We read with interest the article by Strehlau et al. [1] on the diagnosis of two children infected with HIV in a recent issue of AIDS. On the basis of the findings that two breastfed children who showed positive results of nucleic acid amplification tests (NAAT) of HIV within 1 month after birth, became negative NAAT and had no detectable viral load for several months, and finally turned out to be infected with HIV with definite evidence after weaning around 1 year of age, the authors hypothesized that breast milk from the mothers of these two children had HIV-specific or HIV-nonspecific factors that caused the undetectability of HIV in these two children until cessation of breastfeeding. However, we consider that the results should be cautiously interpreted, and that the persistent negativity of repeated HIV tests was possibly true and these two children were likely infected postnatally, because of the two following reasons. First, it has been demonstrated that strong uterine contractions during delivery can cause mother-to-fetus transfusion, leading to transfer of maternal circulation cells and other constituents into fetus [2]. The initial positive results of HIV in the first month of life in these two children [1] were probably caused by the exposure to maternal HIV. In other words, the virus detected in the first month of life was derived from their mothers, but not produced by the babies themselves. The fact that neonatal use of antiretroviral agents, such as zidovudine or nevirapine can prevent mother-to-child transmission indicate that infants have been exposed to, but not yet infected with HIV. Virtually all infants born to HIV-infected mothers are exposed to the maternal virus during birth process, regardless of detectable HIV in newborns; detectability means exposure to a relatively large amount of virions, and undetectability means, a relatively small amount of virions below the detection limits. Second, these two children received breastfeeding, which can increase mother-to-child transmission of HIV [3]. Although HIV was not detectable in the breast milk supernatant of one child's mother, HIV exposure by breastfeeding was highly possible. Breastfeeding requires at least several times per day for a period of several months, during which bloody milk and nipple bleeding are not uncommon. Undetectability of HIV in one milk sample cannot reflect the actual scenarios. Additionally, as the authors mentioned, cell-associated HIV in breast milk is known to be strongly correlated with postnatal transmission [1]. Thus, although the maternal postpartum antiretroviral therapy probably delayed the transmission, the HIV infection in these two children was highly associated with breastfeeding. WHO recommends that breastfed infants should continue antiretroviral prophylaxis for 1 week after complete cessation of breastfeeding [4]. However, these two children did not continue to receive antiretroviral agent while they were breastfed. To differentiate whether these two children had been infected with HIV from early life or around 1 year of age, we consider that determination of anti-HIV titers in the longitudinal serum or plasma samples from the children can provide valuable information. As maternal IgG antibodies can transplacentally transfer to fetus and the maternal antibodies in infants can persist for as long as 18 months, anti-HIV is not tested in children less than 18 months age to define HIV infection. However, based on the half-life of maternal IgG antibodies being 21–24 days, anti-HIV titers in an infant who has not been infected with HIV will progressively decline to an expected level after a known period of time, whereas anti-HIV titers in an infant who has been infected will be relatively constant, or will initially decrease and then increase, or will present an increasing trend, depending upon the time of HIV infection. Thus, observation of dynamic changes of anti-HIV levels in these two children can differentiate the time of HIV infection, if serial serum samples collected at different times are available. Acknowledgements Conflicts of interest There are no conflicts of interest.