Effect of Non-Rotavirus Enteric Infections on Vaccine Efficacy in a ROTASIIL Clinical Trial

ABSTRACT. This study examined the relative proportion of enteric pathogens associated with severe gastroenteritis (GE) among children younger than 2 years in a phase III efficacy trial of the ROTASIIL® vaccine in India, evaluated the impact of co-infections on vaccine efficacy (VE), and characterized the association between specific pathogens and the clinical profile of severe GE. Stored stool samples collected from cases of severe GE in the phase III trial were tested by quantitative polymerase chain reaction using TaqMan™ Array Cards. Etiology was attributed by calculating the adjusted attributable fraction (AF) for each pathogen. A test-negative design was used to estimate VE. The pathogens with the highest AFs for severe diarrhea were rotavirus (23.5%), adenovirus 40/41 (17.0%), Shigella spp./enteroinvasive Escherichia coli, norovirus GII, enterotoxigenic E. coli, and Cryptosporidium spp. A considerable proportion of the disease in these children could not be explained by the pathogens tested. Severe GE cases associated with rotavirus and Shigella spp. were more likely to have a longer duration of vomiting and diarrhea, respectively. Cases attributed to Cryptosporidium spp. were more severe and required hospitalization. In the intention-to-treat population, VE was estimated to be 43.9% before and 46.5% after adjustment for co-infections; in the per-protocol population, VE was 46.7% before and 49.1% after adjustments. Rotavirus continued to be the leading cause of severe GE in this age group. The adjusted VE estimates obtained did not support co-infections as a major cause of lower vaccine performance in low- and middle-income countries.


INTRODUCTION
2][3][4] However, rotavirus remains a significant cause of diarrhea despite substantial reductions in the burden of rotavirus gastroenteritis (RVGE) both globally 1,5,6 and in India. 3,7or example, an effectiveness study carried out in Tanzania observed that rotavirus remained the leading cause of diarrhea requiring hospitalization even 2 years after introduction of the rotavirus vaccine in a national immunization program. 2 Oral rotavirus vaccines, when tested in low-and middleincome countries (LMICs), are consistently found to have lower efficacy compared with trials using the same products in high-income countries.A recent meta-analysis of ROTARIX V R and RotaTeq V R vaccine efficacy (VE) trials categorized by the under-5 mortality rates of the country where the trial was held, showed a distinct demarcation, with lower VEs reported in higher mortality settings. 8In India, phase III clinical trials of ROTAVAC 9 and ROTASIIL 10 vaccines demonstrated VEs against severe RVGE (SRVGE) of 53.6% (95% confidence intervals [CI] -35.0%-66.9%)and 39.5% (26.7%-50.0%),respectively.
Multiple factors have been proposed to account for poor VE in LMICs, including the presence of interfering maternal antibodies from breast milk, impaired immunity owing to nutritional deficiencies, environmental enteric dysfunction, genetic factors, and co-infections. 11,12The presence of co-infecting pathogens may interfere with the uptake and replication of live oral vaccines or may have immunomodulatory effects that can negatively affect immunogenicity and efficacy. 13,14Breakthrough severe gastroenteritis (GE) cases in the trial may have been misclassified as rotavirus cases owing to the presence of rotavirus in the samples but were in actuality caused by pathogens other than rotavirus; this could have led to an underestimation of VE.
Quantitative output from the TaqMan Array Card (TAC) assay (Thermo Fisher Scientific, Carlsbad, CA) has enabled the development of models that have been instrumental in understanding the etiology of disease from samples that could contain multiple pathogens and commensals, even in the absence of disease.This could help us to understand the role of co-infections in the attenuation of rotavirus VE in LMICs, as well as to determine the overall distribution of other significant agents causing severe diarrhea in children from this age group across India. 15The ROTASIIL efficacy trial in India 10 provides an opportunity to examine these questions using this methodology.
The parent BRV-PV (ROTASIIL; Serum Institute of India Pvt Limited, Pune, India) phase III efficacy study was initiated in 2014 and does not reflect the effect of the introduction of the rotavirus vaccine in India's Universal Immunization Program in 2016.

MATERIALS AND METHODS
Study population and sample selection.The BRV-PV phase III trial aimed to measure the clinical efficacy of ROTA-SIIL to prevent SRVGE in children younger than 2 years (ClinicalTrials.govNCT 02133690; Clinical Trial Registry of India [CTRI/2013/05/003667]).Details of this trial and the results obtained have been described. 10Briefly, 7,500 infants were randomized to receive either three doses of BRV-PV vaccine or placebo at 6, 10, and 14 weeks of age and were subsequently followed up until the development of SRVGE, which was the primary endpoint for the trial; these children were later followed up until 2 years of age.Gastroenteritis was defined as the occurrence of at least three watery or looser-than-normal stools within a 24-hour period with or without vomiting, and the Vesikari Scoring System was used to assess severity. 16Gastroenteritis cases with a Vesikari score $11 were classified as severe, and those with a score $15 were classified as very severe.Stool samples were tested for rotavirus by an ELISA assay.Aliquots of these stool samples were stored for further testing.
This analysis was carried out in data obtained from the trial after infants reached 2 years of age, which was the follow-up period in the efficacy study.This ancillary study was designed primarily to estimate the incidence and relative proportion of various etiological agents known to cause GE in the first 2 years of life and to attempt to correlate the severity of diarrhea with the pathogen identified in the analysis and the treatment arm.The endpoint for the original trial was SRVGE cases after the final vaccine dose, whereas in this reanalysis all SRVGE cases after the initial dose (intention-to-treat population [ITT]) were included along with all non-severe RVGE and severe non-RVGE cases.Three groups of samples were included in the study: 1) all rotaviruspositive samples from severe GE that represented the first SRVGE for the participant; 2) all rotavirus-positive samples from GE of any severity irrespective of the timing of sample collection; and 3) all samples from severe GE cases that tested negative for rotavirus.
Sample storage and retrieval.The sample aliquots were stored at The Wellcome Trust Laboratory at Christian Medical College, Vellore, in 2-mL vials at 280 C in freezers that were continually monitored for variations in temperature.At least 1 mL of stool (watery) or 0.5 g (semisolid) was required for inclusion in the study.
Laboratory assays.Total nucleic acid was extracted from the sample using the QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) with a modified protocol that incorporated bead beating in the extraction process. 17Phocine herpesvirus and the MS2 bacteriophage were spiked into each sample prior to extraction as an internal control.
The TAC assay is an arrayed format of duplex quantitative polymerase chain reaction (qPCR) assays in a microfluidic card system, where separate wells contain pathogen or internal control-specific primers and probes (Supplemental Table 1).The TAC assays were carried out using the Quant-Studio TM 12K Flex or a QuantStudio 7 Flex real-time PCR system.Testing for rotavirus was not included in the TAC as these data were already available from the original efficacy study.Each extraction batch included a negative control, which was tested in the TAC panel to exclude contaminants; if the control tested positive for any of the targets, the extraction was repeated for that batch and retested.

STATISTICAL ANALYSES
To estimate GE disease etiology, the attributable fraction (AF) for each pathogen was estimated, adjusting for the effects of other pathogens that act as confounders in this analysis. 18ecause samples from asymptomatic children were not collected in the efficacy trial, models derived from the qPCR reanalysis of the GEMS (Global Enteric Multicenter Study) case-control study 19 were used to calculate pathogen-specific odds ratios (ORs), an approach that has been used in other studies with similar designs. 15,20,21The pathogen-specific ORs were calculated by fitting a multivariable conditional logistic regression model to describe the association between pathogen quantity and diarrhea, adjusting for the presence of other pathogens.Attributable fractions and their variance were calculated as described previously. 15The collection of samples from all GE cases in the ITT population over the period of 2 years enabled the estimation of attributable incidence (AI) for each pathogen, estimated as a measure of its contribution to disease etiology, by dividing the number of pathogen-specific episodes by the person-time observed.
To analyze the relationship between the clinical characteristics of the disease and the disease pathogen, rotavirus detection by antigen ELISA was modeled for presence/absence as a binary variable, whereas pathogens other than rotavirus were modeled as continuous AF values for each episode in a generalized linear mixed-effects model, scaled such that the OR for a one-unit change in AF corresponded to an episode attributable to that pathogen or pathogen category.To account for correlation among multiple episodes of GE in children, generalized estimating equations were used to fit logistic or ordinal logistic regression models.
A test-negative case-control design was used to recalculate the VE after adjusting for the effect of co-infections by various enteropathogens. 4,15,20,21Episodes of severe GE that were RV positive by enzyme immunoassay (EIA) were taken as cases, whereas those who tested negative in severe GE cases were designated as controls; vaccination status, study site, terms for age of the child, and quarter of the calendar year were included as predictors in a logistic regression model.An interaction term was added in between vaccination status and sum of the attributions to nonrotavirus enteropathogens.Vaccine efficacy was calculated as (1 2 ORi) 3 100, where ORi is the exponent of the model coefficient for vaccination status.Vaccine efficacy was calculated again as above but without including the interaction term, which was interpreted as the VE when the sum attribution to non-rotavirus enteropathogens was zero, followed by bootstrapping the data to calculate the difference in VE and variance estimates.

RESULTS
Details of samples selected.Per the criteria for inclusion into the three groups, 2,694 samples were selected to carry out reanalysis (Supplemental Table 2), of which 1,307 samples were from study participants in the vaccine group and 1,387 were from the placebo group.Samples that could not be tested because of unavailability were considered missing at random, and no imputation of values was made for these observations.In all, 2,648 samples were tested by the TAC assay (Figure 1).These consisted of 1,248 samples from the vaccinated group, of which 563 were RV-positive, and 1,364 from the placebo group, of which 712 were RV-positive.
The cases from which the samples were tested were not biased by sex, rotavirus positivity, or vaccination status (Supplemental Table 3).The age of these children at the time of GE ranged from 42 to 734 days, and the mean age for these cases was 364.3 days.Of these cases, 18.7% were SRVGE, with 15.8% having a Vesikari score of 15 or above, and 26.8% required hospitalization.About half of the cases also presented with 2-4 days of vomiting, and severe dehydration was seen in 10.2% of cases.
Among the various pathogens analyzed, viruses were the most common cause of GE with an AI of 64.4 (49.0% of the overall incidence), followed by bacteria with an AI of 18.7 (14.2%) and Cryptosporidium spp. with an AI of 4.3 (3.3%).However, when taken together, these pathogens contributed to only 66.4% of the overall incidence of severe GE, and approximately 33% of GEs were not attributed to any of the pathogens investigated.
Rotavirus was the principal etiology of severe GE, and the difference in AI between the vaccine (24.0; 95% CI: 15.3-25.9)and the placebo group (38.2; 95% CI: 24.5-41.1)was not statistically significant.In the vaccinated group, the burden of adenovirus 40/41 GE was similar to that of rotavirus in the vaccinated group (23.9; 95% CI: 17.5-28.9)(Figure 2).
Attributable incidence of severe diarrhea by age Rotavirus was accountable for the highest contribution to the pathogen-specific AI in most age groups (except for 0-5 months), followed by adenovirus 40/41 (Figure 3).Shigella spp.had a higher AI in children 18-23 months of age, where it was similar to the incidence of adenovirus 40/41 and rotavirus, and an inverse relationship was observed such that as age increased, the AI of rotavirus and adenovirus decreased.Associations between pathogen-specific AF and clinical profile of severe GE.Severe rotavirus cases were associated with Vesikari scores of $15 (OR: 2.2; 95% CI: 1.7-2.8)compared with other pathogens.Severe GE cases attributed to Shigella spp.were more likely to develop diarrhea that persisted for more than 5 days (OR: 3.2; 95% CI: 2.0-5.0).Cryptosporidium spp. as an etiology of diarrhea displayed a profile of very severe GE cases that were more likely to receive a Vesikari score $15 (OR: 2.4; 95% CI: 1.2-4.5),and these patients were more likely to be hospitalized (OR: 1.9; 95% CI: 1.0-3.6)(Figure 4A).Overall, bacterial infections were associated with a longer duration of diarrhea (OR: 5.5; 95% CI: 2.7-11.1)and dehydration (OR: 4.7; 95% CI: 1.2-18.9),whereas protozoan infections were more often associated with severe diarrhea as measured by the Vesikari score (OR: 2.5; 95% CI: 1.3-4.8)and with hospitalization (OR: 1.9; 95% CI: 1.1-3.7).
Effect of co-infections on VE.Vaccine efficacy was calculated from the ITT and per-protocol (PP) populations of study participants who developed severe rotavirus cases followed up until 2 years of age.There were 482 cases of severe or very severe rotavirus cases, of which 446 cases were categorized as falling in the PP group.The VE against SRVGE for the PP population was estimated to be 46.7%(95% CI: 32.7%-57.8%)with this analysis, compared with an efficacy of 39.5% (95% CI: 26.7%-50.0%) in the parent study.When adjusted for co-infections, the efficacy was calculated to be 49.1% (27.6%-64.2%) in PP participants.The difference with CI estimates is 2.4% (95% CI: 214.4% to 15.3%).Adjusted VE estimates were also calculated in the ITT population, and the results showed changes similar to those in the PP population (Table 2).

DISCUSSION
Rotavirus remained the leading cause of severe GE in this analysis among both vaccinated and placebo groups, comparable to results obtained from a corresponding reanalysis of specimens from an efficacy trial of ROTAVAC. 15A similar study in Tanzania examining the impact of the introduction of ROTARIX showed a reduction in diarrhea admissions; however, rotavirus remained the leading pathogen that caused diarrhea in children under 5 years requiring hospitalization. 2 Geographical differences notwithstanding, the impact of rotavirus vaccine in reducing severe GE and preventing hospitalizations has been examined in multiple studies after the availability and introduction of these vaccines to national public health programs, and this effect of vaccine introduction has been confirmed by meta-analyses of these impact studies. 1,5denovirus 40/41 was one of the leading causes of severe diarrhea in this study population.This is consistent with results for adenovirus from the GEMS reanalysis 19 by TAC assay in Kolkata, India, though it was much lower in other sites in Asia and Africa.The two subtypes of adenovirus type F detected in this study (40 and 41) are thought to cause mild diarrhea.Therefore, the high etiological fraction of adenovirus in severe GE cases in children in this analysis and in others with study sites in India 19,22 is unexpected, and this pathogen should receive more focused attention as a leading cause of morbidity in children ,2 years of age.
Shigella spp. was the leading bacterial etiology of severe GE found in this study.This is well corroborated by multiple etiology studies that point to this pathogen as a significant cause of GE in children. 2,15,20,21eta-analyses of norovirus prevalence in GE cases from developing countries have shown that the genogroup NoV GII accounted for more than 90% of norovirus cases. 23,24he AF for norovirus GII in this study was comparable to that of the ROTAVAC efficacy reanalysis in India. 15In the GEMS and MAL-ED (Malnutrition and Enteric Disease Study) reanalysis reports, 19,25 norovirus GII was among the top six causes of diarrhea in children.
Cryptosporidium spp. was the only protozoan among the leading etiologies for severe GE.The adjusted AF for Cryptosporidium in Kolkata, India, reported by the GEMS reanalysis data, matched the estimate in this analysis; however, in a similar reanalysis from Niger, the AI estimated from that population was 10 times higher than that estimated here.
The sum of all pathogen-specific AIs was found to be only 66.4% of the overall AI for severe GE.In contrast, similar reanalyses of samples from rotavirus vaccine efficacy trials from Niger (ROTASIIL) 21 and India (ROTAVAC) 15 were able to account for 95.2% and 84.7% of cases, respectively.This discrepancy could be due to several reasons, including that rotavirus diarrhea was estimated by EIA and not PCR, meaning the rotavirus AI may have been underestimated, or that a notable proportion of diarrhea was caused by pathogens that were not included in the TAC assays.The list of pathogens tested by TAC in this reanalysis was designed to cover the entire spectrum of diarrhea-causing bacteria, viruses, and parasites commonly thought to be prevalent in India and South Asia.The PCR targets were chosen based on their consistent association with the pathogenic strains of the diarrheal agent from the published literature.However, a few uncommon diarrheagenic pathogens were not tested, such as Yersinia spp., Listeria monocytogenes, Clostridium perfringens, and Vibrio parahaemolyticus.In addition, certain PCR targets may be insufficient to detect bacterial pathogens such as the hemolysin A target for Vibrio cholerae (hlyA); other targets such as ompW and ctxA may estimate its prevalence more accurately.Regular and periodic diarrheal surveillance of communities in LMICs is essential for detecting novel and emerging agents causing GE in this age group.
When the AI estimates were stratified by age of the child, the pattern of infections at each stage was apparent (Figure 3).The maximum infection rate per the estimated AIs was from 6 to 18 months.This pattern of diarrhea with age is consistent with the reanalysis from Niger, and the reason could be that children who are freshly weaned are more susceptible to gastrointestinal pathogens, and the burden would gradually decrease over the next 12 months as the child develops immunity to the pathogen after infection.This would have implications for vaccine delivery strategy.For example, current efforts focus on developing a Shigella vaccine to be administered at 1 year of age, and this would not protect against shigellosis in infants. 26evere GE cases associated with rotavirus tend to be associated with a higher Vesikari score and with higher odds of hospitalization.This is comparable with the profile of rotavirus infections in this age group observed in other studies, in which rotavirus infections in children led to very severe diarrhea in 66.7% of children in a birth cohort 27 and a study of hospitalized children in India revealed that rotavirus infections accounted for 36.3% of hospital admissions due to diarrhea in the under 5-year age group. 7indings for the clinical profiles of Shigella spp., adenovirus, ST-ETEC, and Cryptosporidium spp.have parallels in similar analyses in other studies.The ROTAVAC trial reanalysis observed a similar clinical picture for Cryptosporidium spp.(increased hospitalizations), Shigella spp.(longer duration of diarrhea), and adenovirus (association with vomiting).A similar clinical profile for Cryptosporidium and Shigella was also observed in the reanalysis of samples from the ROTASIIL efficacy study in Niger. 21he adjusted VE estimates do not illustrate a statistically significant difference of outcome in efficacy for the vaccine between the mixed infections and infections by rotavirus alone.Other studies have attempted to do the same for ROTARIX 28,29 and the bovine attenuated rotavirus vaccine (RIT 4237) 30 ; however, although the efficacy difference in all these studies ranged from 8.0% to 14.0%, none of these effects were statistically significant, which may be attributable to inadequate sample size.
It should also be noted that the estimated AI of rotavirus for severe GE does not appear to be significantly reduced in the vaccinated group as opposed to the placebo group, as the CIs overlap even though the point estimate in the vaccinated group is lower.Again, this is to be expected because the incidence estimates would incorporate the variance from the attribution of etiology and would provide wider intervals as a result, whereas the VE estimates in the parent study would not.
The results from this analysis failed to demonstrate a significant correlation between co-infections with other pathogens causing severe GE and the efficacy of oral rotavirus vaccines in LMICs.This is comparable to results from the ROTAVAC trial reanalysis from India that showed a 11.3% point estimate increase in VE when adjusted for co-infections, although this was not statistically significant.A study in Botswana also did not find evidence of an effect of co-infections on rotavirus vaccine effectiveness. 28There may be methodological factors that can explain this discrepancy, including misattribution of the etiology of the disease to rotavirus in the vaccinated group in efficacy trials.Co-infections have been hypothesized to interfere with vaccine response through multiple means, such as competition for cell entry and replication or for cellular factors required in transcription, replication, and assembly. 12,13Other biological causes for this lower efficacy could be chronic malnutrition prevalent in children in LMICs, gut inflammation, environmental enteropathy, and interference from antibodies in the system that would lead to a suboptimal immune response.The increased infective load in LMICs due to poor hygienic conditions also could be a factor in the reduced ability of the vaccine to protect against infections when compared with results in high-income settings; a diarrheal state with altered mucosa and rapid gut transit may interfere with re-uptake of the vaccine.Rotavirus replication has been shown to be impaired by enteroviruses, 31 and a study from Bangladesh demonstrated lowered rotavirus seroconversion on concomitant administration of the monovalent rotavirus vaccine and trivalent oral polio vaccine vaccine. 32However, the parent vaccine trial of this study did not find any effect on seroconversion with OPV administration. 10This reanalysis included only targets that were known to cause diarrhea in this age group.The presence of commensals such as enteroviruses could also interfere with efficacy, and further studies in samples with those commensals as targets or a metagenomic analysis in vaccinated cohorts between sites could lead to a better understanding of the interactions between the VE of viruses and organisms present in the gut microbiome.
This study attempted to account for endpoint misclassifications for SRVGE cases in the parent phase III vaccine trial that could lead to lower estimates of VE of ROTASIIL in LMIC settings.This misclassification can be expected to vary inversely with the strength of the association of pathogen quantities in the stool and disease and is therefore even more relevant for enteric pathogens such as norovirus GII and Cryptosporidium spp.Pathogens such as norovirus GII and Campylobacter spp., which are commonly detected from diarrheal stool but less likely to be the cause of diarrhea, 24,33 would be especially susceptible to this misclassification in vaccine trials.Approaches similar to that used in this studytesting for other enteric pathogens in phase III vaccine trials of these pathogens-could allow us to assess this.
This study had several limitations.The samples selected for testing from the parent study did not include the nonsevere non-rotavirus GE cases.This could lead to a bias in the attribution of pathogens for severe GE that usually cause milder infections or could pass through the gastrointestinal tract as commensals (e.g., adenovirus).Another limitation was that rotavirus positivity was estimated by EIA, which is not as sensitive as a molecular assay, and this could underestimate the attribution of rotavirus to severe GE, which would explain the gap between any-cause AI and the combined AI from all measured targets.The use of the Vesikari score to categorize the severity of diarrhea could bias AFs toward viral etiologies, as vomiting, which is a component in the scoring system, is relatively more frequent in viral GE than with bacterial and protozoan etiologies.

CONCLUSION
In summary, the quantitative output of the TAC assay is a useful tool to analyze etiologies of disease in mixed samples that could have commensals in addition to multiple pathogenic organisms.The analysis brings to our attention the multiple etiologies of severe GE in children of a very young age, which include adenovirus, Shigella, norovirus, Cryptosporidium, and ETEC.Future public health and vaccination strategies should be planned based on studies carried out locally to determine the agents of concern.Although the adjusted VE estimate obtained was not sufficient to conclude whether co-infections are indeed the cause of poorer vaccine performance in LMICs, it is likely that this is not a

FIGURE 1 .
FIGURE 1. Selection of samples for testing from the parent ROTASIIL efficacy trial.

FIGURE 3 .
FIGURE 3. Attributable incidence per 1,000 child-years for severe gastroenteritis in children less than 24 months old stratified by age.Dashed lines indicate the 95% confidence intervals for each estimate.C. jejuni/coli 5 Campylobacter jejuni/coli; ST ETEC 5 stable toxin enterotoxigenic Escherichia coli; V. cholerae 5 Vibrio cholerae.
BRV-PV group: AI in participants who were administered the vaccine; Overall burden: overall AI of various pathogens for severe diarrhea; Placebo group: AI in participants who were administered placebo.Dashed lines indicate the 95% confidence intervals for each estimate.C. jejuni/ coli 5 Campylobacter jejuni/coli; ST ETEC 5 stable toxin enterotoxigenic Escherichia coli; V. cholerae 5 Vibrio cholerae.
Attributable incidence of severe diarrhea/1000 child-years FIGURE 2. Attributable incidence (AI) per 1,000 child-years for severe gastroenteritis by study group and overall in children less than 24 months old.

TABLE 2
Vaccine efficacy estimates for the ITT and PP populations 5 intention-to-treat; PP 5 per-protocol; SRVGE 5 severe rotavirus gastroenteritis; VE 5 vaccine efficacy.major contributor, and future trials must include analysis of other factors such as the gut microbiome, interactions with other vaccines, gut inflammation, and malnutrition status.Received May 28, 2023.Accepted for publication January 14, 2024.Published online April 16, 2024.Note: Supplemental material appears at www.ajtmh.org. ITT