A systematic review of interventions to improve postpartum retention of women in PMTCT and ART care

Introduction The World Health Organization recommends lifelong antiretroviral therapy (ART) for all pregnant and breastfeeding women living with HIV. Effective transitioning from maternal and child health to ART services, and long-term retention in ART care postpartum is crucial to the successful implementation of lifelong ART for pregnant women. This systematic review aims to determine which interventions improve (1) retention within prevention of mother-to-child HIV transmission (PMTCT) programmes after birth, (2) transitioning from PMTCT to general ART programmes in the postpartum period, and (3) retention of postpartum women in general ART programmes. Methods We searched Medline, Embase, ISI Web of Knowledge, the regional World Health Organization databases and conference abstracts for data published between 2002 and 2015. The quality of all included studies was assessed using the GRADE criteria. Results and Discussion After screening 8324 records, we identified ten studies for inclusion in this review, all of which were from sub-Saharan Africa except for one from the United Kingdom. Two randomized trials found that phone calls and/or text messages improved early (six to ten weeks) postpartum retention in PMTCT. One cluster-randomized trial and three cohort studies found an inconsistent impact of different levels of integration between antenatal care/PMTCT and ART care on postpartum retention. The inconsistent results of the four identified studies on care integration are likely due to low study quality, and heterogeneity in intervention design and outcome measures. Several randomized trials on postpartum retention in HIV care are currently under way. Conclusions Overall, the evidence base for interventions to improve postpartum retention in HIV care is weak. Nevertheless, there is some evidence that phone-based interventions can improve retention in PMTCT in the first one to three months postpartum.


Introduction
The human immunodeficiency virus (HIV) can be transmitted from the mother to her child during pregnancy and birth through blood, or postnatally through breast milk. With mounting evidence from randomized trials of the safety and efficacy of postnatal maternal antiretroviral drugs in preventing HIV transmission through breast milk [1Á6], the finding that antiretroviral therapy (ART) drastically reduces the risk of sexual transmission [7,8], and the benefits of ART to the patient's health even at high CD4-cell counts [9], the World Health Organization's (WHO) prevention of mother-to-child HIV transmission (PMTCT) guidelines have gradually moved to recommending longer periods of maternal antiretroviral drugs postpartum. In its latest 2015 early-release guidelines, the WHO recommends providing lifelong ART to all pregnant and breastfeeding women living with HIV regardless of CD4cell count or clinical stage, which has also been termed Option B' in earlier guidelines [10]. Countries have moved fast to implement this option, with 18 of 22 priority countries having adopted or scaled-up Option B' to date [11]. Five countries (Malawi, Tanzania, Uganda, Lesotho, and Ethiopia) have nationally implemented lifelong ART for all pregnant and breastfeeding women living with HIV, and an additional eight countries have started to scale up this treatment option.
Pregnant women who test HIV-positive in antenatal care (ANC) are generally initiated on antiretroviral drugs at the antenatal clinic, or after referral from ANC to ART care settings. After a series of prenatal PMTCT visits, the mother is usually expected to attend a number of further postnatal PMTCT visits for both herself and her child. The existing literature shows that a large percentage of women are lost along this cascade [12]. While the evidence on postpartum retention is sparse, depending on the setting and follow-up period, approximately 25Á50% of women on ART at delivery may be lost from care during the postpartum period [13Á15]. This high attrition rate is problematic due to the increased risks of HIV-related morbidity and mortality, HIV transmission, and the development of drug-resistant HIV strains [16,17].
While the move to Option B' simplifies the guidelines for ART by recommending that all pregnant women, regardless of CD4 count or clinical stage, begin lifelong ART, it introduces a new step in the care cascade between PMTCT or maternal and child health (MCH) care services, and long-term ART services. Finding ways to transition women initiating HIV treatment in the antepartum period to lifelong ART postpartum, and retaining them in ART care, will thus be critical to reducing both vertical and sexual HIV transmission, and to improving the mother's health. More broadly, studying these questions is likely to yield lessons that are pertinent to a wide variety of healthcare settings. Other than providing evidence on how retention in long-term care settings can be improved, such studies may also help to answer how a previously stand-alone programme can be effectively integrated into a broader care cascade. In addition, evidence on how retention of postpartum women in HIV care can be increased is likely to apply to other care settings, in which an important motivational factor for the patient to attend care (in this case, reducing the risk of infecting the fetus/newborn with HIV) has ceased to exist.
Focusing on HIV-positive women between birth and five years postpartum, we conducted a systematic review to assess the evidence for interventions that aim to improve (1) retention within PMTCT programmes, (2) transitioning from PMTCT to general ART programmes, and (3) retention in general ART programmes.

Methods
This review was commissioned by the WHO to inform the operational section of the 2015 update of the WHO's consolidated guidelines on the use of antiretroviral drugs. A review protocol exists and can be requested from the authors. . We also searched conference abstracts for the Conference of the International AIDS Society (IAS and AIDS) from 2002 to 2014, and the Conference on Retroviruses and Opportunistic Infections (CROI) in 2014 and 2015. The databases were searched using key words and medical subject headings (MESH) for HIV, PMTCT, retention in care, loss to follow-up, transition, linkage, and pregnancy. The search terms for each database are shown in additional file 1. We screened titles and abstracts for relevance, and then analyzed the full-text versions of potentially relevant articles using the following inclusion criteria: (1) the study population is pregnant or postpartum women (up to a maximum of five years after delivery) living with HIV, (2) the article presents primary quantitative data on the transition from ANC or PMTCT to ART in the postpartum period, or data on retention in PMTCT or ART care of postpartum women living with HIV, and (3) the study evaluates an intervention. The database searches, and analyses of abstracts and articles' full-text versions were not conducted in duplicate. No restrictions were placed on study design, sample size, or publication type. We also did not restrict articles by publication language although the search strategy was only run in English. Finally, the reference lists of all included studies, and relevant review articles and commentaries were screened for additional references. We also contacted the corresponding author of all included studies and relevant ongoing studies, for which we found a published study protocol (see Table 10), to inquire about unpublished data and manuscripts.

Data extraction
The following information was extracted from each included article: author, year of publication, period of data collection, study design, country, study population, unit of randomization (for randomized studies only), sample size by study arm, outcome measure(s), and study results.

Assessment of study quality
We assessed the quality of each included study using the criteria of the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) Working Group [18]. As such, each study was graded as high, moderate, low, or very low quality of evidence. We decided to grade quality by study rather than outcome because of the wide heterogeneity in outcome measures and intervention designs of the included studies. We did not undertake an assessment of risk of bias across studies (e.g., publication bias).

Analysis of extracted data
We extracted or calculated, if possible, the proportion of women who were successfully transitioned to, or retained in ART. We also determined the relative risk (RR) or odds ratio (OR) with 95% confidence intervals (CIs) comparing the rate of transitioning and/or retention between the intervention and control group. Due to the large degree of heterogeneity between study designs, interventions, outcome measures, and the reporting of outcomes, we decided that a metaanalysis was inappropriate.

Results and discussion
The search of the databases retrieved 8324 records with an additional 17 records identified through searching the reference lists of included articles and relevant reviews, and contacting authors ( Figure 1). After removing duplicates, screening abstracts, and full-text reviews, we identified 10 studies that met our inclusion criteria.

Characteristics of included studies
Nine of ten studies included in this review were carried out in sub-Saharan Africa (one study took place in the United Kingdom), seven of which were conducted in East Africa and two in South Africa (Tables 1Á3). All studies were published between 2010 and 2015. Six studies were observational studies and four randomized trials, of which one was a cluster-randomized trial. Three of the studies evaluated interventions that employed text messages and/or phone calls to improve retention, and four studies evaluated the effect of different levels of integration between ANC/PMTCT and ART care on retention.

Quality assessment
Four of the included studies were of high or moderate quality, and six of very low quality (Tables 4Á6). We downgraded all observational studies in this review to very low quality.
Reasons for the downgrade were a high risk of bias and imprecision of results (i.e., wide confidence intervals). The justification for each study's quality grading is shown in Tables 4Á6.

Outcomes with relevant summary measures
Two randomized studies, both of moderate quality according to GRADE criteria, found that the use of text messaging and/or phone calls was associated with improved clinic attendance postpartum (Table 7). Odeny et al. found that two-way text messaging was associated with an increase in the percentage of patients who attended a postpartum visit within eight weeks of delivery (RR: 1.66, 95% CI: 1.02Á2.70) [19], and Kebaya et al. found that biweekly phone calls during the first 10 weeks postpartum were associated with a higher proportion of mother-infant pairs attending the MCH clinic at six weeks (RR: 1.34, 95% CI: 1.07Á1.68) and at 10 weeks postpartum (RR: 1.86, 95% CI: 1.34Á2.58) [20]. A likely reason for which the retention rates in the study by Kebaya et al. are significantly higher than those by Odeny et al. is that the former assessed retention between delivery and six/ten weeks postpartum, while Odeny et al. measured retention during the longer time period between (antepartum) enrolment and eight weeks postpartum. Schwartz et al., an observational study judged to provide a very low quality of evidence, did not find any change in retention in ART at 12 months postpartum due to text messaging and phone calls (RR: 1.03, 95% CI: 0.83Á1.27) [21]. Regarding integration of ANC and ART care, van Lettow et al. found that facilities, which require a referral between ANC and ART for all doses of ART had a higher rate of retention of women at 12 months after ART initiation than facilities where either the first or all doses of ART were provided in the ANC facility [23] (Table 8). Similarly, Weigel et al. reported that a series of more than 10 interventions over three years to improve the transition from ANC to ART was associated with higher retention in HIV care six months after ART initiation (RR: 3.85, 95% CI: 2.10Á7.08) [24]. However, Turan et al., the only included randomized trial of an integration of care intervention, and Stinson et al. identified no effect of integration of care on postpartum retention (Tables 7Á9) [22,25].   programme Retention a at 12-months postpartum HIV'0HIV-positive; ART 0antiretroviral therapy; PMTCT0prevention of mother-to-child HIV transmission; MIP 0mother-infant pair. a Retention was defined as not having missed the last appointment to pick up antiretroviral drugs by more than six weeks, or having transferred out to another healthcare facility.     a Participants in the intervention arm were seen at baseline, week 2, and months 1, 2, 3, 6, 9 and 12, of which the baseline visit and the visits at months 2 and 12 were managed by a doctor. Participants in the control arm where seen at baseline, weeks 2 and 4, and monthly thereafter (all visits were managed by a doctor). b CHWs were responsible for (1) identifying newly pregnant women through household visits; (2) referring newly pregnant women to ANC, (3) reminding women of an upcoming ANC, PMTCT, or baby follow-up appointment, (4) visiting women at home who have missed an ANC, PMTCT, or baby-follow-up appointment, and (5) visiting women at home two weeks before the due date to discuss the birth plan. In Area 2, these activities were overseen by the Millennium Villages Project and closely monitored. In Area 1, they were overseen by the government or other non-governmental organizations. c CHWs registered women in an automatic text message system at the first ANC/PMTCT visit. The system sent an automatic reminder to the CHWs asking them to (1) remind women of an upcoming ANC, PMTCT, or baby follow-up appointment, (2) visit a woman at home who had missed an ANC, PMTCT, or baby follow-up appointment, and (3) visit a woman to discuss her birth plans two weeks prior to her due date.

Currently ongoing studies
restricted to studies that published a protocol in one of the databases searched for this review, which did not include trial registers. On this note, we would also like to point readers to the randomized trial by Yotebieng et al. [29], which was published after this systematic review had been completed.
Eight of the studies in Table 10 are cluster-randomized trials and one is a stepped wedge randomized trial. Four studies are evaluating an intervention that employs peer counsellors, two trials examine the effect of integration of HIV and non-HIV clinical services in the same clinic, and two studies' interventions include male participation in PMTCT. Seven studies assess retention in the first 12 months postpartum, and one study follows women for two years after birth. Two studies only assess short-term retention in MCH care (at six and 12 weeks postpartum).

Summary of the evidence
This review has identified 10 studies that met the inclusion criteria, of which we only judged four to be of high or moderate quality evidence. Two randomized trials found that the group receiving text messaging and/or phone calls had higher postpartum retention. Unfortunately, neither study assessed long-term retention. Odeny et al. found that twoway text messaging led to an increase in attendance of a postpartum visit within eight weeks of delivery [19], and Kebaya et al. found that women living with HIV who received biweekly phone call reminders during the postpartum period were more likely to attend the baby follow-up visit at six weeks and at 10 weeks postpartum [20]. Based on this, we feel that phone-based interventions seem to hold promise in increasing retention in the early postpartum period. Only one of the currently ongoing trials we identified is evaluating a phone-based intervention [34]. It is important to note that the retention rate in the control group in both included trials was low (11.8% at eight weeks postpartum in Odeny et al. [19], and 37.3% at 10 weeks postpartum in Kebaya et al. [20]), which may limit the generalizability of the findings to settings with higher retention and could (partially) explain the lack of effect found by Schwartz et al. [21].  Integration of care may have a variety of benefits, such as higher ART initiation of pregnant women living with HIV; this review, however, focused only on the effect of care integration on postpartum ART retention. The evidence for the impact of integration of care interventions on postpartum ART transition and retention is inconsistent, which is likely due to the heterogeneity of intervention designs, outcome measures, and study populations, as well as the low study quality of the four identified studies. In particular, each of the studies integrated care between ANC and ART services somewhat differently, making the evidence arising from these studies highly context-specific and difficult to generalize to other settings. In addition, only one of the studies [25] included a completely integrated model of care whereby the ANC clinic provides ART to the pregnant women for life, obviating the need for referral to an ART clinic (a care transition that is likely accompanied by a high loss to follow-up) either during pregnancy or in the postpartum period. Only one of the four studies found that integration of care improved retention in HIV care postpartum [24]. However, although the study found a stark increase in the percentage of women retained at six months after ART initiation (65% vs. 17%; RR 3.85, 95% CI: 2.10Á7.08) [24], it suffers from a high risk of bias from the pre-post study design with a large time gap (three years) between the pre-and post-data collection periods. In addition, the study implemented 11 changes aimed at improving ANC to ART linkage during the study period, which makes it impossible to determine which of the measures was effective. The only randomized trial on this topic, a cluster-randomized trial conducted by Turan et al. in   Kenya [22], found no difference in the proportion of women with two or more HIV care follow-up visits in the first six months after testing HIV-positive in ANC between integrated clinics (providing ANC and ART services in the same clinic until 18 months postpartum) and non-integrated clinics (ANC clinics that referred pregnant women for ART). The study, however, was insufficiently powered for this outcome. Nonetheless, the trial found that integration of care can lead to improved transitioning between PMTCT and ART as measured by the higher percentage of ART-eligible women who initiated ART within 12 months of testing HIV-positive in ANC in integrated as compared to non-integrated care clinics (40% vs. 17%; OR 3.22; 95% CI: 1.81Á5.72).
The third randomized trial included in this review, a noninferiority study, found that task-shifting to nurses along with home visits of defaulted patients by peer counsellors did not lead to a change in ART retention in the first 12 months after ART initiation [26]. The potential for the study's intervention programme to have a beneficial effect on retention was limited by the extremely high retention rate in the control group (98%). The only study not carried out in sub-Saharan Africa found that providing financial support for formula feeding to women at certain postpartum clinic visits increased both ART retention and adherence as measured by viral suppression [28]. However, the study, which was carried out in the United Kingdom, has limited generalizability to countries, in which exclusive breastfeeding is the recommended infant feeding option for pregnant women living with HIV. In addition, the financial support provided (a 2015 equivalent of US$930 per woman) is well above what health systems in lowand middle-income countries can currently afford, even if this value is adjusted for purchasing-power-parity.
Concerning the use of monetary incentives to increase postpartum ART retention, we would like to point out that a large randomized trial, published after we had completed this systematic review, evaluated the effect of conditional cash transfers on retention in PMTCT at six weeks postpartum in Kinshasa [29]. The trial found a 12.5% higher probability of retention at six weeks postpartum for women in the intervention compared to the control arm (81% vs. 72%; RR ratio: 1.11; 95% CI: 1.00Á1.24), which was marginally significant (p00.055). The monetary incentive used in this trial (US$5 at the first visit plus US$1 at every subsequent PMTCT visit conditional on attending the scheduled visit and on a few other requirements, such as providing a blood sample for CD4-cell count measurements) is far smaller than the one used in the study in the United Kingdom but may still be considered to be too high for scale-up in many settings in sub-Saharan Africa. Feasibility concerns aside, it is important to consider this cash value in the context of substantial outof-pocket expenditures and opportunity costs incurred by patients who attend HIV care in sub-Saharan Africa, despite ART being free at the point of care [40].
The evidence on approaches to improve the transition between MCH and long-term ART services and retention of women in ART programmes postpartum is currently still very limited, even though a number of relevant randomized trials are currently under way in sub-Saharan Africa. Table 10 provides a non-exhaustive list of currently ongoing trials. Variables that were included in the multivariate regression model are district in which the facility is located, facility type (district hospital, community hospital, health centre, private clinic), whether ART/ PMTCT services are offered on all weekdays or only on certain weekdays, number of women in the study cohort at each facility, number of women in the study cohort per clinical staff, time of adherence counselling (on the day of ART initiation, at the next visit, both on the same as ART initiation and at the next visit), and availability of ART/mother-infant-pair clinic for follow-up. c The denominator is all women not known to have transferred out.
d These values are not given in the original manuscript and were instead calculated by the authors of this review.       However, most of these trials will not yield final results before 2016. This recent surge in research interest on the question of postpartum ART transitioning and retention most likely reflects the policy shift to lifelong ART for pregnant women. The fairly rapid move to Option B' by many countries over the last few years may partially explain the current lag in research evidence on some of the operational questions of implementing lifelong ART for all pregnant and breastfeeding women living with HIV. Unfortunately, only one of the ongoing trials listed in Table 10 assesses retention beyond 12 months postpartum [34]. It is, however, crucial to assess longer term postpartum retention on ART because (1) a major anticipated benefit of lifelong ART is that HIV-positive women will be virally suppressed at the time of subsequent pregnancies, which requires sustained ART adherence; (2) the duration of breastfeeding, a period of high risk of vertical HIV transmission [41], tends to be considerably longer than 12 months in sub-Saharan Africa [42]; and (3) the transition between maternal health services to general ART care, which may occur after 12 months postpartum depending on the country setting, is probably a point in the care cascade at which women are at a particularly high risk of being lost from care. We, therefore, anticipate that the research gap on long-term postpartum ART retention will remain in the coming years.
Learning from research on adult ART retention, adherence, and linkage to HIV care While the evidence base on the retention of women specifically in the postpartum period is limited, it is plausible that interventions increasing ART retention in general also improve postpartum ART retention. Thus, valuable lessons could be learned from studies aimed at improving: (1) retention and adherence in ART programmes, or (2) linkage between different HIV care settings, such as from HIV testing to pre-ART or ART initiation, or from pre-ART care to ART. Regarding adherence, Chaiyachati et al. found in their rapid systematic review that five broad types of interventions, namely cognitive behavioural therapy, education, treatment supporters, directly observed therapy, and active reminder devices (e.g., text messaging), were effective in several settings in improving adherence to ART [43]. Concerning linkage between HIV care settings, a systematic review by Fox et al. identified several intervention types that proved to be effective in improving linkage between HIV testing and enrolment in HIV care with technology-based interventions, such as text messaging and point-of-care CD4-testing, appearing to hold the greatest promise (study under peer review). Point-of-care CD4 testing is, of course, less likely to have a substantial effect for linking pregnant women to ART as they are eligible for ART (or Zidovudine) regardless of CD4-cell count. Another recent systematic review found that food incentives and point-of-care CD4 testing increased the percentage of eligible adults who initiated ART [44]. Thus, although the evidence base for the effectiveness of phone-based interventions in improving postpartum retention comes from only two studies, there is a broader body of evidence that phone-based interventions can increase adult ART retention, and linkage between HIV testing and HIV care.

Learning from non-intervention studies on postpartum retention
An additional important source of evidence that can inform the design of interventions aimed at improving postpartum ART retention is non-intervention studies, usually cohort studies, on factors associated with non-retention in HIV care. Even though some factors associated with non-retention in ART care in general are likely to also apply to pregnant and postpartum women, it is likely that pregnant women face additional barriers specific to their gender and pregnancy status. In a recent systematic review, Hodgson et al. identified a variety of individual and community-level factors associated with non-retention in HIV care among pregnant and postpartum women, including poor understanding of HIV and its treatment, non-disclosure of HIV status to a spouse, and partner involvement in PMTCT [46]. At the community-level, HIV-related stigma was still a major barrier to retention. In addition, Colvin et al. identified a number of health system barriers to HIV care retention of pregnant and postpartum women, including insufficient communication and coordination between different health system actors, poor training of healthcare providers, and a stigmatizing attitude by some healthcare workers [47].
These reviews clearly show that the reasons for nonretention of pregnant and postpartum women on ART are complex and multifaceted, including patient-level factors (knowledge and conflicting priorities, such as caring for family and employment), community-level factors (such as stigma), provider-level factors (such as clinical knowledge and attitudes), and wider health system barriers (such as drug stock outs, access to care, and waiting times). This might suggest that multifaceted interventions that attempt to address as many of these barriers as possible will be most effective in improving postpartum ART retention. However, these interventions also tend to be more costly and complex to implement. While the effect sizes are modest, the two randomized trials on phone calls and/or text messages included in this review showed that simple singular interventions can significantly improve postpartum ART retention [19,20]. Thus, even though multifaceted interventions may be needed to achieve nearperfect retention, it is unclear whether simple singular interventions or more complex multifaceted interventions will prove to be more cost-effective in improving postpartum retention in health systems that currently experience a high rate of loss to follow-up of postpartum women living with HIV. It is, therefore, essential that studies on this topic assess and report the cost of implementing the intervention that is being evaluated. However, the utilitarian cost-effectiveness criterion should not be the only lens through which these interventions are evaluated. After all, stigma towards people living with HIV both by community members and healthcare providers has important human rights implications, which can justify interventions independently of cost-effectiveness considerations.

Limitations
This systematic review has several limitations. First, we may not have identified all studies that have been conducted on this topic as the search strategy was only run in English and we were, therefore, not able to identify studies without an English title or abstract. In addition, although we made an effort to identify grey literature by contacting corresponding authors of the included studies and study protocols, we may not have identified unpublished data on our review questions. Second, due to the wide variation in interventions, study designs, and outcome measures, it was not feasible to summarize the results of the included studies using Forest plots or a meta-analysis. Third, the small number and generally poor quality of the identified studies limited our ability to deduce meaningful conclusions with policy implications.

Conclusions
The evidence base on interventions to improve retention of women in HIV care during the postpartum period is weak, particularly for improving longer term retention on ART. Nevertheless, there is some evidence that phone-based interventions can improve retention in PMTCT in the first one to three months after childbirth. A number of randomized trials are currently under way and expected to publish their results in mid-to late 2016. Our systematic review has identified a weak evidence base on key operational aspects of implementing the WHO's recommendation of lifelong ART for all pregnant and breastfeeding women living with HIV. This study, therefore, highlights the need for more rigorous evaluations of health system interventions to determine the most efficient and effective strategies of providing ART to this important population group.