In-utero exposure to tenofovir is associated with impaired fetal and infant growth

The successful roll out of antiretroviral therapy (ART) regimes to prevent mother-to-child transmission (MTCT) of HIV has resulted in a significant decrease in the number of newly infected children. This major success of MTCT now confronts us with the next important challenge, namely to determine the optimal MTCT regime with the least adverse effects of ART for mother and child [1]. In our study, we started to address this challenge and detected a significant association between prenatal tenofovir exposure and infant growth in HIV-exposed uninfected (HEU) infants [2]. Given that pre and postnatal growth are strongly associated with infant, and adult health, we considered it important to report our results to the scientific community and contribute to the body of knowledge in the field of fetal and early life exposure to HIV and ART. Because AIDS requested a concise version of the submitted manuscript for publication, the response to the correspondence of Le Roux et al.[3] allows us to complement the initial publication with further details on the cohort, including the treatment regimens of the HIV-infected mothers, the use of WHO Child Growth Standards and selection of multivariable linear regression models to examine long-term associations between tenofovir and growth parameters. Mann–Whitney U- and Fisher's exact tests were used for comparisons of, respectively, continuous and categorical data described below. As treatment regimens have changed during the course of the epidemic, Le Roux et al.[3] are concerned about a potential cohort effect. The median date of birth was for both groups was in 2006 (P = 0.39), indicating a lack of potential confounding by the date of inclusion into the study. Le Roux et al.[3] further enquire whether the effect of tenofovir may be confounded by simultaneous exposure to protease inhibitors, as protease inhibitors in general have been implicated with premature delivery [1] and lopinavir boosted ritonavir potentially with impaired bone maturation. In our cohort, tenofovir was less frequently administered together with a protease inhibitor (33.3%) than non-tenofovir regimens (79.6%, P = 0.008). Furthermore, 11% of the regimens containing tenofovir included lopinavir-boosted ritonavir compared with 30.5% of the ART regimens without tenofovir. Collectively, these observations demonstrate the absence of increased use of protease inhibitors or lopinavir-boosted ritonavir in the infants exposed to tenofovir. With regards to gestational age, we did not detect a significant difference between tenofovir-exposed (median = 38 weeks (interquartile range 38–39 weeks) and non-tenofovir-exposed infants (median = 39 weeks, interquartile range 37–40 weeks; P = 0.7). As mentioned by Le Roux et al.[3], recently, the INTERGROWTH-21st project has provided gestational age-related references particularly relevant to studies investigating preterm infants [4]; and prematurity was relatively low in our cohort (12%). Taken together, we believe that the use of the WHO references for weight and height used in our study and the cut off of 2500 g for low birth weight is appropriate, and will increase accessibility to the data [5] as these references continue to be widely used in high HIV prevalence settings. Last, Le Roux et al.[3] comment on a potential reverse effect by adjusting for both prematurity and low birth weight in the multivariable model. We included these variables in the multivariable model investigating long-term postnatal growth to avoid an overestimation of the effect of tenofovir on postnatal growth because of the effect of tenofovir on birth weight. We have repeated the analyses excluding the variable indicating low birth weight as suggested by Le Roux et al.[3] In this new multivariable model, which includes prematurity, maternal CD4+ cell count, the coefficient associating tenofovir exposure with weight-for-age, changes from −0.72 to −0.93 (P = 0.005) indicating an increase in the effect of tenofovir on growth. Similarly, when excluding birth weight from the multivariable model assessing the effect of tenofovir on height-for-age, the coefficient increases from −0.87 to −1.09 (P = 0.004). We conclude therefore that, the multivariable model, as published, did not result in an overestimation or reversal of the effect of tenofovir and is in our view the more conservative approach. As described in the initial manuscript, there have been conflicting reports regarding a potential effect of tenofovir exposure on growth in HEU infants [6–9], which we postulated may be because of differences in patient demographics, including diet and HIV disease. Recently, Jacobson et al.[10] published a large study describing the absence of an effect of tenofovir on height and weight in the offspring of HIV-infected women living in the United States at 2 years of age [10]. Interestingly, in their study prepregnancy BMI was 29.8 kg/m2[10]. In a recent substudy in our cohort, prepregnancy median BMI of the HIV-infected women was 23.7 kg/m2, and none of the women was diagnosed with gestational diabetes and smoking was ceased upon diagnosis of pregnancy (unpublished). As obesity, smoking, and gestational diabetes are major variables impacting fetal growth, more subtle effects of tenofovir may be overshadowed in cohorts where these factors are highly prevalent. We further underscore the possible role of unidentified confounding factors in the title of our publication by stressing the ‘need for follow-up studies’ [2], which is particularly relevant given our limited study size. Disentangling early life exposures that determine health later in life is complex, with, for example, infants with a lower birth weight being at risk for rapid catch-up growth and obesity later in life. We believe that studies investigating larger and long-term cohorts, taking into account confounding factors including diet and BMI, are needed to determine the effects of ART compounds on the growth of HEU infants and to identify the optimal regime to prevent HIV infection in infants. We hope that this report, and the ensuing critical discussion, will accelerate this important process, and thank Le Roux et al.[3] for their comments that enabled us to provide additional information. Acknowledgements Conflicts of interest There are no conflicts of interest.