Our main findings from this study were: 1) among healthy, young, CMV-seropositive children, those who were shedding CMV in saliva or urine tended to continue shedding for at least three months, usually at relatively stable levels; 2) a small subset of children shed at extremely high levels, and are presumably a more likely source of transmission to pregnant women; 3) the highest viral loads (>1×106 copies/mL) were found in saliva; and 4) CMV shedding occurred at higher viral loads among younger children.
Some of the findings in this study were similar to previous observations, but several of our findings were novel. Many studies have examined CMV shedding among healthy children (reviewed in ), but few have collected longitudinal data -. We could not identify any studies that assessed CMV shedding at weekly or daily intervals as we did. In our study, we found that the presence of shedding was highly correlated over time and viral loads were relatively stable over days and weeks, even for as long as three months. This suggests that young children who are documented CMV shedders will probably continue shedding, and therefore may pose an ongoing risk to pregnant women. Our findings further suggest that even seropositive children who are initially not shedding have a reasonable chance of occasionally shedding CMV over the next few months, although usually at much lower viral loads (Figure 4). Our data suggest that children may typically shed CMV for a longer duration than adults, but it is difficult to make a valid comparison because longitudinal studies in adults have only assessed blood rather than saliva or urine . If differences in viral loads or duration of shedding exist, they could be caused by physiological differences between children and adults (e.g., intensity of viral replication, immune system control), by a higher probability of recent infection among children, or by children's more frequent exposures to re-infection from other children. Differences in CMV viral loads have also been found in rhesus macaques, where juveniles had significantly higher viral loads than adults .
Another new finding was the extremely high viral loads in the saliva of healthy young children. Although a number of studies have measured CMV viral loads among children with congenital CMV infection ,, few studies have measured viral loads among healthy children (i.e., not congenitally infected), the group most relevant for transmission to pregnant women. In fact, we were only able to identify two such studies ,, and these were smaller than our study and reported only mean viral loads rather than the entire distribution of viral loads. Nearly half of the 502 CMV DNA-positive saliva specimens in our study contained >100,000 copies/mL and more than one quarter had >1×106 copies/mL (Figure 1). In contrast, CMV viral loads in adults tend to be substantially lower (Stowell et al., companion paper) . Many of the highest viral load specimens were obtained from a relatively small number of children (Figure 3)-perhaps 6 or 7 of the 161 originally screened. This may explain in part why CMV shedding is relatively common  and yet incident infections are relatively uncommon , i.e., the probability of transmission may be low unless viral loads are high.
The higher prevalence of shedding in saliva compared to urine was somewhat surprising. Most previous studies have found higher shedding prevalences in urine, typically 2%-20% higher than in saliva . However, our large difference in shedding prevalence by specimen (40% in saliva vs. 20% in urine using the same limit of detection) cannot be compared to the results from those previous studies; our study design selected for shedders, including a preponderance of saliva shedders, and then sampled from them repeatedly. Our cross-sectional screening study, which is directly comparable to the previously published studies, showed an approximately 3% higher shedding prevalence in saliva, though the difference was not statistically significant (Stowell et al., companion paper). It is unclear why we did not find a higher shedding prevalence in urine, compared with saliva, as reported in previous studies. One possible explanation might be that our study and previous studies had important differences in laboratory procedures-we collected urine using filter paper and tested for CMV DNA using PCR, whereas all previous studies  collected liquid urine and tested for CMV using viral cultures.
On the other hand, the higher viral loads that we found in saliva compared to viral loads in urine were not necessarily unexpected. We identified two previous studies of healthy CMV seropositive children that compared viral loads in saliva to urine ,. In those studies, mean viral loads were about five times higher in saliva.
It is unclear why some seropositive children shed CMV while others did not. It is possible that a few of the younger children may have only had maternal IgG and no infection of their own. Four of the seropositive children who were not shedding at recruitment were younger than 12 months old. One of these children began shedding CMV at high levels later on, suggesting that at most three children could fall into the category of not being infected. For those who truly were infected, the occurrence of shedding may depend on the level of immune control, the particular viral strain, the time since initial infection, or less re-exposure to CMV reinfections.
Our study highlights the potential importance of saliva exposures for child-to-woman transmission of CMV. In addition to saliva sometimes having very high viral loads, women also report behaviors associated with potential exposures to children's saliva more frequently than behaviors associated with potential exposures to urine. According to a survey of more than 2000 women, kissing on the lips, sharing utensils, sharing cups, and sharing food are all common activities for women and children . Women may also be exposed to saliva while wiping children's faces. In contrast, most exposures to urine occur while changing diapers, and most women report that they clean their hands after most diaper changes . Importantly, many saliva exposures allow for direct transfer to mucous membranes, whereas most urine exposures require an intervening step involving hands, thereby reducing the chance for CMV to remain viable .
Another important conclusion from our study is that younger children may pose greater transmission risk than older children because of their tendency to shed CMV at higher levels. This is especially important since younger children also introduce more fluids into the environment (e.g., through drooling, mouthing toys, etc.) than do older children.
Thus, the framing and presentation of CMV prevention messages must take into account the potential risk posed by exposure to young children's saliva. In the past, some messages have emphasized hand washing, primarily as a way to reduce exposure to urine after diaper changing ,,. Although hand washing should remain an important part of prevention messaging, saliva exposures need to be addressed prominently. Intervention studies have shown that messages that highlight ways to reduce exposures to saliva are generally acceptable to women ,.
Our study had several limitations. First, although we had more than 800 patient-visits and nearly 1,700 specimens tested, we followed a relatively small number of children (N = 36). It is possible that the children we followed differ in important ways from the general population of healthy, CMV-seropositive children. This might matter if, because of their age, race/ethnicity, or socioeconomic status, the children in our study were more or less likely than the general population of children to be re-exposed to CMV infection. The second limitation is that our study was not designed to assess risk factors for seroconversion, only risk factors for shedding. Although we could show that saliva can have extremely high levels of CMV DNA, we did not measure the extent to which such viral loads are associated with transmission to another person. A third limitation is that due to the conditions of specimen collection (e.g., in-home collection, filter paper collection, longer-term storage for the daily specimens) we did not routinely place the specimens in viral cultures but used PCR to quantify CMV DNA; thus, the presence of infectious virions was not demonstrated in most cases. This approach would be expected to overestimate the number of infectious particles in specimens. For a small number of specimens we also performed viral cultures and often found evidence for infectious virions but did not quantify them (unpublished data). However, the study conditions were chosen by design because they were the only way we could repeatedly collect specimens at short time intervals. A fourth limitation is that saliva and urine were collected using different materials (swabs vs. filter paper), which complicated between-fluid comparisons because urine had a higher limit of detection (Figure 1). Nevertheless, viral loads were substantially higher in saliva than in urine with or without adjustment for the different limits of detection.