The rate of rotavirus detection was higher by qRT-PCR than by EIA. Rotavirus was detected by qRT-PCR in fecal specimens from an additional 10% of children with AGE who tested negative by EIA. However, several lines of evidence suggest that rotavirus detected by qRT-PCR alone might not have been the causative agent in some patients with AGE. First, C_t values of fecal specimens from children with AGE for whom rotavirus was detected only by qRT-PCR were significantly higher (lower viral loads) than C_t values of specimens from children for whom rotavirus was detected by EIA (36 vs. 18). Second, full genotypes could not be determined for most (88%) specimens for which virus was detected by qRT-PCR only, probably because of the low level of viral shedding. Last, rotavirus vaccine showed limited effectiveness against virus identified by qRT-PCR alone, but this result may partially be a function of the small number of cases in this group. In contrast, 3-dose vaccine effectiveness was high (83%–84%) for children whose samples were positive by EIA and comparable to vaccine effectiveness determined by prelicensure trials and other case–control studies in the United States that similarly identified rotavirus-positive cases by EIA (11–13). These findings, together with the easier implementation of commercial EIAs than qRT-PCRs, support the use of EIA for identifying cases and controls to estimate vaccine effectiveness, even though a few rotavirus infections might be missed by the EIA, particularly in specimens for which C_t values are high (low viral loads). qRT-PCR may be useful for identifying cases and controls during vaccine effectiveness studies. However, because such assays also detect low levels of rotavirus circulating in the population but not associated with illness, further work is needed to define a cutoff C_t value below which the detected virus is likely to cause illness. This C_t value would help to identify the few cases missed by the EIA and exclude cases with low-level background shedding.

Previous studies have compared different methods of detecting rotavirus in fecal specimens. These studies should be directly compared with caution because they used different commercial assays and different PCR techniques; however, trends in patterns of detection can be compared. Similar to researchers in the United Kingdom, we found significantly lower C_t values (higher viral loads) in fecal specimens from patients with AGE that tested positive for rotavirus by EIA than in qRT-PCR–positive specimens from patients with AGE whose feces tested negative for rotavirus by EIA or from healthy controls (7). Among UK children <5 years of age who had AGE, use of conventional RT-PCR increased the rotavirus detection rate from 17% by ELISA to 54% by PCR (6). However, rotavirus also was detected in 23% of healthy controls by PCR, compared with only 1% of those in whom virus was detected by ELISA. In a study in the United States, rotavirus detection rates for patients with AGE were similar between conventional RT-PCR and EIA (53% and 49%, respectively), but in 18% of healthy controls, rotavirus was detected by conventional RT-PCR, whereas no healthy controls were positive by EIA (4). We found a much lower rotavirus detection rate among healthy children (5%) than was found in the previous studies in the United States and United Kingdom (18% and 23%, respectively). The low detection rate among healthy controls in our study also might be partially attributable to the eligibility criteria for healthy controls that required a child to be 14 days without AGE before enrollment and the fecal specimen obtained within 5 days enrollment. In addition, unlike the previous studies, our study was conducted after the introduction of rotavirus vaccine into the US immunization program at a time when rotavirus activity had declined substantially (22–24). The lower detection rate of rotavirus in healthy children in our study may reflect this decrease in rotavirus activity after vaccine introduction; that is, fewer children may have been shedding virus from a previous infection, some may have had an asymptomatic infection, and infected children who had been vaccinated were possibly clearing the virus more quickly.

We detected vaccine virus in 2% of healthy controls, all of whom were either unvaccinated or had not been vaccinated within 4 weeks before illness onset; the source of vaccine virus for these children is unknown. These vaccine strains were detected only by qRT-PCR because no healthy children in whom a vaccine strain was detected were positive for rotavirus by EIA. RV5 virus also was detected in the fecal specimen from 1 unvaccinated child with AGE; the source of vaccine virus for this patient was a recently vaccinated sibling, as described (21). This symptomatic patient was positive by both EIA and qRT-PCR.

Our study had some limitations. First, if children were seen for medical care late in their illness or if specimen collection was delayed, rotavirus might have been the cause of symptoms in some children whose specimens tested negative for rotavirus by EIA but showed low levels of qRT-PCR–detected virus. However, 99% of EIA-negative specimens that had low levels of qRT-PCR–detected virus were collected from children within 7 days after they were brought for treatment, and RV5 was not effective against AGE detected by qRT-PCR only, arguing against this possibility. Second, an internal positive control was not used in this study to monitor for false-negative qRT-PCR results possibly resulting from PCR inhibitors in feces that were carried over into the RNA extracts. We believe that the numbers of such samples would have been small because we detected only 1 EIA-positive, qRT-PCR-negative sample in this study. Third, the enrollment of some healthy controls during an immunization visit resulted in oversampling of children shedding vaccine virus. Although we excluded recently vaccinated children in whom vaccine virus was detected, all detected viruses had to be sequenced to identify children who were shedding vaccine virus. However, in a true random sample of healthy children, we would expect some children to be recently vaccinated, so we might have underestimated the proportion of healthy children in whom vaccine virus can be detected. Last, because these data are from an industrialized country in which rotavirus vaccination is routine, our findings might not apply to developing countries where the severity of infection, rates of asymptomatic viral shedding, and performance of the EIA may differ.

In conclusion, our study, which was performed after rotavirus vaccine was introduced, supports the use of EIA for vaccine effectiveness evaluations in patients with AGE, even though EIA may fail to detect some true rotavirus shedding at lower levels. Although qRT-PCR increases the sensitivity of rotavirus detection, some of these cases may be in children with low-level viral shedding from a resolved or asymptomatic wild-type rotavirus infection and not true disease. The use of qRT-PCR with a cutoff C_t value should be further examined as a possible diagnostic tool in a range of settings, including in developing countries.