Enteropathogen Changes After Rotavirus Vaccine Scale-up

This study was approved by institutional review boards of Instituto Nacional de Salud del Niño, Universidad Peruana Cayetano Heredia, and Asociación Benéfica PRISMA, in compliance with all applicable regulations governing the protection of human participants. Research performed at Naval Medical Research Unit No. 6, Lima, Peru (NAMRU-6) as part of this study was reviewed by the NAMRU-6 IRB and received non-human subjects determination. Parents or legal representatives of all participants provided written informed consent.

Between October 31, 2013, and May 31, 2015, we prospectively enrolled children younger than 5 years accessing care at the national children’s hospital in Lima, Peru, for acute gastroenteritis (cases) or reasons other than gastroenteritis (hospital-based controls). This hospital accepts all children from Lima and its peri-urban surroundings; admissions reflect the demographics and socioeconomic status (SES) of the catchment area. To expand enteropathogen characterization and genetic surveillance among asymptomatic children, we supplemented hospital-based controls with healthy children (community-based controls) from a community cohort¹²  within the hospital catchment area. We defined acute gastroenteritis as the onset of diarrhea or vomiting within 3 days preceding enrollment.¹³  Diarrhea was ≥3 liquid or semiliquid stools within 24 hours¹²  or, for infants younger than 2 months, based on caregiver assessment.¹²  An episode ended after 2 consecutive days without gastroenteritis. Pathogen-associated gastroenteritis occurred when a pathogen was identified in a diarrheal stool, described below. Exclusion criteria were hospitalization for ≥1 month at birth, any congenital defect, birth weight <1500 g, or diarrhea within 30 days preceding enrollment. At enrollment, trained health care workers administered standardized surveys in which participant demographics and SES were assessed by using a modified Progress out of Poverty Index¹⁴ ; obtained history of present illness; determined Rotarix vaccination status by record review; collected anthropometric data; conducted physical examination; measured serum hematocrit; performed ABO and Rhesus typing (Diagast, Loos, France); and collected stool (whole or rectal swab). After enrollment, a study nurse called caregivers daily to prospectively record clinical features of illness, including Clark and Vesikari severity indicators,¹⁵  described below, until 48 symptom-free hours.

We stored stool suspensions (10% wt/vol) at −80°C, prepared in phosphate-buffered saline and thawed at 4°C. We extracted RNA (Qiagen QIAmp viral RNA kit, Valencia, CA), and tested for norovirus GI and GII by real-time polymerase chain reaction (RT-PCR).¹⁶  Samples with a positive result (GI and GII cycle thresholds: 37 and 39, respectively) underwent conventional polymerase chain reaction, amplicon gel purification, capsid gene region C sequencing, and genotyping by using the NoroNet sequence-typing tool.¹⁷  We detected and genotyped sapovirus, aichivirus, and astrovirus as previously described.^18–20  To identify rotavirus, we conducted quantitative RT-PCR on thawed stool specimen suspensions, targeting an 87-bp fragment of the rotavirus group A NSP3 region, and genotyped rotavirus-positive samples using seminested RT-PCR, identifying glycoprotein and protease-cleaved protein genotypes.²¹ 

We cultured and isolated E coli, Salmonella, Shigella, Yersinia, and Campylobacter spp. from fresh stool specimens transported in Cary-Blair medium to Naval Medical Research Unit No. 6.²²  We cultured Campylobacter spp. on Brucella agar with defibrinated sheep blood and Butzler-selective and growth supplement, incubating at 42°C for 48 hours in microaerobic conditions (5% oxygen, 10% carbon dioxide, and 85% nitrogen).²³  We performed polymerase chain reaction for heat labile (LT+) and heat stable (ST+) ETEC, diffuse-adhering E coli (DAEC), enteroaggregative E coli (EAEC), enteropathogenic E coli (EPEC), enteroinvasive E coli (EIEC), and Shigatoxin 1 (STX1)– and Shigatoxin 2 (STX2)–producing E coli on 5 lactose-fermenting colonies morphologically resembling E coli.²²  We identified Plesiomonas spp. using Hugh’s method.²²  We isolated Vibrio cholerae using thiosulfate citrate bile salts sucrose agar.²²  We performed microscopy for ova and parasites on saline wet preparations from fresh stool specimens at Universidad Peruana Cayetano Heredia, using ether sedimentation to identify Giardia spp.²⁴ 

We analyzed data in Stata 15.0 (Stata Corp, College Station, TX), considering P values <.05 statistically significant. We calculated summary statistics (means, SDs, ranges, and percentages, as appropriate) of participant demographics, diarrhea risk factors, and clinical features of illness using World Health Organization (WHO) child growth parameter calculations,²⁵  adding 5% weight to children with moderate and severe diarrhea to adjust for dehydration. We compared continuous variables using the 2-tailed Student’s t test and proportions using Fisher’s exact test, evaluated the association of gastroenteritis with specific pathogens using simple and multiple logistic regression, and estimated the pathogen-specific gastroenteritis burden by calculating the population attributable fraction (AF) using adjusted odds ratios and pathogen prevalence among cases.²⁶ 

During the 18-month study period, we enrolled 1788 participants: 932 medically attended children with gastroenteritis and 856 controls (400 hospital based and 456 community based) without gastroenteritis. Of children with gastroenteritis, 666 had both diarrhea and vomiting, 164 only diarrhea, and 102 only vomiting. Relative to controls, enrollment of children with gastroenteritis was significantly higher from December through May. Children with gastroenteritis and controls had similar blood group and Rhesus factor prevalences. Hospital-based children with gastroenteritis were significantly more likely than controls to have higher SES, improved sanitation, improved water sources, and rotavirus immunizations but had significantly lower mean hematocrit levels, weight-for-height z scores, and weight-for-age z scores. z Score differences resolved after adding 5% of weight to individual case weights for children with moderate to severe dehydration (Table 1, Figs 1–3).

We identified at least 1 organism in 74% of case stools and 42% of control stools. We identified ≥2 organisms in 30% of cases (215 had 2; 53 had 3; 5 had 4; and 2 had 5) and in 8% of controls (30 had 2; 2 had 3; and 1 had 4). Relative to controls, case stools had a significantly higher prevalence of norovirus, rotavirus, sapovirus, EAEC, ETEC, astrovirus, Campylobacter spp., and Shigella spp. Relative to cases, control stools had significantly higher prevalences of Giardia, Entamoeba coli, and Endolimax nana. Norovirus GII was associated with gastroenteritis in all age groups, whereas rotavirus predominated among children aged 12 to 23 and 24 to 60 months, sapovirus among children aged 24 to 60 months, and DAEC, ETEC (LT+), and ETEC (ST+) among children aged 24 to 60 months. Adjusting for age, sex, and SES, multiple logistic regression revealed significant positive associations between gastroenteritis and Shigella spp., norovirus GII, rotavirus, sapovirus, ETEC with LT and/or stable or stable, and astrovirus. Parasite presence was associated with a lower risk of acute gastroenteritis (Tables 1 and 2, Supplemental Table 6).

The Vesikari (20 points total: <11 nonsevere and ≥11 severe) and Clark (24 points total: ≤8 mild, 9–16 moderate, and >16 severe) scales defined 701 (75%) and 9 (<1%) of the cases, respectively, as severe (P < .001). All cases defined as severe by the Clark scale were also defined as severe by the Vesikari scale. However, only 9 (1%) of the Vesikari severe cases were severe on the Clark scale, whereas 653 (93%) were moderate and 40 (6%) were mild. Both mean scores were significantly higher for norovirus GII–positive cases relative to norovirus GII–negative cases (Vesikari [12.7 vs 11.8; P < .001] and Clark [11.7 vs 11.4; P = .016]). No other pathogen revealed a severity difference between children with gastroenteritis infected with that pathogen versus remaining children with gastroenteritis. Children with gastroenteritis identified in the 6- to 12-month age range had significantly higher mean severity scores on both scales relative to children with gastroenteritis in other age groups (Vesikari [12.7 vs 12.0; P = .0002] and Clark [12.0 vs 11.4; P = .0016]). There were no differences in severity based on other characteristics among children with gastroenteritis (Table 3).

We successfully genotyped 282 of 337 (84%) of norovirus and 65 of 71 (92%) of sapovirus isolates from children with gastroenteritis; 31 of 32 (66%) of norovirus and 6 of 6 (100%) of sapovirus isolates from hospital-based controls; and 51 of 74 (67%) of norovirus and 11 of 20 (55%) of sapovirus isolates from community-based controls. The most frequently observed norovirus genotype was GII.4 (68%), followed by genogroup II genotype 6 (GII.6) (10%) and GII genotype 17 (9%). Among those with genotype information, norovirus GI genotype 1, GI genotype 3, GI genotype 5, GII.4, GII.6, GII genotype 12, GII genotype 14, and GII genotype 23 were significantly more frequent in patients. Norovirus GII genotype 8 and GII genotype 13 were only in controls.

Of the 4 sapovirus genogroups, we most frequently observed GI (49%), followed by GII (30%), GIV (11%), and GV (10%). Among those with genotype information, 65 were children with gastroenteritis and 16 were controls. We found GI, GII, GIV, and GV more frequently in children with gastroenteritis (4%, 2%, 1%, and 1%, respectively) than in controls (1%, 0.5%, 0.4%, and 0.2%, respectively). Of the 19 known genotypes, we identified 12. Of these, we most frequently identified GI genotype 2 (32%), GI genotype 1 (15%), GII.4 (17%), GIV genotype 1 (9%), and GV genotype 1 (9%). GI genotypes 1, 2, 3, and 5; GII genotypes 1, 2, 4, 5, and 8; GIV genotype 1; and GV genotype 1 were more frequent in patients than in controls. GI genotype 6 was more frequent in controls.

Rotavirus infections peaked during winter (July to December) (Fig 1). Of 127 rotavirus-positive samples (98 children with gastroenteritis and 28 controls), 81 (64%) had typeable G and P types. Among children with gastroenteritis with typeable rotavirus, we most frequently identified the partially heterotypic strain G12P[8] (54 of 81, 67%). Of these 54, 3 (6%) received only 1 Rotarix dose and 45 (83%) received both doses. Among the 9 children (2 children with gastroenteritis and 7 controls) infected with G1P[8], the fully homotypic strain, 4 (4 controls) had received only 1 dose and 4 (2 children with gastroenteritis and 2 controls) had received both doses of Rotarix. Receiving 1 Rotarix dose resulted in a 1/2-point decreased diarrhea severity trend compared with being unimmunized (median Vesikari score 12.5 vs 13 [both severe]; median Clark score 11.5 vs 12.5 [both moderate]). Likewise, 2 Rotarix doses resulted in a 1-point decreased severity trend in rotavirus-associated cases compared with being unimmunized (median Vesikari score 12 vs 13 [both severe]; median Clark score 11 vs 12 [both moderate]).

This study builds on the findings of the landmark GEMS and MAL-ED study, filling knowledge gaps related to the enteropathogen-specific disease burden, prospectively evaluated diarrhea severity, calicivirus and rotavirus genetic characterization, and rotavirus vaccine impact in hospitalized children in a pathogen-diverse, urban South American setting after rotavirus vaccine implementation.

Six enteropathogens, predominantly viruses, contributed significantly to pediatric medically attended gastroenteritis in this setting. Norovirus GII, rotavirus, sapovirus, and astrovirus bore 47% of the pathogen-attributable burden. Shigella and ST+-producing ETEC, the only bacteria independently associated with gastroenteritis in this population, had a combined AF of only 4.4%. Other organisms, such as Campylobacter spp., Cryptosporidium, Aeromonas, and Vibrio cholerae O1, associated with moderate to severe diarrhea in multicenter studies conducted in Africa and Asia,³  were remarkably not attributable to gastroenteritis in this population or absent entirely (eg, cholera, which Peru curbed with chlorination in the 1990s).²⁷ 

Despite limited detection sensitivity,²⁸  our hospital-based data from Peru attributed the highest burden of hospital-based disease to a similar set of pathogens associated with community-based disease in MAL-ED. The notable exceptions, contributing to community but not hospital diarrhea, were Campylobacter spp. and Cryptosporidium spp. In our hospital population, Campylobacter spp. were twice as prevalent in diarrhea stools relative to controls but failed to maintain independent association when coinfections were considered. Cryptosporidium spp. were also not attributable to diarrhea in our hospital-based population; the nonmolecular MAL-ED analysis associated them more strongly with diarrhea in the first, relative to the second, year of life, and molecular reanalysis of these data yielded a higher attributable incidence. This difference could be due to our detection methods, sample size, or population differences.²  Among our hospital cases, the burden attributable to the predominant pathogen, norovirus GII, was high (29%), compared with only 5% in the community, suggesting it could cause more severe disease in Peru. Norovirus’s importance as a diarrheal pathogen, particularly in Peru, is supported by MAL-ED, which identified norovirus diarrheal stool prevalences ranging from 7% in Tanzania to 33% in Peru and nearly all children in the community in Peru experiencing norovirus-associated diarrhea before age 2.⁴  Sapovirus, attributed to 6% of our hospital disease burden, had the second highest attributable incidence across all community sites in the quantitative molecular MAL-ED reanalysis, and although it was only attributed to facility-based disease in older children (24–59 months) at one site (Kolkata, India) during the original GEMS analysis, the overall detection of sapovirus increased at all sites and all ages during the quantitative molecular reanalysis.^2–4 ,28  Our study found no EPEC association with acute gastroenteritis, despite the known causal relationship between typical EPEC infection and diarrhea in the community-based MAL-ED and other studies,^2,29  although the AF was low in the facility-based GEMS and reanalysis.^3,28  However, we did not determine if our EPEC isolates represented typical or atypical strains, which reveal different virulence characteristics.³⁰  Overall, the heterogeneity of these findings suggests important differences in circulating pathogens, population age, susceptibility, nutritional status, medical comorbidities, study design, detection methods, gastroenteritis definitions, disease severity measures, or a combination.

After scale-up of pediatric vaccines against rotavirus in Peru, norovirus GII has emerged with the highest AF (29%) of all enteropathogens identified in this study. Relative to nonnorovirus-associated cases, cases with norovirus had higher mean Vesikari and Clark severity scores.³¹  These findings are consistent with literature reporting severe norovirus manifestations among immunologically naïve adult travelers to Latin America²²  and military recruits.³²  Our finding that infants aged 6 to 12 month experience relatively severe disease supports this because maternal antibodies have waned by this age.³³  Protecting immunologically naïve individuals is challenging because circulating noroviruses are genetically diverse, as revealed by our identification of 12 different case-associated genotypes, including the predominant GII.4 and temporally clustered GII genotype 17 and GII.6. This suggests that, like influenza and coronavirus, surveillance could inform vaccine development.

After universal monovalent rotavirus vaccine scale-up, rotavirus accounted for the second highest AF (9%) of the medically attended gastroenteritis diseases burden with rotavirus G12P[8] predominance. Since 1998, G12 has been isolated in Africa, Asia, Europe, and the Americas, suggesting global propagation.^34–36  In Brazil and Argentina, G12 was associated with P[6] and P[9], whereas in South Africa and Malawi, it was most frequently associated with P[6] and, less frequently, P[8].^34–39  In Spain, G12 was most commonly associated with P[8],⁴⁰  similar to our findings. An adult outbreak of G12P[8] gastroenteritis in the United States suggests emerging importance.⁴¹  Although the Rotarix vaccine used throughout Latin America and the developing world lacks G12 reassortant rotavirus, it has the same P8 and VP4, and similar VP6 proteins as G12P8, which should elicit protection, demonstrated by 59.1% efficacy against P[8] in Africa.³⁶  Likewise, RotaTeq is 80% efficacious against G12.⁴² 

Sapovirus had the third highest AF (6%) in this population, after norovirus and rotavirus. Hospital-based sapovirus infections were nearly as genetically diverse (all 4 human genogroups and 12 of 19 known human genotypes) as community-based infections (4 genogroups and 14 genotypes).⁴³  This represents more sapovirus genetic diversity than other hospital settings. Only sapovirus GI was detected in Vietnam and Bangladesh,^44,45  and only sapovirus GI and GII were reported in the United States.⁴⁶  In Japan⁴⁷  and in the Philippines,⁴⁸  sapovirus GV and GIV, respectively, were not reported. We detected sapovirus GI more frequently than GII (50% vs 31%), similar to community-based reports from Nicaragua.⁴⁹  For unknown reasons, sapovirus GI may be more pathogenic, being frequently diarrhea associated with lower Cq values than other genotypes.⁴³  Although the sapovirus AF suggests it is a good candidate for multivalent diarrhea vaccine inclusion, its genetic diversity suggests maintaining efficacy could require surveillance.

Our prospectively collected severity data demonstrated strong correlation between the modified 20-point Vesikari and 24-point Clark scales, but because Vesikari has only 2 categories (mild and severe) and Clark has 3 categories (mild, moderate, and severe), most of the severe Vesikari cases scored near the lower limit and were categorized as moderate by Clark. Similar to our findings, analysis of community diarrhea severity data from MAL-ED demonstrated that the MAL-ED–modified Vesikari scale, Clark scale, and Community Diarrhea scores were strongly correlated and consistently predicted care-seeking by caregivers.⁵⁰  If we applied the GEMS severity definitions to our population, all of our children with gastroenteritis would meet GEMS’ nonprospective hospital admission criterion for moderate to severe diarrhea, although only a subset was categorized as severe by the Clark and/or Vesikari scales.³ 

Parasite coinfection was associated with decreased gastroenteritis risk, which might represent a proxy for frequent previous infections and consequently enhanced enteropathogen immunity.³²  Seemingly paradoxical associations of children with gastroenteritis with higher SES scores, improved sanitation, and improved access to water are puzzling and could reflect selection biases between case and control populations, with those of higher SES demonstrating increased care-seeking behavior. Despite their higher SES, the statistically significant, albeit subclinical, decreased hematocrit level among patients is consistent with literature demonstrating that enteric infections alter gut function, decreasing micronutrient absorption and lowering hematocrit.⁵¹ 

Study limitations include potential selection biases demonstrated by higher SES among children with gastroenteritis, testing for only 5 viruses, excluding adenovirus, and using bacterial culture and microscopy, which are less sensitive than molecular methods.²⁸  Pathogens were not found in 26% of diarrhea stools, which is similar to findings of studies with more sensitive evaluation methods but could be related to multiple factors, such as antibiotic use, noninfectious causes of diarrhea, or detection limitations.²  Low rotavirus genotyping success could miss types of public health importance. Calicivirus genotyping limitations include the inability to genotype all specimens, particularly controls, which limits comparison between groups, and partial sequencing of VP2 and VP1 genes without polymerase gene analysis, which cannot identify recombinant strains or genotypic variants.²⁸ 

Two potentially preventable caliciviruses, norovirus and sapovirus, are leading contributors to the severe diarrhea burden in this setting after rotavirus vaccination. In interventions in similar settings with prevalent calicivirus-associated diarrhea, in which these results are generalizable, researchers should consider multivalent calicivirus vaccines. Additionally, rotavirus continues to contribute to disease.

We thank Daniel G. Bausch, Maruja Bernal, Maritza Calderon, Brisaida Cordova-Rafael, Rosio Guerra, Kristen Heitzinger, Mary Carol Jennings, Margaret Kosek, Luis Loayza, C. Giannina Luna, Rina Meza, Yocelinda Meza, Andrew J. Mirelman, Giuliana Montago-Vega, Lisa Marie Nance, Karen Neira, Monica Pajuelo, Karen Perreira, Luis Miguel Pinto, Simon Pollett, Dawn Quigley, Jose Quispe, Laura Rappoport, Erik J. Reaves, Claudio Rocha, Karina Roman-Sanchez, Regan Stiegmann, Hannah E. Steinberg, Vanessa Sarabia, Maria E. Silva, and Manuela Verastegui for their help with study implementation, laboratory diagnosis, epidemiological methods, and/or manuscript review.

AF

attributable fraction

DAEC

diffuse-adhering E coli

EAEC

enteroaggregative E coli

EIEC

enteroinvasive E coli

EPEC

enteropathogenic E coli

ETEC

enterotoxigenic Escherichia coli

GEMS

Global Enterics Multicenter Study

GI

genogroup I

GII

genogroup II

GII.4

genogroup II genotype 4

GII.6

genogroup II, genotype 6

GIV

genogroup IV

GV

genogroup V

LT+

heat labile

MAL-ED

Malnutrition and Enterics Study

RT-PCR

real-time polymerase chain reaction

SES

socioeconomic status

ST+

heat stable

STX1

Shigatoxin 1

STX2

Shigatoxin 2

WHO

World Health Organization