HIV-1 and HIV-2 prevalence, risk factors and birth outcomes among pregnant women in Bissau, Guinea-Bissau: a retrospective cross-sectional hospital study

The human immunodeficiency virus (HIV) remains a leading cause of maternal morbidity and mortality in Sub-Saharan Africa. Prevention of mother-to-child transmission (PMTCT) has proven an effective strategy to end paediatric infections and ensure HIV-infected mothers access treatment. Based on cross-sectional data collected from June 2008 to May 2013, we assessed changes in HIV prevalence, risk factors for HIV, provision of PMTCT antiretroviral treatment (ART), and the association between HIV infection, birth outcomes and maternal characteristics at the Simão Mendes National Hospital, Guinea-Bissau’s largest maternity ward. Among 24,107 women, the HIV prevalence was 3.3% for HIV-1, 0.8% for HIV-2 and 0.9% for HIV-1/2. A significant decline in HIV-1, HIV-2, and HIV-1/2 prevalence was observed over time. HIV infection was associated with age and ethnicity. A total of 85% of HIV-infected women received ART as part of PMTCT, yet overall treatment coverage during labour and delivery declined significantly for both mothers and infants. Twenty-two percent of infants did not receive treatment, and 67% of HIV-2-infected mothers and 77% of their infants received ineffective non-nucleoside reverse transcriptase inhibitors for PMTCT. Maternal HIV was associated with low birth weight but not stillbirth. Inadequate continuity of care and ART coverage present challenges to optimal PMTCT in Guinea-Bissau.


Methods
Study setting. The study was conducted by the Bandim Health Project (BHP) (https ://www.bandi m.org) at the Simão Mendes National Hospital (HNSM) maternity ward located in Bissau. The BHP, a health and demographic surveillance site present in Guinea-Bissau for 4 decades, routinely collects demographic and clinical data on all deliveries at HSNM. This public facility is the principal provider of comprehensive obstetric care in Guinea-Bissau, and approximately 90% of women who deliver at the facility are residents of the country's capital, Bissau.
From June 2008 onwards, opt-out HIV counselling and testing has been offered to women admitted to the maternity ward for the management of labour-and delivery-and/or pregnancy-related complications when HIV tests were available. HIV testing and ART were free of charge and provided by the Guinean Ministry of Health, but all women paid a flat fee of 2000 XOF (3)(4) to give birth at HNSM 14 . All newly diagnosed women and their infants were referred to the Italian-Guinean NGO-clinic, Ceu e Terras, or to other local HIV centres for follow-up, counselling, and further PMTCT ART 13 . Data registration of testing for HIV at delivery ceased in May 2013 due to funding constraints. While HIV testing at labour and delivery was national policy at the time of the study, not all women were tested. We have previously described the characteristics of women tested compared with those not tested 14 . Study design and participants. We conducted a retrospective cross-sectional survey exploring HIV prevalence, risk factors for HIV, treatment provision, and birth outcomes (low birth weight, LBW, and stillbirth), drawing on data routinely collected through the BHP surveillance system at HNSM from June 2008 until May 2013. All women presenting to HNSM for delivery or immediate postpartum care who had been tested for HIV were included in this study.
Data collection. Data were recorded in the HNSM maternity ward registration system and included basic socio-demographic and clinical data. Data cleaning was performed daily (including weekends) by trained research assistants from BHP who also collected supplementary demographic and clinical information using separate case report forms (CRFs). We have previously described the data collected in detail 14,16 . During the study period, it was national policy that hospital midwives offer immediate HIV counselling and testing to all women presenting to the maternity for delivery or, as many women had not attended ANC or did not receive Scientific RepoRtS | (2020) 10:12174 | https://doi.org/10.1038/s41598-020-68806-5 www.nature.com/scientificreports/ PMTCT testing or counselling during pregnancy. Midwives had been trained to complete a short CRF as part of the counselling and testing routine to collect data on previous HIV testing, known sero-status, and ART use 14 . HIV testing. HIV screening was performed using the Determine® HIV-1/2 rapid test (Abbot Diagnostics, Maidenhead, United Kingdom). To confirm infection and to discriminate between HIV types, women with positive and inconclusive screening results were subsequently tested with another rapid test, SD Bioline HIV-1/2 3.0 (Standard Diagnostics, Kyonggi-do, South Korea).
HIV treatment regimens. Provision of PMTCT at the initiation of this survey was guided by the 2006 World Health Organization (WHO) recommendations, i.e., ART prophylaxis in the third trimester (28 weeks) of pregnancy consisting of a regimen of twice daily Zidovudine (AZT), single-dose nevirapine (sd NVP) at onset of labour and Zidovudine plus Lamivudine (AZT + 3TC) for 1 week from birth. Women diagnosed in labour were given a single-dose nevirapine followed by AZT + 3TC for 1 week starting in labour. For women who needed treatment for their own health, triple ART (cART) was initiated as soon as possible. For HIV-2 and HIV-1/2 dually infected mothers, a combination of (AZT + 3TC) + Lopinavir/Ritonavir was recommended. For infants, guidelines recommended sd NVP plus AZT twice a day for either 4 weeks if the mother had received less than 4 weeks AZT prophylaxis prior to labour or for 1 week if the mother had received at least 4 weeks AZT prophylaxis. Infants of mothers receiving cART were recommended AZT twice daily 17 . After 2010, in line with the revised WHO recommendations, all pregnant women with CD4 count levels < 350 cells/mm 3 , irrespective of clinical stage, and women with a WHO clinical stage 3 or 4 infection, were started on triple ART. In addition, women not receiving ART for their own health, received (Option A) prophylaxis with AZT as monotherapy from as early as 14 weeks of pregnancy plus a single dose NVP + AZT + 3TC during labour and delivery followed by AZT + 3TC for 7 days after delivery, or triple ART prophylaxis (Option B) from as early as 14 weeks of pregnancy. The infant regime consisted of daily NVP from birth for a minimum of 4-6 weeks, and until 1 week after all exposure to breastmilk had ended. For infants receiving replacement feeding only daily NVP or sd NVP + daily AZT from birth and until 4-6 weeks of age was recommended 18 . Due to logistical issues, treatment was often started without CD4 cell count measurements based on clinical assessment. Throughout the study period, repeated stock outages and lack of ART at times forced clinicians to switch treatment and often to provide more simple regimes than recommended 9,19 . In 2015, Guinea-Bissau began implementing option B+, i.e., lifelong ART from diagnosis 20,21 .
Statistical methods. Maternity and HIV testing data were entered in password-secured databases (dBase 5.0, dataBased Inc, Vestal, NY, USA; Microsoft Access 2007, Microsoft, Redmond, WA, USA), and datasets were merged using unique birth numbers. The data were analysed using Stata 14.0 (Stata Corporation, College Station, TX, USA). Outcome variables such as LBW and stillbirths were dichotomized (0 = absent, 1 = present). Continuous explanatory variables were grouped categorically. Factors associated with HIV serostatus and birth outcomes were determined using univariate and multivariate logistical regression models. The multivariate analysis was fitted with statistically significant covariates (as determined by Wald's test). Birth outcomes i.e., stillbirths and low birthweight (LBW) were examined according to HIV status and adjusted for significant covariates. In accordance with the WHO criteria, stillbirth was defined as a newborn at or above 1,000 g showing no vital signs immediately after birth (Apgar score = 0) 22 . Low birth weight was defined as birth weight of < 2,500 g 23 . Miscarriages (birth weight < 1,000 g) were excluded. In the birth outcome analysis, "births" pertain to the number of fetuses at risk. Therefore, the number of total births is higher than the number of pregnant women due to multiple pregnancies. In the logistic regression models, adjustment for birth outcomes was made for clustering of twins using a specific pair number. Inconclusive test results were categorized as HIV-negative. Missing values were included in the logistic regression models. Trends over time (calendar year) for HIV prevalence and treatment were determined using Pearson's χ 2 test. Due to a lack of data to confirm self-reported antenatal treatment regimens and reasons for selecting a given treatment at labour, we opted to assess antenatal and treatment at labour separately. A p value of < 0.05 was considered significant.
Ethical considerations. All patient data were fully anonymised before being accessed. The use of government surveillance maternity data was approved by the National Ethical Committee in Guinea-Bissau (CNES-2010-018). A separate ethical approval was obtained for the analysis of HIV data and linkage of delivery and HIV databases (CNES-2011-030). The study was carried out in accordance with Guinea-Bissau Medical Research Ethics Committee requirements and the Helsinki declaration. All participants were counselled and asked for verbal informed consent before data collection and HIV testing. Newly diagnosed women were offered ART at labour and subsequently referred for follow-up. The manuscript was prepared in accordance with the STROBE guidelines (Supplemental Checklist 1).

Risk determinants of HIV-1 and HIV-2 infection.
Based on univariate analysis, we found maternal age, marital status, parity, and calendar year were associated with HIV-1, HIV-2 and HIV-1/2. Surprisingly, education was not significantly associated with HIV status and was therefore excluded from the multivariate analysis. As expected, the risk of HIV-1 increased significantly with age in the multivariate model (  www.nature.com/scientificreports/ ART and PMTCT regimens. Of women diagnosed antenatally (N = 566), 70% (396/566) reported receiving AZT prophylaxis, while 11% (62/566) reported receiving cART. Nineteen percent (108/566) reported that they were not receiving treatment or could not remember. Among women diagnosed in labour and women reporting no antenatal ART, 19% (149/774) received cART, 14% (108/774) received AZT prophylaxis plus a sd NVP + initiation of 3TC + AZT at childbirth, while 52% (408/774) of women, first diagnosed at delivery, received only sd NVP. Fourteen percent (109/774) did not receive any treatment at delivery. For women with HIV-2 (n = 196), 39% (76/196) received prophylaxis AZT plus a single dose-NVP + initiation of 3TC + AZT at childbirth, while 28% (55/196) received sd NVP only. Among women aware of their HIV status prior to labour and delivery, we observed a significant increase in self-reported antenatal treatment coverage from 60% in 2008 to 85% in 2013, p < 0.001. Specifically, there was an increase in women receiving antenatal AZT prophylaxis from 55% in 2008 to 82% in 2013, while there was no overall increase in cART during the same period (Fig. 3a).
For women diagnosed at delivery and HIV-positive women not receiving antenatal ART (excluding women www.nature.com/scientificreports/ already on treatment, n = 458), overall treatment coverage declined from 99% in 2008 to 50% in 2013, p < 0.001. Notably, the proportion of women not receiving treatment at birth increased steadily throughout the study period (Fig. 3b). Figure 3 shows the provision of self-reported antenatal treatment and treatment at birth among HIV-infected women at the HNSM. During the study period, 85% of all HIV-infected women (1,049/1,232) received some form of ART in relation to labour for PMTCT. The majority (34%, n = 418) received option A, 33% (n = 408) received sd NVP, 18% (n = 223) received triple ART and 15% (n = 183) did not receive any treatment. As with women diagnosed at delivery, we observed a steady increase in the number of women not receiving any treatment despite including women diagnosed antenatally (Fig. 4). Among infants of HIV-infected mothers (including twins and triplets), 8% (103/1,275) received the recommended treatment of sd NVP followed by AZT treatment, 70% (891/1,275) received sd NVP, and 22% (281/1,275) did not receive any treatment. Among infants of HIV-2 infected mothers, 77% (156/202) received sd NVP or sd NVP followed by AZT. The number of infants not receiving treatment to prevent mother-to-child transmission increased from 7% in 2008 to 36% in 2013, p < 0.001 (Supplemental Figure 2). Birth outcomes. Seventeen percent (4,213/25,232) of births were LBW, and 10% (2,513/25,256) of babies were stillborn. Factors independently associated with LBW and stillbirths are displayed in Supplemental Table 1. We found no association between antenatal treatment and birth outcomes. Assessing birth outcomes for HIV-1, HIV-2 and HIV-1/2 dually infected women compared with HIV negative women, we found that maternal HIV-1 and HIV-1/2 co-infection were associated with LBW, AOR 1.25 (95% CI 1.03-1.52) and AOR 1.73 (95% CI 1.26-2.37), respectively (Table 3). No significant association was found between HIV status and stillbirth in univariate or multivariate analyses (Table 4).

Discussion
This retrospective cross-sectional study examined HIV-1 and HIV-2 prevalence and factors associated with HIV, PMTCT treatment coverage and adverse birth outcomes at the largest maternity ward in Guinea-Bissau. It shows a decline in HIV-1, HIV-2, HIV-1/2 prevalence among pregnant women presenting for labour during a 5-year  www.nature.com/scientificreports/ period. Women from the Balanta and Fula ethnic groups had a significantly higher risk of HIV-1. Antenatal treatment coverage appears to be improving over time, yet an increase in the proportion of mothers and infants not receiving treatment at birth was observed, inevitably increasing the overall risk of mother-to-child transmission. This is specifically alarming as nearly half of the women were unaware of their HIV status prior to presenting for labour. Over the 5-year period, 15% of HIV-infected mothers and 22% of infants born to HIV-infected mothers did not receive any treatment. In addition, 67% of mothers with HIV-2 received NVP, which is ineffective against HIV-2 infection. Lastly, infants born to mothers with HIV-1 or HIV-1/2 dually infected were at risk of being LBW. The HIV-2 prevalence of 0.8% in this survey was notably lower than in previous surveys among pregnant women, suggesting a continuing decline in HIV-2 prevalence in Guinea-Bissau 8,24-26 . A national survey from 2013 conducted in all 10 regions of the country reported an HIV-2 (including HIV-1 co-infected) prevalence of 1.0% (95% CI 0.5-1.8) 27 . In addition, a recently published study from the Bandim Health Project demographic   www.nature.com/scientificreports/ surveillance site found an HIV-2 prevalence of 1.5% among younger women (aged 15-44 years) 7 . Changes in risk behaviour over the past decades and competitive exclusion have been proposed as explanations 28,29 . Modelling predicts that new HIV-2 cases in the country will cease within a few decades 30 . In contrast, previous surveys  N = 1,232). Overall coverage for cART and option A (AZT + 3TC + sd NVP) remained relatively low throughout the study period. There was a decline in the number of women receiving sd NVP, while the overall number of women not receiving any treatment increased over time. The p value shown was estimated using the chi 2 test for trend.  www.nature.com/scientificreports/ from Guinea-Bissau have shown increasing or stabilized HIV-1 prevalence among a range of populations 7,24,26,31 . Gianelli et al. 8 found that HIV-1 prevalence among pregnant women stabilized between 2003 and 2006. Our study is supported by findings from a recent national survey suggesting that the HIV-1 prevalence has stabilized and may be declining among women of reproductive age 27 . The risk of HIV transmission decreases substantially with ART 32 , and the decline in HIV-1 and HIV-2 may in part reflect the fruits of national interventions, including efforts to strengthen prevention measures during the roll-out of ART. Nevertheless, inadequate health infrastructure, repeated drug stock outages, poor adherence, and high rates of loss to follow in Guinea-Bissau increase the risk of treatment failure 7,10,14,33-35 and thus infectivity, which could reverse current encouraging trends in HIV prevalence. Sustained and sustainable long-term management of affected patients, combined with prevention, is required 10 . The Mandinga and Fula, two important West African ethnic groups, had a significantly higher prevalence of HIV-1 compared with other ethnic groups. A study by Kenyon et al. 36 has suggested that differences in sexual behaviour according to ethnic group may contribute to a large variation in HIV prevalence. Previous research from Guinea-Bissau has suggested that female genital mutilation and polygamy among these ethnicities may play a role 8,37,38 . Furthermore, a recent study from the HIV clinic at HNSM found Fula and Mandinga ethnicity to be associated with late presentation at HIV diagnosis 39 . These findings underline the importance of identifying vulnerable populations, understanding sociocultural motivators, and targeting interventions accordingly.
Overall, self-reported HIV treatment coverage among women accessing antenatal care increased significantly over the course of this study. However, only 47% of women presenting for labour had received antenatal testing and counselling, and only 77% of women were tested for HIV at birth, leaving a large number of women unaware and untreated 14 . HIV-infected women not identified during pregnancy miss out on the benefits of treatment and prevention and remain part of a "hidden epidemic"; they are prone to late diagnosis, poor health outcomes and at risk of infecting their infants 40,41 .
While some national interventions, such as strengthening the early provision of ART as recommended by the WHO, have shown promise 18 , repeated political instability, and inadequate management of HIV tests and medicine stocks appear to have significant negative impacts on the clinical management of HIV in the country 10,14,42 . After a civil war in 1999, Guinea-Bissau suffered five military coups between 2002 and 2012 43,44 , which left the country in a chronic state of political instability 14 . This has hampered development despite national and international efforts and inevitably discouraged donors, as was seen in 2012 after the Global Fund, the World Bank, the African Development Bank and the European Union suspended HIV funding activities in Guinea-Bissau 12,21 . These cuts in funding may explain the increase in HIV-infected women and their infants not receiving ART in labour in 2012 and 2013 as observed in this study, as well as the decline in national ART coverage from 56% in 2011 to 23% in 2014 21 . Repeated drug stock outages were common in Guinea-Bissau 9,10,42 and may have forced doctors and midwives to choose treatments no longer recommended by national guidelines, as reflected by the high proportion of women receiving sd NVP, or no treatment. In addition, a high number of HIV-2-infected patients received NVP, a non-nucleoside reverse transcriptase inhibitor (NNRTI), for PMTCT. This is problematic because NNRTIs are ineffective against HIV-2 due to natural resistance 45 . Periods of drug shortages and the resulting use of NNRTI-based regimes may explain the high levels of NNRTI resistance found in Guinea-Bissau 42,46,47 . Conditions in the country appear to have improved after international donors, led by the Global Fund, have resumed funding activities, which is reflected by a 30% increase (49.5-81.9%) in the number of HIV-positive mothers who received ART for PMTCT in 2014 21 . Current estimates from UNAIDS report that 65% of pregnant women living with HIV in Guinea-Bissau accessed ART in 2017 48 . While these results show promise, overall ART coverage remains low in Guinea-Bissau 48 . Preliminary results from a routine analysis of nationwide HIV Cohort in the country show that only 14% of HIV infected individuals are aware of their HIVstatus and of these only 20% are on treatment underling the urgent need to determine the treatment retention and long-term outcomes of mothers and their infants during and after PMTCT 10 .
This study found an increase in the number of infants not receiving PMTCT. An increase in the number of HIV-infected children and high rates of HIV infection (9.3%) despite PMTCT have been reported in Guinea-Bissau, suggesting poor treatment retention and challenges to PMTCT coverage and follow-up 15 . While an increase in the number of HIV-infected children may be due to better diagnostics, the late diagnosis of children (median age 3.5 years) suggests a large proportion of children die before being diagnosed with HIV 15 .
In line with other studies, women infected with HIV-1 and HIV-1/2 in this study had an increased risk of delivering LBW babies 49,50 . LBW babies have an increased risk of mortality and may experience negative health outcomes during childhood and later in life 51 . Studies have described an increasing risk of LBW with advancing stage of HIV disease 52 , but also according to maternal ART regimens 51 , emphasizing the importance of identifying HIV-infected mothers early in pregnancy and insuring they receive the optimal ART regimen to improve health outcomes, prevent MTCT and minimize the risk of LBW.
In Guinea-Bissau, stillbirths are common. A previous study from the capital Bissau found a community stillbirth rate of 71/1,000 among mothers who did not attend ANC and 36/1,000 among mothers who attended 7 or more ANC 16 . While HIV infection is known to be associated with stillbirths 50 , this study did not find a significant association. Many factors affect the risk of stillbirths, including maternal age, education, marital status and ethnicity, in our setting 16 . In addition, the stage of maternal disease, ART regimen and timing of treatment play a role in the risk of stillbirth and other adverse birth outcomes among HIV-infected mothers 4,53 . Therefore, our analysis may have been limited by our inability to control sufficiently for these factors.
This study is based on data from a large sample of women presenting for labour and delivery at the country's largest maternity ward and thus reflects the challenges to PMTCT services in a setting with limited resources. Our research has several limitations. First, the cross-sectional nature of this study means we are unable to determine trends in incidence and causality. Twenty-three percent of women presenting for labour during the study period were not counselled or tested for HIV. These women were not included in this study due to missing information www.nature.com/scientificreports/ on HIV status presenting a potential selection bias. Data regarding condom use, number of previous partners, and additional high-risk behaviours were unavailable, limiting our risk factor analysis. HIV type discrimination was performed using SD Bioline HIV-1/2 3.0. A study from Guinea-Bissau found that SD Bioline HIV-1/2 3.0 overestimates the number of HIV-1/2 dually infected individuals 54 , and high inter-observer variation was noted 55 , which could have led to misclassification of HIV type in this study. HIV prevalence was determined based on women presenting at the HNSM maternity ward and may differ from women giving birth at other urban health centres or at home. Only half (47%) of women tested at birth reported previous testing during antenatal visits. There were discrepancies between self-reported antenatal test results and results at labour suggesting either insufficient pre and post testing counselling, lack of HIV awareness, or stigma 13 , recall bias or seroconversion. The HNSM maternity ward receives women from all parts of the capital and other regions in Guinea-Bissau for care and treatment of obstetric complications, yet little is known regarding the specific referral process to give birth. The lack of continuity of care and registration systems made it almost impossible for midwives to corroborate self-reported antenatal treatment, which may have been affected by recall bias. In addition, the lack of confidentiality and time to perform pre-and post-test counselling at labour 13 could have limited midwives' and doctors' ability to make treatment decisions regarding the type and length of treatment based on self-reported antenatal treatment regimes.

Conclusion
The prevalence of HIV-1, HIV-2 and HIV-1/2 among pregnant women presenting for labour and delivery at the HNSM declined between 2008 and 2013. The observed increase in the proportion of women and infants not receiving treatment at birth is concerning. Less than half of the women in this study reported attending antenatal counselling and testing, and 23% were not tested in relation at birth, resulting in many mothers unaware of their HIV status and hence deprived of the benefits of treatment and prevention. As Guinea-Bissau continues to take strides towards improving PMTCT testing and ART coverage to mitigate the risk of poor health outcomes of HIV-infected mothers and their infants, further operational research focusing on provision of PMTCT services and continuity of care, retention to PMTCT services, including treatment and patient follow-up, is urgently needed to support future interventions and policy.

Data availability
The original data are stored at the Bandim Health Project, National Institute of Public Health in Guinea-Bissau. The ethical approvals received from The National Committee on Ethics in Health, National Institute of Public Health, Ministry of Public Health, Guinea-Bissau (e-mail: inasagb@gmail.com) do not allow for general data sharing of this dataset. Data access can therefore be granted to individual researchers only after obtaining ethical permission. Any party interested in accessing the relevant data can do so by contacting the Bandim Health Project (e-mail: bandim@ssi.dk). The Bandim Health Project will facilitate the request for ethical permission from the National Committee on Ethics in Health and provide the data on approval. We can confirm that others will be able to access the relevant data in the same manner and form as the authors, provided ethical approval is obtained from Guinea-Bissau's National Committee on Ethics in Health.