ARTICLES

West Nile encephalitis epidemic in southeastern Romania

T F Tsai, F Popovici, C Cernescu, G L Campbell, N I Nedelcu, for the Investigative Team*

Summary
Background West Nile fever (WNF) is a mosquito-borne flavivirus infection endemic in Africa and Asia. In 1996, the first major WNF epidemic in Europe occurred in Romania, with a high rate of neurological infections. We investigated the epidemic to characterise transmission patterns in this novel setting and to determine its origin. Methods Hospital-based surveillance identified patients admitted with acute aseptic meningitis and encephalitis in 40 Romanian districts, including Bucharest. Infection was confirmed with IgM capture and indirect IgG ELISAs. In October, 1996, we surveyed outpatients in Bucharest and seven other districts to estimate seroprevalence and to detect infected patients not admitted to hospital. We also measured the rates of infection and seropositivity in mosquitoes and birds, respectively. Results Between July 15 and Oct 12, we identified 393 patients with serologically confirmed or probable WNF infection, of whom 352 had acute central-nervous-system infections. 17 patients older than 50 years died. Fatality/case ratio and disease incidence increased with age. The outbreak was confined to 14 districts in the lower Danube valley and Bucharest (attack rate 124/100 000 people) with a seroprevalence of 41%. The number of mild cases could not be estimated. WN virus was recovered from Culex pipiens mosquitoes, the most likely vector, and antibodies to WN virus were found in 41% of domestic fowl. Interpretation The epidemic in Bucharest reflected increased regional WNF transmission in 1996. Epidemics of Cx pipiens-borne WNF could occur in other European cities with conditions conducive to transmission.

Introduction
Mosquito-borne West Nile fever (WNF) is an endemic febrile illness in Africa, the Middle East, and southwestern Asia. The flavivirus has also been isolated in Australia and sporadically in Europe but fewer cases in human beings have been recognised. Clinical features are acute fever with severe myalgia, arthralgia, and headache, conjunctivitis, prominent lymphadenopathy, and a roseolar rash, complicated, occasionally, by meningitis or encephalitis.15 Outbreaks with hundreds of cases were reported from Israel in 195057. In 1974, an epidemic in a 2500 km2 area of central South Africa produced tens of thousands of infections.6,7 Except for a small outbreak in southern France in 1962, WNF has not been a publichealth threat in Europe.8 Between July and October, 1996, an epidemic of WNF with mainly neurological infections occurred in Bucharest and the lower Danube valley of southeastern Europe. We described after a field investigation from Sept 28 to Oct 11, 1996, the epidemic, its transmission patterns, and its likely source.

Methods
Romania is divided by the Carpathian range into a region of mountains and high plains to the northwest and a large plain to the south, with the Danube river forming most of the southern border with Bulgaria. Bucharest, the capital city, consists of six administrative sectors and is surrounded by the Ilfov District (figure 1). The national and Bucharest municipal populations were about 23 million and 23 million, respectively, in 1996. In mid-August, 1996, unusually high numbers of acute central-nervous-system infections were reported by clinicians from the two infectious-disease hospitals in Bucharest. In early September, the Centre for Medico-Military Scientific Research, Bucharest, found WN virus haemagglutination-inhibition antibodies in the blood of several patients, and the diagnosis was confirmed by the WHO Collaborating Centre for Haemorrhagic Fevers and Arboviruses, Pasteur Institute, Paris, France. On Aug 20, the Ministry of Health started active surveillance in Bucharest. Suspected cases of WNF were defined as patients admitted to hospital with acute aseptic meningitis, meningoencephalitis, or encephalitis with pleiocytosis. Clinical and epidemiological data were collected and serum and cerebrospinal-fluid samples were obtained. Active hospital-based surveillance was expanded nationwide on Sept 10. Serosurveys were done between Oct 2 and Oct 4 in Bucharest and on Oct 28 and Oct 29 in seven other affected districts to estimate the incidence of acute symptomless and mild WN viral infection. Samples were obtained from blood taken for other medical purposes in general clinics. The numbers of samples were proportional to the sizes of clinic catchment areas and the age distributions in each of the Bucharest clinics, or to the catchment areas of the non-Bucharest clinics. In the Bucharest serosurvey, histories were taken of recent illnesses, including fever, myalgia, headache, rash, joint pain, lymphadenopathy, jaundice, respiratory, gastrointestinal, and central-nervoussystem symptoms and signs. Symptoms among recently infected IgM-positive respondents were compared with those of seronegative controls matched for age and sex. To ascertain whether there had been previous WN viral transmission in the region, serum samples collected in Bucharest, Tulcea, and Dolj for influenza serology during 199396, before the WNF epidemic began, were tested for WN viral antibodies.

Lancet 1998; 352: 76771 See Commentary page xxx

Division of Vectorborne Infectious Diseases, Centers for Disease Control and Prevention, Fort Collins, CO 80522, USA (T F Tsai MD, G L Campbell MD); Bucharest Preventive Medicine Centre, Bucharest, Romania (F Popovici MD); Institute of Virology, Bucharest (Prof C Cernescu MD); and Ministry of Health, Bucharest (N I Nedelcu MD) Correspondence to: Dr T F Tsai

THE LANCET Vol 352 September 5, 1998

ARTICLES

Ukraine Incidence/100 000 0 0110 11100 Bucharest 124

Hungary

Moldavia

Galati

Ukraine

Arges Prahova

Buzau Braila Tulcea

Dimbovita Bucuresti
Olt Dolj Teleorman
Ri

Ialomita Calarasj Black Sea Constanta

Giurgi
ve r

Danube

Bulgaria

Figure 1: WN encephalitis incidence by district, 1996 Viral isolation in cell cultures, suckling mice, or both was attempted with 106 serum, cerebrospinal fluid, or necropsy samples from patients with acute infection; in addition, six serum and 22 cerebrospinal fluid samples were assessed by PCR for WN virus genome. Extracted RNA was reverse transcribed and cDNA was amplified with primers directed against a conserved flaviviral envelope sequence; PCR products were identified by hybridisation with a probe specific for WN virus.9 Viral isolates were identified by PCR, and for a cerebrospinal fluid isolate (96-1030) antigenically, by comparison of its reactivity with that of prototype WN virus (B956) in plaquereduction neutralisation tests against reactivity with reference polyclonal antibodies (dengue 14, WN, yellow fever, and Russian spring-summer encephalitis viruses), monoclonal antibodies (central European and Russian encephalitis viruses), and 11 serum samples taken during convalescence after WNF.10 Serum and cerebrospinal fluid samples were tested for IgM and IgG antibodies to WN virus by capture and indirect EIAs, respectively.10 IgM in clinical samples was captured with an antibody to -chain, WN virus and control antigens were added, and bound antigen was detected with a peroxidase-labelled antiflavivirus antibody. IgG was measured by the addition of samples to plates coated with WN virus and control antigens in alternate wells, and bound IgG was detected with a peroxidaselabelled antibody to human IgG. Acute WN virus infection was confirmed if IgM, IgG, or both classes of antibody to WN virus were present in cerebrospinal fluid, or if there was a conversion in paired serum samples from negative to positive in IgM or IgG EIA. Patients with specific IgM in a single serum sample were classified as having a probable recent infection, and patients with (specific) IgG only in a single serum sample as having had a past infection. Because tick-borne encephalitis is the principal flavivirus infection transmitted in Europe, a subgroup of patients was tested for cross-reactive antibodies to central European encephalitis virus. Mosquito and avian surveys were done between Oct 1 and Oct 9 at 13 sites in Bucharest and the Ilfov District. Adult mosquitoes, collected by aspiration from resting sites, were identified and pooled, and ground suspensions were inoculated on to Vero cells for viral isolation by standard methods.10 Because WNF virus is transmitted in a mosquito-avian cycle, blood samples from wild and domestic birds were tested for the presence of neutralising antibody to WN virus by plaquereduction assay.10 The Ministry of Agriculture was asked for notifications of equine encephalitis cases.

Results
Of 835 patients admitted to hospital with suspected central-nervous-system infection and reported under the surveillance system, 767 (92%) met the definition of WNF cases. Appropriate blood, cerebrospinal fluid, and necropsy samples were available from 441 patients, among whom WN encephalitis was serologically confirmed in 352 (80%). In addition, among 68 patients who did not meet the case definition, 41 (60%) had laboratory-confirmed or probable infection. A total of 393 patients had laboratory-diagnosed WNF. Among these 393 patients, the neurological diagnoses were meningitis (40%), meningoencephalitis (44%), and encephalitis (16%). The onset of illness was typically abrupt, with fever (91% of patients), acute headache (77%), neck stiffness (57%), vomiting (53%), chills (45%), and confusion (34%). Disorientation, disturbed consciousness, and generalised weakness were the predominant signs in patients with encephalitis; some patients had decreased motor power with hypotonia, and others had increased tone, hyper-reflexia, and abnormal reflexes. Ataxia and extrapyramidal signs were recorded in 17% of patients, and cranial-nerve palsies or seizures in smaller proportions. The illness progressed to coma in 13% of cases, and among patients with confirmed or probable infection there were 17 deaths (fatality/case ratio 43%), all in patients older than 50 years. Age-specific fatality/case ratios increased substantially with age, from
THE LANCET Vol 352 September 5, 1998

ARTICLES

50

Clinical incidence

100

Incidence per 100 000

40

80

20

Cases Seroprevalence

30

60

40

10

20

0
8 5 2 9 5 2 9 6 2 9 6 3 0 7 4 ly 1 2 2 g 1 1 2 t t 1 2 3 t 1 Ju July July July Au Aug Aug Aug Sep Sep ept ept ept Oc Oct S S S

Seroprevalence (%)

10

Week of onset Figure 3: WN encephalitis cases by week of onset, Romania

0
0 9 10 1 9 20 2 9 30 3 9 40 4 9 50 5 9 60 6 9 70

Age-group (years) Figure 2: Age-specific incidence and viral-specific IgM seroprevalence of WN encephalitis, Bucharest, 1996

zero in patients younger than 50 years, to 34% in those aged 5059 years, 43% in those aged 6069 years, and 147% in those older than 70 years, without significant differences between men and women. Age-specific attack rates also increased with age (figure 2). Among the 41 patients with WN-virus infections who did not meet the case definition, the diagnoses were acute respiratory infection (11), acute febrile illness (three), monoparesis (one), cerebrovascular disease (one), urinary-tract infection (one), various chronic illnesses (four), and no diagnosis (20). The 326 patients who met the case definition but lacked appropriate samples for testing differed significantly from patients with appropriate samples in age, clinical diagnosis, residence, onset dates, and fatality/case ratio (11%), which was consistent with logistical difficulties in obtaining appropriate samples from patients who died soon after admission to hospital. These differences suggest that untested cases were heterogeneous, and analysed only the 393 patients with laboratory-confirmed infection. The onset of illness in the earliest confirmed case was July 15, 1996. The epidemic peaked in the first week of September and the last confirmed case had onset on Oct 12 (figure 3). The first confirmed cases were reported from Bucharest, and in other districts 3 weeks later. The incidence of cases in the 15 districts in the epidemic area was four per 100 000 people in the general population, without significant differences between the sexes. The epidemic was confined to the Danube plain, southeast of the Carpathian range and, excluding Bucharest, the rate of infection was highest in districts bordering the Danube River (figure 1). The attack rate for Bucharest, 124 per 100 000, was higher than that for all other districts, in which the mean rate of infection was 15 per 100 000. 240 (61%) of 393 patients with

laboratory-confirmed infections lived in urban areas. In Bucharest, however, the rate was significantly higher in the rural Ilfov District (28/100 000) than in urban areas of the region (ten/100 000). In other districts more patients who lived in rural areas were infected, but rates could not be calculated. WN virus genomic sequences were not detected in any serum or cerebrospinal-fluid samples. A cerebrospinalfluid viral isolate (96-1030) recovered in cell cultures was identified as WN virus by PCR. The virus was antigenically indistinguishable from prototype WN virus in plaque-reduction neutralisation-test assays with reference polyclonal and monoclonal flavivirus antibodies. In 11 human serum samples taken during convalescence, however, neutralisation antibody titres were consistently higher for the local WN viral isolate, with a geometric mean ratio of 31; therefore, the isolate could be considered a variety of WN virus. No serological cross-reactivities were detected against central European encephalitis virus in 11 serum samples with IgM antibodies to WN virus (geometric mean titre 10 500) or in eight serum samples with titres of IgG antibody to WN virus of 1:400. Serosurveys showed higher proportions of outpatients with IgG antibodies in Bucharest (39 [41%] of 959) than in the other seven districts (eight [09%] of 868). In Bucharest, the proportions of outpatients with IgG and IgM antibodies were similar in both sexes in all agegroups and regions of the city (data not shown). This finding suggests a uniform risk of infection among city residents (figure 2). At the time of the serosurvey, 2 months after the epidemic peak, only 18 of the 39 Bucharest residents with IgG antibodies also had detectable IgM (19% of the sample). By extrapolation of IgM and IgG prevalence rates to the general population, we calculated that 43 00096 000 people were infected during the epidemic in Bucharest and 31 000 in the other surveyed districts. The estimated clinical-to-subclinical infection ratio was one to 140320 (Bucharest data). In the Bucharest survey, 31 recently infected people and 37 seronegative controls reported similar frequencies of minor illness symptoms. Among serum samples collected before the epidemic (151 from Bucharest, 31 from Tulcea, 40 from Dolj), IgG antibodies to WN virus were found in four Dolj samples. 3

THE LANCET Vol 352 September 5, 1998

ARTICLES

Sylvatic cycle ?Cx modestus Africa Middle East Birds

Urban cycle

Cx pipiens

Human beings

Birds

Birds

?Cx modestus

Cx pipiens

Figure 4: Sylvatic and urban transmission cycles in lower Danube valley and model by which epidemic occurred

Culex pipiens was the predominant mosquito species in Bucharest. All investigated sites were heavily infested, including houses, poultry sheds in the rural Ilfov District, and especially apartment complexes, many of which had leaking pipes and standing water in basements. 3689 adult mosquitoes of seven taxa were collected, 94% of which were Cx pipiens. A virus recovered from a pool of Cx pipiens collected in central Bucharest was identified antigenically as WN virus (H Savage, N Karabatsos, unpublished observation), which resulted in a minimum infection rate of 03/1000. Chickens and other domestic fowl were commonly kept at private residences in Bucharest. Neutralising antibodies to WN virus were detected in 30 (41%) of 73 domestic fowl and in one (8%) of 12 wild birds. No increase in equine encephalitis cases was reported.

Discussion
Although WNF is recognised in Africa, Asia, and Europe, encephalitis cases have been reported previously only from Israel, India, and Pakistan.17,11,12 Transmission of WN virus in at least 11 European countries (France, Cyprus, former Czechoslovakia, Greece, Italy, Portugal, Romania, Spain, Turkey, former Yugoslavia, and Russia) has been surmised from animal and human serosurveys and by the occasional isolation of the virus from human, horse, wild bird, mosquito (Aedes cantans, Anopheles maculipennis, and Cx modestus), and tick (Hyalomma plumbeum and H asiaticum) sources.13 With the exception of a small outbreak in the Rhone delta (La Camargue), France, however, cases of WNF have not been recognised in Europe.8 Cases may have escaped detection because of the mild and non-specific symptoms. WN is only rarely complicated by hepatitis, pancreatitis, or encephalitis.1114 The epidemic in Romania in 1996 was the first important outbreak of WNF recorded in Europe. The true number of cases and deaths may have been higher than recorded because national surveillance was delayed, and because appropriate clinical samples were available from few patients. The paucity of acute-phase samples probably contributed to the poor yield of viral recoveries. The predominance of neurological cases reflected hospital-based surveillance and the lack of case-finding efforts in outpatient clinics. Confirmed infections in patients not meeting the case definition probably represented a mixture of coincidental WN virus infection and other manifestations of the illness, such as myelitis and respiratory infection.15,16 We attempted to identify milder infections retrospectively in the Bucharest survey. With a 6-week delay after the epidemic peak and the nonspecific nature of symptoms, however, we found no difference in minor symptoms between recently infected 4

and seronegative respondents. Furthermore, even if mild illnesses had been ten times more frequent than neurological cases (for an extrapolated incidence of 124/100 000 people in the city), even with 959 respondents, we lacked sufficient statistical power to exclude a much larger epidemic of mild illness over and above the outbreak of neurological infections. The Bucharest serosurvey found nearly identically low infection rates in every age-group, whereas rates of neurological disease increased substantially with age (figure 2). The clinical/subclinical infection ratio (1:140320) showed that higher rates of central-nervoussystem infection in the elderly were due to host susceptibility rather than exposure. Closely related neurotropic flavivirus infections, such as St Louis encephalitis and Japanese encephalitis, have similar patterns, but the pathogenesis of age-related susceptibility has not been elucidated.17 The uniform seroprevalence rates in all age-groups also suggest that the virus had not been present previously in this population or, if infections had occurred before, that they were rare. In the seven non-Bucharest districts surveyed, low seroprevalence rates suggested the absence of widespread endemic transmission. The absence of antibodies to WN virus among Bucharest residents sampled before the outbreak was further evidence of a new epidemic in the city. In this susceptible population, the epidemic affected an extensive area of southeastern Romania and, from anecdotally reported cases in Bulgaria, an even larger area of the Danube delta. Better case-finding probably contributed to the reported high attack rate in Bucharest, but viral transmission may also have been more intense in the urban and periurban Cx pipiens-borne cycles. The absence of cases of WNF north of the Carpathian range suggests that these mountains were a barrier against further spread and transmission. Sporadic cases of encephalitis, serologically diagnosed as WN virus infection, have been suspected previously in Romania.1820 In addition, surveys in the lower Danube valley have shown haemagglutination-inhibition antibody rates to be as high as 18% among certain human populations and 53% among migratory birds, although WN virus was never isolated.2123 Field studies have not differentiated between the existence of a local WN virus transmission cycle, although this may be plausible, and repeated viral introductions by migratory birds, especially from Africa. Reference hyperimmune antibodies did not distinguish the locally recovered (human) WN virus isolate from prototype WN virus, although a slight but consistent antigenic difference was seen in serum samples from recovering patients, which shows that the local strain could be a distinct variety of WN virus. Genetic analysis of WN viral strains recovered from France showed their relation to a viral lineage also including certain strains from Africa and Kunjin virus, a flavivirus transmitted mainly in Australia and Asia, which also produces a febrile exanthem and encephalitis.24 Introduction of a virus from Africa to Europe by migratory birds has precedence in the transfer of Sindbis virus from South Africa to northern Europe and the emergence of Sindbis fever epidemics in Scandinavia in 1982.2526 By analogy, sporadic isolations of WN virus from migratory birds and their ticks in Europe show that this virus could also be transferred in spring migrations.13,8

THE LANCET Vol 352 September 5, 1998

ARTICLES

In Africa, WN virus is principally transmitted in an enzootic cycle among Cx univittatus and birds.6 Many northward migratory routes follow a flyway across the Middle East, Turkey, and the Black Sea, along which the Danube delta is a main avian refuge.27 Frequent introductions of WN virus, with establishment of a local enzootic transmission in the Danube basin seem possible. Cx modestus, identified as an enzootic WN virus vector in southern France, is present in the Danube delta and could mediate local sylvatic transmissions, from which the virus could be transferred to an urban cycle or could infect humans directly (figure 4). Such interactions of sylvatic and urban transmission cycles have been reported previously in the USA for St Louis encephalitis virus.17 Few mosquitoes were collected because of the lateness of the season and only one viral isolate was recovered from mosquitoes. Although the minimum infection rate in Cx pipiens was only 03 per 1000, this species was probably the epidemic vector in Bucharest at least. Cx pipiens is the predominant mosquito species in the city; apartment blocks and houses were heavily infested, as well as domestic environments in rural locations where apparently the mosquitoes feed readily on human beings and birds. WN virus has been isolated previously from the species and its capacity to transmit the virus has been shown.13 The mosquito potentially could have transmitted WN virus from person to person within apartment buildings where flooded basements provide favourable breeding conditions for Cx pipiens.3,28 To prevent future outbreaks we recommended actions to control Cx pipiens in Bucharest. The other circumstances that favoured an epidemic of WNF in Romania in 1996 are unknown. WNF outbreaks in North Africa in 1994 and 1996 suggest a regional increase in enzootic viral transmission.29,30 In some areas of Europe, urban conditions are suitable for the introduction and amplification of WN virus, with the risk of future epidemics that follow the pattern of infrequent but recurrent Cx pipiens-borne St Louis encephalitis epidemics in cities of the eastern USA.17,31
Members of the Investigative Team
RomaniaV Laurentia, L Spantulescu, Bucharest Preventive Medicine Center, Bucharest; V Chitu, S Ruta, G Tardei, Institute of Virology, Bucharest; D Cra ciun, D Nicolaiciuc, D Pit igoi, Ministry of Health, Bucharest; C Ceianu, G Nicolescu, A Ungureanu, L A Vladimirescu, Cantacuzino Institute, Bucharest. FranceV Deubel, B LeGuenno, Centre National de Reference pour les fievres Hemorragiques et les Arbovirus, Paris, France. USAL Han, H Savage, L Tengelsen, Centers for Disease Control and Prevention, Fort Collins; EA Henchal, F Knauert, JA Mangiafico, C Rossi, United States Army Medical Reasearch Institute of Infectious Diseases, Fort Detrick.

Contributors
T F Tsai was responsible for guiding the epidemic investigation, data analysis, and preparation of the paper. Florin Popovici managed the databases and data analysis. Costin Cernescu supervised the laboratory investigation. Grant Campbell supervised the field team and assisted in data analysis. Sergio Nedelcu coordinated surveillance and administered overall operations.

Acknowledgments
We thank Terry Gay, Office of International and Refugee Health, Department of Health and Human Services, USA Public Health Service for administrative assistance, and Peter LaPera, Randall Thompson, and Mary Ann Micka of US Agency for International Development for support.

References
1 Peiris JSM, Amerasinghe FP. WNF. In: Beran GW, ed. Handbook of zoonoses, 2nd edn, section B: Viral. Boca Raton, FL: CRC Press, 1994: 13948. 2 Mitchell CJ. Geographic spread of Aedes albopictus and potential for

involvement in arbovirus cycles in the Mediterranean Basin. J Vector Ecol 1995; 20: 4458. 3 Taylor RM, Work TH, Hurlbutt HS, Rizk F. A study of the ecology of West Nile virus in Egypt. Am J Trop Med Hyg 1956; 5: 579620. 4 Marberg K, Goldblum N, Stork V, Jasinska-Klingberg W, Klingberg M. The natural history of WNF, I: clinical observations during an epidemic in Israel. Am J Hyg 1956; 64: 25969. 5 George S, Prasad SR, Rao JA, Yergolkar PN, Sreenivasaiah Setty CV. Isolation of Japanese encephalitis and West Nile viruses from fatal cases of encephalitis in Kolar district of Karnataka. Indian J Med Res 1987; 86: 13134. 6 McIntosh BM, Jupp PG, Dos Santos I, Meenehan GM. Epidemics of West Nile and Sindbis viruses in South Africa with Culex (Culex) univittatus Theobald as a vector. S Afr J Sci 1976; 72: 295300. 7 Klingberg MA, Jasinska-Klingberg W, Goldblum N. Certain aspects of the epidemiology and distribution of immunity of West Nile virus in Israel. Proceedings of the 6th International Congress on Tropical Medicine and Malaria,. 1959; 5:13240. 8 Panthier R, Hannoun C, Beytout D, Mouchet J. Epidemiologie du virus West Nile: (etude dun foyer en Camargue, 3Les maladies humaines. Ann Instit Pasteur 1968; 115: 43545. 9 Puri B, Henchal EA, Burans J, et al. A rapid method for detection and identification of flaviviruses by polymerase chain reaction and nucleic acid hybridization. Arch Virol 1994; 134: 2937. 10 Tsai TF. Arboviruses. In: Murray PR, Baron EJ, Pfaller MA, Tenover FC, Yolken RH, eds. Manual of clinical microbiology, 6th edn. Washington, DC: ASM Press, 1995: 98096. 11 Varsano M, Basat MB, Rannon L. West Nile virus infection in recent years in Israel. Israel J Med Sci 1982; 19: 11. 12 Flatau E, Kohn D, Daher O, Varsano N. WNF encephalitis. Israel J Med Sci 1981; 17: 105759. 13 Perelman A, Stern J. Acute pancreatitis in WNF. Am J Trop Med Hyg 1974; 23: 115052. 14 George AJ, Lesbordes JL, Georges-Courbot MC, Meunir DMY, Gonzalez JP. Fatal hepatitis from West Nile virus. Ann Inst Pasteur 1987; 138: 23744. 15 Darwish MA, Feinsod FM, Scott RM, et al. Arboviral causes of nonspecific fever and myalgia in a fever hospital patient population in Cairo, Egypt. Trans Royal Soc Trop Med Hyg 1987; 81: 100103. 16 Gadoth N, Weitzman S, Lehmann EE. Acute anterior myelitis complicating WNF. Arch Neurol 1979; 36: 17273. 17 Monath TP, Tsai TF. Flaviviruses. In: Richman DD, Whitley RJ, Hayden FG, eds. Clinical virology. New York: Churchill-Livingstone, 1997: 113386. 18 Miskolczy D, Csiki C, Vendeg V, et al. Epidemic of late summerautumn virus encephalitis in 1955 in Tirgu Mures: so-called mosquito encephalitis. Neurologia Psihiatria Neurochirurgia. 1958; 2: 13947. 19 Duca E, Duca M, Strat E, et al. Investigations on cases of encephalitis transmitted by arthropods in Moldavia. Rev Med Chir Soc Med Nat Iasi. 1963; 67: 399404. 20 Topciu V, Rosiu N, Son G, et al. Consideritions sur quelques cas dencephalite virotique primaire apparus dans le sud-ouest de la Roumanie (rgion du Banat). Arch Roum Path Exp Microbiol 1968; 27: 53138. 21 Dra ga nescu N, Girjabu E, Iftimovici R, Klebleer tudes rolgiques sur la prsence de quelques arbovirus dans la rgion de sylvo-steppe de Tulcea. Rev Roum Virol 1991; 42: 4751. 22 Dra ga nescu N, Girjabu E, Iftimovici R. Investigations s rologiques chez lhomme et chez differentes espces danimaux domestiques de la rgion de sylvo-steppe de Roumanie, pour le dpistage de quelques Togaviridae et Bunyaviridae. Rev Roum Virol 1993; 44: 20710. 23 Dra ga nescu, N. Recherches sur les infections arbovirus. Rev Roum Md Virol 1989, 40: 25769. 24 Berthet FX, Zeller HG, Drouet MT, et al. Extensive nucleotide changes and deletions within the envelope glycoprotein gene of EuroAfrican West Nile viruses. J Gen Virol 1997, 78: 229397. 25 Espmark A, Niklasson B. Ockelbo disease in Sweden: epidemiological, clinical and virological data from the 1982 outbreak. Am J Trop Med Hyg 1984; 33: 120311. 26 Yukio S, Niklasson B, Dalrymple JM, Strauss EG, Strauss JH. Structure of the Ockelbo virus genome and its relationship to other Sindbis viruses. Virology 1991; 182: 75364. 27 Curry-Lindahl K. Bird migration in Africa: movements between six continents, Vol 1. London: Academic Press, 1981. 28 Anonymous. Update: St. Louis encephalitis outbreak, 1994. Louisiana Morbidity Report 1995; 5: 3. 29 Le Guenno B, Bougermouh A, Azzam T, et al. West Nile: a deadly virus? Lancet 1996; 348: 1315. 30 El Harrack M, Le Guenno B, Gounon P Isolement du virus West Nile au Maroc. Virologie 1997; 1: 24849. 31 Lundstrm JO. Vector competence of Western European mosquitoes for arboviruses: a review of field and experimental studies. Bull Soc Vector Ecol 1994; 19: 2336.

THE LANCET Vol 352 September 5, 1998