0.45–3.21||0.72|| || || |
| Catholic||33||18||1.00|| || || || || |
|Travel distance||153|| || || || || || || |
| >30 km||54||18||0.78||0.34–1.79||0.56|| || || |
| ≤30 km||99||22||1.00|| || || || || |
|CD4 count||158|| || || || || || || |
| >100||125||25||2.40||0.78–7.34||0.12|| || || |
| ≤100||33||12||1.00|| || || || || |
|HIV stage||159|| || || || || || || |
| No AIDS||95||30||3.40||1.38–8.37||0.006*||1.72||0.47–6.26||0.41|
| AIDS||64||11||1.00|| || ||1.00|| || |
|Pregnancy||115|| || || || || || || |
| No||62||11||1.00|| || ||1.00|| || |
Baseline characteristics of the treatment cohort
One hundred and twenty-four patients started antiretroviral long-term therapy. Their median was 31 years (range 17–58); 59 (48%) were younger than 30 years. Eighty-eight (71%) of the patients were females; 60 (48%) were without an income-generating occupation, 57 (46%) had primary education (grades 1–4) and 64 (52%) had secondary education (grades 5–7); 88 (71%) were of Luo ethnicity and 94 (76%) of Protestant confession; 45 (36%) of the patients lived more than 30 km away from the hospital. Before starting treatment, the median CD4 cell count was 189/μl (range 15–536/μl); 57 (46%) had AIDS (CDC stage C).
Women were younger than men (median 29 and 35 years, respectively, P = 0.0001), were more often without an income-generating occupation (72% and 42%, respectively, P = 0.002) and fewer women had AIDS (36% and 69%, respectively, P = 0.001). Thirty of 88 (34%) women were pregnant. Pregnant women had higher CD4 cell counts than non-pregnant women (94% and 71%, respectively, >100/μl, P = 0.01) and fewer had AIDS (82% and 53%, respectively, P = 0.006).
The median treatment duration of the 124 patients was 9 months [interquartile range (IQR) 4–12 months]; 97% of those patients who did not drop out were under treatment for at least 2 months.
Adherence and loss to follow-up
The cumulative mean adherence after 2 months of therapy was 85%, indicating that the patients took on average 85% of their pills. Sixty-nine percent of the patients had an adherence of more than 95%. The cumulative mean adherence after 6 months of therapy was 79%; 63% had an adherence of more than 95%.
During the observation period, 34 (27%) patients were lost to follow-up. Of those, 15 (12%) were known to have died and 19 patients (15%) dropped out because of other reasons: four refused treatment without further explanation, four moved away, one suffered from a psychosis, one patient had to hide because of domestic violence, one lost her child and husband and stopped treatment and eight patients could not be traced because they had given a wrong address. Four patients died of tuberculosis, two of cryptococcal meningitis, two of P. jiroveci pneumonia and one patient each died of Kaposi sarcoma, gastroenteritis and wasting syndrome. Three patients died of unknown causes.
The incidence rate of total loss to follow-up was 43.2 per 100 person years (pys), the mortality rate was 19.2 per 100 pys and the incidence rate of drop out for other reasons was 24 per 100 pys. The cumulative probability of remaining alive after 12 months of therapy was 85.4%, while the cumulative probability to stop treatment because of reasons other than death was 18%. Death occurred at 2 months (median; IQR 1–3 months) after treatment initiation as did drop outs because of other reasons (IQR 1–5 months).
A detailed analysis of factors associated with death and drop outs because of other reasons is shown in Table 2. In univariate analysis, variables associated with death were higher age (P = 0.008), lower CD4 cell count (P = 0.009) and AIDS at the beginning of treatment (P = 0.008). In addition, lower mean cumulative adherence after 2 months of treatment was associated with death (P < 0.001). In multvariable analysis, older age (P = 0.04), AIDS at the beginning of treatment (P = 0.03) and lower adherence after 2 months of treatment (P < 0.001) remained independently associated with death. The AR and PAR for older age were 0.64 and 0.47, for AIDS at the beginning of treatment 0.85 and 0.64 and for lower adherence 0.87 and 0.76, respectively.
Table 2. Cox regression (proportional hazard analysis) of factors associated with drop out and death during antiretroviral therapy
|Age||119||0.88||0.80–0.96||0.002**|| || || ||1.07||1.02–1.13||0.008**||1.06||1.01–1.12||0.04|
|Sex||124|| || || || || || || || || || || || |
| Female||88||3.77||0.87–16.3||0.08**|| || || ||0.87||0.29–2.55||0.83|| || || |
| Male||36||1.00|| || || || || ||1.00|| || || || || |
|Occupation||122|| || || || || || || || || || || || |
| No||60||1.91||0.75–4.86||0.17|| || || ||2.20||0.79–6.40||0.15|| || || |
| Yes||62||1.00|| || || || || ||1.00|| || || || || |
|Education||121|| || || || || || || || || || || || |
| Primary||57||2.34||0.90–6.20||0.08**|| || || ||0.62||0.21–1.84||0.39|| || || |
| Secondary||64||1.00|| || || || || ||1.00|| || || || || |
|Ethnic group||102|| || || || || || || || || || || || |
| Other||14||1.01||0.20–4.50||0.99|| || || ||1.19||0.26–5.40||0.82|| || || |
| Luo||88||1.00|| || || || || ||1.00|| || || || || |
|Confession||121|| || || || || || || || || || || || |
| Protestant||94||1.01||0.33–3.07||0.98|| || || ||1.86||0.42–8.23||0.41|| || || |
| Catholic||27||1.00|| || || || || ||1.00|| || || || || |
|Travel distance||122|| || || || || || || || || || || || |
| ≤30 km||77||2.46||0.82–7.40||0.11|| || || ||2.59||0.73–9.20||0.14|| || || |
| >30 km||45||1.00|| || || || || ||1.00|| || || || || |
|CD4 count||123||1.03||0.99–1.01||0.17|| || || ||0.99||0.98–0.99||0.009**|| || || |
|HIV stage||124|| || || || || || || || || || || || |
| No AIDS||67||2.37||0.85–6.58||0.09**|| || || ||1.00|| || ||1.00|| || |
| AIDS||57||1.00|| || || || || ||7.50||1.70–33.1||0.008**||5.83||1.15–29.5||0.03|
|Pregnancy||88|| || || || || || || || || || || || |
| Yes||30||2.03||0.78–5.26||0.15|| || || || || || || || || |
| No||58||1.00|| || || || || || || || || || || |
Younger age (P = 0.002) and lower mean cumulative adherence after 2 months of treatment (P < 0.001) were identified as risk factors for drop outs because of other reasons in univariate analysis. In multivariable analysis, only lower adherence remained independently associated with drop out (P < 0.001).
In this study, we elaborated risk factors predicting treatment denial and loss to follow-up in an ART cohort in a rural, resource-limited African setting. Twenty-two percent of the patients did not start ART even though they had a treatment indication and had undergone thorough counselling about the benefits and implications of therapy. Pregnancy and primary education (compared with secondary education) in women were identified as independent risk factors for not starting therapy. Possibly, pre-treatment counselling was inadequate for some patients. The plenitude and complexity of information given to patients on the course of HIV-infection, treatment goals, risks and benefits of therapy, side effects, resistance formation and adherence issues may have overtaxed patients, especially those with lower levels of education. Pregnant women may have been worried about potential risks of antiretroviral drugs for the unborn child and therefore been reluctant to start treatment. Of note, this ART programme was a component of a PMTCT Programme. Pregnant women were the target group and emphasis in counselling was directed towards risks and benefits of ART during pregnancy, with special regard to reduction of mother-to-child transmission of HIV. In case pregnant women refused to start ART during pregnancy, they received a single dose of nevirapine for prophylaxis of intrapartum transmission, an intervention which is less effective in lowering the risk of HIV transmission than a triple drug combination during pregnancy (WHO 2004, 2005b).
Reasons for treatment denial were not available for most patients. Eleven percent were reported to have died before treatment initiation. It is therefore possible that death as a cause for treatment denial was underreported. It is, however, unlikely that more deaths occurred than reported as neither a low CD4 count nor AIDS at baseline were significant risk factors for treatment denial. On the contrary, in univariate analysis patients with AIDS at baseline were more likely to start treatment than patients without AIDS.
In Uganda, 18% of the patients and in Ivory Coast 48% of the patients were reported to have refused to start ART (Djomand et al. 2003; Kabugo et al. 2005). Reasons were not evaluated, but the fact that patients had to pay for treatment might have played an important role. Evaluations of prevention of mother-to-child transmission programmes in Zambia, Ivory Coast, Uganda and Tanzania showed, however, that a lower level of education in pregnant women was associated with non-intake of antiretroviral drugs (Stringer et al. 2003; Ekouevi et al. 2004; Karcher et al. 2006).
In our study, older age, AIDS at the beginning of therapy and incomplete adherence after 2 months of therapy were independently associated with increased mortality. A direct influence of incomplete adherence on mortality has not yet been examined in other African treatment cohorts but was reported in studies from Spain and Canada (Garcia de Olalla et al. 2002; Hogg et al. 2002; Wood et al. 2003). Usually, incomplete adherence is associated with an unsuppressed viral load (Ledergerber et al. 1999; McNabb et al. 2001; Lanièce et al. 2003; Orrell et al. 2003; Nachega et al. 2004) and failure to suppress viral load during ART was predictive of higher mortality (Lohse et al. 2006). Reasons for older patients to have a higher risk to die during therapy may be manifold: CD4 cell restoration under therapy may be lower in older patients because of an impaired thymic function, as observed in studies from Europe and the United States (Yamashita et al. 2000; Viard et al. 2001; Grabar et al. 2004). Non-HIV-related conditions may be more likely to develop in older patients: in our study, the four patients who presumably died of non-HIV-related causes (gastroenteritis and unknown reasons) were all older than 30. As expected, AIDS at the beginning of the treatment was an independent risk factor for mortality in this study (Egger et al. 2002; Garcia de Olalla et al. 2002; Djomand et al. 2003; WHO 2005a).
Incomplete adherence within the first 2 months of therapy was independently associated with drop out in this study. The first months of therapy are known to be the difficult period for the patients as most of the short-term side effects occur during that time. In addition, patients have to attune to the daily drug intake, possibly secretly in case the family was not informed. Thorough adherence counselling and evaluation of adherence, especially during the first months of therapy, are therefore extremely important particularly for patients who feel uncomfortable with the new treatment. But adherence counselling and evaluation are time-consuming and require trust and sensitivity between patient and caregiver, which is not always feasible in settings where the demand for therapy far exceeds available resources. An association of incomplete adherence and subsequent treatment discontinuation was also demonstrated in studies from Europe and the United States (Mocroft et al. 2001; Li et al. 2005; Yuan et al. 2006).
In our setting, the PAR for disease progression or mortality was highest for patients with low adherence and for those with AIDS at the start of the treatment. Therefore, efforts to increase adherence and to treat patients early, before the onset of AIDS, would be most beneficial.
There are some possible limitations to this study: a misclassification bias may have occurred as eight (42%) patients who dropped out of treatment could not be traced and may have died. Furthermore, defining adherence after drop out or death to be 0% (intention-to-treat analysis) may have resulted in incomplete adherence to be an independent risk factor for these events. However, the analysis was re-run with the last measured adherence carried forward after drop out or death (instead of applying 0%) and incomplete adherence remained an independent risk factor for both events.
As a further limitation, this study was not a randomized trial and the sample size was relatively small. Consequently, additional risk factors might not have been identified. We believe, however, that the observational study design was more adequate to examine treatment denial and loss to follow-up in a resource-limited setting treatment cohort as it better reflects a real life situation than a clinical trial.
This study identified pregnant women and patients with lower level of education to be more likely not to start ART, and older patients, patients with AIDS before starting therapy or with incomplete adherence to be at higher risk for stopping ART. Special attention in ART counselling of these patient groups could further increase adherence and improve the outcome of treatment programmes. Many African HIV-infected pregnant women will be counselled and treated outside specialized PMTCT programmes with ART components. Antenatal care services may feel unable to manage pregnant women on ART and overcrowded ART clinics may be unable to provide the special care and attention which pregnant women on ART need. Our findings underscore the need for harmonization of PMTCT and ART services in the process of scaling up of HIV treatment in Africa because these services are currently not linked in a satisfactory manner.
We are indebted to our colleagues and all staff of the MoH/GTZ PMTCT Programme in Migori District in Kenya. Special thanks go to the nurses Melisa Atieno Ikawa and Selestine Odanga, to the clinical officer Daniel Ochieng and to the laboratory technician Charles Opondo. The project was financially supported by the German Ministry for Economic Cooperation and Development through the project Prevention of Mother-to-Child Transmission of HIV (PN 2001.2029.5).