Background. To design an efficacy trial of a killed oral vaccine against enterotoxigenic Escherichia coli (ETEC) diarrhea in Egyptian children, we derived for ETEC diarrhea an empirical definition that increased the probability that diarrhea associated with excretion of ETEC was caused by the detected ETEC.
Methods. We conducted a cohort study of 397 Egyptian children <24 months old and monitored them until they were 3 years old. Vaccine-preventable (VP) ETEC was defined as ETEC expressing ⩾1 of the toxin- (heat-labile [LT] toxin) and colonization-factor antigens (CFA I, II, and IV) in the vaccine.
Results. Although fecal excretion of VP-ETEC was highly associated with diarrhea, excretion of LT-ETEC per se was not related to diarrhea (adjusted odds ratio [ORJ, 1.16 [95% confidence interval {CI}, 0.90-1.49]). The fecal excretion of antigenic types of VP-ETEC other than LT-ETEC (non-LT VP-ETEC) was highly associated with diarrheal symptoms (OR A, 3.91 [95% CI, 2.78-5.49]; P<.001), and this association was greater for nonbloody than for bloody diarrhea.
Conclusions. Because the vaccine had been anticipated to protect primarily against symptomatic ETEC diarrhea, these results indicate that the primary-outcome definition of ETEC diarrhea for the trial should be restricted to nonbloody diarrheal episodes associated with fecal excretion of non-LT VP-ETEC.
Enteric pathogens are frequently ingested and excreted by persons in developing countries, and ingestion frequently is insufficient to cause diarrheal symptoms; therefore, in such settings, fecal excretion of a particular enteropathogen in a patient with diarrhea can reflect the chance coincidence of 2 common events: diarrhea on the one hand and ingestion and excretion of the enteric pathogen on the other hand [1]. As a result, the detection, in these settings, of the fecal excretion of a pathogen in a patient with diarrhea does not establish per se that the detected pathogen has played a pathogenic role in causing the patient's symptoms.
This diagnostic ambiguity creates potential problems in the evaluation of vaccines against enteric infections in developing countries. Most such vaccines are designed to prevent symptomatic infections. However, the evaluation of a vaccine's preventive effect on all diarrheal episodes associated with the fecal excretion of the target pathogen will include an appreciable fraction of episodes in which symptoms are associated with, but not caused by, the excreted target pathogen. If a vaccine has a greater protective effect against disease-causing infection than against intestinal colonization per se, an analysis of all episodes of diarrhea associated with the fecal excretion of the target pathogen will yield underestimates of vaccine efficacy against symptomatic infections.
A recent efficacy trial of a new vaccine against pediatric enterotoxigenic Escherichia coli (ETEC) diarrhea in Egypt presented us with the challenge of defining diarrheal outcomes that were etiologically related to excreted ETEC. We report the application of an evidence-based, pathogenicity-driven method used to define these outcomes for the trial.
SUBJECTS AND METHODS
Killed oral ETEC vaccine tested in Egypt. The agent under study was an orally administered, formalinized whole-cell ETEC plus recombinant cholera toxin B-subunit (ETEC/rCTB) vaccine that was developed at the University of Goteborg [2, 3]. The vaccine consists of cells that express colonization-factor antigen (CFA)/I, CFA/II, and CFA/IV, the most epidemiologically prevalent CFAs of ETEC, together with rCTB, which is antigen-ically related to the subunit of the heat-labile toxin (LT) of ETEC [4]. On the basis of earlier encouraging results regarding the safety and immunogenicity of this vaccine [5-7], a field trial of the efficacy of a 3-dose regimen of this oral vaccine was planned for Egyptian infants and children 6–18 months old, an age group with a high risk of ETEC diarrhea [8].
Pretrial cohort study of pediatric diarrhea in Abu Homos. To estimate the incidence rates of ETEC diarrhea at the site and the population to be used for the vaccine efficacy trial, we conducted a cohort study. This study also provided data necessary for evaluation of alternative outcome definitions for the trial. The cohort study was conducted in 2 rural villages ∼40 km from the city of Alexandria, in the Abu Homos district of the Beheira Governorate in lower Egypt. The enrollment of children <24 months old began in February 1995, after a census was done. Thereafter, all newborns in the 2 villages were enrolled until September 1997. Eligibility criteria required that the enrolled child have no significant congenital defects or chronic illnesses. Before enrollment, written informed consent was obtained from the parent or guardian of each child. Human-subject guidelines of the US Departments of Defense and Health and Human Services were followed throughout the study.
Children were followed from the time of enrollment until age 36 months or February 1998, whichever came first. Surveillance was conducted by twice-weekly visits by trained field-workers to the homes of the enrolled children. Children reported to have had loose or liquid stools had a rectal swab and a fresh stool specimen collected by the field-worker. When a mother or caretaker reported that the child had loose or liquid stools, a physician examined the child to assess the degree of dehydration and other measures of disease severity, and arranged for appropriate therapy. For visits during which, because of the absence of the child or the caretaker, histories could not be obtained, another visit was attempted the next day.
In addition, cross-sectional surveys of all children in the cohort were conducted at 2-month intervals over the 3-year period of the study. In these surveys, rectal swabs were collected and stool specimens were obtained and processed from all children, regardless of the occurrence of diarrheal symptoms, in a manner identical to that for the twice-weekly longitudinal surveillance. Demographic events such as deaths and out-migrations, which were censoring events for follow-up, were ascertained during these visits to the homes. A total of 180 children <24 months old were enrolled at the start of surveillance, and an additional 217 newborns were added to the cohort thereafter; 95% of the scheduled twice-weekly visits were successfully completed. Rectal-swab specimens collected in the field were immediately placed in Cary-Blair medium and, along with the stool specimens, were transported on ice packs to the Abu Homos field laboratory, where they were refrigerated. Twice each week, the specimens were transported on ice packs for microbiological evaluation at the laboratories of US Naval Medical Research Unit-3 in Cairo. Rectal swabs were plated on MacConkey's media, and 5 lactose-fermenting colonies were evaluated for both LT and heat-stable toxin (ST) by GM1-ganglioside ELISAs [9, 10]. Colonies that were positive for either LT or ST were further evaluated for CFAs by dot-blot tests with monoclonal antibodies specific for CFA/I, CFA/II (CS1, CS2, and CS3) and for CFA/IV (CS4, CS5, and CS6) [11, 12]. Conventional methods were used for isolation of Campylobacter, Salmonella, Shigella, and Vibrionaceae from rectal-swab specimens and for detection of rotavirus antigen in stool specimens [13-15].
Definitions.A diarrheal day was defined as the passage of ⩾3 loose or liquid stools during any 24-h period (in addition, for breast-fed infants, the mother had to state that the stools were less formed or more frequent than usual) or ⩾ 1 loose or liquid stool with the presence of visible blood. A diarrheal episode was considered to have begun on the first day of loose or liquid stools after ⩾3 consecutive nondiarrheal days and to have ended on the first diarrheal day that was followed by ⩾3 consecutive nondiarrheal days. Pathogens, including ETEC, were considered to be related to a diarrheal episode if they were detected on any day of the episode. An ETEC diarrheal episode denoted an episode in which ⩾ 1 tested lactose-fermenting colony from a fecal specimen expressed either LT or ST. A vaccine-preventable (VP) ETEC episode (an episode associated with ETEC expressing CFA or toxin antigens contained in the vaccine) was one in which every detected ETEC colony expressed either LT or CFA/I, CFA/II, or CFA/IV. A pure ETEC episode was one in which all detected ETEC had the same Toxin and CFA phenotypes. The intensity of ETEC excretion was measured in terms of the number of the 5 tested lactose-fermenting colonies that were ETEC. Diarrheal episodes were considered to be severe if ⩾1 of the following features was present by history or examination: (1) a history of ⩾1 generalized convulsions, (2) a history of drowsiness to the point that the subject no longer could take fluids, (3) a history of vomiting ⩾3 times within a 24-h period since the onset of diarrhea, (4) inability to awaken on examination, or (5) evidence of ⩾1 objective sign of dehydration on examination (e.g., sunken fontanelle, sunken eyes, absence of tears during crying, tenting of skin, parched oral mucous membranes, or weak-to-absent radial pulse); all other episodes were classified as nonsevere. Diarrhea was classified as bloody if the caretaker reported visible blood in stool and as nonbloody if no blood was described. An episode of ETEC was considered to have a copathogen if Shigella species, Campylobacter jejuni, or rotavirus was isolated in a fecal specimen collected during the episode. Seasons of the year were divided into the cool (November-April) and warm (May-October) months.
Strategies for evaluation of the pathogenic roles of ETEC in diarrheal episodes. The field trial of the efficacy of ETEC/ rCTB began in 1998, after several phase 2 trials in descending age groups of Egyptian adults and children had found the vaccine to be safe and immunogenic [5-7, 16]. The trial targeted healthy infants and children 6-18 months old who were randomized to receive, at 2-week intervals, 3 oral doses of either the ETEC/rCTB vaccine or a placebo that consisted of a killed K12 strain of E.coli whole cells. After they had received an assigned regimen, participants were monitored for 1 year, by twice-weekly home visits, for the detection of ETEC diarrhea, by surveillance methods identical to those in the preceding cohort study. The goal of the trial was to assess the protective efficacy, over a 1-year period of follow-up, of a complete course of vaccine against VP-ETEC diarrheal episodes. The potential age range of follow-up for the trial was 6-30 months.
Because the goal of the trial was to determine whether the ETEC/rCTB vaccine prevented symptomatic episodes that were etiologically attributable to VP-ETEC, we investigated various strategies for refining the definition of VP-ETEC diarrhea in ways that increased the strength of the pathogenic relationship between detected VP-ETEC organisms and the occurrence of diarrheal symptoms. To discern features for refining this definition, we conducted nested case-control analyses of the pretrial cohort study. Case patients had episodes of diarrhea detected during the cohort study. Control subjects were children selected during the bimonthly surveys who did not have diarrheal episodes within 7 days of the survey visit. The pathogenicity of VP-ETEC was assessed by comparing the prevalence of fecal excretion of VP-ETEC in the case patients and in the control subjects. These associations were expressed as odds ratios (ORs), with ORs >1 denoting a pathogenic relationship and progressively larger ORs corresponding to progressively stronger pathogenic relationships. Because a single diarrheal episode could have been evaluated on the basis of multiple rectal-swab specimens if the episode spanned several visits, only the results of the first rectal swab of each case were considered in the case-control comparisons, to maintain comparability with the control subjects, from whom only 1 such specimen was obtained.
In these analyses, we considered several features for refining the definition of VP-ETEC diarrhea; these included characteristics of the diarrheal episode (clinical severity and diarrhea character) and characteristics of the isolated VP-ETEC (toxin phe-notype, expression of a CFA, and intensity of excretion). We also evaluated whether the presence or absence of a fecal copathogen, in addition to VP-ETEC, modified the pathogenic relationship between VP-ETEC and diarrheal symptoms, because, a priori, we reasoned that episodes in which such copathogens were present should be excluded from the trial outcomes.
To simulate the conditions of the field trial, the case-control analyses were restricted to children 6–30 months old, the age range of follow-up for the field trial. We excluded case patients and control subjects who had had <5 lactose-fermenting colonies tested for toxin and CFA expression, because we wanted to use uniform criteria to judge the intensity of excretion, and we calculated this intensity on the basis of the number of colonies that were ETEC. Of the 2624 age-eligible diarrheal episodes in children who were potentially eligible as case patients, 423 were excluded because the specimens had <5 tested colonies; this left 2201 eligible diarrheal episodes for analysis. Of the 2295 age-eligible visits to control subjects during the cross-sectional surveys, 414 were excluded because the specimens had <5 tested colonies; this left 1881 eligible visits to control subjects for analysis.
Because children could be selected more than once as either a case patient or a control subject, we used, for this repeated sampling of cases and controls, logistic-regression models that used a robust variance estimator to adjust the variance of coefficients [17]. The method is the same as the generalized estimating equations method under the identity working correlation. To obtain crude ORs for associations in these models, case-control status was taken as the dependent variable, and a variable for ETEC excretion was fitted as the independent variable. The coefficient for the ETEC-excretion variable in these models was exponentiated to estimate ORs, and 2-tailed Pvalues and 95% confidence intervals (CIs) for the ORs were estimated using SEs of the coefficients. To obtain estimates of ORs adjusted for important confounding variables (OR A), we refitted the models after forcing 2 additional independent variables in the models: age (in months) and season at the time of selection. All analyses were performed by the SAS macro GLIMMIX (version 6.12; SAS Institute).
RESULTS
Of the 2201 eligible episodes of diarrhea observed among children 6–30 months old, 610 (28%) were associated with the fecal excretion of ETEC, and 321 (15%) were associated with the fecal excretion of VP-ETEC. Most episodes of ETEC diarrhea were associated with the excretion of pure ST-ETEC (61%); lesser proportions of ETEC episodes were associated with the excretion of pure LT-ETEC (24%), pure LT/ST ETEC (7%), or mixtures of ETEC-toxin phenotypes (8%). Among episodes associated with the excretion of a single ETEC-toxin phenotype, the expression of CFA/I, CFA/II, or CFA/IV was most prevalent among those with either pure LT/ST-ETEC (49%) or pure ST-ETEC (29%) and was uncommon among those with pure LT-ETEC (6%). Bloody diarrhea and severe diarrhea were noted in 3% and 31% of VP-ETEC episodes, respectively.
As shown in table 1, the association between the fecal excretion of ETEC and diarrheal symptoms varied by the toxin phenotype of the excreted ETEC. Whereas associations were significant for pure ST-ETEC (OR A, 2.04 [95% CI, 1.69–2.47]; P<.001) and pure LT/ST-ETEC (OR A, 4.21 [95% CI, 2.04–8.71]; P< .001), there appeared to be no relationship between the excretion of pure LT-ETEC and the occurrence of diarrheal symptoms (OR A, 1.16 [95% CI, 0.90–1.49]). Correspondingly, although there was a significant association between the excretion of VP-ETEC and diarrheal symptoms (OR A, 1.89 [95% CI, 1.54–2.33]; P<.001), removal of pure LT-only ETEC from the definition of VP-ETEC (a step that left only non-LT VP-ETEC) markedly augmented the strength of this association (OR A, 3.91 [95% CI, 2.78–5.49]; P<.001).
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children.
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children.
It remained possible that an analysis of either the coex-pression of a CFA or the intensity of ETEC excretion might reveal subsets of pure LT-ETEC episodes for which there were pathogenic associations; however, the excretion of pure LT-ETEC remained unassociated with diarrheal symptoms, regardless of the coexpression of a vaccine CFA (table 2) and regardless of the intensity of its excretion (table 3). By contrast, the other subcategories of VP-ETEC (LT/ST-CFA positive, LT/ST-CFA negative, and ST-CFA positive) each exhibited a strong association with diarrheal symptoms (table 2). Although there was a suggestion that the association between the fecal excretion of non-LT VP-ETEC and diarrheal symptoms increased with an increasing intensity of excretion (table 3), a significant association was observed even when only 1 positive colony was detected (OR A, 2.44 [95% CI, 1.17–5.10]; P<.05).
Association between the fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children, by toxin and colonization-factor antigen (CFA) phenotypes of vaccine-preventable (VP) ETEC.
Association between the fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children, by toxin and colonization-factor antigen (CFA) phenotypes of vaccine-preventable (VP) ETEC.
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrhea, in a cohort of Egyptian children, by intensity of ETEC excretion
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrhea, in a cohort of Egyptian children, by intensity of ETEC excretion
As shown in table 4, pure LT-ETEC was not associated with the occurrence of diarrheal symptoms, even if cases considered were restricted to episodes of nonbloody diarrhea (OR A, 1.16 [95% CI, 0.90–1.50]) or to patients with severe symptoms (OR A, 1.14 [95% CI, 0.80–1.62]). The association between the excretion of non-LT VP-ETEC and the occurrence of diarrheal symptoms was similar in severe (OR A, 3.86 [95% CI, 2.56–5.82]; P< .001) and nonsevere (OR A, 3.94 [95% CI, 2.75–5.64]; P< .001) diarrhea, but it appeared to be more strongly associated with nonbloody (OR A, 3.96 [95% CI, 2.82–5.57]; P< .001) than with bloody (OR A, 2.62 [95% CI, 1.04–6.58]; P< .05) diarrhea.
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children, by clinical features of the diarrheal episode
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children, by clinical features of the diarrheal episode
Finally, as shown in table 5, the fecal excretion of VP-ETEC, without the excretion of copathogens, was significantly associated with the occurrence of diarrheal symptoms (OR A, 1.81 [95% CI, 1.46–2.24]; P<.001); the same held true for the excretion of non-LT VP-ETEC (OR A, 3.87 [95% CI, 2.70–5.55]; P<.001). The isolation of pure LT-ETEC from feces, without the isolation of copathogens, was not associated with the occurrence of diarrheal symptoms (OR A, 1.09 [95% CI, 0.84–1.40]).
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children, by excretion of enteric copathogens
Association between fecal isolation of enterotoxigenic Escherichia coli (ETEC) and the occurrence of diarrheal symptoms, in a cohort of Egyptian children, by excretion of enteric copathogens
DISCUSSION
The results of the present study provide an empirical basis for the formulation of criteria for pathogenicity-driven definitions of end points in a trial of the efficacy of an ETEC vaccine in infants and young children in Egypt. We have shown that the detection of pure LT-ETEC in a diarrheal episode in our target population did not denote a pathogenic role for the detected organismsand that this remained true regardless of (1) additional restrictions on the definition of pure LT-ETEC, in terms of the coexpression of a CFA or the intensity of fecal excretion, and (2) whether diarrhea was limited only to nonbloody episodes or those which were clinically severe. As expected, limiting the definition of pure LT-ETEC to exclude instances in which copathogens were excreted revealed no pathogenic association. Because the isolation of pure LT-ETEC in feces was not associated with diarrheal symptoms in the setting and population of the trial and because the vaccine to be tested in the trial was expected to provide protection against only symptomatic LT-ETEC, our analyses indicate that there was no justification for including diarrhea associated with the excretion of pure LT-ETEC in the primary-outcome definition for the trial.
Our analyses also demonstrate that the excretion of VP-ETEC, when it was redefined to exclude pure LT-ETEC, was associated with diarrhea regardless of either the intensity of fecal excretion of ETEC or the severity or clinical character of diarrhea; however, because the association was borderline for bloody diarrhea, which is not the noninvasive syndrome known to be caused by ETEC, and because bloody diarrhea was described in only 5 (3%) of non-LT VP-ETEC episodes, we decided to limit the case definition to include only nonbloody episodes in the outcome definition of the trial. Finally, because the vaccine was not expected to protect against diarrhea caused by non-ETEC copathogens, it was reassuring that the excretion of non-LT VP-ETEC was associated with diarrheal symptoms even when such enteric copathogens were not also isolated.
Several potential limitations of our study merit discussion. First, it should be stressed that our methods apply only to situations in which it is desirable to study outcome events that are pathogenically related to the isolation of the targeted pathogen. There may be situations in which this goal is not applicable, such as trials of vaccines that are designed to reduce the transmission of a pathogen by prevention of asymptomatic colonization. Second, the exact definition derived by the method used in our analysis may have been idiosyncratic to the target population for our trial and may not be applicable to other trials of the same vaccine in different target populations; for example, in the present analysis, it was found that the isolation of pure LT-ETEC was not associated with diarrhea, and it was concluded that this entity should be excluded from the primary-outcome definition for the trial. Although a similar lack of pathogenicity of LT-ETEC has been reported for certain other settings in the developing world [18], we cannot exclude the possibility that similar studies performed in other populations would find LT-ETEC to be associated with diarrhea. Two earlier vaccine trials demonstrated that the oral ingestion of cholera toxin subunit protected against diarrhea associated with LT-ETEC, an observation that strongly suggests that LT-ETEC was a cause of diarrhea in the populations in those trials [4, 19]. Variability in the pathogenicity of LT-ETEC in different studies could be the result of several features that can differ between populations, such as the inoculum to which persons are usually exposed and the level of preexisting immunity to the target pathogen. In addition, because LT-ETEC organisms may be heterogeneous with respect to a variety of other virulence factors, the inherent pathogenicity of LT-ETEC may vary from place to place. Indeed, convincing evidence that at least certain strains of LT-ETEC are pathogenic comes from both common source-outbreak investigations and human volunteer-challenge studies [20, 21]. Clearly, in settings in which LT-ETEC is pathogenic, it would be important to include LT-ETEC diarrhea as a primary-outcome event for trials of CTB-con-taining vaccines against ETEC, such as the vaccine that we tested in Egypt. Third, the results of our analysis pertain to the type of surveillance used in our study; for example, we cannot exclude the possibility that LT-ETEC would have been associated with diarrhea had our surveillance been treatment-center based, an approach that might have detected cases that were clinically more severe than those detected in our household-based surveillance.
Counterbalancing these limitations are several strengths of the method that we used. First, the method enables the definition of outcome events in a fashion whose rationale is transparent. Second, the method provides definitions of outcomes in a manner that is based on empirical data rather than on arbitrary assumptions. Third, the data required for these empirical definitions derived from the method are often collected routinely, as part of the preparation of a site for a vaccine trial and, hence, do not entail a great deal of additional expense.
A noteworthy feature of the method that we used is its wide potential applicability. Its usefulness could extend to case definitions for other infections in which the isolation of a pathogen does not, per se, implicate the pathogen as the cause of a patient's symptoms; for example, an analogous approach has been used for the definition of clinical episodes of malaria among febrile persons in malaria-endemic areas [22, 23], and additional examples include other enteric infections and nonenteric mucosal infections. Moreover, the method's utility is not limited to trials of preventive interventions, such as vaccines. The method is also potentially applicable to the definition of clinical infections in studies of therapeutic interventions, as well as in observational epidemiological studies.
In summary, our data illustrate the use of an evidence-based, pathogenicity-driven method for the definition of outcomes in trials of vaccine efficacy. The application of the method to a trial of the efficacy of a killed oral vaccine against ETEC in Egyptian infants and children resulted in the exclusion of episodes of pure LT-ETEC diarrhea from the outcome definition, a result that had not been anticipated before our analyses. These findings illustrate that definitions for outcomes in phase 3 vaccine trials need not be arbitrary and that, in situations in which the vaccine is targeted to symptomatic infections, they can be based on analyses of pathogenicity that use epidemiological data that are customarily obtained as part of the preparation of a field site for a vaccine trial. Outcome definitions that fail to consider the pathogenicity of organisms whose detection does not necessarily imply an etiologic role for a patient's symptoms can dilute the vaccine protection measured by a trial.
