Optimizing the World Health Organization algorithm for HIV vertical transmission risk assessment by adding maternal self-reported antiretroviral therapy adherence

Vertical transmission (VT) of HIV remains an unacceptable 12% in many Global Plan priority countries, where 90% of the world's pregnant women are living with HIV [1]. Since implementation of Option B+ strategy, [2] antiretroviral therapy (ART) coverage among pregnant women in eastern and southern Africa increased from 84 to 92% [3]. However, the goal of eliminating new paediatric infections by 2030 [4] will likely be compromised by multiple challenges in the VT-prevention cascade, including access to adequate postnatal prophylaxis among HIV-exposed infants [1, 5‐11].

VT prevention guidelines have evolved recently. Extended nevirapine (NVP) postnatal prophylaxis and dual or triple antiretroviral (ARV) prophylaxis have shown to halve VT among HIV-exposed infants compared with short courses of AZT or NVP alone [12‐17]. The latest World Health Organization (WHO) guidelines recommend enhanced post-natal prophylaxis with daily zidovudine (ZDV) and NVP for the first 6 weeks of life for HIV-exposed infants at high-risk of HIV acquisition, followed by an additional 6 weeks if high-risk and breastfeeding. HIV-exposed infants at low-risk of VT should receive 4–6 weeks of prophylaxis with daily NVP (or twice-daily AZT) [18].

High-risk infants have mothers diagnosed with HIV at delivery and are either: 1.) not on ART 2.) started treatment within 4 weeks before delivery, or 3.) had a plasma viral load (VL) > 1000 copies/mL in the 4 weeks before delivering [18]. WHO designed an algorithm to assess HIV-exposed infants risk for HIV acquisition at delivery and to identify high-risk infants eligible for enhanced post-natal prophylaxis [19]. The implementation of the WHO algorithm is challenging due to the paucity of required information for risk stratification [20]. The ‘high-risk criteria’ assume VL result availability, adequate maternal adherence, and early and sustained viral suppression after 4 weeks of ART [18]. However, VL testing coverage and monitoring, especially during pregnancy, is challenging in low-and-middle-income countries (LMIC). Many countries have scaled VL testing via point of care or dried blood spot testing. However, in routine operating conditions, many patients who are tested either don’t receive results or they are extremely delayed [1, 21, 22]. Women with poor adherence or interrupted ART during pregnancy or breastfeeding have viral rebound and increased risk of VT [23‐26]. In fact, peripartum viral suppression varies from 30 to 98% in different sub-Saharan African settings, and despite an estimated 92% ART coverage among pregnant women, nearly 30% of new pediatric infections in East and Southern Africa in 2018 were related to ART interruptions during pregnancy or breastfeeding [3, 10].

We retrospectively characterized post-natal prophylaxis coverage among infants living with HIV in sub-Saharan Africa and assessed whether the inclusion of readily accessible information such as mothers’ self-reported ART adherence into the WHO algorithm improved the identification of infants with perinatal HIV infection when maternal VL resulted unavailable.

A total of 184 infants living with HIV (n = 96 male) and their mothers were studied, 112 (61%) from South Africa, 64 (35%) from Mozambique and 8 (4%) from Mali (Fig. 1). A total of 143 (78%) women were known HIV-positive and on ART at delivery. However, 29 women (16%) had their HIV diagnosis at delivery and were not on ART and 12 (7%) were known to be HIV-positive prior to delivery but were not on ART. Among them, only 17/143 (12%) had a VL result within 4 weeks before delivery, 13/17 (76%) of them had VL ≥1000 copies/ml. Therefore, most known HIV-positive women on ART (126/143, 88%) did not have a VL result. Of them, 99/126 (79%) had been on ART for over 4 weeks and 27/126 (21%) under 4 weeks. A total of 45/99 (45%) mothers reported suboptimal adherence to ART.

Using the WHO algorithm, 81/184 (44%) of infants with HIV were defined as high risk of VT with VL available data. Following the inclusion of the self-reported adherence risk factor applied to 99 mothers on ART greater than 4 weeks without VL test result, an additional 45 high-risk infants were identified. In total, the modified algorithm classified 126/184 (68%) infants as high risk of MTCT – assuming that infants born to mothers with suboptimal adherence are high risk (Fig. 2). The modified algorithm, identified higher number of high risk children than the WHO algorithm (p = 0.0001).

Among the 81 at high risk of MTCT according to WHO algorithm, 40% (32/81) received enhanced post-natal prophylaxis, 30% (24/81) standard post-natal prophylaxis, 22% (18/81) did not received any post-natal prophylaxis and 9% (7/81) did not have information about the post-natal prophylaxis.

In table 1, we compared clinical and sociodemographic characteristics that could have acted as confounding factors in the risk of VT, between participants classified as high-risk by the WHO algorithm and the additional participants classified as high risk when maternal adherence was incorporated into the algorithm. We did not find any differences in age, sex, maternal WHO status at delivery, maternal health, or social adverse events.

In our study, nearly 90% of mothers of infants diagnosed with HIV did not have a VL result at delivery; a result on which the WHO algorithm relies. When applied in our cohort, the WHO algorithm identified 44% of the infected infants as high-risk for VT. When we added mothers’ self-reported ART adherence to the WHO algorithm, the proportion of high-risk infants increased to 68%, suggesting that suboptimal adherence is associated with high risk. This modification of the algorithm increased the number of infected infants being classified as high-risk and eligible for enhanced post-natal prophylaxis.

This study also showed that information is often missing when assessing VT risk assessment according to the WHO algorithm, which jeopardizes early prescription of enhanced post-natal prophylaxis [15]. Only 12% of mothers on ART had a VL result 4 weeks prior to delivery. Similarly, a cross-sectional study of programmatic data in Zimbabwe revealed that the risk of MTCT using the WHO algorithm could not be determined in 90% of HIV-exposed infants due to lack of data [20]. A more recent study of 2080 HIV-exposed infants from Zimbabwe showed that 80% of mothers did not have a VL result between 28 weeks gestation to delivery [30]. Therefore, other factors should be considered to establish the risk of VT. Our results suggest that self-reported maternal adherence to ART could be an additional clinical factor to be considered in order to improve the algorithm to assess the risk of VT in the absence of VL. We also evaluated other socio-clinical factors that could potentially affect the risk of MTCT, such as maternal WHO stage, level of education, employment status or adverse social events between the two algorithms. We didn’t find any difference among the participants newly classified as high risk of VT according to maternal ART adherence and the participants classified by the WHO algorithm as high risk of VT. However, further studies evaluating additional socio-clinical variables could inform more targeted algorithms to improve prevention of VT.

ART duration over 4 weeks, as incorporated in the algorithm, does not justify the assumption of viral suppression or the low VT risk. Although many countries have recently changed from EFV-based maternal ART to Dolutegravir, which has been shown to have superior early virologic suppression [31], adherence remains a paramount. Poor maternal adherence or interrupted ART during pregnancy and breastfeeding may cause high VL and subsequent increased transmission risk [23‐26]. As such, infants are misclassified as ‘low-risk’ and do not receive appropriate e-PNP. The use of point-of-care devices for VL may be used at delivery and provide results in under 90 minutes, however, these devices are not widely available in LMIC [32]. Modification of the existing WHO risk evaluation algorithm to include a maternal self-reported adherence may significantly benefit countries without a consolidated laboratory network capacity to ensure optimal VL monitoring. Investigating the mothers’ self-reported ART adherence at delivery is an easy and low-cost intervention to guide nurses in identifying high-risk infants eligible for enhanced post-natal prophylaxis in the absence of VL result.

In our cohort, 45% of mothers on ART for over 4 weeks without a VL result reported sub-optimal adherence. The actual proportion might be higher, since self-reported adherence is a reliable method with low cost and good specificity [33] widely used in clinical practice, but tends to overestimate adherence behavior compared with other methods [9]. Low self-reported adherence during pregnancy and breastfeeding is associated with viremia and virological treatment failure [34, 35]. The high number of women who self-disclosed ART-adherence difficulties in our study suggests that in the absence of VL result, time on ART is not enough to evaluate VT risk. Effective interventions to support adherence during pregnancy must be considered.

Our findings showed that 30% of HIV-exposed infants at high-risk of HIV infection according to WHO algorithm had no access to enhanced post-natal prophylaxis, which can be partially due to the recent guideline implementation in Mozambique [28]. A study in Zimbabwe also showed low enhanced post-natal prophylaxis coverage rates among HIV-exposed infants [20]. We also found a high proportion (17%) of HIV-exposed infants who alarmingly did not receive any PNP at all. Further studies should evaluate enhanced post-natal prophylaxis compliance and algorithm feasibility in the clinical setting, including HIV-exposed children at high risk of VT who never become infected. Nevertheless, our results suggest that besides modifying the WHO algorithm and training health staff on its correct application, further efforts are needed to ensure access to timely maternal VL testing and infant enhanced post-natal prophylaxis.

Our results have implications for policy, practice and public health. This study provided a comprehensive evaluation of the WHO risk assessment algorithm for VT. On one hand, we identified that the coverage of enhanced postnatal prophylaxis among infants with high risk of VT transmission was not optimal. On the other hand, our results suggested that adding maternal adherence to ART could increase the coverage of enhanced postnatal prophylaxis in high-risk infants, thus reinforcing WHO recommendations in clinical practice. The application of this simple measure would be especially beneficial in settings with less structural capacity to obtain viral load results. It would allow the allocation of limited resources towards prevention measures for children at high risk and inform interventions to improve VT prevention.

This study has several limitations. First, we included only infants confirmed to have HIV infection, and retrospectively reviewed their VT risk classification. For this reason, we could not calculate the specificity of the two algorithms. Second, the mothers’ ART adherence was self-reported at recruitment and recall bias, as well as social desirability bias, may have led to over-reporting of good adherence. We defined 1 month prior to study enrollment as the time interval for observing the number of missed doses in order to improve the validity of past reporting. We based this on the fact that a shorter time frame allows the respondent to more easily recall an event rather than having to recall a behavior over a large period of time. However, further studies evaluating a standardized method and definition of self-reported adherence would be important to generalize our results on ART adherence and validate the inclusion of the self-reported adherence to ART in the algorithm in other contexts. Third, we didn’t find difference in the confounders factors analyzed, however other potential confounding factors such as type of infant feeding, home delivery or maternal drug resistance have not been accounted for. Fourth, the relatively small sample size can compromise the generalizability of the results. Further studies with larger sample size are needed to validate this results in populations from different settings.

In conclusion, incorporation of maternal self-reported adherence in the WHO algorithm for VT risk assessment improves the identification of infants eligible for enhanced post-natal prophylaxis and should be considered in the algorithm. A study including HIV-exposed uninfected infants is needed to assess whether the modified WHO algorithm adequately identifies infants who do not need enhanced post-natal prophylaxis and remain HIV-free after breastfeeding.