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.

Methods

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.

Results

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.

To determine the proportion of children with Shigella infection presenting as outpatients who were admitted to the inpatient unit, we extrapolated from data from the systematic surveillance system. 83,402 patients came to the icddr, b Dhaka hospital for care during the study period, of whom 55,644 (67%) were <15 years. Shigella was isolated from a stool or rectal swab sample of 209 (9.1%) of the 2,292 patients <15 years of age entered in the 4% systematic sample of outpatients. If this proportion is extrapolated to all outpatients <15 years of age, an estimated 5,074 outpatients were children <15 years of age infected with Shigella. Seven-hundred ninety-two inpatients with Shigella infection <15 years made for an admission rate of 16% for outpatients infected with Shigella, a rate of admission significantly (P<0.001) higher than the 10% rate for children not infected with Shigella. Shigellosis accounted for 14% (792/5,745) of patients <15 years of age admitted to the inpatient unit.

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).

Table 1

Isolation of Shigella Species by Age Group of Patients Admitted to the Dhaka Hospital of the ICDDR, B.

Age group	S. dysenteriae type 1 (n = 165)	S. dysenteriae type 2–10(n = 28)	S. flexneri (n = 555)	S. boydii (n = 85)	S. sonnei (n = 30)	P for overall comparison
<1 year	31 (19)	6 (21)	217 (39)	34 (40)	18 (60)	<0.001*
1 to 5 years	108 (65)	15 (54)	259 (47)	38 (45)	11 (37)	<0.001†
>5 to <15 years	18 (11)	3 (11)	28 (5)	5 (6)	1 (3)	0.069
≥15 years	8 (5)	4 (14)	51 (9)	8 (9)	0	0.117

Values are n (% of patients in age group).

P<0.040; S. dysenteriae type 1 versus S. flexneri, S. boydii or S.sonnei; S. dysenteriae types 2–10 versus S. sonnei; S. flexneri versus S. sonnei.

P<0.006: S. dysenteriae type 1 versus S. flexneri, S. boydii or S. sonnei.

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.

Table 2

Admission Clinical Characteristics of 792 Inpatients <15 Years by Species of Shigella.

Characteristic	S. dysenteriae type 1 (n = 157)	S. dysenteriaetype 2–10 (n = 24)	S. flexneri(n = 504)	S. boydii(n = 77)	S. sonnei(n = 30)	P for overallcomparison	All patients withnon-S. dysenteriaetype 1 infection(n = 635)	P for S. dysenteriae type 1 versus non-S. dysenteriae type 1 infection
Age, m (median, 25th, 75th quartiles)	24 (13, 43)	26 (10, 50)	13 (7, 30)	13 (6, 30)	8 (4, 15)	<0.001*	13 (6, 30)	<0.001
Male	92 (59)	15 (63)	285 (57)	43 (56)	18 (60)	0.958	361 (57)	0.719
Duration of illness, d (median, 25th, 75th centiles) †	6 (4, 9)	2 (1, 10)	5 (3, 14)	5 (3, 15)	5 (3, 10)	0.051‡	5 (3, 14)	0.405
Severe dehydration	4 (3)	0	23 (5)	3 (4)	3 (10)	0.304	29 (5)	0.362
Temperature ≥ 38.0°C	107 (68)	12 (50)	265 (53)	27 (35)	15 (50)	<0.001§	319 (50)	<0.001
Weight for age, % of median (mean ± SD) || ¶	62±14	65±7	57±15	55±13	59±16	<0.001**	57±14	<0.001
Pedal edema	27 (17)	4 (17)	88 (18)	11 (14)	2 (7)	0.603	105 (17)	0.842

Values are n (%) unless noted.

P<0.020: S. dysenteriae type 1 versus S. flexneri, S. boydii, or S. sonnei; S. dysenteriae types 2–10 versus S. flexneri, S. boydii or S. sonnei; S. flexneri versus S. sonnei. P<0.008: S. dysenteriae type 1 versus non-S. dysenteriae type 1.

Duration of illness data were missing for 2/157 patients in the S. dysenteriae type 1 group, 5/504 patients in the S. flexneri group, and 1/77 patient in the S. boydii group.

P<0.009: S. dysenteriae types 2–10 versus S. dysenteriae type 1, S. flexneri, S. boydii or S. sonnei.

P<0.007: S. dysenteriae type 1 versus S. flexneri or S. boydii; S. flexneri versus S. boydii.

Weight-for-age was calculated as a percentage of the United States National Center for Health Statistics median weight-for-age17.

Weight-for-age data were missing for 5/157 patients in the S. dysenteriae type 1 group, 1/24 patient in the S. dysenteriae type 2–10 group, and 11/504 patients in the S. flexneri group.

P<0.002: S. dysenteriae type 1 versus S. flexneri or S. boydii; S. dysenteriae types 2–10 versus S. flexneri or S. boydii.

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 Shigella-infected 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.

Table 3

Intestinal Manifestation of Shigellosis in 792 Inpatients <15 Years by Species of Shigella.

Clinical characteristic	S. dysenteriaetype 1 (n = 157)	S. dysenteriae. type 2–10 (n = 24)	S. flexneri(n = 504)	S. boydii(n = 77)	S. sonnei(n = 30)	P – overall comparison	All patients with non-S. dysenteriae type 1 infection (n = 635)	P for S. dysenteriae type 1 versus non-S. dysenteriae type 1 infection
Grossly bloody stool *	122 (78)	5 (21)	180 (36)	19 (25)	6 (20)	<0.001†	210 (33)	<0.001
Number of stools in 24 h before admission, (median, 25th, 75th centiles) ‡	25 (12, >100)	10 (6, 15)	12 (7, 20)	10 (6, 15)	10 (7, 15)	<0.001§	11 (7, 20)	<0.001
Rectal prolapse ||	79 (52)	2 (8)	86 (18)	4 (5)	1 (3)	<0.001¶	93 (15)	<0.001
Abdominal examination
Decreased or absent bowel sounds	4 (3)	0	5 (1)	3 (4)	0	0.215	8 (1)	0.268
Abdominal distention	21 (13)	2 (8)	57 (11)	18 (23)	2 (7)	0.036**	79 (12)	0.856
Abdominal tenderness	41 (26)	3 (13)	27 (5)	5 (7)	1 (3)	<0.001††	36 (6)	<0.001
Stool microscopic examination
Leukocyte count ‡‡
≤10 per high-powered field	10 (7)	1 (4)	64 (14)	25 (38)	3 (12)	<0.001§§	93 (16)	<0.001
11–50 per high-powered field	35 (24)	10 (44)	151 (33)	21 (31)	15 (63)		197 (35)
>50 per high-powered field	101 (69)	12 (52)	241 (53)	21 (31)	6 (25)		280 (49)
Erythrocyte count §§
0 per high-powered field	11 (7)	8 (35)	97 (21)	32 (48)	4 (17)	<0.001|| ||	141 (25)	<0.001
1–10 per high-powered field	33 (23)	7 (30)	188 (41)	24 (36)	17 (71)		236 (41)
11–50 per high-powered field	38 (26)	6 (26)	113 (25)	7 (10)	2 (8)		128 (22)
>50 per high-powered field	64 (44)	2 (9)	60 (13)	4 (6)	1 (4)		67 (12)

Values are n (%), unless noted.

Stool character data were missing for 1/157 patient in the S. dysenteriae group, 1/504 in the S. flexneri group, and 1/77 in the S. boydii group.

P<0.001:S. dysenteriae type 1 versus S. dysenteriae type 2–10, S. flexneri, S. boydii or S. sonnei.

Stool frequency data were missing for 3/157 patients in the S. dysenteriae type 1 group, 9/504 patients in the S. flexneri group, 5/77 patients in the S. boydii group, and 1/30 patient in the S. sonnei group.

P<0.001: S. dysenteriae type 1 versus S. dysenteriae type 2–10, S. flexneri, S. boydii, or S. sonnei.

Rectal prolapse data were missing for 6/157 patients in the S. dysenteriae type 1 group, 13/504 patients in the S. flexneri group, 3/77 patients in the S. boydii group, and 1/30 patient in the S. sonnei group.

P<0.040: S. dysenteriae type 1 versus S. dysenteriae type 2–10, S. flexneri, S. boydii, or S. sonnei; S. flexneri versus S. boydii.

P = 0.006, S. flexneri versus S. boydii.

P<0.015: S. dysentereriae type 1 versus S. flexneri, S. boydii, or S. sonnei.

Stool leukocyte count data were missing for 11/157 patients in the S. dysenteriae type 1 group, 1/24 patient in the S. dysenteriae type 2–10 group, 48/504 patients in the S. flexneri group, 10/77 patients in the S. boydii group, and 6/30 patients in the S. sonnei group.

P<0.020: S. dysenteriae type 1 versus S. flexneri, S. boydii, or S. sonne; S. dysenteriae type 2–10 versus S. boydii; S. flexneri versus S. boydii, or S. sonnei; S. boydii versus S.sonnei.

Stool erythrocyte count data were missing for 11/157 patients in the S. dysenteriae type 1 group, 1/24 patient in the S. dysenteriae type 2–10 group, 46/504 patients in the S. flexneri group, 10/77 patients in the S. boydii group, and 6/30 patients in the S. sonnei group.

P<0.050: S. dysenteriae type 1 versus S. dysenteriae type 2–10, S. flexneri, S. boydii, or S. sonnei; S. dysenteriae type 2–10 versus S. sonnei; S. flexneri versus S. boydii, or S. sonnei; S. boydii versus S.sonnei.

Extra-intestinal Manifestation

On admission, or during their hospital stay, children infected with S. dysenteriae type 1, in comparison with children infected with other species or serotypes of Shigella, were significantly (P<0.035) more likely to have leukemoid reaction (22% versus 2%), HUS (8% versus 1%), hyponatremia as manifested by a serum sodium <126 mmol/l (58% versus 26%), hypoproteinemia <60 gm/l (81% versus 62%), and any neurological abnormality (24% versus 16%). Features that approached statistical significance were convulsions during hospitalization (8% versus 4%, P = 0.079); and unconsciousness (13% versus 8%, P = 0.121) (Table 4).

Table 4

Extra-intestinal Manifestations of Shigellosis in 792 Inpatients <15 Years by Species of Shigella.

Characteristics	S. dysenteriaetype 1 (n = 157)	S. dysenteriae type2–10 (n = 24)	S. flexneri(n = 504)	S. boydii(n = 77)	S. sonnei(n = 30)	P – overallcomparison	All patients with non-S. dysenteriae type 1infection (n = 635)	P for S. dysenteriae type 1 versus non-S. dysenteriae type 1 infection
Altered neurological state
Convulsion
Documented in hospital	13 (8)	2 (8)	26 (5)	0	0	0.052*	28 (4)	0.079
By history (before admission)	4 (3)	2 (8)	12 (4)	5 (7)	1 (3)	0.178	20 (3)	1.0
Unconsciousness	20 (13)	4 (17)	45 (9)	2 (3)	2 (7)	0.083	53 (8)	0.121
Any neurologic abnormality	37 (24)	8 (33)	83 (17)	7 (9)	3 (10)	0.009†	101 (16)	0.032
Leukemoid reaction (peripheral blood white cells >40,000/mm3) ‡	32 (22)	0	10 (2)	0	2 (7)	<0.001§	12 (2)	<0.001
Hematocrit, (mean ± SD) ||	36±7	37±5	34±6	34±6	35±7	0.001¶	34±6	0.001
Creatinine (mmol/L; median & interquartile) **	94 (78, 135)	107 (72, 160)	84 (67, 112)	88 (73, 116)	77 (68, 137)	0.024††	86 (68, 114)	0.003
Hemolytic-uremic syndrome ‡‡	13 (8)	0	4 (1)	2 (3)	2 (7)	<0.001§§	8 (1)	<0.001
Severe hyponatremia (<126 mmol/L) || ||	86 (58)	3 (13)	129 (27)	14 (21)	5 (18)	<0.001¶¶	151 (26)	<0.001
Hypoglycemia (<2.2 mmol/L) ***	7 (19)	1 (20)	21 (21)	5 (36)	0	0.521	27 (21)	0.797
Hypoproteinemia (<60 gm/L) †††	42 (81)	4 (44)	82 (65)	12 (48)	5 (71)	0.030‡‡‡	103 (62)	0.011
Bacteremia §§§	15 (13)	2 (12)	47 (14)	5 (11)	5 (25)	0.647	59 (14)	0.858

Values are worst (most abnormal) during hospital stay. Values are n (%).

P<0.040: S. dysenteriae type 1 versus S. boydii, S. flexneri versus S. boydii.

P<0.05: S. dysenteriae type 1 versus S. boydii; S. dysenteriae types 2–10 versus S. flexneri, S. boydii or S. sonnei.

Blood leukocytes data were missing for 10/157 patients in the S. dysenteriae type 1 group, 26/504 patients in the S. flexneri group, 6/77 patients in the S. boydii group, and 1/30 patient in the S. sonnei group.

.P<0.001: S. dysenteriae type 1 versus S. dysenteriae types 2–10, S. flexneri, S. boydii.

Hematocrit data were missing for 4/157 patients in the S. dysenteriae type 1 group, 23/504 patients in the S. flexneri group, and 5/77 patients in the S. boydii group.

P<0.019: S. dysenteriae type 1 versus S. flexneri or S. boydii; S dysenteriae type 2–10 versus S. flexneri or S. boydii.

Serum Creatinine data were missing for 53/157 patients in the S. dysenteriae type 1 group, 9/24 patients in the S. dysenteriae types 2–10 group, 304/504 patients in the S. flexneri group, 51/77 patients in the S. boydii group, and 20/30 patients in the S. sonnei.

P = 0.001: S. dysenteriae type 1 versus S. flexneri.

Diagnosis of hemolytic-uremic syndrome required three criteria to be met: 1) an absolute packed cell volume <20%; or 2) a decrease in absolute packed cell volume of >10% in 24 hours; or 3) ≥0.5% schistocytes on a peripheral blood; and 4) serum creratinine, >180 mmol/L.

P<0.050: S. dysenterie type 1 vesus S. flexneri; S. flexneri versus S. sonnei.

Serum sodium data were missing for 9/157 patients in the S. dysenteriae type1 group, 34/504 patients in the S. flexneri group, 9/77 patients in the S. boydii group, and 2/30 patients in the S. sonnei.

P<0.001: S. dysenteriae type 1 versus S. dysenteriae type 2–10, S. flexneri, S. boydii or S. sonnei.

Blood glucose data were missing for 121/157 patients in the S. dysenteriae type 1 group, 19/24 patients in the S. dysenteriae types 2–10 group, 402/504 patients in the S. flexneri group, 63/77 patients in the S. boydii group, and 25/30 patients in the S. sonnei group.

Serum protein data were missing for 105/157 patients in the S. dysenteriae type 1 group, 15/24 patients in the S. dysenteriae types 2–10 group, 378/504 patients in the S. flexneri group, 52/77 patients in the S. boydii group; and 23/30 patients in the S. sonnei group.

P<0.035: S. dysenteriae type 1 versus S. dysenteriae types 2–10, or S. boydii.

Blood culture data were missing for 42/157 patients in the S. dysenteriae type 1 group, 5/24 patients in the S. dysenteriae types 2–10 group, 172/504 patients in the S. flexneri group, 32/77 patients in the S. boydii group; and 10/30 patients in the S. sonnei group.

Outcome

Mortality in patients with Shigella infection was high - 83 (10%) of the 792 patients less than 15 years died (Table 5). Mortality was high for all species and serotypes, except for those infected with S. dysenteriae type 2–10, only 1 (4%) of whom died. There were no deaths in the 71 patients 15 years or older infected with Shigella.

Table 5

Outcome of 792 patients <15 Years by Species of Shigella.

Age group	S. dysenteriae type 1 (n = 157)	S. dysenteriae types 2–10 (n = 24)	S. flexneri (n = 504)	S. boydii (n = 77)	S. sonnei (n = 30)	P, overall comparison	All patients with non-S. dysenteriae type 1 infection (n = 635)	P for S. dysenteriae type 1 versus non-S. dysenteriae type 1 infection
Discharged improved	97 (62)	19 (79)	331 (65)	59 (77)	19 (63)	0.131	428 (67)	0.215
Discharged against medical advice	32 (20)	4 (17)	91 (18)	8 (10)	7 (23)	0.374	110 (17)	0.436
Referred	11 (7)	0	29 (6)	2 (3)	0	0.262	31 (5)	0.387
Died in hospital	17 (11)	1 (4)	53 (11)	8 (10)	4 (13)	0.861	66 (10)	0.989

Values are n (% of patients by species).

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).

Table 6

Features Predictive of Death in 792 Patients <15 Years with Shigellosis in a Multiple Logistic Regression Analysis.

Characteristic	Odds Ratio	95% Confidence Interval	P
Age, m	0.966	0.947–0.986	0.001
Number of stools before admission	0.976	0.961–0.991	0.002
Percent of median weight-for-age *	0.957	0.932–0.983	0.001
Serum sodium <126 mml/L	3.75	1.93–7.29	<0.001
Convulsion	14.60	5.31–40.13	<0.001
Unconsciousness	45.40	17.48–117.96	<0.001

Median weight-for-age is based on United States National Center for Health Statistics standard17.

Analysis was limited to 451 (86%) of the 525 patients discharged improved, and 65 (78%) of the 83 patients who died for whom information on each of the variables entered into the final iteration of the multiple logistic regression analysis was available.

Variables significant (P<0.05) in the bivariate analysis but excluded from the multiple logistic regression analysis because information was available for a limited number of the 516 patients were blood glucose, available for 127 (25%) patients, and serum protein, available in 163 (32%) patients.

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 community-based 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.

Table 7

Summary of Previous Studies Examining Risk Factors for Death in Patients with Shigellosis.

Study Author, Reference	Country	Study Date	Age of Patients	Patients, n	Deaths, n (%)	Shigella Species	Predictors of Death	Odds ratio	Comment
Bennish [16]	Bangladesh	1983	≤10 years	201	67 (34)	23% S. dysenteriae type 1,4% S. dysenteriae type 2–10,62% S. Flexneri, 6% S.boydii, 5% S. sonnei	Age in months	0.969	Case-control study conducted in hospitalized patients in an urban area
							Serum protein (g/l)	0.945
							Altered consciousness	4.80
							Thrombocytopenia	9.29
Mitra [34]	Bangladesh	1980	≤5 years	46	23 (50)	Not recorded	Female	4.30	Case-control study conducted in hospitalized patients in a rural area
							Signs of lower respiratory infection	24.0
							Severe malnutrition (<60% of weight for age)	8.90
Huskins [20]	Bangladesh	1984–1988	≤3 months	121	26 (21)	9% S. dysenteriae type 1,3% S. dysenteriae type 2–10,59% S. flexneri, 21% S. boydii,8% S. sonnei	Gram-negative bacteremia	14.61	Prospective study of infants and young children in an urban hospital. Study patients were a subset of patients in the current study
							Ileus	93.83
							Decreased bowel sounds	86.81
							Serum sodium (mmol/L)	0.884
							Serum protein concentration (g//L)	0.819
							Number of erythrocytes on stool microscopic examination (high power field)	0.931
Nathoo [32]	Ziwakhanmbabwe	1993–1994	1 month to 12 years	312	95 (30)	100% S. dysenteriae type 1	Temperature (<36.0°C)	2.12	Study conducted in two urban tertiary referral hospitals. Data from 264 patients collected prospectively; data on 48 patients collectedRetrospectively
							Severe dehydration	1.70
							Serum sodium (<120 mmol/L)	1.57
							Serum potassium (>5.5 mmol/L)	1.41
							Urea (>8 mmol/L)	1.74
							Abdominal distention	1.67
Legros [31]	Rwanda	1994	All ages	849	108 (13)	100% S. dysenteriae type 1	Severe dehydration	2.79	Prospective study conducted in 10 rural hospitals
							Pedal edema	2.20
							Age (<5 years or >50 years)	3.22
							Use of nalidixic acid	8.66
Van den Broek [33]	Bangladesh	1993–1999	≤4 years	200	100 (50)	19% S. dysenteriae type 1, 81% S. flexneri	Altered consciousness	2.60	Case control study in urban hospital restricted to malnourished children (<60% of median weight-for- age)
							Pneumonia	2.50
							Hypoglycemia (<3 mmol/L)	7.80
							Temperature (<36.0°C)	5.70
Khan, this study	Bangladesh	1987–1988	<15 years	792	83 (10)	20% S. dysenteriae type 1, 3% S. dysenteriae type 2–10, 63% S. Flexneri, 10% S. boydii, 4% S. sonnei	Age, m	0.966	Prospective study in an urban hospital
							Number. of stools before admission	0.976
							Percent of median weight-for-age	0.957
							Serum sodium, (<126 µmol/l)	3.75
							Convulsion	14.60
							Unconsciousness	45.40

All studies used multivariate logistic regression analysis to determine factors predictive of death with the exception of Mitra (reference 34), which used bivariate analysis.

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 illness - hypoproteinemia, alterations in consciousness, and HUS - were well established. Preventing the more serious intestinal and extra-intestinal 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 extra-intestinal 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.