8562

Saturday 3 October 1987

BRAZILIAN PURPURIC FEVER:EPIDEMIC PURPURA FULMINANSASSOCIATED WITH ANTECEDENTPURULENT CONJUNCTIVITIS

BRAZILIAN PURPURIC FEVER STUDY GROUP*

Summary In late 1984, 10 children in a small, ruraltown in Brazil had high fever associated

with vomiting and abdominal pain. Within 12-48 h of theonset of fever, purpura developed associated with vascularcollapse and peripheral necrosis. All 10 children died.

Cerebrospinal fluid examinations did not suggest meningitisand, when done, tests were negative for Neisseria

meningitidis. Other culture, serological, and necropsyexaminations did not reveal a cause. Case-finding uncoveredanother cluster of similar illness in children in a second townand sporadic cases in five other cities. Two case-controlstudies demonstrated that children who became ill were

significantly more likely than control children to have had

*The principal institutions, their directors, and other individuals whocontributed to this study are: Institute Aldolfo Lutz, São Paulo, Brazil(general director, Pedro Paulo Chieffi; immunologist, Augusta K. Takeda;microbiologists, Maria Cristina de Cunto Brandileone and Maria LuciaCecconi Tondella); São Paulo Ministry of Health, Centre for Epidemiology(director, Alexandre Vranjac; epidemiologists Carlos Alberto Macharelli andPedro Veneziani); Promissao General Hospital, São Paulo (medical director,Antonio Gelas); Parana Ministry of Health, Epidemiology Division,Londrina (director, Helio Camargo); Hospital Emilio Ribas (general director,Vasco de Lima); and Meningitis and Special Pathogens Branch, Centers forDisease Control, Atlanta, Georgia, USA (chief, Claire V. Broome;epidemiologists David Fleming and Seth Berkley; microbiologists, GeorgeCarlone and Loretta Carson; pathologists Leo Gorelkin and Francis

Chandler).This report was prepared by Dr David Fleming and Dr Seth Berkley,

CDC, Atlanta.

conjunctivitis during the month before illness. This

conjunctivitis was purulent, preceded the onset of moresevere disease by 3-15 days, and had resolved before feverbegan. Although no conjunctival cultures were obtainedfrom case-children, Haemophilus aegyptius was the mostcommon pathogen isolated from other conjunctival culturesduring the epidemic. This organism was also isolated from anon-aseptic skin scraping from 1 case child. A 25-

megadalton plasmid distinguished the H aegyptius isolatesepidemiologically associated with illness from otherBrazilian conjunctival isolates. Brazilian purpuric fever is anewly recognised syndrome of epidemic purpura fulminansassociated with antecedent purulent conjunctivitis, possiblycaused by H aegyptius.

Introduction

PURPURA fulminans is an acute syndrome consisting ofrapidly progressive, disseminated purpura associated withvascular collapse and hypotensive shock. This illness usuallyoccurs in children as a rare complication of bacterial sepsis,most commonly meningococcaemia. A delayed form offulminant purpura not associated with sepsis and presumedto be related to a preceding viral or bacterial illness has alsobeen reported.!

In late 1984, 10 children in Promissao, a small town in theinterior of the state of Sao Paulo, Brazil, died after afulminant purpuric illness. No cause was identified. Therehad been no cases of meningitis during this period, and noincrease in illnesses or deaths was noted in older children oradults. In January, 1985, the Sao Paulo Ministry of Healthinvited the Pan American Health Organisation (PAHO) andthe Centers for Disease Control (CDC) to collaborate in anevaluation of the outbreak. The subsequent investigation

758

resulted in the characterisation of an apparently newsyndrome-epidemic purpura fulminans associated withantecedent purulent conjunctivitis. This syndrome, giventhe working name of Brazilian purpuric fever (BPF), isdescribed in this report.

Methods

Case DefinitionAnalysis of the characteristics of the illness affecting the 10

children whose deaths prompted the investigation led theformulation of a clinical case definition. This definition, used in thesecond outbreak investigation in Londrina, included children aged10 years of age or younger with an unexplained, febrile ( > 38-5°C)illness without meningeal signs or symptoms that progressed within48 h to hypotensive shock associated with purpura. When the linkbetween severe disease and antecedent conjunctivitis was identified,a more specific case definition for identifying sporadic illness wasdeveloped (table i).

’

Case FindingCases were identified retrospectively in Promissao and Londrina

through physicians’ and hospital reports. To determine whetherother but unrecognised clusters of illness had occurred, deathcertificates from 1983 and 1984 for residents from selected cities inwestern Sao Paulo State were reviewed for diagnoses compatiblewith the svndrome. including Duroura fulminans, haemorrhagicfever, meningococcaemia, sepsis, and cause of death unknown.Active surveillance for sporadic cases was initiated in January, 1985,and was facilitated by widespread media attention.

Collection of Clinical and Pathological DataMedical records of affected children were reviewed and the

physicians who had provided primary care were interviewed. Whenpossible, interviews with patients, physical examination of patients,and necropsies were done or attended by one of the principalinvestigators.

Clinical laboratory testing, done at the discretion of the primaryphysician, was limited by the remoteness of the two towns. Mostpathological material was obtained at necropsy and was fixed in10% formalin or frozen in liquid nitrogen. Tissue slides werereviewed by pathologists at the University of Sao Paulo in RibeiraoPreto and Botucatu, at the Instituto Adolfo Lutz in Sao Paulo, atCDC, and at the US Armed Forces Institute of Pathology.

Laboratory TestingSeveral laboratories evaluated clinical and microbiological

material. Clinical laboratory examinations were done at local orregional facilities in Sao Paulo and Parana. Conjunctival cultureswere processed, bacterial isolates were identified, and virus serologywere done in Brazil. Quantitative assays for endotoxin (Limulusamoebocyte lysate system2), IgG and IgM assays for haemorrhagicfevers,3 and attempts at viral, rickettsial, and bacterial isolation from

TABLE I-CASE DEFINITION FOR BRAZILIAN PURPURIC FEVER

necropsy tissues were done at CDC. Plasmid analysis4 andbiotyping of Haemophilus spp were also done at CDC; conjunctivalstrains and the one skin isolate were identified as H aegyptius bybiochemical and outer-membrane protein analysis.5 Laboratoryinvestigators did not know the source of specimens sent forendotoxin assay and plasmid analysis.

Epidemiological EvaluationMembers of the families of case children in Promissao were

intensively questioned about events affecting the child during themonth before death. Hypotheses generated by these interviewswere tested in a case-control study. Case children were frequencymatched by age with control children randomly selected from thetown’s public health clinic records, estimated to include 75-90% ofall children in Promissao. Parents were asked about events duringOctober, November, and December, 1984, the epidemic period.

Physicians caring for children in Londrina were directlyinterviewed by one of the investigators. Data for the limitedcase-control study in Londrina were obtained by interviewers whowere unaware of hypotheses being tested.

Results

Outbreaks

Promissao (population 20 430) is a small rural town 470km west of Sao Paulo. Between Oct 14 and Dec 15,1984,10children in Promissao had a rapidly progressive illness thatbegan with the acute onset of fever, usually accompanied byvomiting and abdominal pain. These children ranged in agefrom 3 months to 8 years (median 30 months); 4 weremale and 6 were female. 12-48 hours after the onset offever, petechiae developed associated with hypotension andshock. All 10 died. At the time of death, purpuric lesionswere prominent, most often affecting the face (figure) andextremities. These lesions were usually accompanied bycyanosis or early necrosis of peripheral tissues, notably thehands, feet, nose, and ears. The median time from onset ofsymptoms to death was 1 day; the estimated attack rate of

Purpura in child with BPF. -

759

TABLE II-SUMMARY OF CLINICAL LABORATORY EVALUATIONS OF

CHILDREN WITH BRAZILIAN PURPURIC FEVER 1984-85

this illness in Promissao was 2-3 per 1000 children aged 0-9years. Meningitis was not present on clinical examination oron cerebrospinal fluid (CSF) analysis (see below). Nomeningococcal disease or other unspecified meningitis wasrecorded in any town resident during the last quarter of1984.

During this same time, 11 other children in Promissao fellill with fever, vomiting, and abdominal pain of unknownorigin. These suspect cases were admitted to hospital butdisseminated purpura did not develop and they all survived.2 had purpuric lesions. The dates of onset for the suspectcases clustered towards the end of the outbreak, at a timewhen all children with fever of undetermined cause were

being admitted to hospital and treated promptly. Hospitaltreatment of cases and suspect cases was similar, consistingof broad-spectrum antimicrobials (usually ampicillin andchloramphenicol), intravenous fluids, and systemiccorticosteroids.Review of death records uncovered no other clusters of

purpuric deaths in the western half of Sao Paulo State.However, a cluster of cases indistinguishable from those inPromissao had been reported from the city of Londrina, 250km south-west of Promissao in the State of Parana. From

February to July, 1984, 13 children, aged 6 months to 110years (median 36 months), had had fever, vomiting, andabdominal pain that quickly progressed to purpura,hypotension, and shock of unknown origin. Meningitis wasnot present clinically or by CSF examination, and necrosisof the extremities was a common finding, especially insurvivors. 7 children died.

Sporadic Cases

6 additional sick children, clinically indistinguishable butseparated by 1-6 months in time from outbreak cases, wereidentified in Promissao (2) and Londrina (4). Subsequently,a case definition for sporadic illness was developed (table 1)to see if unrecognised cases were occurring elsewhere in SaoPaulo State. 9 children meeting this definition of BPF withonsets between October, 1984, and February, 1985, wereidentified in six towns, all within 100 km of Promissao.

Laboratory EvaluationOf the 38 children (12 in Promissao, 17 in Londrina, and 9

in other towns) with BPF 27 (71 %) died. Clinical laboratorydata are summarised in table 11, the most remarkable

findings being a profound thrombocytopenia and prolonged

TABLE III-SUMMARY OF NEGATIVE TESTS FOR INFECTIOUS

CAUSES FROM CHILDREN WITH BRAZILIAN PURPURIC FEVER:

BRAZIL 1984-85

*A Vero or HLF pus.tHEK, HLF, HEP2, PMK, enterovir-uses.tAcute sera for cases; paired sera for suspect cases.§Congo, Rift Valley, Ebola, Lassa, Marburg, Junin, Machupo.

prothrombin time that usually accompanied the purpuricphase of the illness.CSF Gram stains and cultures from case children were

uniformly negative. CSF counterimmunoelectrophoresis(CIE) was negative in 4 cases for Neisseria meningitidis(groups A, B, C, Y, W135), H influenzae b, and

Streptococcus pneumoniae. For 9 cases and 2 suspect casesnegative blood cultures were noted in the medical records.However, we could not exclude the possibility that over-the-counter antimicrobials, freely available in Brazil, had beenadministered to these children before hospital admission.Acute serum was available from 5 case children and acuteurine from 3 case children. All were negative by latexagglutination testing for N meningitidis (A, B, C, Y, W 135),H influenzae b, and S pneumoniae antigens. A skin scrapingfrom a purpuric lesion from one child who died grewH aegyptius; however, this scraping was obtained non-aseptically. A summary of other negative diagnostic tests ispresented in table III.The mean quantitative endotoxin level was 675 pg/ml

(range 1000-1600) in acute sera from 8 children with BPFand 2 case-children with suspected BPF. In contrast, themean level was 25 pg/ml (range 0-100) in similarlyprocessed sera from 7 healthy control children fromPromissao and it was 70 pg/ml (range 0-400) in acute serastored at CDC from 5 children with meningococcal and 1child with pneumococcal disease.

Pathology

Necropsy material was available from 6 children who diedfrom BPF (3 epidemic and 3 sporadic cases). Petechial andpurpuric lesions of the skin and mucous membranes werethe most prominent findings. Cerebral and pulmonaryoedema and adrenal haemorrhage were also present. Underlight microscopy, the most consistent finding was a

haemorrhagic diathesis compatible with purpura fulminans.This was characterised by major adrenal haemorrhage,ecchymosis in the upper dermis with focal intravascular

760

fibrin deposition, and pulmonary congestion with diffusealveolar damage and focal haemorrhage. There was

pronounced lymphocyte depletion in the spleen and lymphnodes. Meningitis was not present either by gross or

microscopic examination, and there was no evidence ofacute vasculitis. No viral-like inclusions were seen in any ofthe tissues, nor were bacteria demonstrated by haematoxylinand eosin, Dieterle, or Brown-Hopps tissue Gram stains.Immune staining of pathological material with sera fromconvalescent children was unrevealing.

Epidemiological InvestigationThe case-control study in Promissao enrolled 10 case

children, 10 suspect case children, and 20 control children.There were no significant differences between case or

suspect case and control children with regard to severalpotential risk factors, including history of chronic illnesses,exposure to children with BPF, recent immunisation

history, parental occupation, family socioeconomic status,method of sewage disposal, source of water, and exposure toanimals. Exposure to a known exogenous toxin was notsuggested by the clinical syndrome, and no potential sourceof an unknown toxin affecting either this town or thechildren who died was identified. No evidence was found to

support mercury, ergot, atropine, or pesticide exposure.Although case children tended to live around the peripheryof the town, this pattern was not greatly different from thatin control children.One strong association with illness was identified. Case

children and suspect case children were more likely thancontrols to have recently had conjunctivitis (18/20 cases andsuspect cases vs 9/20 controls; odds ratio = 11, p = 0-006Fisher’s exact test). Characteristically, this conjunctivitiswas improving or had resolved by the time more severesymptoms occurred. Parents of case children had not usuallyassociated this conjunctivitis with more severe disease.The association between BPF and antecedent

conjunctivitis was assessed in a second study in Londrina.Londrina’s hospital serves 500 000 people, and severalpeople with culture-proven meningococcal disease had beenadmitted to hospital during the BPF epidemic. To assess thepotential bias, inherent in the Promissao study, of

comparing ill cases with well controls we compared childrenin Londrina with meningococcal disease with children withBPF. Parents of 13 of 15 children with negative cultures forN meningitidis and presumed BPF reported a history ofconjunctivitis preceding more severe illness compared withonly 1 of 8 cases with proven meningococcal disease (oddsratio-45, p= 0-0001).The association of conjunctivitis with BPF was found

retrospectively, and no cultures or Gram stains of

conjunctivitis -material were obtained from any child inwhom BPF subsequently developed. The conjunctivitispreceding BPF was described in retrospect as purulentrather than mucoid or haemorrhagic. There was a time lagbetween onset of conjunctivitis and onset of more severedisease (median 7 days, range 15 days). Local physiciansreported that purulent conjunctivitis had been epidemic atthe time of the BPF outbreaks both in Promissao and inLondrina. However, in contrast to previous epidemicsof haemorrhagic viral conjunctivitis, the epidemicconjunctivitis associated with BPF was reported to haveaffected young children predominantly.During the outbreak of BPF in Promissao, conjunctival

specimens were obtained from 24 children with epidemicconjunctivitis. Gram stains of 17 of these specimens

demonstrated a neutrophilic exudate with toxic granulation,suggesting a bacterial cause. H aegyptius, the organismisolated from the skin scraping of 1 acutely ill case child, wasisolated from 7 of the 24 conjunctival swabs obtained duringthe Promissao epidemic, making this bacterium the onlypathogen commonly isolated.To investigate further the possible association between

BPF and H aegyptius, we compared 19 control isolates ofH aegyptius obtained from a systematic survey of cases ofconjunctivitis in western Sao Paulo State in February,1985,6 with the isolates obtained from Promissao during theepidemic period and with isolates from family contacts ofchildren with sporadic BPF. All 9 isolates of H aegyptiusassociated with BPF contained a 25 megadalton plasmid. Incontrast, only 7 of 19 control isolates contained this plasmid(p = 0-002).

Discussion

This report has summarised the characteristics of a

purpuric illness occurring in southern Brazil in 1984 andearly 1985. This illness, called Brazilian purpuric fever,affected young healthy children, occurred both sporadicallyand in clusters, and was fulminant and often fatal. The dataavailable at this time suggest that BPF is a distinct and

previously unrecognised illness.Study of BPF was hindered by the remote location of

cases, the short time that children were ill before death, andthe retrospective nature of much of the data collection.Although many potential causes were investigated, not allcases could be tested against all potential pathogens andsome of these cases could potentially represent otherillnesses. However, given the clinical presentation,fulminant nature, and epidemic pattern of BPF, few knownillnesses could mimic these cases when considered as a

group. The time from onset of symptoms to death was tooshort to implicate a known viral haemorrhagic fever.

Dengue shock syndrome was considered; however,haematocrits during the acute phase of the illness were notabnormal, no dengue was reported in the area, and dengueserological tests on suspect case children were negative.Epidemic meningococcal disease (specifically, epidemicmeningococcaemia) seemed the most likely explanation atfirst. However, several lines of evidence led us away fromthis diagnosis. Meningococcal disease is common in Brazil,yet physicians experienced in diagnosing N meningitidisinfection felt that BPF was a separate entity. More directly,many of the case children were negative by cultureor antigen detection for N meningitidis. From an

epidemiological standpoint, meningococcaemia occurs

concurrently with meningococcal meningitis duringepidemics, and comprises only 0-35 % of all N meningitidisdisease.7 In Promissao, no meningitis, meningococcal or

. otherwise, occurred during the epidemic period. In

Londrina, background meningococcal disease was

recorded, but BPF was shown to be epidemiologicallydistinct from proven N meningitidis infection in respect ofhistory of preceding purulent conjunctivitis.

Several conclusions about the nature of BPF can bedrawn. Children, predominantly those aged 1, are most atrisk. The disease has both epidemic and sporadic forms.Because no ill child had contact with another known case,either BPF is not transmissible from person to person orthere is a subclinical or asymptomatic carrier state. Thefulminant nature of BPF strongly suggests toxin-mediatedillness; the clinical presentation suggests that the vascularendothelial cell may be one site of action. High levels of

761

circulating endotoxin measured in acute sera raise the

possibility of an endotoxin-mediated illness. A role forendotoxin is supported by-the resemblance between BPFand purpura fulminans caused by bacterial sepsis.

Purulent conjunctivitis distinct from subsequent, moresevere symptoms appears to be a major risk factor fordevelopment of BPF. The data available indicate that

conjunctivitis associated with BPF may be bacterial in originand several lines of evidence suggest a role for H aegyptius:H aegyptius (H influenzae biotype III) is a recognised causeof epidemic conjunctivitis;8 H aegyptius was the mostcommon pathogen isolated from children with

conjunctivitis in Promissao during the epidemic of BPF andwas also isolated from a non-aseptically obtained skinscraping from 1 case child; in contrast to H aegyptius fromconjunctivitis unrelated to BPF, H aegyptiusepidemiologically associated with BPF contained a 25

megadalton plasmid, an epidemiological marker

distinguishing strains that otherwise have no reason todiffer; in contrast to findings in children in Brazil with nohistory of conjunctivitis, H aegyptius can frequently beisolated from normally appearing conjunctiva of childrenwith a recent history of conjunctivitis.6Any explanation of how an antecedent conjunctivitis

predisposes a child to BPF must account for three unusualfeatures of this association: the apparent low attack-rate forchildren with conjunctivitis, the interval of days to weeksbetween onset of conjunctivitis and onset of fever, and thefact that conjunctivitis is often resolving or gone by the timemore severe symptoms occur. One hypothesis is that, undercertain circumstances, specific strains of H aegyptius mayelude host defences and cause invasive disease. If the

acquisition of H aegyptius carriage were manifested byconjunctivitis, the epidemiology of BPF would be similar tothat of invasive H influenzae type b disease, in which

acquisition of carriage may be asymptomatic.The geographical distribution of BPF is uncertain and its

potential for spread unknown. In sporadic form, BPF wouldmost possibly be diagnosed as "culture-negative"meningococcaemia. In Brazil, surveillance for additionalcases has been established and systematic collection ofclinical data is continuing. In other tropical countries,however, this illness may be recognised only with anincreased level of clinical and epidemiological suspicion.

Correspondence to Dr Lee H. Harrison or Dr Claire V. Brocine,Meningitis and Special Pathogens Branch, Center for Infectious Diseases,Centers for Disease Control, Atlanta, Georgia 30333, USA.

REFERENCES

1. Spicer TE, Rau JM Purpura fulminans. Am J Med 1976; 61: 566-71.2 Sullivan JD, Valois FW, Watson SW. Endotoxin. the Limulus amebocyte lysate

system In. Bernheimer AW, ed. Mechanisms m bacterial toxicology. New York:Wiley, 1976, 218.

3. Johnson KM, Elliot LH, Heymann DL. Preparation of polyvalent viral

immunofluorescent intracellular antigens and use in human serosurveys. J ClinMicrobiol 1981; 14: 527-29.

4. Meyers JA, Sanchez D, Elwell LP, Falkow S. Simple agarose gel electrophereticmethod for identification and characterization of plasmid deoxyribonucleic acid.J Bacteriol 1976; 127: 1529-37.

5 Carlone GM, Sottnek FO, Plikaytis BD Comparison of biochemical and outermembrane protein profiles of Haemophilus aegyptius and Haemophilus influenzaebiotype III. J Clin Microbiol 1985; 22: 708-13.

6. Fleming DW, Berkley SF, Broome CV, Brandileone MC, Waldman EA, and theBrazilian Purpuric Fever Study Group. Haemophilus aegyptius conjunctivitisassociated with Brazilian purpuric fever. Program and abstracts of the 86th annualmeeting of the American Society for Microbiology (Washington, DC, 1986) abstrC2.

7. Sippel JE. Meningococcemia. CRC Crit Rev Microbiol 1981; 8: 267-302.8 Buehler JW, Holloway JT, Goodman RA, Sikes RK. Gnat sore eyes: Seasonal acute

conjunctivitis in a Southern State. South Med J 1983; 76: 587-89

HAEMOPHILUS AEGYPTIUS BACTERAEMIAIN BRAZILIAN PURPURIC FEVER

BRAZILIAN PURPURIC FEVER STUDY GROUP*

Summary Brazilian purpuric fever (BPF) is a

fulminant, often fatal childhood illness thatwas first recognised in 1984. An outbreak in Serrana, SãoPaulo State in March to May, 1986, resulted in 11 cases.Haemophilus aegyptius was isolated from normally sterilebody fluids in 10 children (9 from blood and 1 from

cerebrospinal fluid contaminated with blood), consistentwith a direct role for H aegyptius in the pathogenesis of BPF.The ability to define cases by positive blood culturespermitted an evaluation of the spectrum of illness of thisdisease. 5 culture-positive cases were clinically similar tothose previously described; the other 5 had milder illnesswithout petechial or purpuric skin manifestations at the timethe bacterium was isolated. Blood cultures were a sensitivemeans of diagnosing BPF; cultures were positive in 5 of 6patients that met the full clinical case definition. Treatmentof conjunctivitis did not appear to prevent BPF. However,children treated with intravenous antimicrobials early in thesystemic illness had a trend toward better survival,suggesting that early therapy may prevent progression of theillness.

Introduction

Brazilian purpuric fever (BPF), first recognised in 1984 inthe small town of Promissao, Sao Paulo State, Brazil,1presents in children aged 1-10 years with the acute onset offever, commonly associated with vomiting and abdominalpain. Previous investigations have demonstrated an

association with a preceding purulent conjunctivitis.Without therapy, the illness rapidly progresses with thedevelopment of petechiae followed by purpura, vascularcollapse, and death an average of 24-48 h after the onset ofsymptoms. Extensive epidemiological and laboratoryinvestigations ruled out meningococcal disease but did notestablish a definitive aetiology. We report here the isolationof Haemophilus aegyptius from both conjunctivae andnormally sterile sites of patients with BPF, suggesting adirect role for this agent in the pathogenesis of the illness.The ability to define cases by positive blood culture alsopermitted evaluation of therapy and expanded our

knowledge of the clinical spectrum of this illness.

Methods

In March, 1986, a cluster of BPF cases occurred in the small townof Serrana, Sao Paulo State. As a result of previous outbreaks ofBPF, clinical and epidemiological protocols had been developed forevaluating and managing suspected cases. Children seeking medicalcare at the health clinic in Serrana with fever without an obviouscause of infection were identified, especially those with a history of

*The principal institutions, their directors, and other individuals whocontributed to this study are: Instituto Adolfo Lutz, São Paulo, Brazil(general director, Pedro Paulo Chieffi; microbiologists, Maria Cristina deCunto Brandileone and Mituka Kuku); São Paulo Ministry of Health Centrefor Epidemiology (director Alexandre Vranjac; epidemiologist, GrazielaAlmeida da Silva); São Paulo Ministry of Health, Sanitary District ofRibeirao Preto (director, Jôao Germano Neto; epidemiologist, Ligia ReginaKerr Pontes); University of São Paulo, Faculty of Medicine, Ribeirao PretoCampus, Hospital das Clinicas (professor of paediatrics, Gutenberg MeloRocha); and Meningitis and Special Pathogens Branch, Centers for DiseaseControl, Atlanta, Georgia, USA (chief, Claire V. Broome, epidemiologistsSeth Berkley and Lee Harrison microbiologist, Gloria Ajello).This report was prepared by Dr Seth Berkley and Dr Lee Harrison, CDC,

Atlanta.