The Journal of Infectious Diseases    1999;179:230-233
© 1999 by the Infectious Diseases Society of America. All rights reserved.


Risk Factors for West Nile Virus Infection and Meningoencephalitis, Romania, 1996

Linda L. Han, Florin Popovici, James P. Alexander, Jr., Velea Laurentia, Leslie A. Tengelsen, Costin Cernescu, Howard E. Gary, Jr., Nicolae Ion-Nedelcu, Grant L. Campbell, and Theodore F. Tsai

Centers for Disease Control and Prevention, Atlanta, Georgia, and Fort Collins, Colorado; Public Health Inspectorate of Bucharest, and Institute of Virology, and Ministry of Health, Bucharest, Romania

Submitted 27 May 1998; revised 31 August 1998.

In 1996, an epidemic of 393 cases of laboratory-confirmed West Nile meningoencephalitis occurred in southeast Romania, with widespread subclinical human infection. Two case-control studies were performed to identify risk factors for acquiring infection and for developing clinical meningoencephalitis after infection. Mosquitoes in the home were associated with infection (reported by 37 [97%] of 38 asymptomatically seropositive persons compared with 36 [72%] of 50 seronegative controls, P and, among apartment dwellers, flooded basements were a risk factor (reported by 15 [63%] of 24 seropositive persons vs. 11 [30%] of 37 seronegative controls, P = .01). Meningoencephalitis was not associated with hypertension or other underlying medical conditions but was associated with spending more time outdoors (meningoencephalitis patients and asymptomatically seropositive persons spent 8.0 and 3.5 h [medians] outdoors daily, respectively, P .01). Disease prevention efforts should focus on eliminating peridomestic mosquito breeding sites and reducing peridomestic mosquito exposure.


     Presented in part: 35th annual meeting of the Infectious Diseases Society of America, September 1997, San Francisco (abstract 163).
     Verbal consent was obtained from study participants. The investigation was conducted in accordance with human experimentation guidelines of the Romanian Ministry of Health and with guidelines of the Centers for Disease Control and Prevention for studies conducted in rapid response to public health emergencies.
     Financial support: United States Agency for International Development.

     Reprints or correspondence: Dr. Grant L. Campbell, Division of Vector-Borne Infectious Diseases, Mailstop P-02, Centers for Disease Control and Prevention, Fort Collins, CO 80522-2087 ().

     West Nile virus (WNV), a mosquitoborne flavivirus closely related to St. Louis encephalitis (StLE) and Japanese encephalitis (JE) viruses, is endemic to Africa, the Middle East, and southwest Asia [1, 2]. The virus is transmitted between Culex mosquitoes and wild birds, with incidental infections occurring in humans [1, 2]. Risk factors for human WNV infection and disease have not been examined previously. Between 15 July and 12 October 1996, 393 cases of laboratory-confirmed WNV meningoencephalitis (ME) occurred in southeastern Romania, of which 286 (73%) occurred in the capital city of Bucharest [3]. The highest attack rates and fatality rates occurred among the elderly, a pattern similar to that seen with StLE, JE, and tickborne encephalitis [1–4]. By extrapolation from a serosurvey in which 39 (4%) of 959 city residents had WNV IgG antibody, >90,000 Bucharest residents were infected during the epidemic [3]. This was the largest epidemic of WNV-associated neurologic disease ever reported, the first to occur predominantly in an urban setting, and only the second to occur in Europe [3]. The unprecedented magnitude and novel setting of this epidemic provided an opportunity to conduct case-control studies to examine risk factors for acquiring WNV infection and, among persons infected with WNV, risks for developing ME.

Materials and Methods

     Case definitions.     A serosurvey was conducted in Bucharest between 2 and 4 October 1996 [3]. Serum samples were collected from 959 persons whose ages and locations of residence were representative of the city's population and who required phlebotomy for reasons unrelated to the serosurvey. An uninfected person was defined as a serosurvey participant who was WNV IgG–negative. Since WNV appeared to have been newly introduced to Bucharest with this epidemic [3], an asymptomatically infected person was defined simply as a serosurvey participant who had WNV-specific IgG antibodies. A person with WNV ME was defined as any Bucharest resident hospitalized with acute meningitis, ME, or encephalitis with pleocytosis during the epidemic period and who had either serum WNV IgM, WNV IgG seroconversion, or WNV antibodies in the cerebrospinal fluid.

     Case-control study 1: risk factors for acquiring WNV infection.     To identify risk factors for acquiring WNV infection among Bucharest residents, we compared asymptomatically infected with uninfected persons identified in the serosurvey. ME patients were excluded from the infected group so that risk factors for asymptomatic and symptomatic infection could be examined separately. For the first 25 asymptomatically infected persons identified, 50 uninfected persons, selected with a random-number generator, were matched within broad age groups (0–25 years, 26–65 years, >65 years). Two of the 25 asymptomatically infected persons were subsequently excluded; 1 was found to live outside Bucharest, and 1 could not be located for interview. An additional 15 IgG-positive persons were subsequently identified and enrolled, completing the final case group of 38 asymptomatically infected persons. These persons remained well-matched within broad age groups with the 50 uninfected persons. A standardized questionnaire was used by local medical epidemiologists to interview subjects in person about environmental conditions and behaviors during the epidemic period, including housing type and age, flooding of basements, presence of rainwater collection containers, domestic fowl, and mosquitoes in or near the home, and amount of time spent outdoors, in general, or in gardens, in particular. Study participants responded to the survey without advance knowledge of their infection status.

     Case-control study 2: risk factors for developing ME among WNV-infected persons.     To identify risk factors for developing ME among infected persons, we compared WNV ME patients with asymptomatically infected persons. The 38 asymptomatically infected persons described above comprised the control group. The case group consisted of 50 WNV ME patients, selected with a random-number generator from nonfatal WNV ME cases and matched within broad age groups to the first 25 asymptomatically infected persons identified. These 50 WNV ME patients remained well-matched within broad age groups with the final control group of 38 asymptomatically infected persons. Local medical epidemiologists used a standardized questionnaire to interview participants in person about medical conditions (including hypertension, cardiovascular disease, cerebrovascular disease, and neurologic conditions), smoking habits, environmental conditions, and behaviors during the epidemic period.

     Serology.     Serum samples were tested by isotype capture ELISA for IgM antibodies and by indirect ELISA for WNV IgG antibodies [5]. To detect IgM, diluted serum was added to plates coated with anti–human &mgr; chain antibody, followed by WNV antigen and labeled antiviral antibody. To detect IgG, WNV-infected cell culture lysates were coated onto microtiter plates, and then diluted human serum, peroxidase-labeled anti–human IgG, and substrate were added. Absorbance signals from specimen wells and negative and positive control wells were read. Specimens with absorbances of >5 SD above the average absorbance of the negative controls were considered positive.

     Data analysis.     Analyses were performed using Fisher's exact or &khgr;2 tests for categoric data and the Wilcoxon rank sum test for ordinal data. Stratified analyses were done using the Mantel-Haenszel &khgr;2 test.


     Risk factors for acquiring WNV infection.     Asymptomatically infected persons and uninfected persons were similar with respect to age (medians, 37 and 46 years, respectively), sex (34% males in both groups), level of education, type of occupation, length of residence in Bucharest, and location of residence within the city (all P > .20). Compared with uninfected persons, infected persons were more likely to report mosquitoes in the home (P more mosquito bites per day during the epidemic (medians, 4 vs. 3 bites per day [interquartile ranges: 3–9 vs. 1–4 bites per day], among infected and uninfected persons, respectively, Wilcoxon two-sample test, P among apartment dwellers, were more likely to have had a flooded basement (P = .01; table 1). The association of infection with mosquitoes in the home remained after controlling for age, residence in the rural agricultural sector of the city, flooding of the basement, number of daily mosquito bites, and amount of time spent outdoors and was consistently found in nearly all subgroups examined. The associations of infection with mosquito bites and flooded basements remained after controlling for age, residence in the agricultural sector, and other variables but were not consistently present in all subgroups examined. No association was found between infection and housing type, age of home, amount of time spent outdoors, exposure to domestic fowl, or the presence of rainwater collection containers around the home.

Table 1.     Potential risk factors for acquiring West Nile virus infection, Bucharest, Romania, 1996.

     Risk factors for developing ME.     Persons with WNV ME and seropositive persons without ME were similar with respect to age (medians, 47 and 37 years, respectively), sex (36% vs. 34% male, respectively), level of education, type of occupation, length of residence in Bucharest, and location of residence within the city (all P > .20). Hypertension, other underlying medical conditions, and smoking history were no more prevalent among ME patients than among asymptomatically infected persons (table 2). ME patients, compared with asymptomatically infected persons, spent more time outdoors daily during the epidemic period (medians, 8 vs. 3.5 h [interquartile ranges: 3–12 vs. 2–6 h], respectively, Wilcoxon two-sample test, P and were more likely to have spent >6 h/day outdoors during the epidemic period (P .01; table 2). This association remained after controlling for age, residence in the agricultural sector of the city, number of daily mosquito bites, and daily hours spent in a garden or on a balcony and was found consistently in nearly all subgroups examined. Both groups spent similar periods of time in a garden or on a balcony (medians, 2 vs. 2.5 h [interquartile ranges: 0–12 vs. 0–10 h] among ME patients and controls, respectively, Wilcoxon two-sample test, P = 0.44). However, ME patients were more likely than asymptomatically infected persons to have spent >6 h/day in a garden or on a balcony (P = .02; table 2). This association remained after controlling for age, number of hours spent outdoors, and other variables but was not consistently found in all subgroups examined.

Table 2.     Potential risk factors for developing meningoencephalitis (ME) among West Nile virus–infected persons, Bucharest, Romania, 1996.


     In the first part of this study, we found WNV infection to be associated with specific features of the residential environment, namely, presence of mosquitoes indoors and flooded apartment building basements. The association of infection with mosquitoes in the home suggests a peridomestic route of transmission consistent with the ecology of Culex pipiens-pipiens. This species typically is found in urban settings, is the predominant mosquito species in Bucharest, and was the likely epidemic vector [3]. The abundance of mosquitoes indoors undoubtedly reflected the scarcity of home air conditioners and window screens in Bucharest, factors that have been associated with risk of acquiring StLE virus infection [6]. The association of infection with flooded apartment building basements likewise is consistent with the breeding habits of Cx. pipiens. During the epidemic period, many apartment building basements were flooded with sewage leaking from poorly maintained plumbing, creating a high-organic-content habitat favorable for Cx. pipiens larvae and almost certainly contributing to the dense populations of resting mosquitoes, which sometimes blackened the walls and ceilings of apartment building hallways and entryways [2, 3]. Similar circumstances led to a Cx. pipiens-quinquefasciatus–borne StLE outbreak in New Orleans in 1994 (Michael Carroll, New Orleans Mosquito Control, personal communication). Another common feature of the peridomestic environment, the presence of domestic fowl, might have been expected to have been associated with risk of acquiring infection, since 42% of fowl were estimated to have become infected during this epidemic [3]. The absence of such an association is understandable because most domestic fowl do not achieve serum viral levels adequate for WNV amplification [7].

     In the second part of this study, we found no association between WNV ME and hypertension, other underlying medical conditions, or cigarette smoking. These conditions are putative risk factors for severe illness following infection with other neurotropic flaviviruses, theoretically promoting neurologic disease by disruption of the blood-brain barrier [2, 8, 9]. Our inability to detect any association between ME and these medical conditions may have been a consequence of small sample size. However, the low prevalence of these conditions among ME patients suggests that these chronic medical illnesses were not important contributors to the development of ME. Rather, outdoor activity was the only significant risk factor for the development of ME. The reason for this association is unclear. Possibly, the greater amount of time spent outdoors resulted in more bites from infected mosquitoes, leading, in turn, to a viral inoculum sufficient to cause neuroinvasive infection. Although only a small fraction of mosquitoes was infected during the epidemic [3], even bites from uninfected mosquitoes could have potentiated infection via local host responses to components of mosquito saliva [10]. However, the absence of an association between ME and the number of daily mosquito bites is inconsistent with a dose-response effect (although this may just reflect small sample size or inaccurate recall). Another possibility is that outdoor exposure may have been increased during hot intervals of the summer, when the extrinsic incubation period, the time required for a mosquito to become infectious after taking an infectious meal, is shortenedpotentially affecting the infectivity of mosquito bites acquired during those intervals. Outdoor exposure also may have been increased during a period when a vector other than C. pipiens was active. Alternatively, the association between greater amount of time spent outdoors and ME could be due to an associated behavior, such as physical activity, which may have contributed directly to increased disease severity, as in poliomyelitis [11]. This explanation seems unlikely because advanced age, the most clearly defined risk for flaviviral encephalitis, is generally associated with a more sedentary lifestyle.

     We considered two potential sources of bias but doubt that either would have created an artifactual association between ME and time spent outdoors. First, WNV ME patients were aware that they had a mosquito-transmitted illness and may have been more likely than controls to report exposures or behaviors perceived to be associated with mosquitoes, such as spending time outdoors. However, such recall bias is unlikely, since other mosquito-associated factors, namely, the number of daily mosquito bites or presence of mosquitoes in the home, were not associated with ME.

     Second, our findings may have reflected a misclassification bias, since we defined infection (presumably recent) on the basis of IgG positivity. By using this definition, we may have included both false-positive (IgG-seropositive persons who were infected prior to this epidemic) and false-negative subjects (newly infected persons who had blood drawn before they developed an IgG response) in our sample. Although IgM positivity would have been a more specific indicator of recent infection, it also would have been relatively insensitive because sera were collected 12 weeks after the epidemic began and IgM in infected persons may have declined. However, it is unlikely that the positive associations can be explained by misclassification bias, since results of both case-control analyses were unchanged when we included only IgM-seropositive persons in the asymptomatically infected group. Furthermore, misclassification of newly infected IgG-negative persons as uninfected only would have made the comparison groups more similar and would have led us to underestimate or miss associations between putative risk factors and WNV infection, but it would not have created nonexistent associations. Inclusion of subjects with false-negative results in our sample, therefore, would not account for the observed associations.

     In summary, an epidemic of WNV ME that was unprecedented in magnitude, location, and setting presented an opportunity to investigate potential risk factors for infection and neurologic disease. We found that increased exposure to the mosquito vector, particularly in the home and its immediate environs, was associated with risk of acquiring WNV infection. In addition, we found that risk of developing ME was not associated with any underlying medical condition but was correlated with outdoor activity. Therefore, efforts to prevent and control future urban WNV epidemics should focus on reduction of mosquito breeding sites and limiting human exposure to mosquitoes in domestic and peridomestic settings.


     We gratefully acknowledge the staff epidemiologists of the Public Health Inspectorate of Bucharest for conducting the interviews and E. Henchal and C. Rossi, US Army Medical Research Institute of Infectious Diseases, Frederick, Maryland, for providing the WNV serologic test kits.



1.  Hayes CG. West Nile fever. In: Monath TF, ed. The arboviruses: epidemiology and ecology. Vol. 5. Boca Raton, FL: CRC Press, 1989:59–88. First citation in article
2.  Monath TP, Tsai TF. Flaviviruses. In: Richman DD, Whitley RJ, Hayden FG, eds. Clinical virology. New York: Churchill-Livingstone, 1997:1133–86. First citation in article
3.  Tsai TF, Popovici F, Cernescu C, Campbell GC, Nedelcu NI, the Investigative Team. An epidemic of West Nile encephalitis in southeastern Romania. Lancet 1998; 352:767–71. First citation in article | Medline
4.  Aebi C, Tsai TF. Tick-borne encephalitis. In: Feigin RD, Cherry JD, eds. Textbook of pediatric infectious diseases. Vol. 2. 4th ed. Philadelphia: WB Saunders, 1998:2003–9. First citation in article
5.  Sharp TW, Wallace MR, Hayes CG, et al. Dengue fever in U.S. troops during Operation Restore Hope, Somalia, 1992–1993. Am J Trop Med Hyg 1995; 53:89–94. First citation in article | Medline
6.  Henderson BE, Pigford, CA, Work T, Wende RD. Serologic survey for St. Louis encephalitis and other group B arbovirus antibodies in residents of Houston, Texas. Am J Epidemiol 1970; 91:87–98. First citation in article | Medline
7.  Taylor RM, Work TH, Hurlbut S, Rizk F. A study of the ecology of West Nile virus in Egypt. Am J Trop Med Hyg 1956; 5:579–620. First citation in article
8.  Marfin AA, Bleed DM, Lofgren JP, et al. Epidemiologic aspects of a St. Louis encephalitis epidemic in Jefferson County, Arkansas, 1991. Am J Trop Med Hyg 1993; 49:30–7. First citation in article | Medline
9.  Tsai TF, Canfield MA, Reed CM, et al. Epidemiological aspects of a St. Louis encephalitis outbreak in Harris County, Texas, 1986. J Infect Dis 1988; 157:351–6. First citation in article | Medline
10.  Osorio JE, Godsey MS, Defoliart GR, Yuill TM. LaCrosse viremias in white-tailed deer and chipmunks exposed by injection or mosquito bite. Am J Trop Med Hyg 1996; 54:338–42. First citation in article | Medline
11.  Morens DM, Pallansch MA, Moore M. Polioviruses and other enteroviruses. In: Belshe RB, ed. Textbook of human virology. St. Louis: Mosby Year Book, 1991:427–97. First citation in article