Download Diagnostic Accuracy of Stool Assays for Inflammatory Bacterial

MAJOR ARTICLE
Diagnostic Accuracy of Stool Assays
for Inflammatory Bacterial Gastroenteritis
in Developed and Resource-Poor Countries
Christopher J. Gill,1,2,4 Joseph Lau,2 Sherwood L. Gorbach,3,4 and Davidson H. Hamer1,4
1
Center for International Health and Development, Boston University School of Public Health, and 2Division of Clinical Care Research, 3Department
of Family Medicine and Community Health, and 4Division of Geographic Medicine and Infectious Diseases, Tufts–New England Medical Center,
Boston, Massachusetts
Because acute bacterial gastroenteritis is often inflammatory, rapid stool assays that detect intestinal inflammation might be used to distinguish between bacterial and nonbacterial gastroenteritis. We performed metaanalyses to determine the discriminatory power, in developed and in resource-poor countries, of rapid stool
assays that test for lactoferrin, fecal leukocytes, fecal erythrocytes, and occult blood. In developed countries,
the area under the summary receiver operating characteristic curve (AUC/SROC) was 0.89 for fecal leukocytes
and 0.81 for occult blood. In resource-poor countries, the AUC/SROC was 0.79 for lactoferrin, 0.72 for fecal
leukocytes, 0.63 for occult blood, and 0.61 for fecal erythrocytes. In developed countries, positive and negative
likelihood ratios (LR+ and LR⫺, respectively) for fecal leukocytes were 4.56 and 0.32 when a threshold of 15
cells/high-power field was used, compared with 2.94 and 0.6 in resource-poor countries; for lactoferrin, LR+
was 1.34 and LR⫺ was 0.17 in resource-poor countries when the threshold was an agglutination rating of “+”
and a dilution of 1:50. In developing countries, rapid stool assays performed poorly, whereas in developed
countries, tests for fecal leukocytes, lactoferrin, and occult blood were moderately useful and could identify
patients who were more likely to benefit from empirical antibiotic therapy.
More than 1 billion cases of acute diarrhea occur worldwide each year [1, 2]. Given the limited availability of
rapid pathogen-specific tests and the 24–48 h required
to process stool cultures, patients with acute community-acquired gastroenteritis are often treated empirically with antibiotics. However, most acute diarrhea is
caused by pathogens or agents that do not respond to
antibiotics, such as viruses, preformed bacterial entero-
Received 14 January 2003; accepted 28 March 2003; electronically published
22 July 2003.
Financial support: Agency for Healthcare Research and Quality (National
Research Service Award training grants 5T32AI0738 and T32 HS00060-09).
Reprints or correspondence: Dr. Christopher J. Gill, Center for International
Health and Development, Boston University School of Public Health, 710 Albany
St., Boston, MA 02118 ([email protected]).
Clinical Infectious Diseases 2003; 37:365–75
2003 by the Infectious Diseases Society of America. All rights reserved.
1058-4838/2003/3703-0008$15.00
toxins, and most parasites [1]. Misapplied antibacterial
therapy places patients at risk of medication-related
adverse events, contributes to the emergence of antimicrobial resistance, and represents a needless expenditure of limited resources in poorer countries. Reliable
identification of patients with acute gastroenteritis that
would be amenable to antibacterial therapy at time of
first contact remains a formidable but worthy clinical
challenge.
Invasive forms of bacterial gastroenteritis often cause
intestinal inflammation, whereas most forms with viral,
parasitic, and toxin-mediated etiologies do not [3].
Consequently, rapid stool assays that identify by-products of intestinal inflammation in stool might identify
patients who are more likely to benefit from antibiotics.
Rapid stool assays currently available include microscopic examination for leukocytes and erythrocytes,
guaiac-based fecal occult blood tests, and a rapid im
Stool Assays for Bacterial Gastroenteritis • CID 2003:37 (1 August) • 365
munologic assay for fecal lactoferrin (Leukotest; Techlabs) [3–5].
Published estimates of the diagnostic accuracy of these tests
vary considerably. Although an earlier meta-analysis concluded
that the lactoferrin assay was the preferred test [6], this analysis
did not provide summary estimates of the various assays’ sensitivities and specificities or the area under the summary receiver operating characteristic curve (AUC/SROC), both of
which are needed to judge the overall value of diagnostic tests.
Perhaps more important, this analysis combined studies from
both resource-poor and developed countries without considering how the disease spectrum prevalent in each setting might
influence the diagnostic performance of the tests. Accordingly,
we conducted meta-analyses of studies that have evaluated the
diagnostic accuracy of rapid stool assays for distinguishing bacterial from nonbacterial gastroenteritis. In addition to determining the most accurate test, we wished to understand how
the performance of these assays varied with the prevailing epidemiologic setting in which the study was conducted.
METHODS
Search strategy. We sought studies that measured the diagnostic accuracy of tests for lactoferrin and occult blood and
microscopy for fecal leukocytes and erythrocytes for patients
who presented with acute, community-acquired bacterial gastroenteritis of ⭐7 days’ duration. We searched the MEDLINE
database and the Cochrane Database of Abstracts of Reviews
of Effectiveness for studies published from 1966 through 2002
and conducted secondary and tertiary searches for additional
articles within the bibliographies of retrieved articles [6]. Relevant studies were evaluated using the criteria listed below.
Failure to satisfy any of the inclusion criteria resulted in exclusion from our analysis. Inclusion criteria were the following:
1. Peer-reviewed journal article
2. Published in English
3. Primary data study (i.e., no review articles, meta-analyses, systematic reviews, or expert opinion papers, except as
sources for additional bibliographic references)
Exclusion criteria were the following:
1. Published only in abstract form
2. Published only as a letter to the editor
3. Personal communication
4. Study of patients with autoimmune inflammatory bowel
disease
5. Study of Clostridium difficile colitis
6. Limited to specific pathogens
7. Did not include cultures for Campylobacter species
Although most of these criteria are self-explanatory, several
366 • CID 2003:37 (1 August) • Gill et al.
require clarification. Studies that focused on C. difficile were
excluded because the vast majority of cases of C. difficile enteritis are hospital rather than community acquired. Campylobacter species are common causes of inflammatory enteritis
in both developed and resource-poor countries [7–10]. Recent
statistics show that Campylobacter species cause 38%–50% of
all cases of culture-positive community-acquired gastroenteritis
in the United States [11–14]; however, surprisingly, campylobacters were not recognized as enteric pathogens before the late
1970s [15]. Consequently, many early rapid stool assay studies
were conducted before Campylobacter cultures became routine
and risked misclassifying the positive assay results for the substantial proportion of patients who had campylobacteriosis as
false positive, resulting in inaccurate estimates of sensitivity and
specificity.
Studies retained after application of inclusion and exclusion
criteria were next evaluated using the following validity criteria.
These were designed to select studies with a focus on community-acquired gastroenteritis that were amenable to metaanalysis and to minimize the potential for bias:
1. Data collection was prospective.
2. Case definition was provided.
3. Case definition was compatible with our case definition
(acute-onset diarrhea of ⭐7 days’ duration).
4. Study focused on community-acquired diarrhea.
5. Study provided either sensitivities and specificities or
raw data allowing these to be calculated.
6. Case and control subjects all had diarrhea and were
drawn contemporaneously from the same source population.
7. Stool culture was used as the reference standard.
8. Test thresholds for positivity were defined for each assay. (For fecal WBCs and RBCs, this is the number of cells per
high-power field [HPF]. For lactoferrin, the test threshold is
varied both by adjusting the ratio of stool to saline [dilution
range, 1:50 to 1:200] and by judging the subjective intensity
of the agglutination [rated “+,” fine agglutination visible against
a milky background; “++,” agglutination visible against a white
background, with a visible rim; or “+++,” agglutination visible
against a clear background, with a clear rim]. Occult blood tests
are assumed to provide dichotomous results: any color change
vs. no color change.)
Reference standard. To learn which stool assays are useful
for identifying cases of diarrhea that are theoretically amenable
to empirical antibiotic therapy, we defined our “disease-positive” reference standard as a stool culture on which any of the
following pathogens grew: Campylobacter species, Shigella species, Salmonella species, Yersinia enterocolitica, non-cholerae
Vibrio species, Plesiomonas shigelloides, Aeromonas hydrophila,
and diarrheagenic Escherichia coli (these were enterotoxigenic
E. coli [ETEC] in almost all cases). Stool assays demonstrating
viruses, C. difficile toxin, or parasites or parasite ova or which
had no growth were classified as “disease negative.” Patients
coinfected with pathogens from “disease-positive” and “disease-negative” categories were excluded from our analysis, because which pathogen is responsible for a positive test result
cannot be determined (!1% of subjects were excluded from
our study for this reason); patients infected with multiple pathogens from the “disease-positive” category were considered to
be disease positive.
Sensitivity analysis for ETEC. The fact that some forms
of bacterial enteritidis are noninflammatory and some forms
of nonbacterial enteritidis are highly inflammatory may cause
assays to misclassify certain pathogens. ETEC cause acute antibiotic-treatable diarrhea that is noninflammatory and hence
unlikely to cause a positive assay result [3]. Because ETEC are
common agents of bacterial gastroenteritis, we did analyses in
which cases caused by ETEC were categorized alternately as
disease positive and disease negative to see how this affected
test performance.
Statistical analysis. True-positive, false-positive, true-negative, and false-negative values were used to calculate sensitivities, specificities, and likelihood ratios (LRs). We calculated
95% CIs according to Fleiss’ method [16]. For studies that
provided only summarized sensitivity and specificity data but
no raw data, we used the cohort size and culture-positivity rate
to back-calculate true-positive, false-positive, true-negative, and
false-negative values.
Positive LR values (LR+) were calculated using the formula
[sensitivity/(1 ⫺ specificity)]; negative LR values (LR⫺) were
calculated using the formula [(1 ⫺ sensitivity)/specificity].
Multiplying pretest odds for a disease by LR+ or LR⫺ (depending on whether the test result was positive or negative)
yields posttest odds for that disease. By definition, a test with
an LR+ or LR⫺ of 1.0 is worthless, because it does not alter
posttest odds. Although there is no absolute threshold, an LR+
of 2.0 and an LR⫺ of 0.5 are considered the minimum useful
values [17].
We only performed meta-analysis when ⭓3 data points existed for a test in a given epidemiologic setting. Using the
MetaTest software program [18], we grouped studies for analysis depending on whether they were conducted in a developed
or a resource-poor country, and all summary analyses were
confined to one or the other classification. The SROC method
was our primary analytic tool. We calculated AUC/SROCs according to Moses’ method, by extrapolating curves to the corners of the SROC [19]. As with LRs, there is no absolute AUC/
SROC threshold that defines a “good” test. However, an AUC/
SROC of 1.0 defines a “perfect” test, and an AUC/SROC of
0.5 defines a “useless” test [20]. To reduce the influence of
studies that reported multiple cutoff points, we plotted unweighted curves [19, 21]. Where possible, we calculated com-
bined sensitivities and specificities independently, using random effects [22], on the basis of the most commonly used test
thresholds, and we calculated summary LRs using these pooled
sensitivities and specificities.
RESULTS
From 13000 screened abstracts, we identified 49 candidate studies. Of these, we excluded 27 a priori, including 11 [23–33]
from the earlier meta-analysis [6]. Reasons for exclusion were
as follows: 11 studies tested assays against single pathogens [7,
23, 29–31, 34–39], 11 did not include Campylobacter cultures
[24–28, 33, 40–44], and 5 were abstracts or letters to the editor
[32, 45–48]. Of the remaining 22 studies, 8 violated validity
criteria, including 4 [4, 49–51] from the 1996 meta-analysis
[52]: sensitivities and specificities could not be obtained from
5 [4, 50, 53–55], 5 lacked a case definition [51, 53–56], 3
included cases and controls from separate populations [4, 49,
53], 2 analyzed experimentally induced diarrhea [4, 49], and
1 was a retrospective analysis [51] (several studies violated multiple criteria).
The remaining 14 studies [52, 57–69] included 1 article that
contained 2 distinct data sets, on pediatric and adult cohorts,
collected in different years [65]. We analyzed both cohorts separately, bringing the total to 15 studies and 7161 patients (table 1).
Lactoferrin assay. In the only study that focused on a
developed country and included lactoferrin assays [62], lactoferrin assays provided useful diagnostic information, particularly when the results were negative (table 2). In studies that
focused on resource-poor countries, LR+ values were all !2.0
at the +/1:50 dilution threshold. At the +++/1:50 dilution
threshold, the LR+ of 11.32 from the study by Ruiz-Pelaez and
Mattar [69] was the highest LR obtained for studies in either
epidemiologic setting, although the 95% CIs were wide and the
sensitivity was low (7%).
SROC analysis could not be performed, because only a single
study that focused on a developed country included lactoferrin
assays. In resource-poor countries, the overall AUC/SROC for
lactoferrin was 0.79 (figure 1), compared with 0.51 at the common +/1:50 threshold (curve not shown). At the common
+/1:50 threshold, a positive test result provided little diagnostic
information, although a negative test result substantially decreased the odds that the diagnosis would be bacterial gastroenteritis (table 3).
Fecal occult blood tests. In studies that focused on developed countries, all LR+ values for occult blood tests were
12.0, and all LR⫺ values were !0.5. Conversely, in studies that
focused on resource-poor countries, all LR+ values were !2.0,
and all LR⫺ values were 10.5 (table 4). The combined sensitivities and specificities and the LRs were also superior in studies
conducted in developed countries (table 3). Similarly, the AUC/
Stool Assays for Bacterial Gastroenteritis • CID 2003:37 (1 August) • 367
Table 1.
analysis.
Overview of rapid stool assays used to diagnose inflammatory bacterial gastroenteritis in studies included in a meta-
Type of rapid stool
assay performed
Setting, reference
Year
Country
Alvarado [59]
1983
United Kingdom
Thorson et al. [60]
1985
United States
DeWitt et al. [70]
1985
Loosli et al. [57]
1985
Age range
FLT
OBT
Lacto
FEM
All ages
MB
—
—
—
All ages
MB
—
—
—
United States
!4 years
MB
—
—
—
Switzerland
115 years
NAa
HEMO
—
—
⭓18 years
Developed countries
Siegel et al. [58]
1987
United States
GS
HEMO
—
—
Silletti et al. [62]
1996
United States
NS
MB
SER
LEUK
—
Savola et al. [63]
2001
United States
NS
MB
—
—
—
1983
Bangladesh
SWM
—
—
SWM
Resource-poor countries
Stoll et al. [66]
50% of subjects were
⭐3 years old; 70% were
⭐14 years old
Ascher and Edusada-Corpus [67]
1991
Philippines
⭐6 years
MB
—
—
—
Huicho et al. [52]
1993
Peru
⭐2 years
MB
HEMO
—
—
Scerpella [64]
1994
Mexico
“Adults attending summer
school”
TS
HEMO
LEUK
—
McNeely et al. [65]
1996
Mexico
“Adult college students”
MB
HEMO
—
—
3 months–7 years
TS
HEMO
—
—
Huicho et al. [68]
1997
Peru
⭐2 years
MB
HEMAS
LEUK
—
Ruiz-Pelaez and Mattar [69]
1999
Colombia
6 months–13 years
MB
NS
LEUK
—
NOTE. FEM, fecal erythrocyte microscopy; FLT, fecal leukocyte test; GS, Gram stain; HEMAS, Hemascreen (Alpha Laboratories); HEMO, Hemoccult (Smith
Kline Diagnostics); Lacto, lactoferrin assay; LEUK, Leukotest (Techlabs); MB, methylene blue; NA, not applicable; NS, not stated; OBT, occult blood test; SER,
Seracult (Propper Manufacturing); SWM, saline wet mount; TS, trichrome stain.
a
No threshold was defined for the FLT, and therefore we only analyzed OBT data from that study.
SROC for occult blood tests in studies from developed countries
greatly exceeded that in studies from resource-poor countries
(figure 2).
Microscopy for fecal leukocytes. In developed countries,
there was substantial heterogeneity in the distribution of sensitivities and specificities of microscopy for fecal leukocytes
(table 5). No consistent trend emerged that related the test
threshold (cells/HPF) to any of the diagnostic parameters.
However, the diagnostic accuracy of microscopy for fecal leukocytes tended to be higher in studies from developed countries
than in studies from resource-poor countries.
In the study conducted in Bangladesh [66], the LR+ and
LR⫺ at the 11-cell/HPF cutoff point were 1.0 and 1.0, respectively, which means that the test provided literally no diagnostic
information at this threshold. Of 3558 patients in that study,
at least 1 leukocyte/HPF was found in samples from 3487
(98%), which shows that intestinal inflammation may be ubiquitous in some cohorts and some epidemiologic settings.
The study by Ascher and Edusada-Corpus [67] generated an
LR+ of 26.0, almost 8-fold higher than the next highest LR+.
It may be relevant that this study also had the lowest culture-
368 • CID 2003:37 (1 August) • Gill et al.
positivity rate of studies from resource-poor countries, a feature
that lowers the potential for false-negative results and thus
increases the test’s specificity and LR+. In addition, the study
population may not have been truly representative of resourcepoor countries, because it consisted of children of US military
personnel stationed at a US base in the Philippines.
As can be seen in figure 3, the AUC/SROCs for studies from
developed countries greatly exceeded those for studies from
resource-poor countries (0.89 vs. 0.72). This trend persisted at
the most common threshold of 15 cells/HPF (AUC/SROC, 0.85
and 0.78, respectively). Pooled sensitivities and specificities were
also higher in studies from developed countries than in studies
from resource-poor countries (table 3). Notably, the largest
study—and the one that found the lowest diagnostic power for
microscopy for fecal leukocytes—did not test at the 15-cell/
HPF threshold, and thus it is not included in this calculation,
which will be biased upward as a result [66].
The Bangladesh study of 3558 mostly pediatric patients accounted for 62% of all patients from studies from resourcepoor countries and 50% of patients from studies in both epidemiologic settings combined [66]. Exclusion of the Bangladesh
Table 2.
Evaluation of the lactoferrin test for diagnosis of inflammatory bacterial gastroenteritis.
Setting, reference
Cohort
size
Proportion
of subjects
with positive
a
cultures, %
416
3
121
92
Lactoferrin test
Threshold,
agglutination/
dilution
Sensitivity
(95% CI)
Specificity
(95% CI)
LR+ (95% CI)
LR⫺ (95% CI)
+/1:50
0.92 (0.67–0.99)
0.79 (0.74–0.82)
4.33 (3.41–5.48)
0.10 (0.08–0.11)
24
+/1:50
0.97 (0.83–0.99)
0.15 (0.09–0.24)
1.14 (1.02–1.27)
0.23 (0.14–0.37)
59
+/1:50
0.63 (0.50–0.75)
0.47 (0.32–0.63)
1.20 (0.86–1.66)
0.78 (0.55–1.12)
+/1:200
0.41 (0.29–0.54)
0.79 (0.64–0.89)
1.94 (1.03–3.63)
0.75 (0.61–0.93)
Developed countries
Silletti et al. [62]
Total no. of subjects
for this category
416
Resource-poor countries
Huicho et al. [68]
Scerpella et al. [64]
Ruiz-Pelaez and Mattar [69]
Total no. of subjects
for this category
NOTE.
a
b
b
453
24
+/1:50
1.00 (0.97–1.00)
0.31 (0.26–0.36)
1.45 (1.35–1.55)
0.00 (0.00–0.00)
++/1:50
0.43 (0.34–0.52)
0.89 (0.85–0.92)
3.91 (2.86–5.36)
0.64 (0.58–0.71)
+++/1:50
0.07 (0.03–0.13)
0.99 (0.98–1.00)
11.32 (2.84–45.11)
0.94 (0.90–0.99)
666
LR+, positive likelihood ratio; LR⫺, negative likelihood ratio. +, ++, and +++ indicate degree of agglutination.
No. of subjects with a bacteria-positive stool culture/total no. of participants in that study.
Subjects from this study were only counted once in calculating the total no. of subjects in the category.
study yielded a new AUC/SROC for studies from resourcepoor countries of 0.80, dominated by the 5 data points from
the study by Ruiz-Pelaez and Mattar [69]. The AUC/SROC for
that study alone was 0.81, compared with 0.62 for the study
by Stoll et al. [66] alone (curves not shown).
Microscopy for fecal erythrocytes. All data on fecal erythrocytes came from a single study conducted in Bangladesh [66].
As shown in table 6, the diagnostic accuracy of microscopy for
fecal erythrocytes was poor at all thresholds; summary sensitivity and specificity analyses could not be performed. The
AUC/SROC was 0.61, which suggests that this is a poor test
for identifying bacterial gastroenteritis (data not shown).
Effect of reclassifying ETEC as “disease negative.” Five
studies, with a total of 4257 patients, all from resource-poor
countries, included patients infected with ETEC. All 5 studies
tested microscopy for fecal leukocytes, 4 included occult blood
tests, 1 included the lactoferrin assay, and 1 included microscopy for fecal erythrocytes. Reclassifying ETEC as “disease
negative” did not have a consistent effect on the magnitude or
direction of the recalculated LR+ or LR⫺ values (LR+ values
increased in 3 studies and decreased in 2; LR⫺ values decreased
in 3 studies and increased in 2). Most of these changes were
!1.0 and thus of minimal clinical relevance. The largest changes
were seen in the adult cohort of the studies by McNeely et al.
[65], in which the LR+ for occult blood tests increased by 1.22,
and the LR+ for microscopy for fecal leukocytes at a threshold
of 15 cells/HPF increased by 1.69; by Stoll et al. [66], in which
the LR+ for microscopy for fecal leukocytes at a threshold of
150 cells/HPF increased by 1.39, and the LR+ for microscopy
for fecal erythrocytes at a threshold of 150 cells/HPF increased
by 2.12; and by Scerpella et al. [64], in which the LR+ for
microscopy for fecal leukocytes at a threshold of 11 cell/HPF
decreased by 1.23. Changes in LR⫺ were !0.2 in all cases.
DISCUSSION
This meta-analysis demonstrates that, in developed countries,
tests for fecal leukocytes, lactoferrin, and occult blood are helpful for identifying patients who are more likely to have bacterial
gastroenteritis and who might, theoretically, benefit from empirical antibiotic therapy. By contrast, rapid stool assays were
far less powerful when performed in resource-poor countries.
Overall, microscopy for fecal leukocytes and lactoferrin tests
were superior to occult blood tests and examination for fecal
erythrocytes. In resource-poor countries, occult blood tests and
examination for fecal erythrocytes were essentially worthless,
although data on the latter came from a single study and may
not be representative of resource-poor countries in general [66].
In contrast with the findings of an earlier meta-analysis [6],
we were unable to conclude that the lactoferrin assay was clearly
superior to examination for fecal leukocytes. However, from a
practical standpoint, the relative speed and simplicity of the
lactoferrin test, compared with stool microscopy, may be the
deciding factor in choosing which test to adopt. This point also
pertains to the use of occult blood tests in developed countries:
we believe that the convenience and low cost of this test compellingly support its continued use, even though it is less powerful than either the lactoferrin test or microscopy for fecal
leukocytes. Conversely, the ability to vary thresholds represents
a significant advantage associated with use of the lactoferrin
Stool Assays for Bacterial Gastroenteritis • CID 2003:37 (1 August) • 369
Figure 1. Summary receiver operating characteristic (SROC) curve for
use of the fecal lactoferrin assay to diagnose bacterial gastroenteritis in
resource-poor countries. The area under the curve is 0.79. Only a single
study that focused on a developed country included lactoferrin assays,
and this was insufficient for plotting an SROC curve. Reference nos. are
shown in brackets, and the test threshold, expressed as agglutination
(rated “+,” “++,” or “+++”)/dilution, is shown to the right. The width of
the oval is proportional to the square root of the no. of persons without
disease in the cohort; the height of the oval is proportional to the square
root of the no. of persons with disease in the cohort.
test or examination for fecal leukocytes, compared with occult
blood tests, because test thresholds can be adjusted to better
optimize performance on the basis of the local epidemiology
of acute diarrheal disease.
Table 3.
There are several possible explanations for the superiority of
the performance of rapid stool assays in developed countries
over that in resource-poor countries. One is that technological
factors reduce the diagnostic accuracy of the reference standard
in resource-poor settings. The resulting misclassification due
to false-negative stool culture results would likely bias any assessment of these tests’ accuracy towards the null.
Spectrum bias provides another possible explanation. Spectrum bias refers to the ways in which the accuracy of a diagnostic test varies depending on such factors as who performs
the test, where and how the study was conducted, the characteristics of the study population, disease severity, and the
varying prevalence and virulence of diarrheagenic pathogens in
different regions of the world [71]. Any or all of these factors
may make it difficult to infer cause and effect when evaluating
a test. For example, if a previously healthy child in a developed
country presents with gastroenteritis and has detectable fecal
leukocytes and a stool culture positive for Salmonella enterica,
it is reasonable to assume that the 2 findings are linked. Ascribing this causal relationship is more difficult in settings where
diarrheal illness is frequent and the causes diverse and multifactorial. Was the assay result positive because of the Shigella
species found in the child’s stool culture that day? Or was it
the result of the child’s preexisting vitamin A deficiency [72],
protein deficiency enteropathy [73], or the Blastocystis hominis,
hookworm, or Strongyloides stercoralis colonizing the child’s
intestines? In situations in which cause and effect are blurred,
rapid stool assays are probably too nonspecific to provide much
clinically useful information. It is sadly ironic that these assays
are least useful in settings where the incidence of and morbidity
from diarrhea is greatest and where a rapid, inexpensive diagnostic test would, presumably, be most valuable.
Even though the stool assays performed better in studies
Pooled analyses of selected rapid stool assays, by epidemiologic setting.
Setting, test
No. of
studies
Test threshold
Sensitivity
a
(95% CI)
Specificity
a
(95% CI)
LR+
b
LR⫺
b
Developed countries
Fecal occult blood test
3
Any color change
0.71 (0.36–0.91)
0.79 (0.40–0.96)
3.38
0.37
Microscopy for fecal
leukocytes
6
15 cells/HPF
0.73 (0.33–0.94)
0.84 (0.50–0.96)
4.56
0.32
Lactoferrin test
3
+/1:50c
0.95 (0.48–1.00)
0.29 (0.17–0.46)
1.34
0.17
Fecal occult blood test
6
Any color change
0.44 (0.32–0.57)
0.72 (0.60–0.82)
1.57
0.78
Microscopy for fecal
leukocytes
8
15 cells/HPF
0.50 (0.33–0.67)
0.83 (0.74–0.89)
2.94
0.60
Resource-poor countries
d
NOTE. Pooled analysis was conducted when ⭓3 studies were present at a given threshold. HPF, high-power field; LR+,
positive likelihood ratio; LR⫺, negative likelihood ratio. +, ++, and +++ indicate degree of agglutination.
a
Pooled sensitivity and specificity were calculated independently using a random-effects model.
95% CIs were not calculated for pooled LRs.
Agglutination/dilution.
d
This summary LR⫺ is unrealistically low; its calculation included a study that reported a sensitivity of 100% [69], and, thus,
an LR⫺ of 0.00.
b
c
370 • CID 2003:37 (1 August) • Gill et al.
Table 4.
Evaluation of fecal occult blood tests for diagnosis of inflammatory bacterial gastroenteritis.
Cohort size,
no. of
subjects
Proportion
of subjects
with positive
a
cultures, %
Loosli et al. [57]
111
32
Siegel et al. [58]
113
42
Silletti et al. [62]
416
3
Setting, reference
Fecal occult blood test
Threshold
(test)
Sensitivity
(95% CI)
Specificity
(95% CI)
LR+ (95% CI)
LR⫺ (95% CI)
+ (Hemoccult)
0.80 (0.64–0.90)
0.63 (0.52–0.73)
2.17 (1.57–3.00)
0.32 (0.25–0.39)
+ (Hemoccult)
0.87 (0.75–0.94)
0.58 (0.46–0.69)
2.07 (1.53–2.80)
0.22 (0.18–0.28)
+ (Seracult)
0.31 (0.13–0.58)
0.96 (0.94–0.98)
7.75 (4.51–13.32)
0.72 (0.56–0.93)
Developed countries
Total no. of subjects
for this category
640
Resource-poor countries
92
59
+ (Hemoccult)
0.33 (0.22–0.47)
0.82 (0.7–0.91)
1.81 (0.92–3.57)
0.82 (0.67–0.99)
Huicho et al. [52]
Scerpella et al. [64]
280
40
+ (Hemoccult)
0.39 (0.31–0.49)
0.64 (0.56–0.71)
1.08 (0.87–1.35)
0.95 (0.82–1.10)
Huicho et al. [68]
121
24
+ (Hemascreen)
0.79 (0.62–0.90)
0.50 (0.40–0.60)
1.59 (1.23–2.04)
0.41 (0.32–0.53)
McNeely et al. [65]
Adult cohort
887
48
+ (Hemoccult)
0.26 (0.22–0.31)
0.85 (0.82–0.88)
1.79 (1.43–2.24)
0.86 (0.82–0.91)
Pediatric cohort
140
26
+ (Hemoccult)
0.59 (0.43–0.74)
0.65 (0.55–0.74)
1.70 (1.25–2.31)
0.62 (0.50–0.77)
453
24
0.39 (0.31–0.49)
0.79 (0.74–0.83)
1.86 (1.49–2.33)
0.77 (0.69–0.86)
Ruiz-Pelaez and Mattar [69]
Total no. of subjects
for this category
NOTE.
a
NS
1973
LR+, positive likelihood ratio; LR⫺, negative likelihood ratio; NS, not stated.
No. of subjects with a bacteria-positive stool culture/total no. of participants in that study.
from developed countries, none of these tests were particularly
powerful. Most LR+ values were 3–8, and most LR⫺ values
were 0.5–0.1. For comparison, an earlier—and now obsolete—HIV antibody test had an LR+ of 327 and an LR⫺ of
0.014 [74]. Rapid stool assays will be most useful for clinical
decision-making when the cause of the diarrhea is truly ambiguous (i.e., in situations in which the pretest probability of
bacterial gastroenteritis approaches 50%, or 1:1 odds). In this
situation, a 3–8-fold increase or decrease in the odds of disease
usefully informs the decision to administer or withhold em-
Figure 2. Summary receiver operating characteristic curve for use of fecal occult blood test to diagnose bacterial gastroenteritis in developed
countries (A; area under the curve [AUC], 0.81) and resource-poor countries (B; AUC, 0.63). Reference nos. are shown in brackets. The width of the
oval is proportional to the square root of the no. of persons without disease in the cohort; the height of the oval is proportional to the square root
of the no. of persons with disease in the cohort.
Stool Assays for Bacterial Gastroenteritis • CID 2003:37 (1 August) • 371
156
223
55
Siegel et al. [58]
DeWitt [70]
Savola et al. [63]
Thorson et al. [60]
b
b
121
180
Huicho et al. [68]
Ascher and Edusada-Corpus [67]
b
a
0.20 (0.13–0.28)
0.03 (0.01–0.08)
120
150
15
15
15
15
0.83 (0.64–0.93)
0.69 (0.51–0.83)
0.29 (0.22–0.38)
0.35 (0.22–0.51)
0.28 (0.24–0.32)
0.47 (0.38–0.56)
110
15
0.63 (0.53–0.71)
0.17 (0.15–0.20)
150
0.92 (0.85–0.96)
0.38 (0.35–0.41)
120
15
0.60 (0.57–0.63)
⭓1
0.98 (0.97–0.99)
⭓1
0.50 (0.37–0.63)
0.87 (0.71–0.95)
0.35 (0.19–0.55)
0.85 (0.68–0.94)
1.00 (0.93–1.00)
0.80 (0.74–0.86)
0.31 (0.13–0.58)
0.97 (0.93–0.99)
0.60 (0.50–0.69)
0.84 (0.78–0.89)
0.75 (0.66–0.82)
0.85 (0.82–0.88)
0.98 (0.96–0.99)
0.97 (0.95–0.98)
0.91 (0.88–0.94)
0.84 (0.80–0.88)
0.43 (0.38–0.48)
0.95 (0.94–0.96)
0.79 (0.77–0.80)
0.54 (0.52–0.56)
0.02 (0.02–0.03)
0.89 (0.76–0.96)
0.46 (0.28–0.65)
0.96 (0.92–0.98)
0.88 (0.81–0.92)
0.68 (0.56–0.78)
0.84 (0.78–0.88)
0.94 (0.91–0.96)
26.0 (10.84–62.38)
1.71 (1.27–2.32)
1.83 (1.28–2.62)
1.39 (0.97–2.01)
1.86 (1.49–2.33)
1.39 (0.67–2.89)
6.79 (3.67–12.57)
5.22 (3.68–7.39)
3.94 (3.04–5.10)
1.61 (1.45–1.79)
3.40 (2.88–4.01)
1.81 (1.67–1.96)
1.30 (1.24–1.37)
1.00 (0.99–1.01)
4.75 (1.86–12.12)
1.61 (1.09–2.37)
7.73 (3.97–15.07)
6.87 (4.26–11.07)
3.13 (2.19–4.45)
4.89 (3.59–6.66)
5.17 (3.25–8.20)
LR+ (95% CI)
Microscopy for fecal leukocytes
Sensitivity (95% CI) Specificity (95% CI)
110
⭓1
15
15
15
13
12
⭓1
Threshold,
cells/HPF
HPF, high-power field; LR+, positive likelihood ratio; LR⫺, negative likelihood ratio.
13
24
40
26
48
24
23
59
56
10
17
42
43
3
Proportion
of subjects
with positive
cultures, %a
No. of subjects with a bacteria-positive stool culture/total no. of participants in that study.
Subjects from this study were only counted once in calculating the total no. of subjects in the category.
NOTE.
5711
280
Huicho et al. [52]
Total no. of subjects for this category
887
140
Pediatric cohort
453
3558
92
Adult cohort
McNeely et al. [65]
Ruiz-Pelaez and Mattar [69]
Stoll et al. [66]
Scerpella [64]
Resource-poor countries
1339
113
Alvarado [59]
Total no. of subjects for this category
416
376
Silletti et al. [62]
Developed countries
Cohort size,
no. of
subjects
Evaluation of microscopy for fecal leukocytes for diagnosis of inflammatory bacterial gastroenteritis.
Setting, reference
Table 5.
0.17 (0.15–0.20)
0.52 (0.41–0.66)
0.84 (0.75–0.94)
0.87 (0.72–1.05)
0.85 (0.80–0.90)
0.99 (0.96–1.03)
0.83 (0.77–0.89)
0.59 (0.53–0.65)
0.44 (0.40–0.49)
0.20 (0.17–0.22)
0.87 (0.85–0.90)
0.78 (0.6–0.82)
0.74 (0.71–0.78)
1.00 (0.77–1.30)
0.56 (0.47–0.66)
0.28 (0.18–0.44)
0.68 (0.56–0.83)
0.17 (0.15–0.20)
0.00 (0.00–0.00)
0.23 (0.22–0.26)
0.74 (0.57–0.95)
LR⫺ (95% CI)
Figure 3. Summary receiver operating characteristic curve for use of microscopic examination for fecal leukocytes to diagnose bacterial gastroenteritis
in developed countries (A; area under the curve [AUC], 0.89) and resource-poor countries (B; AUC, 0.72). Reference nos. are shown in brackets, and
the test threshold, in leukocytes per high-power field (HPF), is shown to the right. The width of the oval is proportional to the square root of the no.
of persons without disease in the cohort; the height of the oval is proportional to the square root of the no. of persons with disease in the cohort.
pirical antibiotic therapy. Merely interpreting positive results
of an assay as synonymous with bacterial gastroenteritis without
first considering the pretest odds of disease is overly simplistic
and likely to result in clinical errors.
A meta-analysis can only provide useful insights if the studies
it contains are valid and representative of the diseases and
patients involved. Although we pursued a conservative approach to selecting studies for analysis, the possibility remains
that our selection process resulted in a biased summary estimate. Publication bias is a prevailing threat to the validity of
any meta-analysis and results when studies reporting a positive
result are more likely to enter the literature than are those that
find a negative result. Some authorities recommend the use of
funnel plots to detect publication bias, theorizing that large
studies with negative results are more likely to be published
than are small ones [75]; however, the validity of this approach
remains controversial [76]. We note, however, that the study
with the largest sample size [66] also documented the weakest
performance for rapid stool assays. If publication bias contributed to this discrepancy, then the already unimpressive performance of these tests in resource-poor countries probably
overestimates their actual value.
Our meta-analysis identified several areas requiring further
research. First, evidence about the lactoferrin assay’s accuracy
in the developed world derived from a single trial [62]. Although the assay performed well in this instance, it may be
presumptuous to assume that this finding is typical and generalizable to other populations in developed countries. Studies
evaluating the lactoferrin test for children living in developed
countries are conspicuously absent. This is unfortunate, because
children are particularly susceptible to diarrheal disease and its
complications and because age also contributes to spectrum
bias.
Second, the optimum thresholds for the lactoferrin and fecal
Table 6. Evaluation of microscopy for fecal erythrocytes for diagnosis of inflammatory
bacterial gastroenteritis in 3558 subjects in Bangladesh.
Threshold,
cells/HPF
Sensitivity (95% CI)
Specificity (95% CI)
LR+ (95% CI)
LR⫺ (95% CI)
⭓1
0.53 (0.50–0.56)
0.64 (0.62–0.66)
1.47 (1.39–1.56)
0.73 (0.70–0.77)
110
0.22 (0.19–0.25)
0.91 (0.90–0.92)
2.44 (2.16–2.76)
0.86 (0.83–0.88)
150
0.06 (0.05–0.08)
0.98 (0.97–0.98)
3.00 (2.31–3.90)
0.96 (0.94–0.98)
NOTE. Culture results were positive for 23% of subjects. HPF, high-power field; LR+, positive likelihood
ratio; LR⫺, negative likelihood ratio. Data are from Stoll et al. [66].
a
No. of subjects with a bacteria-positive stool culture/total no. of participants in that study.
Stool Assays for Bacterial Gastroenteritis • CID 2003:37 (1 August) • 373
leukocyte tests in each epidemiologic setting could not be defined given available data. It quite possible that a uniform
threshold is inappropriate. In areas where diarrheal disease is
hyperendemic and where the background “noise” of weakly
positive test results is high, our analysis indicates that higher
thresholds are probably necessary. To define optimum thresholds, future studies should routinely evaluate multiple cutoff
points, and thresholds should be optimized to suit the local
epidemiology.
Third, we cannot be certain that the resource-poor nations
in this analysis are truly representative of resource-poor countries from other parts of the world, notably sub-Saharan Africa.
Similarly, our search yielded no appropriate studies of rapid
stool assays in areas that are not easily classified as either “developed” or “resource poor,” such as China, former Soviet Union nations, and the Middle East.
Last, randomized trials found that empirical treatment with
fluoroquinolones reduces the duration of diarrheal symptoms
by ∼1–2 days [77–79]. It would be extremely helpful to know
whether this modest benefit would improve if a rapid stool
assay were used to identify which patients should receive
antibiotics.
19.
Acknowledgments
20.
We thank Michael Callahan, Donald Thea, Sydney Rosen,
and William MacLeod (Center for International Health and
Development, Boston University School of Public Health, Boston), for offering their advice during the preparation of this
manuscript.
21.
10.
11.
12.
13.
14.
15.
16.
17.
18.
22.
23.
24.
References
1. Mead PS, Slutsker L, Dietz V, et al. Food-related illness and death in
the United States. Emerg Infect Dis 1999; 5:607–25.
2. Bern C, Martines J, de Zoysa I, Glass RI. The magnitude of the global
problem of diarrhoeal disease: a ten-year update. Bull World Health
Organ 1992; 70:705–14.
3. Guerrant RL, Shields DS, Thorson SM, Schorling JB, Groschel DH.
Evaluation and diagnosis of acute infectious diarrhea. Am J Med
1985; 78:91–8.
4. Guerrant RL, Araujo V, Soares E, et al. Measurement of fecal lactoferrin
as a marker of fecal leukocytes. J Clin Microbiol 1992; 30:1238–42.
5. Uchida K, Matsuse R, Tomita S, Sugi K, Saitoh O, Ohshiba S. Immunochemical detection of human lactoferrin in feces as a new marker
for inflammatory gastrointestinal disorders and colon cancer. Clin
Biochem 1994; 27:259–64.
6. Huicho L, Campos M, Rivera J, Guerrant RL. Fecal screening tests in
the approach to acute infectious diarrhea: a scientific overview. Pediatr
Infect Dis J 1996; 15:486–94.
7. Black RE, Levine MM, Clements ML, Hughes TP, Blaser MJ. Experimental Campylobacter jejuni infection in humans. J Infect Dis 1988;
157:472–9.
8. Tauxe RV, Pegues DA, Hargrett-Bean N. Campylobacter infections: the
emerging national pattern. Am J Public Health 1987; 77:1219–21.
9. Centers for Disease Control and Prevention. Foodborne Diseases Active
374 • CID 2003:37 (1 August) • Gill et al.
25.
26.
27.
28.
29.
30.
31.
32.
33.
Surveillance Network, 1996. MMWR Morb Mortal Wkly Rep 1997;
46:258–61.
Blaser MJ. Epidemiologic and clinical features of Campylobacter jejuni
infections. J Infect Dis 1997; 176(Suppl 2):S103–5.
Centers for Disease Control and Prevention. Incidence of foodborne
illnesses—FoodNet, 1997. MMWR Morb Mortal Wkly Rep 1998; 47:
782–6.
Centers for Disease Control and Prevention. Incidence of foodborne
illnesses: preliminary data from the Foodborne Diseases Active Surveillance Network (FoodNet)—United States, 1998. MMWR Morb
Mortal Wkly Rep 1999; 48:189–94.
Centers for Disease Control and Prevention. Preliminary FoodNet data
on the incidence of foodborne illnesses—selected sites, United States,
1999. MMWR Morb Mortal Wkly Rep 2000; 49:201–5.
Centers for Disease Control and Prevention. Preliminary FoodNet data
on the incidence of foodborne illnesses—selected sites, United States,
2000. MMWR Morb Mortal Wkly Rep 2001; 50:241–6.
Altekruse SF, Stern NJ, Fields PI, Swerdlow DL. Campylobacter jejuni—an emerging foodborne pathogen. Emerg Infect Dis 1999; 5:
28–35.
Fleiss JL. An introduction to applied probability: statistical methods
for rates and proportions. New York: John Wiley & Sons, 1981.
Jaeschke R, Guyatt GH, Sackett DL. Users’ guides to the medical literature. III. How to use an article about a diagnostic test. A. Are the
results of the study valid? Evidence-based Medicine Working Group.
JAMA 1994; 271:703–7.
Lau J. MetaTest version 0.6. Boston: Tufts–New England Medical Center, 1997.
Moses LE, Shapiro D, Littenberg B. Combining independent studies of
a diagnostic test into a summary ROC curve: data-analytic approaches
and some additional considerations. Stat Med 1993; 12:1293–316.
Walter SD. Properties of the summary receiver operating characteristic
(SROC) curve for diagnostic test data. Stat Med 2002; 21:1237–56.
Irwig L, Tosteson AN, Gatsonis C, et al. Guidelines for meta-analyses
evaluating diagnostic tests. Ann Intern Med 1994; 120:667–76.
DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin
Trials 1986; 7:177–88.
Ronsmans C, Bennish ML, Wierzba T. Diagnosis and management of
dysentery by community health workers. Lancet 1988; 2:552–5.
Harris JC, Dupont HL, Hornick RB. Fecal leukocytes in diarrheal illness. Ann Intern Med 1972; 76:697–703.
Peirce JE, DuPont HL, Lewis KR. Acute diarrhea in a residential institution for the retarded: usefulness of fecal leukocyte examination.
Am J Dis Child 1974; 128:772–5.
Guerrant RL, Moore RA, Kirschenfeld PM, Sande MA. Role of toxigenic and invasive bacteria in acute diarrhea of childhood. N Engl J
Med 1975; 293:567–72.
Pickering LK, DuPont HL, Olarte J, Conklin R, Ericsson C. Fecal
leukocytes in enteric infections. Am J Clin Pathol 1977; 68:562–5.
Korzeniowski OM, Barada FA, Rouse JD, Guerrant RL. Value of examination for fecal leukocytes in the early diagnosis of shigellosis. Am
J Trop Med Hyg 1979; 28:1031–5.
Stoll BJ, Glass RI, Huq MI, Khan MU, Banu H, Holt J. Epidemiologic
and clinical features of patients infected with Shigella who attended a
diarrheal disease hospital in Bangladesh. J Infect Dis 1982; 146:177–83.
Speelman P, McGlaughlin R, Kabir I, Butler T. Differential clinical
features and stool findings in shigellosis and amoebic dysentery. Trans
R Soc Trop Med Hyg 1987; 81:549–51.
Hossain MA, Albert MJ. Effect of duration of diarrhoea and predictive
values of stool leucocytes and red blood cells in the isolation of different
serogroups or serotypes of Shigella. Trans R Soc Trop Med Hyg 1991;85:
664–6.
Paccagnini S, Fontana M, Ceriani R, et al. Occult blood and faecal
leucocyte tests in acute infectious diarrhoea in children. Lancet 1987;1:
442.
Jindal N, Arora R, Bhushan B, Arora S. A study of infective aetiology
34.
35.
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
50.
51.
52.
53.
54.
55.
56.
of chronic diarrhoea in children in Amritsar. J Indian Med Assoc
1995; 93:169–70.
Maki M, Maki R, Vesikari T. Faecal leucocytes in campylobacterassociated diarrhoea in infants. Acta Paediatr Scand 1979; 68:271–2.
Yong WH, Mattia AR, Ferraro MJ. Comparison of fecal lactoferrin
latex agglutination assay and methylene blue microscopy for detection
of fecal leukocytes in Clostridium difficile–associated disease. J Clin
Microbiol 1994; 32:1360–1.
Iida T, Naka A, Suthienkul O, Sakaue Y, Guerrant RJ, Honda T. Measurement of fecal lactoferrin for rapid diagnosis of enterohemorrhagic
Escherichia coli infection. Clin Infect Dis 1997; 25:167.
Cimolai N. Fecal leukocytes in Escherichia coli O157:H7 enteritis. J Clin
Pathol 1990; 43:347.
Bouckenooghe AR, Dupont HL, Jiang ZD, et al. Markers of enteric
inflammation in enteroaggregative Escherichia coli diarrhea in travelers.
Am J Trop Med Hyg 2000; 62:711–3.
Tarr PI, Neill MA, Clausen CR, Watkins SL, Christie DL, Hickman
RO. Escherichia coli O157:H7 and the hemolytic uremic syndrome:
importance of early cultures in establishing the etiology. J Infect Dis
1990; 162:553–6.
Koplan JP, Fineberg HV, Ferraro MJ, Rosenberg ML. Value of stool
cultures. Lancet 1980; 2:413–6.
Nelson JD, Haltalin KC. Accuracy of diagnosis of bacterial diarrheal
disease by clinical features. J Pediatr 1971; 78:519–22.
Jonas A, Yahav J, Soudry A. Clinical features of viral and bacterial
gastroenteritis in hospitalized children. Isr J Med Sci 1982; 18:753–9.
Beltinger J, Walther R, Bardhan P, Mahalanabis D, Gyr K. Immunological testing for occult blood in patients with acute infectious diarrhea: can it improve the specificity of the guaiac test? Dig Dis Sci 1997;
42:366–71.
Patwari AK, Deb M, Dudeja M, Jayasheela M, Agarwal A, Singh P.
Clinical and laboratory predictors of invasive diarrhoea in children less
than five years old. J Diarrhoeal Dis Res 1993; 11:211–6.
Vogtlin J, Stalder H, Hurzeler L, et al. Modified guaiac test may replace
search for faecal leukocytes in acute infectious diarrhea. Lancet 1983;
2:1204.
Quiroga T, Garcia P, Goycoolea M, Potin M, Rodriguez L, Vial P. Fecal
lactoferrin as a marker of fecal leukocytes. J Clin Microbiol 1994; 32:
2629–30.
Vaishnavi C, Bhasin DK, Singh K. Fecal lactoferrin assay as a costeffective tool for intestinal inflammation. Am J Gastroenterol 2000;
95:3002–3.
Guyot J, Gonvers JJ, Pyndiah N, Heitz M. Which screening test in
acute diarrhea [abstract]? Gastroenterology 1985; 88:1408.
Miller JR, Barrett LJ, Kotloff K, Guerrant RL. A rapid test for infectious
and inflammatory enteritis. Arch Intern Med 1994; 154:2660–4.
Fontana M, Zuin G, Paccagnini S, et al. Simple clinical score and
laboratory-based method to predict bacterial etiology of acute diarrhea
in childhood. Pediatr Infect Dis J 1987; 6:1088–91.
Choi SW, Park CH, Silva TM, Zaenker EI, Guerrant RL. To culture or
not to culture: fecal lactoferrin screening for inflammatory bacterial
diarrhea. J Clin Microbiol 1996; 34:928–32.
Huicho L, Sanchez D, Contreras M, et al. Occult blood and fecal
leukocytes as screening tests in childhood infectious diarrhea: an old
problem revisited. Pediatr Infect Dis J 1993; 12:474–7.
Mikhail IA, Fox E, Haberberger RL Jr, Ahmed MH, Abbatte EA. Epidemiology of bacterial pathogens associated with infectious diarrhea
in Djibouti. J Clin Microbiol 1990; 28:956–61.
McIver CJ, Hansman G, White P, Doultree JC, Catton M, Rawlinson
WD. Diagnosis of enteric pathogens in children with gastroenteritis.
Pathology 2001; 33:353–8.
Echeverria P, Sethabutr O, Pitarangsi C. Microbiology and diagnosis
of infections with Shigella and enteroinvasive Escherichia coli. Rev Infect
Dis 1991; 13(Suppl 4):S220–5.
Bardhan PK, Albert MJ, Alam NH, Faruque SM, Neogi PK, Mahal-
57.
58.
59.
60.
61.
62.
63.
64.
65.
66.
67.
68.
69.
70.
71.
72.
73.
74.
75.
76.
77.
78.
79.
anabis D. Small bowel and fecal microbiology in children suffering
from persistent diarrhea in Bangladesh. J Pediatr Gastroenterol Nutr
1998; 26:9–15.
Loosli J, Gyr K, Stalder H, et al. Etiology of acute infectious diarrhea
in a highly industrialized area of Switzerland. Gastroenterology 1985;
88:75–9.
Siegel D, Cohen PT, Neighbor M, et al. Predictive value of stool examination in acute diarrhea. Arch Pathol Lab Med 1987; 111:715–8.
Alvarado T. Faecal leucocytes in patients with infectious diarrhoea.
Trans R Soc Trop Med Hyg 1983; 77:316–20.
Thorson SM, Lohr JA, Dudley S, Guerrant RL. Value of methylene
blue examination, dark-field microscopy, and carbol-fuchsin Gram
stain in the detection of Campylobacter enteritis. J Pediatr 1985; 106:
941–3.
DeWitt TG. Acute diarrhea in children. Pediatr Rev 1989; 11:6–13.
Silletti RP, Lee G, Ailey E. Role of stool screening tests in diagnosis of
inflammatory bacterial enteritis and in selection of specimens likely to
yield invasive enteric pathogens. J Clin Microbiol 1996; 34:1161–5.
Savola KL, Baron EJ, Tompkins LS, Passaro DJ. Fecal leukocyte stain
has diagnostic value for outpatients but not inpatients. J Clin Microbiol
2001; 39:266–9.
Scerpella EG, Okhuysen PC, Mathewson JJ, et al. Evaluation of a new
latex agglutination test for fecal lactoferrin in travelers’ diarrhea. J
Travel Med 1994; 1:68–71.
McNeely WS, Dupont HL, Mathewson JJ, Oberhelman RA, Ericsson
CD. Occult blood versus fecal leukocytes in the diagnosis of bacterial
diarrhea: a study of US travelers to Mexico and Mexican children. Am
J Trop Med Hyg 1996; 55:430–3.
Stoll BJ, Glass RI, Banu H, Huq MI, Khan MU, Ahmed M. Value of
stool examination in patients with diarrhoea. Br Med J (Clin Res Ed)
1983; 286:2037–40.
Ascher DP, Edusada-Corpus R. Clinical and laboratory predictors of
bacterial diarrhea in a tropical environment. Mil Med 1991; 156:74–6.
Huicho L, Garaycochea V, Uchima N, Zerpa R, Guerrant RL. Fecal
lactoferrin, fecal leukocytes and occult blood in the diagnostic approach
to childhood invasive diarrhea. Pediatr Infect Dis J 1997; 16:644–7.
Ruiz-Pelaez JG, Mattar S. Accuracy of fecal lactoferrin and other stool
tests for diagnosis of invasive diarrhea at a Colombian pediatric hospital. Pediatr Infect Dis J 1999; 18:342–6.
DeWitt TG, Humphrey KF, McCarthy P. Clinical predictors of acute
bacterial diarrhea in young children. Pediatrics 1985; 76:551–6.
Boyd JC. Mathematical tools for demonstrating the clinical usefulness
of biochemical markers. Scand J Clin Lab Invest Suppl 1997; 227:46–63.
Bloem MW, Wedel M, Egger RJ, et al. Mild vitamin A deficiency and
risk of respiratory tract diseases and diarrhea in preschool and school
children in northeastern Thailand. Am J Epidemiol 1990; 131:332–9.
Patwari AK. Diarrhoea and malnutrition interaction. Indian J Pediatr
1999; 66(Suppl):S124–34.
Matter L, Germann D. Detection of human immunodeficiency virus
(HIV) type 1 antibodies by new automated microparticle enzyme immunoassay for HIV types 1 and 2. J Clin Microbiol 1995; 33:2338–41.
Egger M, Davey Smith G, Schneider M, Minder C. Bias in meta-analysis
detected by a simple, graphical test. BMJ 1997; 315:629–34.
Tang JL, Liu JL. Misleading funnel plot for detection of bias in metaanalysis. J Clin Epidemiol 2000; 53:477–84.
Wistrom J, Jertborn M, Ekwall E, et al. Empiric treatment of acute
diarrheal disease with norfloxacin: a randomized, placebo-controlled
study. Swedish Study Group. Ann Intern Med 1992; 117:202–8.
Pichler HE, Diridl G, Stickler K, Wolf D. Clinical efficacy of ciprofloxacin compared with placebo in bacterial diarrhea. Am J Med 1987;
82(Suppl 4A):329–32.
Goodman LJ, Trenholme GM, Kaplan RL, et al. Empiric antimicrobial
therapy of domestically acquired acute diarrhea in urban adults. Arch
Intern Med 1990; 150:541–6.
Stool Assays for Bacterial Gastroenteritis • CID 2003:37 (1 August) • 375