Gastrointestinal and Extra-Intestinal Manifestations of Childhood Shigellosis in a Region Where All Four Species of Shigella Are Endemic

Objective To determine the clinical manifestations and outcome of shigellosis among children infected with different species of Shigella. Methods We identified all patients <15 years infected with Shigella admitted to the icddr, b Dhaka hospital during one year. Study staff reviewed admission records and repeated the physical examinations and history of patients daily. Results Of 792 children with shigellosis 63% were infected with S. flexneri, 20% with S. dysenteriae type 1, 10% with S. boydii, 4% with S. sonnei, and 3% with S. dysenteriae types 2–10. Children infected with S. dysenteriae type 1, when compared to children infected with other species, were significantly (P<0.05) more likely to have severe gastrointestinal manifestations: grossly bloody stools (78% vs. 33%), more stools in the 24 h before admission (median 25 vs. 11), and rectal prolapse (52% vs. 15%) - and extra-intestinal manifestations - leukemoid reaction (22% vs. 2%), hemolytic-uremic syndrome (8% vs. 1%), severe hyponatremia (58% vs. 26%) and neurologic abnormalities (24% vs. 16%). The overall fatality rate was 10% and did not differ significantly by species. In a multiple regression analysis young age, malnutrition, hyponatremia, lesser stool frequency, documented seizure, and unconsciousness were predictive of death. Conclusions Both severe intestinal disease and extra-intestinal manifestations of shigellosis occur with infection by any of the four species of Shigella, but are most common with S. dysenteriae type 1. Among these inpatient children, the risk of death was high with infection of any of the four Shigella species.


Introduction
Shigellosis remains a major cause of morbidity and mortality among children in developing countries, and is also an important cause of morbidity in industrialized countries. [1,2,3,4,5] Clinical descriptions of bacillary dysentery were published soon after the identification by Shiga in 1897 of the organism now known as Shigella dysenteriae type 1, [6] and subsequent identifications of other species of Shigella by Flexner, Sonne, and Boyd [7].
Most clinical descriptions of shigellosis have focused on a single complication, or on the one or two serotypes that are prevalent in a single clinical setting, thus making direct comparisons of clinical manifestations of infections caused by different serotypes difficult. [8,9] Few regions have endemic infection with all four species of Shigella simultaneously, or the ability to identify and study the problem if they do.
Bangladesh is an exception. Although severe dysentery and extra-intestinal manifestations -including hemolytic-uremic syndrome (HUS), [10] leukocytosis [11] and intestinal obstruction [12] -are thought to be more frequent with S. dysenteriae type 1 infection as a consequence of its toxin production, [13] the relative frequency of other complications -such as convulsions, [14] hypoglycemia, [15] and sepsis [16] -is either unknown or thought to be more frequent in species of Shigella other than S. dysenteriae type 1. In this report we compare clinical manifestations -both intestinal and extra-intestinal -and outcome in 792 children admitted with Shigella to an urban diarrhea disease treatment centre in Bangladesh.

Ethics Statement
This study was approved by the Ethical Review Committee (ERC) of the icddr, b, which waived the need for signed informed consent as all information was obtained from the standard medical record or as part of standard medical care. Information was recorded on a case report form that was expunged of any identifiers that would have linked the information to an individual patient.

Patient Recruitment
The study was conducted at the Dhaka hospital of the icddr, b in Dhaka, Bangladesh, which provides care to patients with diarrhea. Most of the patients are treated as outpatients, or in a short stay (,24 hours) ward, where the focus is on hydration. Approximately 6% of patients who have more complicated disease are admitted to an inpatient ward where more intensive diagnostic and therapeutic care is available. Stool or rectal swab samples for identification of common enteric bacterial pathogens, including Shigella, are obtained from a systematic 2% sample of outpatients (during this study and until 1995 it was 4% sampling), and on all patients admitted to the inpatient unit.
Patients described in this study were enrolled in the 12-months from March 1987 through February 1988. Patients infected with Shigella were identified by daily review of the microbiology laboratory records of admission stool or rectal swab samples.

Information Obtained
A systematic set of information was obtained from all patients who had Shigella isolated from a stool or rectal swab sample. Using a standard case report form study physicians obtained demographic and historical information, and physical examination findings, by reviewing the charts of patients infected with Shigella and by interviewing patients, or their parents or guardians, to confirm or complete the history of illness obtained by the admitting physician. Study physicians also recorded the results of all laboratory tests that were done, interval histories, and the results of repeat physical examinations. All patients who were still in the hospital at the time Shigella was isolated had physical examinations repeated by study physicians, and daily interval histories in addition to those done by the ward physician. Clinical management was under the direction of the attending physician.
Patient outcome was classified as discharged improved, discharged against medical advice, transferred to another health facility, or died in the hospital. Nutritional status was determined using United States National Center for Health Statistics growth charts. [17] All laboratory tests were performed using standard methods previously described [18].

Statistical Analysis
Data were entered onto a computer database using StatPac Gold Version 3.2 (Walonick Associates, Minneapolis, Minnesota). Data analysis was performed using the Statistical Package for Social Sciences, versions 12.0 for Windows, (SPSS, Chicago, IL), and EpiInfo 2003 version 3.3.2 (Centers for Disease Control and Prevention, Atlanta, Georgia, USA). In analyzing differences by species and serotype, we compared S. dysenteriae type 1 infected patients with all other Shigella-infected patients, and also compared the groups individually. The significance of differences in proportions was tested by the chi-square test with continuity correction, or Fisher's exact test if an expected cell size was ,5. The significance of differences between continuous variables in two groups was tested with Students t test if the data were normally distributed, or the Mann-Whitney U test if the data were not normally distributed. For normally distributed continuous variables involving three or more groups an analysis of variance (ANOVA) was first used to test the significance of differences. Differences between individual groups were then tested for significance using Scheffe's procedure if the overall F statistic was #0.20. For non-normal continuous variables involving three or more groups the Kruskal-Wallis test was used, and the Mann-Whitney U test was used to compare differences between two groups.
Factors independently predictive of death were determined using a multiple logistic regression analysis. Variables that in bivariate analysis had a P,0.10 when tested for association with death, or that could biologically be plausibly associated with death (infected with Shigella dysenteriae type 1 and stool frequency before admission) were entered into the regression equation and eliminated in a backward stepwise fashion if the probability associated with the likelihood ratio statistic exceeded 0.05. Blood glucose and serum protein results, though associated with death in bivariate analyses, were excluded from the multiple logistic regression analysis because these tests were performed on only a limited number of patients. Multivariate odds ratios and confidence intervals in the final logistic regression equation were calculated from the coefficient of the multiple regression models.

Patient Population
During the one-year study Shigella was identified in a stool or rectal swab sample of 863 (14%) of the 6,290 patients admitted to the inpatient unit. This analysis focuses on the 792 (92%) of the 863 inpatients with Shigella infection who were children less than 15 years.

Species of Shigella Isolated from Inpatients
S. flexneri was the most common species isolated, accounting for 504 (64%) of the 792 Shigella infections in inpatients ,15 years. S. dysenteriae type-1 was isolated from 157 (20%) of the patients with culture-confirmed shigellosis ( Table 1).
Three-hundred six (39%) of the children ,15 years of age with shigellosis were below 1 year of age (including 5 neonates), 431 (54%) were 1-5 years, and 55 (6.9%) were .5 years. S. dysenteriae type 1 infection was significantly (P,0.001) less common, and S. flexneri significantly more common (P,0.001), in infants than in older children (Table 1). Median age for those infected with S. dysenteriae type 1 was 24 months versus 13 months for those infected with S. flexneri or S. boydii and 8 months for those infected with S. sonnei (P,0.020 for all comparisons between S. dysenteriae type 1 and each of the three other species). Patients infected with S. dysenteriae types 2-10 (median age 26 months) were significantly older (P = 0.018), and those infected with S. sonnei were significantly younger (P = 0.002) than those infected with S. flexneri.

Admission Demographic and Clinical Characteristics
Admission demographic and clinical features that differed significantly (P,0.05) by infecting species in addition to age were body temperature and weight-for-age (Table 2). Children infected with S. dysenteriae type 1 were significantly more likely to be febrile than those infected with S. flexneri, and were significantly better nourished.

Intestinal Manifestations of Infection
Children with S. dysenteriae type 1 infection had a more severe colitis than children infected with other species or serotypes of Shigella. Those infected with S. dysenteriae type 1 more commonly had grossly bloody stools (78% versus 33%), more erythrocytes and leukocytes on microscopic examination of stool, more often had rectal prolapse (52% versus 15%) or abdominal tenderness (26% versus 6%), and a higher number of stools in the 24 hours before admission to hospital (median 25 versus 11) (P,0.001 for all comparisons with other species) ( Table 3). Severe dehydration was rare in these children, occurring in only 33 (4%) of Shigellainfected patients. Of those who were severely dehydrated, the median creatinine concentration was 118 mmol/l, compared to 87 mmol/l in children without severe dehydration.

Factors Associated with Mortality in a Multiple Logistic Regression Analysis
Variables included in the regression analysis were: age; stool frequency and duration of illness before admission; admission dehydration status and abdominal distention; weight for age; leukemoid reaction; infection with S. dysenteriae type 1 serotype; hyponatremia; documented seizure and unconsciousness. Factors significantly (P,0.05) associated with death were younger age, lower stool frequency before admission, poor nutrition, hyponatremia, documented seizure and unconsciousness (Table 6).

Discussion
This is the largest study to date comparing clinical manifestations of shigellosis in all four species of Shigella. The results of this study emphasize both the distinctiveness of infections with S. dysenteriae type 1, and the ability of all species of Shigella to cause intestinal and extra-intestinal manifestations, and death, in a susceptible population.
In the 792 children in this study, children with S. dysenteriae type 1 infection were older than those with the most common infection, S. flexneri. This appears to not simply result from a selection bias of who is admitted to the inpatient unit, as during the study period outpatients with S. dysenteriae type 1 infection (median age 24 months) were also older than outpatients with S. flexneri (median age13 months). This older age of patients with S. dysenteriae type 1 infection is consistent with previous findings from communitybased studies in Bangladesh and our previous study of infant shigellosis. [19,20] It is not clear why S. dysenteriae type 1 affects older children when compared to S. flexneri. The age difference might result from differing patterns of transmission, but both are known to be transmissible with a very low inoculum. It may be that infection with one serotype of S. flexneri confers subsequent immunity against infection with other serotypes of S. flexneri, but provides lesser or no immunity to infection with other Shigella species. Most evidence suggests, however, that immunity is serotype, and not species, specific. [21] With multiple serotypes of S. flexneri circulating, the probability of exposure to S. flexneri in the first year of life might simply be greater than for S. dysenteriae type 1. Exposure to S. dysenteriae type 1 may also be less sustained because it occurs in periodic epidemics, and cohorts of children may not be exposed every year.
S. dysenteriae type 1 is known to more often cause clinical dysentery than infections with other species of Shigella. [19,22] The results of this study make clear the clinical manifestations and magnitude of those differences. S. dysenteriae type 1 infected patients were more than twice as likely to have grossly bloody stools when compared to patients infected with other species of Shigella, and S. dysenteriae type 1 infection was more than three times as likely to cause rectal prolapse. Patients with S. dysenteriae type 1 infection had a stool frequency before admission more than twice that of other groups, and were more likely to have abdominal tenderness and fever. All of this is consistent with the role of Shiga toxin, which, among Shigella, is produced only by the S. dysenteriae type 1 serotype, in exacerbating mucosal destruction and in intensifying the inflammatory response, [13] including eliciting increased production of pro-inflammatory cytokines [23,24].
Dehydration was not a common problem in these patients with shigellosis. Only 4% had severe dehydration diagnosed clinically, and serum creatinine concentrations were consistent with the clinical assessments. Thus despite the identification of three enterotoxins in Shigella -Shiga toxin (only found in S. dysenteriae type 1) [25] and Shigella enterotoxin 1 (ShET-1) and Shigella enterotoxin-2 (ShET-2) (both found predominantly in S. flexneri 2a) [26,27] -that in animal models cause fluid secretion. The effect on fluid secretion of these toxins in infections in humans is clearly much less than for cholera toxin, or Escherichia coli heat stable or heat labile toxins, which much more commonly cause dehydration.
The hypoproteinemia was most severe in patients infected with S. dysenteriae type 1 and most likely results from the more severe colitis and protein loss in the gut that occur with S. dysenteriae type 1 infection. [28] The severity of the hypoproteinemia in the S. dysenteriae type 1 infected group is all the more remarkable given that by anthropometry they were better nourished than patients infected with other species of Shigella. This latter finding is at odds with a previous study of outpatients with shigellosis at the icddr, b [19] and may reflect a selection bias in types of patients admitted to the inpatient unit -with older, better nourished patients being more likely to be admitted if they have severe clinical dysentery.
Other extra-intestinal manifestations also more common in S. dysenteriae type 1 infection included alterations in consciousness, severe hyponatremia, leukemoid reaction, and HUS. Both  leukemoid reaction and HUS are thought to be related to Shiga toxin production. In this study they occurred significantly more frequently in S. dysenteriae type 1 infection, though they were not restricted to S. dysenteriae type 1 infections. Thirteen patients with S. dysenteriae type 1 infection and eight patients with non-S. dysenteriae type 1 infections met the criteria for HUS. Although the etiologic trigger for HUS is presumed in most cases to be infection with Shiga toxin-producing organisms (either S. dysenteriae type 1, or Shiga toxin-producing Escherichia coli) there are childhood cases of HUS that are not associated with Shiga toxin-producing organisms. [29] Further studies are needed to understand more about the pathogenesis and clinical course of these patients with HUS caused by non-S. dysenteriae type 1 strains of Shigella.
Severe hyponatremia, though more often found with S. dysenteriae type 1 infection, was common in infections with all species of Shigella. This suggests that it is not solely a Shiga-toxin related phenomenon. We have previously shown that it is associated with altered consciousness in this group of patients. [14] The cause of hyponatremia during shigellosis is uncertain, but preliminary studies suggest that it may be due to inappropriate secretion of antidiuretic hormone [30].
During the one year of this study the death rate in patients with shigellosis was high -83 (10%) of the 792 children less than 15 years admitted to the inpatient unit died. Another 5% were transferred to another facility because of complications that could not be treated at the icddr, b Dhaka hospital, or left the hospital against medical advice. Although the outcome of these patients was not known, patients in both groups were often severely ill, and it is likely that a number of them also died. None of the 71 patients 15 years or older died, including 15 older than 60 years. Most of the factors predictive of death in this study -young age, poorer nutritional status, lower serum sodium, lower stool frequency, convulsions and unconsciousness -have been associated with death in shigellosis in previous studies (Table 7). [16,20,31,32,33,34] The two studies examining risk factors for death that were not done in Bangladesh (one in Rwanda, one in Zimbabwe, and both restricted to S. dysenteriae type 1-infected patients) both found severe dehydration as predictive of death. [31,32] In this and other studies dehydration was neither common nor associated with death. For reasons we cannot explain, stool frequency was inversely associated with risk of death.
The high death rate seen in patients with shigellosis in developing countries such as Bangladesh contrasts dramatically with the rarity of deaths from shigellosis found in industrialized countries. [33,35] Given that deaths occurred with all Shigella species, including those species -S. sonnei and S. flexneri -that cause infection in industrialized countries -the differential rates in mortality between rich and poor countries are most likely due to differences in host factors (including underlying nutritional status) and health care. [36] We have shown that early treatment (within 72 hours of the onset of illness) of S. dysenteriae type 1 virtually eliminates the risk of developing HUS. [37] In this study 68% of patients were ill for more than 72 hours before coming for care, and by that time many of the complications of illnesshypoproteinemia, alterations in consciousness, and HUS -were well established. Preventing the more serious intestinal and extraintestinal manifestations of shigellosis, and death, requires both earlier treatment of clinical disease and improvements in nutritional conditions of children. Neither of these recommendations is easy to implement, as antimicrobial resistance in Shigella is increasingly common, [38] the health care systems required to provide early care continue to struggle in many developing countries, and the same conditions of poverty that lead to high rates of shigellosis also lead to the high rates of malnutrition seen in these poor communities [39].
Since Kiyoshi Shiga first identified the etiologic agent of bacillary dysentery in 1897 [40] there have been a number of descriptions of the clinical manifestations of illness caused by Shigella. [1,9,41,42] Many of these reports have focused on Shigella infections and complications, usually rare, that occur at sites outside the gut, including the eye [41,43] or the genitourinary tract. [41,44,45] We did not actively search for infections at these sites, nor did we have the longer term follow-up that would allow us to accurately ascertain the risk of post-infectious complications, such as Reiter's syndrome, that have been infrequently reported to occur during shigellosis [46].
This study demonstrates that extra-intestinal manifestations of shigellosis are not confined to infection with S. dysenteriae type 1, as is commonly thought. The spectrum of severe intestinal and extraintestinal complications that occur with all species of Shigella make this a challenging illness to treat effectively once complications have developed, especially in resource constrained settings and in malnourished hosts. Early treatment with an effective antimicrobial agent is likely to lessen both morbidity and mortality, especially the development of severe intestinal and extra-intestinal manifestations of illness. But increasing resistance to commonly used antimicrobial agents also makes this option an increasing challenge.