Distribution of Eosinophilic Meningitis Cases Attributable to

During November 2004-January 2005, 5 cases of eosinophilic meningitis (EM) attributable to Angiostrongylus cantonensis infection were reported in Hawaii. To determine if this temporal clustering reflected an increased incidence, we ascertained EM and A. cantonensis cases by systematic review of statewide laboratory and medical records for January 2001-February 2005 and generalized the data to population estimates. We identified 83 EM cases; 24 (29%) were attributed to A. cantonensis infection, which was included in the discharge diagnoses for only 2 cases. Comparison of A. cantonensis infection incidence rates (per 100,000 person-years) for the baseline (January 2001-October 2004) and cluster (November 2004-February 2005) periods showed statistically significant increases for the state as a whole (0.3 vs. 2.1), the Big Island of Hawaii (1.1 vs. 7.4), and Maui County (0.4 vs. 4.3). These findings underscore the need to consider the diagnosis of A. cantonensis infection, especially in the state of Hawaii.

E osinophilic meningitis (EM) is a rare clinical entity characterized by meningeal infl ammation and eosinophilic pleocytosis in the cerebrospinal fl uid (CSF) (1-7). Among the infectious causes of EM, Angiostrongylus cantonensis is the most common worldwide. A. cantonensis, the rat lungworm, was fi rst described in rats in 1935, in Canton, China. The parasite was fi rst postulated to cause human infection in a fatal case in 1944 in Taiwan and was confi rmed to be pathogenic for humans through investigations in the early 1960s in Hawaii (8)(9)(10)(11)(12).
During November 2004-January 2005, 1 parasitologically confi rmed and 4 presumptive cases of A. cantonensis infection were reported to the Hawaii State Department of Health. The 5 cases included 3 from the Big Island of Ha-waii and 2 from Oahu; 1 Oahu case was in a visitor to Hawaii whose lumbar puncture (LP) was performed elsewhere. Recognition of these 5 index cases prompted multifaceted investigations of epidemiologic, clinical, and environmental aspects of EM/A. cantonensis infection in Hawaii.
To assess whether the unusual temporal clustering of case reports refl ected an increased incidence of EM/A. cantonensis infection, we ascertained cases through comprehensive review of statewide laboratory and medical records. Although investigations of EM/A. cantonensis infection in various Hawaiian Islands have been described since the 1960s (4)(5)(6)(9)(10)(11)(12)(13)20,30,35), to our knowledge, this is the fi rst study to systematically ascertain cases and determine regional incidence rates in this manner.   *If a patient had >1 LP, the LP considered in the analyses was the one that met criteria for EM and had the highest absolute eosinophil count. †The exposure period was defined as the 45-d period before the symptom-onset date (if unknown, the date of the LP). ‡Potential cases of EM were excluded if the eosinophilic pleocytosis was potentially attributable to blood and thus was difficult to evaluate (e.g., traumatic LP, grossly bloody CSF, or intracranial hemorrhage). For CSF specimens with >500 erythrocytes/mm 3 , the leukocyte and eosinophil criteria had to be met after using a correction ratio of a decrease of 1 leukocyte for every 500 erythrocytes. §The symptoms and signs included headache, neck stiffness or nuchal rigidity, visual disturbance, photophobia or hyperacusis, cranial nerve abnormality (e.g., palsy), abnormal skin sensation (e.g., paresthesia, hyperesthesia), sensory deficit, nausea or vomiting, documented fever, increased irritability (if <4 y of age), and bulging fontanelle (if <18 mo of age).

Control and Prevention (CDC); this case was 1 of the 5 index cases that prompted the investigation.
Our case defi nitions for EM and A. cantonensis infection are provided in Table 1. If the inclusionary criteria for EM were met, we reviewed the patient's medical record to obtain additional information regarding the EM and to categorize cases of EM by known or likely cause (e.g., A. cantonensis infection). The information collected during chart review included basic demographic data, pertinent dates (e.g., birth, hospitalization, travel, symptom onset, and LP), medical history, medications, clinical manifestations, additional laboratory and radiologic results, and discharge diagnoses. Because the primary focus of the study was A. cantonensis infection, if, at the time of the LP, the patient had intracranial hardware (i.e., a well-established cause of EM) or was <2 months of age (i.e., angiostrongyliasis was epidemiologically unlikely), we collected only demographic data and discharge diagnoses.
We attributed cases of EM to A. cantonensis infection only if this diagnosis was epidemiologically and clinically plausible and no other possible cause of EM was identifi ed. Examples of possible alternative causes included CNS infection with other microbes, reaction to foreign material in the CNS (e.g., intracranial hardware or myelography dye), medications (e.g., intrathecal vancomycin or gentamicin), neoplasms, multiple sclerosis, and neurosarcoidosis (1-7). The study neurologist (J.J.S.) facilitated fi nal selection and classifi cation of cases of EM and A. cantonensis infection by reviewing the available case data and ensuring that the inclusionary and exclusionary criteria were applied consistently and objectively.

Statistical Analysis and Human Subjects Protection
Data entry was performed with Epi Info version 2002 (CDC, Atlanta, GA, USA), and data analyses were conducted with SAS version 9.1 (SAS Institute, Cary, NC, USA). Two-tailed p values were calculated by using the Fisher exact test for binary variables and the Wilcoxon test for continuous variables. Linear and quadratic regression models were evaluated to assess whether eosinophilic pleocytosis varied with time (i.e., the interval from symptom onset to LP). We calculated incidence rates by generalizing hospital-based frequency data to the population at large for various periods and counties in Hawaii using the US Census Bureau's annual population estimates for 2001-2004 (the estimate for 2004 also was used for January and February 2005) (39). We used Poisson regression analyses to compare county-specifi c annual rates. We defi ned the 46-month period of January 2001-October 2004 as the baseline period and the 4-month period of November 2004-February 2005 as the cluster period. CDC's policies with regard to human study participants were followed in this investigation.
For the 22 case-patients with known symptom onset dates, the median interval from onset to LP was 3 days (range 0-48 days); the 2 longest intervals were 14 days (2 patients) and 48 days (1 patient). When a linear regression model was applied to data for the intervals <14 days, the longer the interval (between symptom onset and LP), the higher the CSF eosinophil percentage and absolute eosinophil count (p = 0.001 and 0.005, respectively). Compared with patients with other causes of EM, A. cantonensis casepatients had signifi cantly higher CSF leukocyte counts (median 573/mm 3 vs. 304/mm 3 , p = 0.03) and absolute eosinophil counts (median 120/mm 3 vs. 14/mm 3 , p<0.001); they also tended to have higher eosinophil percentages (median 15.0% vs. 12.0%), but the difference was not statistically signifi cant (p = 0.08). The temporal distribution of the 24 cases included 15 (63%) during the baseline period (3-5 cases per year) and 9 (38%) during the cluster period ( Figure 2). The mean number of A. cantonensis cases per month increased from 0.3 in the baseline period to 2.3 in the cluster period, whereas the mean monthly rates for cases of EM with other causes were essentially unchanged (1.2 and 1.0, respectively). Thus, the proportion of EM cases attributed to A. cantonensis increased from 21% (15/70) for the baseline period to 69% (9/13) for the cluster period. The A. cantonensis incidence rates for the state as a whole increased from 0.3 per 100,000 person-years in the baseline period to 2.1 in the cluster period (p<0.001) (Figure 2).
The geographic distribution of the 24 cases included 3 counties and 4 islands: Honolulu County (Oahu Island; n = 11 cases, including the case in the visitor), Hawaii County (Big Island of Hawaii; n = 9, including the parasitologically confi rmed case), and Maui County (n = 4, including 3 cases associated with Maui Island and 1 with Lanai). Although the absolute number of cases was highest for Honolulu, the county-specifi c incidence rates (per 100,000 person-years) for the study period as a whole were higher for Hawaii (1.4) and Maui (0.7) than Honolulu (0.3) (Figure 3). The case-patients were signifi cantly more likely to have been in Hawaii County than Honolulu County (risk ratio 4.6, 95% confidence interval 1.9-11.1); the comparison between Hawaii and Maui Counties was not signifi cant (data not shown). The increases in annualized incidence rates (cases/100,000 person-years) from the baseline to the cluster periods were statistically signifi cant for Hawaii County (1.1 vs. 7.4; p<0.001) and Maui County (0.4 vs. 4.3; p = 0.03) but not for Honolulu County (0.2 vs. 1.0; p = 0.07) (Figure 3).

Discussion
This study was prompted by an unusual temporal clustering of 5 reported cases of EM/A. cantonensis infection from 2 Hawaiian Islands during November 2004-January 2005. Our primary goal was to assess whether these voluntarily reported cases refl ected an increased incidence. To accomplish this, we used a laboratory-and hospital-based approach to ascertain symptomatic cases of EM and A. cantonensis infection. To our knowledge, this is the fi rst study to systematically determine incidence rates of EM and A. cantonensis infection for the entire state of Hawaii or any angiostrongyliasis-endemic area. We determined that the incidence of angiostrongyliasis was higher during the cluster period (November 2004-February 2005) than the baseline period (January 2001-October 2004). The overall fi ndings of our study support conclusions specifi c for Hawaii but also highlight general principles about EM and A. cantonensis infection. In addition, our study may serve as a useful model in other settings. Surveillance of regional laboratory data, coupled with investigation of medical records of case-patients, may help identify temporal and geographic trends for angiostrongyliasis or other diseases.
Our data underscore that EM is an uncommon entity: <1% of the patients whose CSF data were reviewed fulfi lled the laboratory criteria for EM. This diagnosis is commonly missed or dismissed, but the presence of eosinophilic pleocytosis is abnormal and should prompt consideration of both infectious and noninfectious causes. In our study, intracranial hardware was the most frequently identifi ed cause of EM (42% of 83 cases). Although the presence of hardware or other foreign material in the CNS is a well-established cause of EM, the possibility of associated bacterial infection should be considered (2,6). In our study, EM also was associated with confi rmed cases of bacterial and viral meningoencephalitis, as well as idiopathic cases (no microbial etiology identifi ed) in infants evaluated because of fever or failure to thrive.
We found that a substantial proportion of the EM cases in Hawaii were attributable to A. cantonensis infection (29%) and that the proportion was 3-fold higher during the  cluster than during the baseline period. This rate increase was particularly notable in Hawaii and Maui Counties. Despite the fact that 23 of the 24 cases were clinically defi ned, the likelihood of misclassifi cation was low. By defi nition, none of the case-patients had another possible cause of EM identifi ed. In most angiostrongyliasis-endemic areas, parasitologic confi rmation is unusual, and a presumptive diagnosis is typical. Furthermore, Hawaii is hyperenzootic for infection with A. cantonensis but not Gnathostoma spinigerum or Baylisascaris procyonis, 2 other parasites commonly associated with EM. Our confi dence that the A. cantonensis cases were correctly classifi ed as such is further increased by the fi ndings of other components of our multifaceted investigations, which included comprehensive epidemiologic and clinical characterization of patients, with longitudinal evaluation of clinical status and sequelae (N. Hochberg, unpub. data).
One of the limitations of our laboratory/hospital-based study is the likelihood that we underestimated the numbers of cases of EM and A. cantonensis infection. By defi nition, we did not include persons who were asymptomatic, were not medically evaluated, did not have an LP, did not have CSF data that met specifi ed criteria for EM (e.g., if the LP was performed early or late in the course of infection, few eosinophils might have been noted), or did not meet conservative epidemiologic and clinical criteria. In addition, cases of EM/angiostrongyliasis that were associated with exposures in Hawaii but were diagnosed elsewhere were not systematically ascertained. Cases diagnosed after the end of the study period (February 2005) were not included (specifi cally, 2 cases reported in March and April 2005 that were associated with Hawaii County). Their existence, however, lends even more credence to the temporal clustering of cases in late 2004-early 2005.
A second limitation is that we cannot exclude the possibility that the temporal increases in frequency of cases were artifactual (e.g., refl ected heightened awareness of A. cantonensis infection or decreased thresholds for performing LPs). However, the investigation was prompted by clustering of 5 voluntary case reports during November 2004-January 2005, when EM and A. cantonensis infection were not reportable conditions, and included a parasitologically confi rmed case. In addition, for patients who accessed healthcare and had an LP, our methods for case ascertainment were not dependent upon clinicians considering or listing EM or A. cantonensis infection in discharge diagnoses. Our methods were systematic, statewide, and unbiased.
We recognize the limitations and the utility of the incidence data. We calculated incidence rates by generalizing relatively small numbers of cases to the population estimates for particular periods in the state and the pertinent counties. Adjusting the frequency data for the sizes of populations and the durations of periods facilitated comparisons between counties, periods, and causes of EM. The cases of EM not attributed to A. cantonensis served as a useful internal control for the conclusion that the incidence of angiostrongyliasis increased: the incidence of A. cantonensis infection was signifi cantly higher during the cluster period, whereas the incidence of the other EM cases did not increase.
In conclusion, we demonstrated the utility of a comprehensive, laboratory/hospital-based approach for statewide surveillance of EM and A. cantonensis infection in Hawaii. We found a cluster of angiostrongyliasis cases between November 2004 through February 2005 primarily centered in Hawaii and Maui Counties. Furthermore, EM and A. cantonensis infection were often not included in the discharge diagnoses for the case-patients. Our study therefore underscores the need to educate clinicians in Hawaii and elsewhere about EM and its causes, most notably A. cantonensis infection, a potentially severe but preventable infection. Improved detection and reporting may facilitate recognition of clusters of cases and prompt investigations that yield valuable insights about the epidemiologic and clinical characteristics of A. cantonensis infection. Dr Hochberg is an infectious disease fellow at Emory University. She is also a guest researcher with the Division of Parasitic Diseases at CDC, where she was an Epidemic Intelligence Service Offi cer at the time of this study. Her research currently focuses on the epidemiology of parasitic diseases.