Astrovirus Diarrhea in Egyptian Children

Transcription

1 685 Astrovirus Diarrhea in Egyptian Children Abdollah B. Naficy, 1 Malla R. Rao, 1 Jennifer L. Holmes, 2 Remon Abu-Elyazeed, 3 Stephen J. Savarino, 3 Thomas F. Wierzba, 3 Robert W. Frenck, 3 Stephan S. Monroe, 2 Roger I. Glass, 2 and John D. Clemens 1,4 1 Pediatric Infectious Diseases and Vaccines Section, Epidemiology Branch, National Institute of Child Health and Human Development, National Institutes of Health, Bethesda, Maryland; 2 Viral Gastroenteritis Section, Respiratory and Enteric Viruses Branch, National Center for Infectious Diseases, Centers for Disease Control and Prevention, Atlanta, Georgia; 3 United States Naval Medical Research Unit 3, Cairo, Egypt; 4 International Vaccine Institute, Seoul, Korea This study describes the epidemiology of astrovirus diarrhea among a population-based cohort of 397 children aged!3 years residing in rural Egypt from 1995 to The agespecific incidence s of astrovirus diarrheal per person-year were 0.38 for infants aged!6 months, 0.40 for those aged 6 11 months, 0.16 for those aged months, and 0.05 for those aged months. The overall incidence of astrovirus diarrhea was the same as that of rotavirus diarrhea, 0.19 per person-year. Astrovirus infection was pathogenic and associated with severe dehydration in 17% of the cases. The most frequent serotype was HAstV-1, and, in order of decreasing frequency, HAstV-5, HAstV-8 and HAstV- 3, HAstV-6, HAstV-4, and HAstV-2. In determining whether astrovirus diarrhea was associated with a reduced incidence of subsequent disease, there was evidence to suggest HAstV- 1 homotypic immunity but not heterotypic immunity. Because we observed 38% of the incidence of astrovirus diarrhea to occur in infants aged!6 months, a candidate astrovirus vaccine would have to confer immunity very early in life. Since the initial visualization of astroviruses by electron microscopy in 1975, evidence supporting their importance as causes of childhood diarrhea has been slow to accrue [1, 2]. Although they were rarely seen by electron microscopy in fecal specimens, a seroprevalence study in England reported that 15 (71%) of 21 children aged 3 4 years had antibody to the virus, which indicates a high prevalence of infection but without defined disease [3]. The authors considered low viral pathogenicity or insensitive diagnostic methods as possible reasons for astroviruses not having been more widely recognized at the time. Subsequent improvements in diagnostic methods have led to Received 1 March 2000; revised 20 May 2000; electronically published 17 August Written informed consent was obtained from each child s parent or guardian, and the human experimentation guidelines of the US Departments of Defense and Health and Human Services were followed throughout the study. The views expressed in this article are those of the authors and do not necessarily reflect the official policy or position of the Department of the Navy, Department of Defense, or the US government. Financial support: Naval Medical Research and Development Command (work units M00101.HIX.3421 and M00101.PIX.3270), National Institute of Child Health and Human Development (interagency agreement Y1-HD ), World Health Organization Global Programme for Vaccines and Immunization, World Health Organization Control of Diarrheal Diseases Programme, and Oak Ridge Institute for Science and Education. Reprints or correspondence: Dr. Abdollah B. Naficy, Epidemiology Branch, National Institute of Child Health and Human Development, Rm. 7B03, 6100 Executive Blvd., Rockville, MD (naficy@nih.gov). The Journal of Infectious Diseases 2000;182: by the Infectious Diseases Society of America. All rights reserved /2000/ $02.00 an increasing awareness that astroviruses are a relatively important cause of childhood diarrhea. Studies using EIAs have reported astrovirus prevalence s of 4% for children with diarrhea attending day care centers [4], 7.3% for a cohort of ambulatory Guatemalan children [5], and 2.3% 8.6% for children with diarrhea who presented to treatment centers [6 10]. Most of these studies found astroviruses to be significantly less prevalent among asymptomatic control subjects, and some found them to be the second-most-frequent cause of diarrhea, after rotavirus. In addition, astroviruses have been associated with diarrhea in the immunosuppressed (including those with human immunodeficiency virus infection), persistent diarrhea, and diarrheal outbreaks in schools, nosocomial settings, and nursing homes [11, 12]. Of the 8 distinct astrovirus serotypes identified to date, serotype 1 (HAstV-1) appears to be the most common, and HAstV-5, HAstV-6, and HAstV-7 are the least common [13 17]; however, this observation may depend on location [18]. Precise epidemiological data based on recent advances in the molecular characterization of astrovirus serotypes are needed to provide a framework for determining future research priorities and for identifying potential stgies for the control of astrovirus diarrhea. The objective of this prospective population-based cohort study of Egyptian children was to determine age-specific incidence s of diarrhea caused by astroviruses, the association between excretion of astroviruses and the occurrence of diarrheal symptoms, sociodemographic and environmental factors associated with the risk of astrovirus diarrhea, the astrovirus serotypes in circulation, and whether

2 686 Naficy et al. JID 2000;182 (September) an astrovirus diarrheal episode was associated with a reduced incidence of a subsequent episode. Methods Study population. In January 1995, following a house-to-house census, a cohort of children aged!24 months was assembled from 2 villages (village 830 and village 820) in Abu Homos, a rural district located in the Nile Delta in northern Egypt. Baseline sociodemographic and household hygiene information was collected during the census. Thereafter, newborns in households within the census area were continuously enrolled in the cohort. During the study period, a total of 397 children were recruited. At the time of enrollment in the cohort, 211 children were aged!28 days and thereby constituted the birth cohort. Surveillance. Active surveillance of the cohort with twiceweekly home visits commenced in February 1995 and terminated at the end of the study period, in February If the child had diarrhea, he or she was examined by a study physician who obtained a fecal specimen (stool sample) and classified clinical dehydration, according to the World Health Organization criteria as none, some, or severe [19]. When the physician deemed it necessary, symptomatic children were referred to a village clinic for further evaluation and rehydration therapy. Deaths and other losses to follow-up were recorded at the twice-weekly visits. All individuals in the cohort, irrespective of symptoms, were surveyed once every 2 months, at which time a fecal specimen (stool sample) was collected. At each home visit, information regarding the child s diet and breast-feeding status was obtained. Definitions. A diarrheal day was defined as the occurrence of 3 nonformed stools (or 1, if bloody) in a 24-h period. In addition, if the child was breast-fed and the stool was not bloody, the mother had to report an increase in frequency or a reduction in consistency of the stools, compared with what she considered to be normal. A diarrheal episode was defined as beginning on the first diarrheal day after 3 consecutive nondiarrheal days and ending on the first diarrheal day to be followed by 3 consecutive nondiarrheal days. An astrovirus diarrheal episode was defined as a diarrheal episode in which astrovirus was detected in a fecal specimen collected during the episode. A rotavirus diarrheal episode was defined as a diarrheal episode in which rotavirus was detected in a fecal specimen collected during the episode. Breastfeeding was defined as the receipt of any breast milk, irrespective of other components of the diet. Laboratory evaluations. All fecal specimens were transported in ice packs to the Abu Homos field laboratory, where they were refriged. Twice weekly, the fecal specimens were transported in ice packs to the laboratories of US Naval Medical Research Unit 3 in Cairo, where they were kept frozen at 70 C and were subsequently transported on dry ice to laboratories at the Viral Gastroenteritis Section of the Centers for Disease Control and Prevention in Atlanta, Georgia. Stool suspensions were prepared as 10% extracts and were tested for astrovirus and rotavirus by use of commercial enzyme-linked immunosorbent assay kits (IDEIA Astrovirus; DAKO Diagnostics, Ely, UK; and Rotaclone; Meridian Diagnostics, Cincinnati, OH). IDEIA Astrovirus has been tested on 213 fecal specimens collected from patients presenting with gastroenteritis to 4 centers in the United Kingdom. The reference method used to determine astrovirus infection was electron microscopy (EM), and discrepant results were resolved by immune electron microscopy (IEM). The overall sensitivity and specificity of IDEIA Astrovirus, compared with that of EM and IEM, was 100% (IDEIA Astrovirus [package insert]; DAKO Diagnostics). All astrovirus strains were genotyped, which correlates with serotype, by reverse transcriptase polymerase chain reaction (RT-PCR) using the MON244 and MON245 primers [16]. Among the fecal specimens collected during the bimonthly surveys, only specimens collected during the first year of surveillance were tested for astrovirus and rotavirus. Analyses. Crude diarrhea incidence s and age-specific incidence s were calculated by dividing the number of by the number of person-years at risk. The variables evaluated as potential predictors of astrovirus incidence were sex, village, season (warm vs. cool), maternal education (any vs. none), crowding (no. of persons in household/no. of rooms in house), presence of a household latrine, whether the house had electricity, and breast-feeding (any vs. none). Breast-feeding was treated as time-varying. Univariate analyses were carried out for all exploratory variables, and crude incidence s were compared using the Mantel-Haenszel test statistic for density follow-up studies [20]. The independent associations between variables under study and incidence of astrovirus diarrhea were estimated by multivariate analyses using Poisson regression models. The generalized estimating equations method was used to account for intrachild correlation [21]. A statistical significance level of 0.10 was selected for inclusion of variables in multivariate modeling. Associations, expressed as relative s, were derived from exponentiating the model coefficients. Standard errors of the coefficients were used to derive 95% confidence intervals (CI). In all analyses, statistical significance was defined as P!.05 (2-tailed). To identify factors associated with severe disease in children with of astrovirus diarrhea (virulence), we compared children who had clinical evidence of at least some dehydration with children who had no clinical evidence of dehydration. Comparisons were statistically appraised with the x 2 test or with Fisher s exact test, if the population was sparsely distributed. A case-control analysis was used to evaluate the association between excretion of astroviruses and the occurrence of diarrheal symptoms (pathogenicity). This analysis was restricted to the first year of surveillance, because only the first-year specimens from the bimonthly surveys were tested for astrovirus. The sampling frame for case patients was any individual with an episode of diarrhea, and the sampling frame for control subjects was all individuals without diarrhea in each bimonthly stool survey. Both case patients and control subjects with a history of having astrovirus detected in their stools were excluded (either as a case patient or as a control subject). To avoid bias from repeated fecal sampling during a single diarrheal episode, only the first fecal specimen obtained during each episode was included in these analyses. The independent association between the initial detection of astrovirus in fecal specimens and diarrheal was evaluated using a multiple logistic regression model, taking case-control status as the dependent variable and fitting the detection of astrovirus in fecal specimens and potentially confounding variables as the independent variables. The generalized estimating equations method was used to account for intrachild correlation [21]. The coefficient for detection of astro-

3 JID 2000;182 (September) Astrovirus Diarrhea in Egyptian Children 687 Table 1. Annual age-specific incidence s ( per person-year) of all-cause, astrovirus, and rotavirus diarrhea, Abu Homos, Egypt, Age group, months Days of follow-up Diarrhea Astrovirus Rotavirus , , , , Total 232, virus in fecal specimens was exponentiated to estimate the adjusted odds ratio. To determine whether an astrovirus diarrheal episode was associated with a reduced incidence of a subsequent episode during the period of observation, the incidence of a second astrovirus diarrheal episode was compared with the incidence of a first episode by use of the Cox proportional-hazards model [22, 23]. This analysis was restricted to the birth cohort. Results During the study period, 95% of the twice-weekly scheduled visits were completed, and 1 stool specimen was collected during 2863 (82%) of the 3477 diarrheal. Of the bimonthly scheduled survey visits, 99% were completed, and a stool specimen was obtained from 3947 (98%) of the 4041 visits. Losses to follow-up recorded at the twice-weekly visits included 13 deaths and 10 out-migrations. During the study period, 123 astrovirus diarrheal occurred in 106 children; 13 children each had 2, and 2 children each had 3. The overall annual incidence s of all diarrheal and of astrovirus and rotavirus diarrheal were 5.46, 0.19, and 0.19 per person-year, respectively (table 1). The age-specific incidence s of astrovirus diarrheal per person-year were 0.38 for infants aged!6 months, 0.40 for those aged 6 11 months, 0.16 for those aged months, and 0.05 for those aged months (table 1). Astrovirus diarrhea occurred more frequently during the warmer season, with a peak incidence during the months of February July (figure 1). The clinical characteristics of the 123 astrovirus diarrheal included fever (33%), vomiting (20%), some degree of dehydration (48%), and severe dehydration (17%). The median duration of an episode was 2 days (range,!1 31 days). By comparison, of the 118 rotavirus diarrheal, 63 (53%) were characterized by some degree of dehydration and 28 (24%) by severe dehydration. After excluding 4 of astrovirus and rotavirus coinfection, there was no significant difference between the proportion of astrovirus and rotavirus diarrheal characterized by some dehydration (55/119 vs. 59/114, respectively; P p.40) or severe dehydration (20/119 vs. 27/114, respectively; P p.19). At least 1 copathogen was identified in 54 of the 123 astrovirus diarrheal (38 enterotoxigenic Escherichia coli, 13 Campylobacter, 8 Shigella, and 4 rotavirus). There was no significant difference between the proportion of astrovirus diarrheal with and without copathogens characterized by some dehydration (28/54 vs. 31/69, respectively; P p.45) or severe dehydration (9/54 vs. 12/69, respectively; P p.92). Crude relative s for the associations between selected sociodemographic or environmental factors and the incidence ( per person-year) of astrovirus diarrhea were statistically significant for village, season, and breast-feeding (table 2). The variables included in the final multivariate model were age, village, season, maternal education, crowding, and breastfeeding. Relative s for the associations between each variable and the incidence of astrovirus diarrhea, adjusted for the aforementioned variables, and their 95% CIs are shown in table 2. The variables found to be significantly associated with the incidence of astrovirus diarrheal were village and season. No sociodemographic or environmental factor was found to be significantly associated with severe disease in children with of astrovirus diarrhea. Severe dehydration was observed in 20 (19%) of the 106 first of astrovirus diarrhea, compared with 1 (6%) of the 17 second or third of astrovirus diarrhea (relative odds, 3.72; 95% CI, ; P p.3). In evaluating astrovirus pathogenicity, after controlling for age, village, and season, the relative odds of excreting astrovirus in children having diarrhea, compared with children having no diarrhea, was 1.94 (95% CI, ; P p.01). Of the astroviruses isolated from each of the 123 diarrheal Figure 1. s ( per person-year) of astrovirus diarrhea by calendar month, Abu Homos, Egypt,

4 688 Naficy et al. JID 2000;182 (September) Table 2. Crude and adjusted relative s for the associations between selected sociodemographic or environmental factors and the incidence ( per person-year) of astrovirus diarrhea, Abu Homos, Egypt, Variable (n p 123) ( per person-year) Crude relative Adjusted relative a 95% confidence interval Sex Male Female Village Season Warm Cool Maternal education Any None persons in household/ no. of rooms in house ! Latrine Yes No Electricity Yes No Breast-feeding Yes No a Adjusted for age, village, season, maternal education, crowding, and breast-feeding., 40 were nontypeable (2 because of an inadequate amount of stool). Of the remaining 83, the most frequent type was HAstV-1 ( n p 36), and, in order of decreasing frequency, HAstV-5 ( n p 13), HAstV-8 and HAstV-3 ( n p 10 each), HAstV-6 ( n p 6), HAstV-4 ( n p 4), and HAstV-2 ( n p 3). One episode consisted of a mixed infection of HAstV-6 and HAstV- 8. There were too few to compare clinical severity by each infecting astrovirus serotype. There was, however, no significant difference between HAstV-1 and all other astrovirus serotypes combined in the proportion of diarrheal characterized by severe dehydration (5 of 36 vs. 8 of 46, respectively; P p.67). The hazards ratio of of astrovirus diarrhea in those with a prior history of astrovirus diarrhea, compared with those who had no prior history of astrovirus diarrhea, was 1.12 (95% CI, ; P p.75). During the study period, there were 16 children in the birth cohort who had an episode of HAstV-1 diarrhea. Of these 16 children, none had a subsequent episode of HAstV-1 diarrhea, although 1 had a subsequent episode of HAstV-3 diarrhea, and 1 had 2 subsequent of astrovirus diarrhea, an HAstV-5 and a nontypeable astrovirus. Because the Cox proportional-hazards model assumptions were not met, the incidence of a second HAstV-1 diarrheal episode was compared with the incidence of a first episode using the Mantel-Haenszel test statistic, which yielded a relative of zero ( P p.32). Discussion In our study population of rural Egyptian children, we found astrovirus infection to be pathogenic, as common as rotavirus infection, and associated with severe dehydration in 17% of the cases (with or without copathogens). It has been shown that the diagnostic sensitivity for astrovirus improves significantly if EIA is substituted by RT-PCR. In an astrovirus diarrhea outbreak in a day care center, astrovirus was detected in 32% of stool specimens screened by RT-PCR, as opposed to only 10% by EIA [24]. Because we used an EIA to screen fecal specimens for the presence of astrovirus (although not the same assay as used in the study by Mitchell et al. [24]), and because we were unable to collect a stool specimen from 18% of diarrheal, it is possible that we have underestimated the true incidence of astrovirus diarrhea in this population. Nevertheless, the incidence of astrovirus diarrhea in this population was significantly higher than that in an earlier study of Mexican children observed from birth until 18 months of age [18], and is consistent with seroprevalence studies in Europe and the United States, which suggest that, by the age of 10 years, almost all children have been infected with astrovirus [25 27]. A limitation of this study was our inability to serotype 40 of the astroviruses detected by IDEIA Astrovirus. Other than 2 specimens in which there was an inadequate amount of stool, we believe that the most likely explanation was low concen-

5 JID 2000;182 (September) Astrovirus Diarrhea in Egyptian Children 689 tration of virus in the specimens that precluded both direct typing in the stool and adaptation to cell culture. Nevertheless, other possibilities, such as IDEIA Astrovirus false positives, suboptimal stool extraction methods resulting in low RNA yields, and primer selection that was inadequate for the strains detected in this geographic area cannot be entirely ruled out. The intent of this study was not to evaluate the specificity of the IDEIA Astrovirus, and the RT-PCR used in this study was designed for typing astrovirus strains and is not the most sensitive RT-PCR system available for astrovirus detection. Our inability to identify any potential risk factors whose modification might control diarrheal disease due to astrovirus infection, coupled with previous observations that astrovirus diarrhea occurs with comparable frequency among children in both developed and developing countries, suggest that, as is true for rotavirus diarrhea, improvements in hygiene and living conditions alone are unlikely to be sufficient preventive measures in this population. In assessing the potential role for immunization in the control of astrovirus diarrhea, we were interested in evaluating the evidence for natural immunity following an episode of astrovirus diarrhea. The observed decrease in incidence of astrovirus diarrhea with increasing age does suggest the likelihood of acquired immunity following infection. During the period of observation in this study, an episode of astrovirus diarrhea was not associated with a reduced incidence of a subsequent episode, which suggests the absence of heterotypic immunity; however, the observation that no child in the birth cohort who had an episode of HAstV-1 diarrhea suffered a subsequent episode with the same serotype, despite continued circulation of HAstV-1 in the environment, suggests that there may be homotypic immunity against subsequent astrovirus diarrhea. Also of note was the diminution in the frequency of severe dehydration observed in children with second and third astrovirus diarrheal, compared with children who were having a first episode. Although not statistically significant, these latter observations suggest that, as with rotavirus, immunization against astrovirus may be more effective in preventing infection caused by the same serotype contained in the vaccine and more effective in preventing severe disease rather than infection. Further epidemiological studies in different settings are needed to determine the burden of astrovirus disease, especially the risk of severe disease in hospital settings, and the different serotypes in circulation. In conducting these studies, it will be important to use the most sensitive and refined diagnostic methods available, to avoid underestimating the true incidence of astrovirus disease. The results of such studies are required to determine whether vaccine development against astrovirus diarrhea should be considered a public health priority. Because we observed 38% of the incidence of astrovirus diarrhea to occur in infants aged!6 months, if a candidate astrovirus vaccine is to be incorpod into the national Expanded Programme on Immunization (World Health Organization), it will have to confer immunity very early in life. Acknowledgments We wish to thank Dr. Sahar Abd El Samad, Dr. Manal El Sayed, and the staff of the Abu Homos Field Research Center, for their contributions to field and laboratory work; Dr. Mahmoud Abu El Nasr and Dr. Badria Z. Morsy of the Egyptian Ministry of Health and Population, for their support and advice; and John O Connor of the Centers for Disease Control and Prevention, for editorial assistance. References 1. Appleton H, Higgins PG. Viruses and gastroenteritis in infants. Lancet 1975;1: Madeley CR, Cosgrove BP. 28-nm particles in faeces in infantile gastroenteritis. Lancet 1975;2: Kurtz J, Lee T. Astrovirus gastroenteritis age distribution of antibody. Med Microbiol Immunol 1978;166: Lew JF, Moe CL, Monroe SS, et al. Astrovirus and adenovirus associated with diarrhea in children in day care settings. J Infect Dis 1991;164: Cruz JR, Bartlett AV, Herrmann JE, Caceres P, Blacklow NR, Cano F. Astrovirus-associated diarrhea among Guatemalan ambulatory rural children. J Clin Microbiol 1992;30: Herrmann JE, Taylor DN, Echeverria P, Blacklow NR. Astroviruses as a cause of gastroenteritis in children. N Engl J Med 1991;324: Kotloff KL, Herrmann JE, Blacklow NR, et al. Medical importance of astrovirus as a cause of diarrhea in Baltimore children. Pediatr Infect Dis J 1992;11: Steele AD, Basetse HR, Blacklow NR, Herrmann JE. Astrovirus infection in South Africa: a pilot study. Ann Trop Paediatr 1998;18: Unicomb LE, Banu NN, Azim T, et al. Astrovirus infection in association with acute, persistent and nosocomial diarrhea in Bangladesh. Pediatr Infect Dis J 1998;17: Bon F, Fascia P, Dauvergne M, et al. Prevalence of group A rotavirus, human calicivirus, astrovirus, and adenovirus type 40 and 41 infections among children with acute gastroenteritis in Dijon, France. J Clin Microbiol 1999;37: Glass RI, Noel J, Mitchell D, et al. The changing epidemiology of astrovirusassociated gastroenteritis: a review. Arch Virol 1996;12(Suppl): Unicomb LE, Banu NN, Azim T, et al. Astrovirus infection in association with acute, persistent, and nosocomial diarrhea in Bangladesh. Pediatr Infect Dis J 1998;17: Kurtz JB, Lee TW. Human astrovirus serotypes. Lancet 1984;2: Lee TW, Kurtz JB. Prevalence of human astrovirus serotypes in the Oxford region, , with evidence for two new serotypes. Epidemiol Infect 1994;112: Noel J, Cubitt D. Identification of astrovirus serotypes from children treated at the Hospitals for Sick Children, London, Epidemiol Infect 1994;113: Noel JS, Lee TW, Kurtz JB, Glass RI, Monroe SS. Typing of human astroviruses from clinical isolates by enzyme immunoassay and nucleotide sequencing. J Clin Microbiol 1995;33: Carter MJ, Willcocks MM. The molecular biology of astroviruses. Arch Virol 1996;12(Suppl): Guerrero ML, Noel JS, Mitchell DK, et al. A prospective study of astrovirus diarrhea of infancy in Mexico City. Pediatr Infect Dis J 1998;17: World Health Organization Diarrhoeal Disease Control Program. The treatment of diarrhoea: a manual for physicians and other senior health workers [WHO/CDR/95.3 rev. 3]. Geneva: World Health Organization, 1995: 1 7.

6 690 Naficy et al. JID 2000;182 (September) 20. Klienbaum DG, Kupper LL, Morgenstern H. Epidemiologic research: principles and quantitative methods. Belmont, CA: Lifetime Learning, 1982: Liang KY, Zeger SL. Longitudinal data analysis using generalized linear models. Biometrika 1986;73: Prentice RL, Williams BJ, Peterson AV. On the regression analysis of multivariate failure time data. Biometrika 1981;68: SAS Institute. SAS/STAT software: changes and enhancements through release Cary, NC: SAS Institute, Mitchell DK, Monroe SS, Jiang X, Matson DO, Glass RI, Pickering LK. Virologic features of an astrovirus diarrhea outbreak in a day care center revealed by reverse transcriptase polymerase chain reaction. J Infect Dis 1995;172: Kriston S, Willcocks MM, Carter MJ, Cubitt WD. Seroprevalence of astrovirus type 1 and 6 in London, determined using recombinant virus antigen. Epidemiol Infect 1996;117: Koopmans MPG, Bijen MHL, Monroe SS, Vinje J. Age-stratified seroprevalence of neutralizing antibodies to astrovirus types 1 to 7 in humans in the Netherlands. Clin Diagn Lab Immunol 1998;5: Mitchell DK, Matson DO, Cubitt WD, et al. Prevalence of antibodies to astrovirus types 1 and 3 in children and adolescents in Norfolk, Virginia. Pediatr Infect Dis J 1999;18: