The study was approved by institutional review boards of Pontificia Universidad Catolica del Ecuador (Quito, Ecuador) and the Centers for Disease Control and Prevention (Atlanta, Georgia). Written informed consent was obtained from each child's parent or guardian.

The study was conducted in Quinindé, a town in Esmeraldas Province in tropical, coastal Ecuador. A diarrhea surveillance cohort of 194 children was recruited between March and December 2009 from within a birth cohort, the ECUAVIDA cohort, which is described in detail elsewhere [27]. Briefly, the cohort of 2404 children from the District of Quinindé was recruited around the time of birth at the public hospital (Hospital Padre Alberto Buffoni [HPAB]) in Quinindé between November 2006 and December 2009. A subset of the larger cohort was included in a group monitored under diarrhea surveillance: children entering the birth cohort between March 2009 and December 2009 who lived within the municipal boundaries of Quinindé were enrolled for the surveillance study and were followed up to 3 years of age. Inclusion criteria into the surveillance study were as follows: (1) healthy and aged <14 days, (2) maternal stool sample availability, (3) maternal age of ≥17 years, and (4) residence in Quinindé. Routine telephone calls were made twice weekly to mothers to determine whether the child in the surveillance sample had diarrhea, and medical records of visits to the cohort pediatric clinic at HPAB were collected. All families were advised on the importance of surveillance and were asked to report all cases of diarrhea to the clinic or directly to the study coordinator in Quinindé. During each telephone call, the mother or caregiver was also asked about any illness at present; respiratory symptoms, fever, diarrhea, or other signs or symptoms in the child or other members of the household were recorded. If diarrhea was reported, the family was asked to take the child to the clinic the same day for assessment of severity, treatment, and collection of a stool specimen. If this was not possible, a physician visited the child's house for evaluation and sample collection either the same day or the next day. Stool collection from asymptomatic children occurred during routine home visits at ages 7, 13, 24, and 36 months and during routine clinic visits at ages 3, 18, and 30 months. A child was considered asymptomatic if they did not have diarrhea for at least 2 weeks before the routine home visit. During home visits at ages 7, 13, 24, and 36 months, child length/height (measured using infantometers/stadiometers) and weight (measured using digital pediatric scales) were recorded.

All collected stool samples were aliquoted into 2-mL tubes at the FEPIS laboratory in Quinindé and stored at 4°C. RNA extraction was done using the Qiagen QIAamp Viral RNA Mini Kit, and extracted RNA was stored at −20°C prior to analysis. Real-time reverse transcription polymerase chain reaction (RT-qPCR) was used to detect GI and GII norovirus RNA [28]. All cycle threshold (Ct) values of <40 were considered positive; we also repeated all analysis with Ct cutoffs of <35 for GI viruses or <37 for GII viruses. The latter approach reduced the overall rates of infection but did not affect any of the patterns or observed associations (data not shown). Positive samples were genotyped by sequence analysis [29].

A diarrhea sample was defined as a stool sample obtained from a child who had ≥3 liquid or semiliquid stools in 24 hours, with a time of collection of <14 days before the telephone call or consultation with the clinic/hospital. When a child presented to the clinic with diarrhea, stool samples and clinical data were collected after physician evaluation. Norovirus diarrhea was considered an episode of diarrhea in which norovirus RNA was detected in stool by RT-qPCR. Asymptomatic norovirus infection was defined as a stool sample from an asymptomatic child in which norovirus was detected.

ABO blood group determination of capillary blood samples was done by agglutination with standard antisera. To ascertain secretor status, saliva samples were collected from children >3 years of age into Oragene Discover OG575 kits (DNA Genotek, Canada), and DNA was extracted by following the manufacturer's instructions. Secretor status genotyping was determined by PCR amplification and pyrosequencing of FUT2 at nucleotide 428 (G>A) as previously described [6].

The prevalence and 95% confidence intervals (CIs) of norovirus infection among diarrhea cases and asymptomatic control samples was estimated according to a binomial distribution. We calculated the cumulative incidence of diarrhea, norovirus infection, and disease by Kaplan–Meier survival analysis. Cumulative incidence estimates were generated at 12, 24, and 36 month of age.

The follow-up time for an individual child was divided according to the number of previous infections. For example, the period between birth and first infection was identified as 0 previous infections and the period between first infection and second infection was identified as 1 previous infection. The number of previous infections was then modeled as a time-varying covariate. To examine the association of previous infection episodes, individual-level, and family-level characteristics with norovirus infection and disease risk, we fitted multilevel mixed-effect Poisson regression models. The individual child (with multiple observation periods) was considered the level for the random effect under the assumption of independent covariance. We calculated P values for associations with norovirus infection and disease rates and genogroup-specific infection rates by the Wald test. All regression analyses were controlled for age (in 6-month intervals up to 1 year and in yearly intervals thereafter) and sex.

A total of 194 children were recruited and followed up in the surveillance cohort. The 194 children were followed for 206 658 person-days. The range of follow-up was 211–1305 days, with >90% of children followed for at least 2 years. The numbers of children followed to at least 6 months, 1 year, 2 years, and 3 years of age were 194 (100%), 193 (>99%), 183 (94%), and 135 (70%), respectively. A total of 151 children (77%) were recruited on the first day of life, and all were recruited before day 14 of age. Relevant characteristics of study children, their mothers, and their living environment are provided in Table 1. All children lived in an urban environment in the town of Quinindé and tended to be of higher socioeconomic level than the cohort in general (50% were defined as being in the high socioeconomic tertile). A high proportion of children lived in a household with at least 2 other children of households (88 [45%]) and had a flushing toilet (122 [63%]), but relatively few had access to potable piped water (25 [13%]). Blood group and secretor genotypes were determined at 3 years of age in 178 children (92%; Table 1): most children (111 [62%]) were blood group O, and 158 (88%) were secretor positive. The proportions, by FUT2 genotype, were as follows: SeSe (homozygote secretor-positive children), 79 (44%); Sese⁴²⁸ (heterozygote secretor-negative children): 79 (44%), and se⁴²⁸se⁴²⁸ (homozygote secretor negative children): 21 (12%). Table 1.

Characteristics of Surveillance Birth Cohort (N = 194)

Over the 3-year study period, 1454 stool samples were collected, of which 438 (30%) were collected during an episode of acute diarrhea, while 1016 (70%) samples were collected from children without diarrheal symptoms for at least 14 days. There was a median of 7 samples (range, 1–20 samples) collected per child. There were no major demographic differences in children with ≥7 samples collected, compared with those with <7 samples collected, with the exception of children with older mothers, who had fewer samples collected (P = .008).

Over the 3-year period, we detected a mean (±SD) of 2.26 ± 2.30 diarrheal cases per child (range, 0–12 cases): 44 children (23%) had 0 episodes of diarrhea, 85 (44%) had 1–2 episodes, 50 (26%) had 3–5 episodes, and 15 (8%) had >5 episodes. The cumulative incidence of diarrhea in the surveillance cohort, by number of episodes over the 36 months of follow-up, is shown in Figure 1A. Figure 1.

Cumulative incidence of diarrhea (A), norovirus infection (B), and norovirus diarrhea (C) from birth to age 36 months. First event, black; second event, red; third event, green; fourth event, blue; and fifth event, orange.

A total of 154 samples (61%) were GII positive, 85 (34%) were GI positive, and 13 (5.2%) had GI/GII mixed infections. Factors associated with incidence of infection with GI and GII norovirus are shown in Tables 3 and 4. There was a tendency for higher risk among secretor-positive children for GII infection (although the difference was not statistically significant) but no apparent association with GI, after control for age and sex. Similarly, previous norovirus infections, regardless of genogroup, were not associated with the incidence of subsequent infection. The only factor associated with infection was the reduced incidence of GII infections observed during the wet season (December through June; P < .001). An association in the opposite direction was observed for GI, but this was nonsignificant. Table 3.

Association Between Previous Infections, Sociodemographic Characteristics, and Host Characteristics and Norovirus Genogroup I and II Infection Risk

We were able to determine the genotype for 106 samples (Figure 3). The most common genotype was GII.4 (27 infections [25%]), followed by GII.1 (20 [19%]), GI.3 (9 [8%]), GII.2 (11 [10%]), and a tentative new GII genotype (GII.23; 6 [6%]). We observed 12 GI/GII mixed infections (11%), including 4 due to GII.4 (4%). Among the GII.4-positive samples (including those from mixed infections), the following variants were observed: New Orleans 2009, 21 samples; Den Haag 2006, 5; Sydney 2012, 4; and Osaka 2008, 1. The genotype distribution among diarrheal stools and healthy stools was similar. Figure 3.

Norovirus genotype distribution of infections in secretor-negative (A) and secretor-positive (B) children. Abbreviation: G, genogroup.