Virologic and immunologic failure, drug resistance and mortality during the first 24 months postpartum among HIV-infected women initiated on antiretroviral therapy for life in the Mitra plus Study, Dar es Salaam, Tanzania

We found an unexpectedly high rate of treatment failure and drug resistance in this cohort of HIV-infected women initiated on ART for life during pregnancy in urban Dar es Salaam, Tanzania. The lowest proportion with virologic treatment failure (defined as ≥400 copies/mL), 16% was found 3 months post-delivery, but thereafter increased to 22% at month 6, 61% at month 12 and 86% at 24 months after delivery. Thus, the great majority of women initiated on ART during pregnancy failed to stay on treatment despite needing it for their own health. A high proportion (56%) of mothers with detectable viral load also had resistance mutations at 12 months postpartum. Despite of this, there was a significantly lower two-year maternal mortality rate in this study, 5.9% (95% CI 2.5-13.7%), compared to the 29.9% (95% CI 13.1-46.9%) mortality rate among mothers with ≤200 CD4 cells/μL at enrolment in the previous Petra PMTCT trial at the Dar es Salaam site who only received short-course ART perinatally [31]. The long-term mortality rates in the currently analyzed subgroup with such high failure rates can only be speculated upon.

A review of 89 studies on virologic outcomes among adult African patients on ART showed an overall treatment success defined as a viral load of <400 copies/mL of 78% after 6 months of ART, in 76% after 12 months and in 67% after 24 months [23], but the between study discrepancies were large, varying from 51% to 97% at 12 months and from 46% to 90% at 24 months [23]. The 6 months virologic failure rate in our study (defined as detectable viral load ≥400 copies/mL) was similar to that in the reviewed African studies [23] but higher at 12 and 24 months than in these studies [23]. One possible reason for the high virologic failure rates at 12 and 24 months postpartum in the women in the present study could be that they all participated in a PMTCT study in which women with CD4 counts ≤ 200 cells/μL who were supposed to continue with ART for life, met women with CD4 counts >200 cells/μL who stopped ART at 6 months as per study protocol. The possibility that women shared information in the waiting room and were influenced by peers to discontinue ART cannot be excluded. On the other hand, repeated counseling efforts to promote adherence to life-long ART was carried out at each follow-up visit, more so than in routine ART care. The fact that treatment failure increased after 6 months is likely related to the recommended cessation of breastfeeding at the same point in time. Efforts to adhere to ART among the Mitra Plus women put on ART for life were probably better during the breastfeeding period because women were highly motivated to prevent transmission of HIV to their infants, as supported by the drug adherence self-reports and by qualitative in-depth interviews performed to explore the reasons for treatment failure among women who were to stay on life-long ART [32].

Previous studies of HIV-infected women in resource-limited countries have found high attrition between diagnosis of HIV infection in pregnancy-related services and initiation of ART among those eligible for ART for life [33-35]. A meta-analysis on adherence to ART during and after pregnancy showed that drug adherence was higher during the antenatal period (74%) compared to postnatal period (53%) [36]. Long-term follow up in a community-based ART program in South Africa showed that women who initiated ART during pregnancy had a significantly higher risk of loss to follow-up compared to non-pregnant women [37] while another study from 6 resource-limited SSA countries found similar retention rates and CD4 count responses in HIV-infected women who initiated ART during pregnancy and other adults [38].

We defined virologic failure as a viral load ≥400 copies/mL while previous studies of virologic failure in resource-limited settings often have used a cut-off of 1000 copies/mL [23]. On the other hand, in most high-income settings, virologic failure is defined as a viral load above 20 HIV-RNA copies/mL. Had we used the higher cut-off (≥1000 copies/mL) in our study, the proportion of women with virologic failure at 12 and 24 months postpartum would have remained very high, 53% and 82%, respectively.

As expected virologic failure preceded immunologic failure as a warning indicator of treatment failure and was also associated with self-reported adherence. Several other African studies have also found that CD4 cell count monitoring is a poor indicator for treatment failure [39-43]. A study in a large ART program in the general population in Nigeria showed that immunologic criteria missed almost half of patients with virologic failure [40].

In our study, self-reported adherence correlated but also seriously underestimated the proportion with virologic failure at month 24 postpartum. Several studies have shown that virologic failure in resource-limited countries is associated with incomplete adherence [44-46] and is a strong predictor of resistance [29,44]. Virologic failure has also been shown to be associated with prior NVP-based PMTCT treatment [45,47]. In our study 34% of the women had first line drug resistance at 12 months postpartum, which corresponds to 56% of the women with detectable viral load. Although we did not conduct any baseline analysis to rule out transmitted drug resistance, all women in the Mitra Plus study were treatment naïve. However, a study involving treatment naïve HIV-infected pregnant mothers in Dodoma and youth in Dar es Salaam, showed that the prevalence of ART primary resistant mutations was 11.9% and 9% respectively [48,49]. Assuming similar proportions at baseline among treatment-naïve women enrolled in our Mitra Plus study, it is most likely that the high proportion of drug resistance detectable at 12 months was secondary to suboptimal drug adherence.

The majority of the women with drug resistance in our study had multiple resistance mutations, most commonly M184V, K103N and Y181C. The same resistant mutations were also found in infants of women in the Kisumu PMTCT study [50]. In the Malawi PEPI trial multiclass drug resistance strains were detected in 29.7% of infants of women who were initiated on antiretroviral therapy postpartum [51]. A review of acquired HIV drug resistance in adults in resource-limited settings, including six studies from SSA showed resistance in 11% of patients on ART for 12–23 months [24]. However, a high frequency of drug resistance after one year of first line ART similar to that in our study was reported from Cameroon, 46% [52] and Togo 24.5% [53]. A study of HIV-1 drug resistance after first-line ART failures in six sub-Saharan countries showed that 70% of those with virologic failure (1000 copies/mL) after one year of ART had drug resistant mutations [54]. Accumulation of drug resistance mutations may limit the response to both first and second- line ART regimens and lead to the transmission of drug resistant HIV [55].

Despite close monitoring and follow-up as part of our research study including home visits to all clients who had missed more than two consecutive appointments, 21% (15/73) of the women on ART for life and eligible for follow up after delivery did not return for their 24-month visit. Loss to follow-up and sub-optimal postpartum ART adherence has been major drawbacks to the efforts of combating MTCT of HIV in resource-limited regions [33-36]. This makes it pertinent to discuss how feasible it will be to implement the 2013 WHO guidelines to initiate all HIV-positive pregnant women on ART for life (Option B+). A study done in Malawi, which was the first country to implement the Option B+ recommendations, has shown that 17% of women were already lost to follow-up within six months post-delivery [56].

Thus, there is a clear need for rapid and affordable viral load and drug resistance testing and finding feasible ways to improve drug adherence and minimize loss to follow-up in resource-limited settings.

The main limitation of this study is the small sample size limiting us from observing significant associations so as to draw more firm conclusions. A major strength of the study is the prospective follow-up for 24 months and the availability of viral load and drug resistance information.