This analysis was based on a group of HIV-infected pregnant women on HAART at conception but with detectable VL. Most of our study population could be considered as having sub-optimal viral suppression [12
], as nearly two-thirds had been on HAART for at least 6 months, theoretically sufficient time for achievement of viral suppression [20
]; however, three-quarters had baseline VL below 3.93 log10
and could thus be classified as having partial control of HIV replication (complete suppression of VL is the aim for PMTCT). Although most women remained on their pre-pregnancy HAART regimens throughout pregnancy, a quarter interrupted treatment in early pregnancy and a further fifth changed their HAART regimens. Our initial ITT analysis indicated a constant VL for the first 10 weeks of pregnancy, with a subsequent significant decrease until delivery of around 13% per week. As the ITT model included women who switched or interrupted HAART, potentially resulting in improved virological control or virological rebound respectively [13
], we ran further models excluding the HAART modification group and/or the interruption group, with estimated significant declines of between 7% and 15% log10
copies/ml per week. In all models, NNRTI-based HAART at conception was associated with significantly lower VL throughout pregnancy.
] recommend that HIV-infected women planning a pregnancy are offered pre-conception counselling, which provides the opportunity to switch from antiretroviral drugs that may be associated with toxicity [12
] and to optimize treatment through adherence assessment/support and/or regimen modification in order to attain a stable, maximally suppressed VL prior to conception. However, a considerable proportion of pregnancies among HIV-infected women are unintended. European studies have indicated prevalences of unplanned pregnancy of 51% to 58% [10
] and pre-conception counselling rates of only 25% among HIV-positive women [24
]; in the USA, underuse of effective contraception has been documented among HIV-positive women [25
] and 83% of pregnancies in a study of HIV-positive youth were unplanned [26
]. It was not recorded whether the women in our study had planned their pregnancies, but it is likely that a substantial proportion were unintended, consistent with their detectable VL.
We showed a significantly declining VL during pregnancy, both for women remaining on their conception regimens and for those switching HAART regimens. We have previously documented declining VL in untreated women from our study, which may in part be due to pregnancy-related haemodilution, which could also explain some of the VL decline seen here [19
]. For the women remaining on the same HAART, the VL decline may be partly explained by improved adherence after awareness of the pregnancy, but this could not be verified as adherence data were unavailable. Three-quarters of pregnant women in the American PACTG 1025 study reported perfect adherence to HAART (no missed doses 4 days before all study visits), with those antiretroviral-naïve before the pregnancy more than twice as likely to be perfectly adherent compared with antiretroviral-experienced women [27
]. Women in our study with existing adherence problems may have been motivated to improve adherence once aware of their pregnancy and/or may have received more intensive/effective adherence support from their care providers compared with that received before pregnancy [28
For the fifth of women who switched HAART regimens, their HIV physicians may have determined the need for a switch through adherence assessment and resistance testing, in accordance with clinical guidelines which state that women on failing HAART should have their regimens changed to maximize the likelihood of achieving undetectable VL by delivery [12
]. This group may also include women for whom the caring physician may have had more concerns regarding viremia control and immunosuppression, suggested by higher VL in women with lower CD4 counts. Of note, treatment modifications took place at a median of 21 weeks gestation, possibly due to delayed notification of pregnancy to a woman’s HIV physician and/or testing lag times. In the non-pregnant HIV-positive population, delays between detection of virological failure and switching to a new regimen are not uncommon, with a six month delay reported for a third of such patients in a recent UK national audit [31
]. The lack of a significant difference between the switching and the continuing HAART groups in terms of HIV RNA slopes in model two was confirmed by the almost identical slope in the model which excluded the modifying HAART group (model three). The fact that treatment change did not occur until around half-way through pregnancy may have contributed to this finding. Women undergoing treatment modifications had significantly lower baseline CD4 counts than those continuing with the same HAART regimen (on average over 100 cells/mm3
lower), most likely reflecting treatment guidelines and an example of confounding by indication [13
]. Although we adjusted for baseline CD4 count in the model, we lacked data on drug resistance and could not investigate whether women switching treatment not only had poorer immune status, but also more resistance than those remaining on the same regimens.
Two key factors predictive of HIV RNA levels in pregnancy were baseline CD4 count and type of HAART at conception. In the ITT model, women with CD4 counts <500 cells/mm3
had significantly higher VL than other women, although in subsequent models this lost significance for the women with moderate CD4 counts (200-499 cells/mm3
). Women conceiving on NNRTI-containing regimens had consistenly lower VL than those on PI-containing HAART. We previously showed among antiretroviral-naïve women that those initiating HAART with a NNRTI-based regimen reached an undetectable VL more rapidly than those starting with a PI-based regimen (mainly non-boosted) [15
]. Several studies have indicated that very high adherence levels are required to maintain virological suppression on PI-containing regimens (above 90%), higher than those required for NNRTI-containing HAART [32
]; this could help explain our finding here that women conceiving on PI-based HAART had significantly reduced slope of VL decline versus those on NNRTI-based HAART. A further potential explanation of the differences by HAART type could relate to pharmacokinetics: some studies have reported low concentrations of nelfinavir and other PIs in the third trimester [35
], while studies on plasma nevirapine concentrations have generally shown no significant differences between pregnant and non-pregnant women [38
Treatment interruption in pregnancy is not recommended by current guidelines as it may result in viral rebound and subsequent increased risk of immune and/or clinical deterioration, in addition to an increased MTCT risk [12
]. A recent study from Italy examining the impact of interruption of therapy during pregnancy, found that interruption in the first trimester was associated with a 10-fold increased risk of MTCT, after adjusting for maternal VL, type of therapy and other factors [39
]. Most women who interrupted therapy in our study population did so in the earlier years of the study [data not shown] and it is likely that this practice is becoming increasingly rare, underscored by the recent Italian results. Switches away from efavirenz due to concerns regarding terato-embryogenic toxicity risk are often advocated for pregnant women [12
] and were also seen here.
Study limitations include the observational nature of the data and lack of information on adherence, drug resistance, HIV subtype and immunological and virological patterns before pregnancy. Our ability to explore potential reasons behind the similar VL declines in the women who switched treatment and in those who remained on their conception regimens was therefore constrained. A limitation of our analyses was the inability to adjust for baseline VL, as women had their first measurement at different gestational ages. However the statistical approach used was able to account for the variability in intercepts between women which should reduce some of the bias from not adjusting for the baseline VL implicitly. However we modelled appropriately the left-censored HIV RNA measurements and avoided imputation with midpoints, which can result in biased parameter estimates and their standard errors [18
]. An additional limitation was a relative overestimation of the effects of drugs that were used in HAART regimens in pregnant women during earlier years (e.g. nelfinavir), which have been largely substituted by more potent PIs today. Although antiretroviral-experienced, our study population had not yet accumulated long durations of treatment and most had partial control of viremia and thus the generalizability of our findings, for example, to highly treatment experienced pregnant women with high VL at conception, remains uncertain.
Our findings come from a “real life” setting, and indicate a variety of practices regarding the management of women conceiving on HAART with detectable VL in this Western European setting. The existence of this group of pregnant women highlights the need for improved monitoring of and support for treated women before they become pregnant, as well as during pregnancy itself. Clinical concerns during pregnancy include attempts at improving virological control to avoid MTCT and improve maternal health, whilst minimizing potential adverse effects on the fetus and mother, including the risk of exposure to potentially teratogenic drugs. The MTCT rate in our study population was just over 1%; concern regarding the risk of pregnant women on failing HAART developing drug resistance, which could potentially be vertically transmitted [40
] is a key factor behind recommendations for switching regimens as soon as virological failure is determined. However, there is a lack of consensus regarding when to change HAART for virological failure in non-pregnant adults [13
]; few studies are aimed at evaluating optimal antiretroviral management during pregnancy for women on HAART at conception, while clinical trial data on efficacy of different HAART regimens in pregnancy are lacking. This context may partly explain the finding here that only a fifth of women underwent a treatment switch in pregnancy. Future controlled studies are needed to obtain information on mechanisms for VL decrease and on optimal clinical management of HAART in pregnant women, who are already on treatment at conception, particularly those with accumulated resistance.