Consequences of Shigella infection in young children: a systematic review

Highlights • Systematic review of longitudinal Shigella outcomes in children.• Shigella is associated with continued diarrhea and linear growth faltering.• There is a need for standardized measurement and reporting of Shigella outcomes.


Introduction
Shigella is a highly transmissible enteric pathogen, which causes an estimated 68,0 0 0 deaths in children aged < 5 years each year [1] and is indirectly responsible for an additional 13,600 deaths from Shigella -associated linear growth faltering or stunting [2] . The mortality rates from Shigella have declined substantially over the last few decades due to the apparent disappearance of the highly virulent Shiga toxin-producing Shigella dysenteriae 1 serotype, measles vaccination, antibiotics, improvements in nutritional status, and economic development [3][4][5] . Despite these gains, antibiotic resistance to first and secondline antibiotics that have historically been effective in reducing disease severity, diar- [ 17 , 18 ] and from potential decreased economic/earning potential from the longer-term outcomes of Shigella [19] .
Based on the clinical severity, disease burden, links to longerterm outcomes, and the emergence of antimicrobial resistance, Shigella is a priority for vaccine development in the target population of young children living in LMICs [20] . Vaccines targeting the most common Shigella flexneri serotypes and Shigella sonnei are in development [ 21 , 22 ]. As pediatric Shigella vaccines move toward licensure and policy makers consider vaccine introduction, there is a need to synthesize evidence on the long-term consequences of Shigella to aid global and country decision-making to support vaccine adoption [ 20 , 23 ]. We conducted a systematic review of the consequences of Shigella infection among children in LMICs to help characterize the potential value of a Shigella vaccine.

Methods
We conducted a systematic review following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines [24] to identify literature on the consequences of Shigella infection in children aged < 5 years in LMICs. We aimed to gather data on the breadth of sequelae attributable to Shigella infection among young children, including but not limited to diarrhea persistence, linear growth faltering, ponderal growth faltering, neurodevelopmental delay, economic impacts, immune response, and systemic and enteric inflammation. In addition to characterizing the evidence and direction of effect, we sought to identify evidence gaps that could be addressed in future research studies.

Search strategy and selection criteria
We searched PubMed and Embase for articles published from January 01, 1980 to December 12, 2022 that indicated longitudinal follow-up of children after detection of Shigella in fecal samples or blood by any laboratory method. We included terms that described LMICs, as well as the names of all countries categorized as LMICs by the World Bank in 2020 (see Appendix 1 for full search strings).
We included clinical trials and observational studies that followed up at least five children with Shigella detected for any duration beyond 1 hour, regardless of symptoms. We restricted to studies conducted in LMICs that reported outcome data for children aged < 5 years (0-60 months) to focus on the population with the highest morbidity and mortality burden attributed to Shigella [1] . We excluded cross-sectional studies and outcomes that were assessed contemporaneously with Shigella detection. Conference abstracts were included if they met other inclusion criteria and contained outcome data. We translated non-English publications using DeepL Translator (Cologne, Germany) or Google Translate.
Two reviewers (FA, MD, or TL) independently screened the title and abstract of each article for eligibility using Covidence (Veritas Health Innovation, Melbourne, Australia). Any disagreements were resolved by a third reviewer (PP) or through group discussion and consensus. If a decision could not be made using the information available in the abstract or if no abstract was available, the article was passed to full-text review. The same methods (dual review and conflict resolution using Covidence) were used during full-text review. The review's International prospective register of systematic reviews registration number is CRD42021241169 ( link ).

Data analysis
The summary data were abstracted from full-text reports of included publications. We abstracted information on the original study design and methodology (e.g., length of follow-up, inclusion criteria), the location of study, the number of children and/or stools with Shigella detected, laboratory method of detection, Shigella species identified, co-infections, and funding source. For each outcome identified, we abstracted the method of measurement, time point of measurement or duration of follow-up, any adjustment variables, and the effect estimate. All longitudinal outcomes were abstracted except mortality because this outcome was recently summarized in a systematic review of case fatality rates for common diarrheal pathogens [25] . Clinical characteristics and outcomes reported only at medical presentation or study enrollment were not abstracted because it was not possible to determine temporality in relation to Shigella detection. Data from randomized trials were abstracted for each randomization arm; the measures of excess risk comparing randomization arms were not abstracted unless they compared children with and without Shigella detected.
Because all data in this review were treated as a cohort study ( Shigella as the exposure), we did not feel it would be relevant to assess the risk of bias for the original study design (e.g., randomized control trial) nor would it be possible to uniformly apply a risk of bias assessment tool to the variety of designs included in this review because many questions are not suited to our included outcomes. Instead, we conducted a quality assessment of included studies using a modified version of a composite quality construct based on the Strengthening the Reporting of Observational Studies in Epidemiology statement [26] , which was developed and implemented previously [27] . In this assessment, each article was awarded points (10 maximum) for satisfying components of the methods section of the Strengthening the Reporting of Observational Studies in Epidemiology statement checklist, which includes an assessment of effort s to address potential sources of bias (Appendix 2). A rating of 'poor' was assigned to articles with zero to four points, 'fair' with five to seven points, and 'good' with eight to 10 points. As part of our quality assessment, we reviewed information contained within a given publication, as well as the text of referenced articles as needed.
Data abstraction was performed by a single reviewer (FA, MD, or TL) and quality checks were performed on a random subset of the data (20%). The study data were collected and managed using Research Electronic Data Capture tools hosted at the University of Washington Institute of Translational Health Sciences [ 28 , 29 ]. We performed a descriptive summary of the study characteristics and longitudinal outcomes. The definitions of acute and persistent diarrhea were accepted from included studies, but the review adapted the distinction of < 14 and ≥14 days, distinguishing the two as described in WHO diarrhea treatment guidelines [30] . We intended to conduct a meta-analysis for any outcomes that were reported consistently by more than two studies. Due to heterogeneity in the measurement methods, comparison groups, and follow-up duration, we report a narrative summary of the evidence for each outcome.

Results
Our final search identified 2627 potentially eligible records from PubMed and Embase after deduplication ( Figure 1 ). We completed the dual review of titles and abstracts passing 368 (14%) publications to full-text review, of which 52 met the inclusion criteria ( Figure 1 ). The 316 studies excluded at full-text review are described in Appendix 3. The key characteristics of the 52 included articles are shown in Table 1 and summarized in Table 2 . The data on Shigella outcomes were collected in 20 different countries; although 56% (n = 29) of the publications were from studies conducted at least partially in Bangladesh. There were 13 publications from studies conducted on the African continent. Five publications reported data from multiple countries either collected as part of the Etiology, Risk Factors, and Interactions of Enteric Infections and RCT, randomized controlled trial; qPCR, quantitative polymerase chain reaction; MSD, moderate-to-severe diarrhea; WAZ, weight-for-age z-score. a The number of children with Shigella detected was not specified in some studies; see Appendix 4 for the # of Shigella -positive stools or diarrhea episodes attributable to Shigella , which were used to verify inclusion criteria of 5 + children with Shigella . b The months (if available) and years of participant enrollment. c The age range of enrolled children for whom outcomes were measured/reported.

Figure 1. Study selection (preferred reporting items for systematic reviews and meta-analyses [PRISMA] diagram).
a Studies that were excluded for "no follow-up of Shigella cases/cross-sectional outcomes only" include some studies that were longitudinal in nature, but presented outcomes cross-sectionally such that the likelihood of longitudinal outcomes given Shigella infection could not be determined (e.g., given all children with an outcome, the percent of children that had Shigella infection) either from direct interpretation of tables or through back calculations. Abbreviations: LMIC, low-or middleincome country.
Publications included a median of 66 children with Shigella , ranging from five to 2172 ( Table 2 ). Of note, some of the included studies did not specify the number of children with Shigella but provided other information that made it possible to estimate the number of children with Shigella as being five or more (Appendix 4). Although most publications were among children with Shigella diarrhea only, nine (17%) publications also included Shigella detected in asymptomatic patients ( Table 1 ). The study setting and initial inclusion criteria varied widely, such as malnourishment, current diarrhea, participation in birth and community-based cohorts or randomized controlled trials, admittance to hospitals, and presentation at health care facilities. Culture was the most common primary Shigella detection method (71%), followed by quantitative polymerase chain reaction (21%; Tables 1 , 2 ). Most studies were rated 'good' quality (n = 35; 67%), followed by 'fair' quality (n = 16; 31%) and 'poor' quality (n = 1; 2%) (Appendix 5).
The most commonly reported outcomes of Shigella were related to diarrhea (n = 20) and linear growth (n = 14). Other anthropometric measures, such as ponderal growth (e.g., change in weight-for-height z-score [WHZ]) or weight gain (e.g., change in weight or weight-for-age z-score [WAZ]), were reported in 10 studies ( Table 2 ). In each of these categories, fewer than three studies reported on the same outcome using a similar comparison group, thus precluding meta-analyses.

Diarrhea outcomes
There were three general categories of measurement among studies of diarrhea outcomes: duration of diarrhea measured continuously (n = 9); duration of diarrhea measured categorically ( < 7 days, 7-< 14 days, ≥14 days) and presented as corresponding percentages, odds ratios (ORs), and risk ratios (n = 11); and characteristics of subsequent diarrhea episodes (both Shigella and unspecified) that occurred after diarrhea-free days (n = 3). The measurement details are summarized in Table 3 .
Briefly, based on three studies, between 11% and 25% of children with Shigella diarrhea went on to develop prolonged diarrhea (duration 7-< 14 days) [ 9 , 33 , 34 ], with no statistically significant difference in risk by age (1 year vs 2 years) or co-infection status [9] . Six studies reported on persistent diarrhea (duration ≥14 days) and in these studies, 2.9-61.1% of children with Shigella diarrhea developed persistent diarrhea [ 9 , 35-39 ]. Two of these studies reported on risk factors of diarrhea persistence among Shigella diarrhea cases, with a statistically significantly higher likelihood of persistence among children who were malnourished (malnourished: 19.2% vs well-nourished: 3.2%) [36] , had blood in stool (bloody: 30% vs nonbloody: 19%) [39] , or had multidrug-resistant Shigella (multidrug resistant: 66% vs not multidrug resistant: 20%) [39] . Of note, a study comparing likelihood of persistent diarrhea between children with Shigella-positive diarrhea compared with Shigellanegative diarrhea found Shigella to be significantly associated with persistent diarrhea (relative risk: 1.83; 95% confidence interval [CI]: 1.91, 2.81) [39] . Similarly, another study reported a longer duration of diarrhea in children with Shigella diarrhea than those with other causes of diarrhea (OR of duration longer than 3 days: 1.4; 95% CI: 1.0-2.0) [40] . Across the studies, the continuously measured mean duration of diarrhea ranged from 2 to 22.2 days, with substantial variation by intervention status in trials and anthropometric groups [ 33 , 40-47 ]. There was wide heterogeneity in the information presented on subsequent new diarrhea episodes ( Table 3 ) [ 9 , 33 , 48 ].

Growth outcomes
Six of 14 studies meeting the inclusion criteria found a statistically significant decrease in linear growth associated with Shigella in diarrheal [ 11 , 12 , 49 , 50 ] and nondiarrheal [ 11 , 51 , 52 ] stools ( Table 4 ). There was substantial heterogeneity in measurement time points (ranging from 21 days to 8 years) and comparison groups ( Table 4 ). Linear growth was commonly operationalized as the mean change in the length-for-age z-score (LAZ) between two time points (n = 3) or the difference in LAZ between two groups, defined by presence/absence of Shigella or high/low quantity of Shigella (n = 7). The effect estimates from these studies are summarized in Figure 2 . The differences in LAZ comparing high with low Shigella prevalence in nondiarrheal stools ranged from -0.14 (95% CI: -0.27, -0.01) at 2 years to -0.32 (95% CI: -0.56, -0.08) at 6-8 years [ 11 , 52 ]; the mean differences in LAZ per attributable episode of Shigella diarrhea ranged from -0.12 (95% CI: -0.26, 0.03) [53] to 0.05 (95% CI: -0.15, 0.25) [54] . Two studies reported on the impact of Shigella diarrhea on linear growth at 3 months after diarrhea: one study found a statistically significant average loss of -0.03 (95% CI: -0.05, -0.00) in LAZ [11] , whereas another study found no difference in the 3-month LAZ associated with Shigella quantity during the diarrheal episodes [55] . In GEMS, Shigella episodes not treated with antibiotics led to greater declines in linear growth than treated episodes among children aged < 24 months [12] . Another study found that Malawian children with Shigella detected at age 18 months had, on average, 0.39 lower LAZ at 24 months than children without Shigella detected [51] . George   Abbreviations: CI, confidence interval; OR, odds ratio; RR, relative risk; SE, standard error; SEM, Standard error of the mean; CODA, a diarrheal severity score (Community Diarrhea). a "Illness" was presumed to mean diarrhea because stool samples were taken when diarrheal episodes were detected. b Based on results in Taylor et al. [34] Table 3 (there is a discrepancy in number of children with Shigella spp. isolated on day 0 in the erythromycin group reported in results text and in Table 3 ). et al. [18] found Shigella infection to be associated with a two-fold increase in the odds of stunting (defined as height-for-age z-score < -2) at 9 months of follow-up (OR: 2.01; 95% CI: 1.02, 3.93) [50] , and Black et al. [ 7 , 8 ] reported a statistically significant association between the periods of Shigella diarrhea and change in height-forage compared with a village standard between the beginning and end of the study period [49] .
Additional anthropometric outcomes are summarized in Appendix 6. Seven studies assessed the ponderal growth and weightfor-age, four of which did not have a comparison group without Shigella infection nor with low levels of Shigella [ 45 , 46 , 48 , 56 ]. The MAL-ED study found no significant difference in mean WHZ or WAZ between children with high (90th percentile) and low (10 th percentile) Shigella prevalence in nondiarrheal stools [11] . Two studies reported on children enrolled in the Bangladesh site of GEMS: one found children with Shigella infection had significantly lower WHZ (-0.11; 95% CI: -0.21, -0.001) than children who were Shigella -negative after 60 days of follow-up [57] , whereas the other found no significant difference in the odds of wasting (WHZ < -2) or underweight (WAZ < -2) at the 9-month follow-up [50] (Appendix 6).

Cost of diarrhea episode
Five publications estimated the cost of a Shigella diarrhea episode ( Table 5 ) [ 17 , 18 , 57-59 ]. In one of these studies, across seven sites, the mean total household out-of-pocket cost (including inpatient and outpatient medical costs, transportation, and prescriptions) was $10.61 (converted from local currency to 2012 US dollars), ranging from $4.92 in Mozambique to $17.18 in Mali [59] . This same study found no statistically significant difference in the cost between Shigella diarrhea and other pathogens. A study from China, which additionally included self-reported out-of-pocket expenses for overnight stays, estimated the mean cost to be $22 for children aged 0-1 year and $31 for 2-5 years, which represented 12% and 18% of the average monthly income, respectively [58] . One study from Bangladesh found Shigella episodes to cost an average of 5.7% (range < 1-78%) of the household monthly income [18] . Al-  Abbreviations: CI, confidence interval; LAZ, length-for-age z-score; HAZ, height-for-age z-score; NS, not specified; OR, odds ratio; SD, standard deviation; SE, standard error a Represents the number of children enrolled in the study because the number with Shigella was not specified (results reported as Escherichia/ Shigella) .

Figure 2.
Mean change or difference in LAZ by comparison group and duration of follow-up. * Shigella prevalence or quantity was assessed over a 24-month period. "High" was defined as 90 th percentile and "low" as 10 th percentile. Abbreviations: CI, confidence interval; LAZ, length-for-age z-score. though there was heterogeneity in measurement and adjustment factors across studies, a large proportion of costs were associated with hospitalization or inpatient care.

Enteric and systemic inflammation
Three studies reported on the longitudinal markers of gut and/or systemic inflammatory response among children with Shigella (Appendix 7). In a study of children with Shigella treated with antibiotic therapy and randomly assigned to 14 days of zinc supplementation or control, there were no significant differences in concentrations of innate mediators (myeloperoxidase, superoxidase, nitrate) and cytokines (interleukin-2, interferon-γ ) in stool or released from mitogen-stimulated mononuclear cells within or between treatment groups over 30 days of follow-up [60] . Stool interleukin-1ß concentrations and serum C-reactive protein levels significantly decreased at days seven and 30 in both groups [60] . Over 2 years of follow-up, Schnee et al. [45] found diarrhea attributable to Shigella to be associated with elevated C-reactive protein levels (increase of 0.24 [95% CI: 0.03, 0.49] per diarrhea episode).

Discussion
The World Health Organization recently articulated the need for evidence synthesis of long-term morbidities associated with key enteric pathogens, such as Shigella [23] . In this systematic review, we document the consequences of Shigella infection and disease in children aged < 5 years living in LMICs. We found evidence that Shigella was associated with linear growth faltering and persistent diarrhea [ 9 , 11 , 12 , 16 , 39 , 51 , 52 ]. There was a substantial economic impact on families with children suffering from Shigella diarrhea [ 17 , 18 , 57-59 ]. Heterogeneity in measurement and presentation of outcomes and differences in comparison groups between studies prohibited quantitative synthesis of the data, highlighting the need for standardizing methods for characterizing and reporting on enteric pathogen sequelae.
Shigella is a well-known cause of diarrhea, with moderate and severe forms of diarrhea constituting a substantial financial burden on health care systems and families. Our systematic review added to this evidence base by highlighting the consequences of Shigella diarrhea. Notably, children with Shigella diarrhea had an average duration of illness of 2-22 days, with wide variation [ 33 , 40 , 41 , 43 , 44 , 46 , 47 , 49 , 69 ], and children with acute Shigella diarrhea were more likely to develop persistent diarrhea than children with acute diarrhea caused by other pathogens [39] . Longer diarrhea duration is associated with poorer health outcomes, including mortality, stunting, and wasting [ 72 , 73 ], and poses a greater burden on health care systems due to its increased need for facilitybased care. In the few studies that included the economic consequences of Shigella, all were focused on the cost of Shigella diarrhea borne by families, which ranged from 1% to 78% of the monthly household income [ 18 , 58 ]. With Shigella infections likely having impact on a child's health, even in the absence of diarrhea, assigning an economic value to a Shigella vaccine will require additional data estimating the financial impact of Shigella sequelae beyond diarrhea.
We found Shigella to have modest and inconsistent effects on linear growth. Children who fall off their linear growth trajectories are at substantial risk for stunting, a precursor to poorer school performance, cognitive development, and reduced earning potential [74][75][76] . The greatest differences in LAZ were observed in the MAL-ED cohort study evaluating cumulative asymptomatic Shigella infections occurring over the first 24 months of life and their impact at 2, 5, and 6-8 years of life, with magnitudes ranging from -0.32 to -0.14 [ 11 , 52 ]. These magnitudes are expected to be high because they are comparing the extremes of Shigella infection burden: a high burden of attributable diarrhea episodes (90 th percentile) with a low burden (10th percentile). To the best of our knowledge, there is no established threshold for what loss in LAZ translates to an increased risk of stunting or even more deleterious outcomes, such as impaired cognitive development and poor school performance. If infection rather than disease is primarily re-sponsible for growth faltering from Shigella, then Shigella vaccines will need to induce sterilizing immunity to expect a growth benefit from the vaccine-a tall order for any vaccine. More realistically, and similar to the rotavirus vaccines, a Shigella vaccine will prevent more severe presentations of Shigella diarrhea [77] ; we found Shigella diarrhea to have modest and inconsistently statistically significant effects on linear growth, which may be due to confounding by antibiotic use. Notably, the GEMS study found a statistically significant mean loss in LAZ of 0.06-0.17 after an untreated Shigella MSD episode in infants and toddlers, respectively [12] , magnitudes of association consistent with studies of asymptomatic Shigella included in this review [ 11 , 52 ]. Shigella vaccine trials including linear growth as a secondary outcome, as has been suggested by recent study design consensus statements [ 20 , 78 ], will be best suited to estimate a causal association between Shigella and linear growth deficits.
One of the key pathways by which Shigella and other enteric pathogens are hypothesized to impact linear growth is through EED. EED is a syndrome characterized by inflammation and impaired function of the small intestine and has been associated with stunting among children [13] . The biomarkers of EED, such as myeloperoxidase, may be an intermediate marker of Shigella 's impact on linear growth, and therefore may be important targets for vaccine probe studies to estimate more quickly the impact of Shigella on growth. Although we only identified one study that looked at the biomarkers for EED longitudinally [60] , we note that the indicators for EED may be measured cross-sectionally at the time of acute infection and such outcomes would not have met inclusion criteria. Therefore, our review of longitudinal consequences is not well suited to examine the cross-sectional associations between Shigella and EED.
This review was subject to several limitations in addition to those already discussed. To inform the value proposition for soon to be available Shigella vaccines, it was valuable to limit this review to outcomes reported among children aged < 5 years with confirmed Shigella detection. However, excluding studies that did not disaggregate children with clinically compatible illness but without Shigella confirmation may have disproportionally excluded studies from certain time periods or settings with limited diagnostic capacity. For example, publications from the 1980s summarizing dysentery epidemics suspected to be caused by shigellosis rarely reported outcomes among the subgroup of children with cultureconfirmed Shigella . In addition, some potentially relevant growth data by Lee and colleagues [79] were excluded because all reported results were aggregated with children aged > 60 months. However, this study found a similar magnitude of change in linear growth (-0.081 cm) per Shigella diarrhea episode as another study included in this review, Black et al. [49] , and thus, inclusion would not have changed our conclusions. Although the unpublished data were beyond the scope of this review, an individual-level reanalysis of included studies could provide valuable information on linear growth faltering associated with shigellosis. Finally, this review focused exclusively on LMICs, based on children living in these settings having the highest burden of Shigella morbidity and mortality.
Our ability to meta-analyze these data was limited by substantial heterogeneity in comparison groups between studies and in how outcomes were measured. To illustrate the variability in comparison groups, we've summarized in Figure 3 some common ways children with Shigella were defined (in blue) and possible comparison groups (in green). The interpretation of the results is dependent on the combination of the two groups and may or may not be comparable between studies. In practice, it may not be possible to distinguish between the first two blue boxes (whether diarrhea is attributed to Shigella or to another pathogen) particularly in studies that do not test for multiple pathogens. Moreover, it was common for studies to not have any comparison group (particularly Abbreviations: qPCR, quantitative polymerase chain reaction. for diarrhea outcomes), which limits our ability to make conclusions regarding Shigella consequences, relative to other pathogens or to absence of diarrhea. The comparability of findings was further limited by study heterogeneity in pathogen confirmation techniques (e.g., varying sensitivity of polymerase chain reaction vs culture), assessment of costs, and adjustment for co-infections and confounding factors, including antibiotic use and differences in the standard of care over time and by setting. Furthermore, host factors, such as age, malnutrition, HIV, and measles, are all established risk factors for poor Shigella outcomes and although these characteristics were included in our descriptions of the studies, without a formal meta-analysis, we were unable to statistically assess their contribution to outcomes. The strengths of this systematic review include the wide time span of the reviewed literature and extensive breadth of the outcomes assessed. This review identified several evidence gaps, including lack of data on neurodevelopmental outcomes, relatively short follow-up periods for shigellosis, and limited geographic diversity in study locations. Future trials of Shigella -specific vaccines or treatments with long-term follow-up will ultimately be best positioned to document Shigella consequences.

Declaration of interests
The authors have no competing interests to declare.

Role of the funding source
This study was completed by the Strategic Analysis, Research, and Training (START) Center at the University of Washington. START is a collaborative effort with, and is supported by, the Bill and Melinda Gates Foundation (grant # OPP1155935). The funder of the study proposed the study design but had no role in data collection or data analysis. TL was supported, in part, by the University of Washington Biostatistics, Epidemiologic, and Bioinformatic Training in Environmental Health (BEBTEH) Training Grant (grant # NIEHS 5T32ES015459).

Ethical approval statement
Approval was not required. No human subjects were involved in this research.