American Journal of Epidemiology Copyright © 1992 by The Johns Hopkins University School of Hygiene and Public Health All rights reserved
Vol. 136. No. 11 Printed in U.S A.
Risk Factors for Lyme Disease in a Small Rural Community in Northern California
Robert S. Lane,1 Stephen A. Manweiler,1-2 Harrison A. Stubbs,3 Evelyne T. Lennette,4 John E. Madigan,5 and Paul E. Lavoie6
A 1-year prospective study of risk factors for seropositivity to and contraction of Lyme disease among members of a small rural community (population, approximately 150) was conducted in northwestern California in 1988-1989. The initial rate of seropositivity for Borrelia burgdorferi for 119 current or former residents ranged from 15 to 20% among three laboratories, with statistically significant interlaboratory agreement. Questionnaires were completed by 93 current residents at entry and 80 residents a year later to evaluate the association of serologic status with 20 categorical and 47 continuous variables. Seropositive subjects had resided in the study area about 2 years longer, were bitten by unspecified biting flies more often, and were less likely to have engaged in hiking than seronegative subjects. One of 59 seronegative subjects seroconverted a year later (annual incidence = 1.7%). The cumulative frequency of seropositivity for Lyme disease in the study population was >24%. Of 83 subjects examined physically, 13 were diagnosed as having definite and 18 as having probable Lyme disease. The seropositivity rate was significantly higher (38.7%) among individuals with definite/probable Lyme disease than in asymptomatic subjects (13.5%). Subjects who were seronegative or free of Lyme disease reported nearly as many tick bites as subjects who were seropositive or had a diagnosis of the disease. Age, time spent outdoors in the fall multiplied by a clothing index, and woodcutting were significantly associated with Lyme disease in logistic regression analyses. Am J Epidemiol 1992;136:1358-68. Borrelia; Lyme disease; ticks
In the United States, autochthonous cases of Lyme disease were reported in 43 states through 1988; six northeastern states (New York, New Jersey, Connecticut, Pennsylvania, Rhode Island, Massachusetts) and Wisconsin, Minnnesota, and California collectively accounted for 95 percent of the total number of cases for 1987 and 1988 (1).
However, factors that potentially may place people at risk for contracting the disease in this country have been assessed only in Connecticut, Massachusetts, New Jersey, and New York (2-9). Although these studies have produced inconsistent findings, various leisure or work-related outdoor activities that lead to greater tick exposure have been
Received for publication October 21, 1991, and in final form July 17, 1992. Abbreviations: ACIF, anticomplement indirect immunofluorescence assay; IFA, indirect immunofluorescence assay. 1 Department of Entomological Sciences, University of California-Berkeley, Berkeley, CA. 2 Current address: Biosys, Palo Alto, CA. 3 Program in Biostatistics, Biomedical and Environmental Health Sciences, School of Public Health, University of California-Berkeley, Berkeley, CA. 4 Virolab, Inc., Berkeley, CA. 5 Department of Medicine, School of Veterinary Medi-
cine, University of California-Davis, Davis, CA. 6 California Pacific Medical Center, San Francisco, CA Reprint requests to Dr. Robert S. Lane, Department of Entomological Sciences, 201 Wellman Hall, University of California-Berkeley, Berkeley, CA 94720. This research was supported in part by grants to Dr. Lane from the National Institutes of Health (AI22501) and from the Alpha Omicron Pi Foundation. We thank Phyllis and Lynn, whose support and help not only facilitated but made possible this study, and Elfnede DeRock, Judith A Pascocello, Kenji W Takeda, Mary M. Herrmann, and Mariko I Yasuda for technical assistance
1358
Risk Factors for Lyme Disease
implicated as risk factors. Time spent outdoors and pet ownership may also be risk factors for Lyme disease in northern California (10). In 1986, one of the authors (J. E. M.), while providing veterinary care in a small rural community located near Ukiah, California, learned that a number of the human inhabitants were afflicted with Lyme disease. A preliminary serologic survey was subsequently conducted to determine the prevalence of antibodies to Borrelia burgdorferi among members of this community (L. Szucs, University of California-Davis, unpublished data, 1990). This survey, which involved fewer than half of the estimated 150 people residing there, suggested that the cumulative frequency of Lyme disease was comparable to that reported previously for other small communities in areas of the northeastern United States in which the disease is hyperendemic (4, 5, 7). Consequently, a 1-year prospective epidemiologic investigation of the community near Ukiah was undertaken to evaluate the potential risk factors for Lyme disease. We hypothesized that antibody titers to B. burgdorferi and the acquisition of Lyme disease would be associated with pet ownership (in particular, dog ownership) and time spent outdoors. Ecologic studies were performed concurrently to identify the tick vectors and vertebrate reservoirs of B. burgdorferi. Veterinary studies were also conducted to determine the seroprevalence of B. burgdorferi in dogs, cats, horses, and other domestic animals as well as clinical manifestations in these animals that might be associated with Lyme disease. Here we present our epidemiologic findings on the cumulative frequency and incidence of, and the risk factors for, Lyme disease and seropositivity to B. burgdorferi in the human population. MATERIALS AND METHODS The study area
This investigation was conducted in a small community in Mendocino County, several kilometers northwest of Ukiah, California. The residents inhabit a former cattle
1359
ranch that was subdivided into 40-acre (16.2 hectares) parcels and sold during the early 1970s. The property encompasses approximately 5,300 acres (2,145 hectares) of rolling hills interspersed with ravines and covered predominantly by woodland-grass, dense woodland, and grassland. Lesser amounts of rock outcrops, wet meadows, and chaparral are present. The range of elevation is 305781 m. The climate is mediterranean, with cool, moist winters and hot, dry summers. Snow occurs occasionally, but usually melts within a few days. There is an abundance of wildlife on the ranch, including native Columbian black-tailed deer (Odocoileus hemionus columbianus), exotic fallow deer (Cervus dama), feral pigs (Susscrofa), blacktailed jackrabbits (Lepus californicus), several species of rodents, insectivores, and carnivores, many species of birds, western fence lizards (Sceloporus occidentalis), northern alligator lizards (Gerrhonotus multicarinatus), and other reptiles. Seroepidemiology
In September 1988, a community meeting was held at the ranch to explain the purpose of the study and to draw blood from volunteers after obtaining informed consent (n = 99 current and 20 former residents, some of whom had blood drawn on subsequent dates). Ninety-three current residents also completed a questionnaire that elicited information about several demographic factors (age, sex, length of residency), 67 potential risk factors (e.g., exposure to ticks and other arthropods during the last 2 years; pet ownership; the presence of rodents, deer, and other wildlife in the vicinity of homes; the average time spent outdoors in leisure or work-related activities; the type of clothing worn in different seasons; and the use of repellents), and clinical manifestations of the disease (11). Participants were asked to respond over a time-interval that ranged from the previous 2 years (number of tick or insect bites) to the duration of their residency (other variables). In total, 20 categorical and 47 continuous variables were evaluated. One variable, the
1360
Lane et al.
warm weather clothing index, requires detailed explanation. This index ranged from 0 to 6, with larger values indicating less protective clothing usually worn in warm weather. The index for males was constructed by scoring 0, I, and 2 for long sleeves, short sleeves, and no shirt, respectively; pants were scored similarly. The index for females was scored likewise, with a blouse and dress as possible options for a shirt and pants; they also had the same opportunity as males to respond to "no shirt," etc. The score for pants or a dress was doubled, so that the index score consisted of the shirt or blouse score plus two times the pants or dress score. This index was multiplied by the average number of hours spent outdoors per week each season to derive a seasonal index of potential exposure. The sera were stored at -74°C, blind coded, and split three ways. They were tested by conventional indirect immunofluorescence assay (IFA) or by anticomplement indirect immunofiuorescence assay (ACIF) in three of our laboratories (R. S. L., E. T. L., J. E. M.) and by immunoblotting in two of them (R. S. L., E. T. L.) (II). One year later, new blood samples from 80 current residents who had participated in the initial serologic survey were tested for antibodies to B. burgdorferi in two laboratories (R. S. L., E. T. L.). Details of the IFA, ACIF, immunoblotting (Western blot analysis), and control group have been presented previously (II). In each laboratory, the cutoff titer was the lowest serum titer below which 99 percent of the group of control sera (n = 115) would be considered negative. Thus, the minimum test specificity was preset at 99 percent and standardized among the three laboratories. Using this criterion, the cutofftitersweresetat 1:128, 1:8, and 1:256 in the laboratories of R. S. L., E. T. L., and J. E. M., respectively. Of these, the cutoff titer of 1:8 was ACIF based, whereas the titers of 1:128 and 1:256 were IFA based. To be considered seropositive for B. burgdorferi infection, a subject's serum specimen had to test positive by IFA in at least two of the three laboratories and by immunoblotting. A positive immunoblot was defined as
one in which the serum contained antibodies that were reactive to polypeptides having molecular weights (Mr) of approximately 31,000/32,000, 34,000/35,000, and 40,000/ 41,000. In a few cases, however, only one of the Mr 31,000/32,000 and Mr 34,000/ 35,000 polypeptides was present along with the Mr 40,000/41,000 polypeptide. Clinical manifestations
Of the 99 current residents from whom blood was drawn initially, 83 were examined physically by one of the authors (P. E. L.) within several months of entry to the study to determine whether they met the Centers for Disease Control's 1988 standardized case definition for exposure to Lyme disease in a county in which the disease is endemic. Subjects having erythema migrans alone or, in the absence of erythema migrans, involvement of one or more organ systems (i.e., cardiac, neurologic, or arthritic complications) as well as laboratory confirmation were included as cases. Subjects were classified according to the following system: 1) definite Lyme disease (i.e., the subject met the standardized case definition and the disease had been diagnosed previously by a physician); 2) probable Lyme disease (i.e., the subject had chronic multisystemic disorders consistent with the clinical spectrum of Lyme disease or had involvement of a single system and a history of an erythema migrans type skin lesion); or 3) no Lyme disease. Each subject was evaluated clinically only once in the present study. The physical examination was retrospective, strictly investigational, and relied predominantly upon histories given by participants. The few pathologic signs that were observed at entry were limited to rare oligoarticular or subtle neurologic abnormalities (e.g., areflexia in a child). Evaluations provided by the primary care physicians of individuals who had a previous diagnosis of Lyme disease were relied on to exclude potential confounding medical diagnoses. However, because the main focus of this study was epidemiologic and not clinical, neither history since entry nor treatment were reviewed.
Risk Factors for Lyme Disease
Statistical analyses
All data analyses were performed on a personal computer using SPSS/PC+ (SPSS Inc., Chicago, Illinois) for descriptive statistics and preliminary tests of group differences and PECAN (SERC, Seattle, Washington) for logistic regression analysis (12). Univariate analyses of group differences employed Student's / test and the MannWhitney U test (13) for continuous and quantitative variables and the chi-square test (14) for categorical variables, with Yates' correction for continuity for fourfold tables. The measure of association between various risk factors and definite/probable Lyme disease was the odds ratio, which approximates the relative risk (13). Logistic regression analysis was employed to adjust for and evaluate potential confounding variables. All variables included in logistic regression analyses are specified in the results section. The kappa statistic (14) was employed to describe the agreement between clinical determination of Lyme disease and seropositivity. The serologic results for the 93 current residents who completed the entry questionnaire were linked with their responses to evaluate potential risk factors by both univariate and multivariate analyses. These analyses were not repeated with questionnaire data gathered 1 year after entry because the small sample size at entry had diminished by 14 percent, and only one of the subjects who was retested had seroconverted. Furthermore, we believe that the baseline responses of the subjects provide the most accurate representation of potential risk factors for B. burgdorferi infection with the least amount of recall bias. Of the 83 subjects who were examined physically for evidence of Lyme disease, only 13(15.7 percent) reported having had a previous diagnosis of the disease, an inadequate sample for statistical purposes. Therefore, subjects having definite and probable Lyme disease were combined (n = 31) under a single category, probable Lyme disease, and compared statistically with subjects who did
1361
not have Lyme disease (n = 52) in the ensuing analyses. RESULTS Serologic status of subjects 1 year after entry
Serum specimens obtained 1 year after entry from 80 of the 93 residents who completed the entry questionnaire were tested in the laboratory of R. S. L., and all but four of these specimens were also tested in the laboratory of E. T. L. In the former laboratory, 16 of 17 subjects whose sera were positive at entry were still positive, and three of 63 sera (4.8 percent) from subjects who were seronegative at entry were positive by the IFA. Among the apparent seroconversions, however, one specimen was positive by immunoblot at entry, and the remaining two specimens could not be confirmed by immunoblotting. Thus, none of these three subjects was considered to have seroconverted. The remaining sera that were positive by the IFA test were confirmed by immunoblotting. A representative sample (n 20) of the 61 sera that were negative by the IFA were also evaluated and confirmed as negative by immunoblotting. Overall, the concordance between the IFA and immunoblotting was 94.7 percent (36 of 38 test results agreed), and the intralaboratory agreement for the 76 sera that were tested twice was 100 percent. In the laboratory of E. T. L., all 17 subjects found to be seropositive by the ACIF at entry were also seropositive when new blood samples were tested 1 year after entry, and one of the 59 subjects (1.7 percent) who were seronegative at entry had a positive titer 1 year later that was confirmed by immunoblotting. Agreement between the ACIF and immunoblotting was 98.7 percent (n = 76). Serum specimens were tested only once, so data concerning intralaboratory agreement are unavailable. The interlaboratory agreement was 96.1 percent (73 of 76) for the IFA and 89.5 percent (34 of 38) for immunoblotting. One of the four discrepant immunoblot test re-
1362
Laneetal.
suits was positive in one laboratory and partially reactive in the other (i.e., to the genusspecific Mr 40,000/41,000 polypeptide). Descriptive epidemiology
Initial analyses of all 67 variables for their association with serologic status yielded only three variables that were significant at the 5 percent level. Seropositive individuals were more likely to have resided in the study area longer (12.2 vs. 10.0 years, p = 0.032), were more likely to have been bitten by biting flies (p - 0.048), and were less likely to have participated in hiking as an outdoor activity (p = 0.006). The latter finding is based on five of 14 seropositive subjects who reported hiking versus 62 of 64 seronegative subjects. The seropositivity rate among subjects who had been bitten by biting flies was 36.7 percent (11 of 30) versus 15.1 percent (eight of 53 subjects) among those who reported that they had not been bitten. Of the 83 study subjects who were examined physically, 31 (37 percent) had a clinical diagnosis of probable Lyme disease (table
1). Overall, 19 of the subjects (23 percent) were seropositive, 12 of whom had a diagnosis of probable Lyme disease. Among these 31 subjects, the 12 who were seropositive were compared statistically with the 19 seronegative individuals for all 67 baseline variables. Only one of these variables, years of residence, approached statistical significance. On average, seropositive subjects tended to have resided longer (11.5 ± 3.9 years (mean ± standard deviation)) than seronegative subjects (8.5 ± 4.6 years). Characteristics of the study subjects are summarized in tables 1 and 2. Many of the other variables that were examined but are not tabulated here (« = 20) involved pets (dog, cat, horse), both individually and collectively, with respect to duration of ownership, use of insecticides, and, for dogs and cats, access to the interior of the home. Other variables omitted from the tables include certain warm weather clothing scores (shirt or blouse and pants or dress), and an additional warm weather clothing index that weighted pants or dress equally with shirt or
TABLE 1. Characteristics of study subjects, Ukiah area, California, 1988-1989
Characteristic
No. of males No. of females No. seropositive No. bitten by Ticks Winter Spring Summer Fall Fleas Mosquitoes Biting flies (unspecified) No. engaging in outdoor activities Hiking Gardening Woodcutting Sunbathing Bird watching Jogging Horseback riding
Lyme disease probable (" = 31)
Lyme disease unlikely (n = 52)
Total In — AT (n — BJ,
No.
%
No.
%
No.
%
12 19 12
38.7 61.3 38.7
31 21 7
59.6 40.4 13.5
43 40 19
51.8 48.2 22.9
11/25 19/25 6/25 8/25 21/30 30/31
44.0 76.0 24.0 32.0 70.0 96.8
7/27 20/27 10/27 6/27 35/52 47/52
25.9 74.1 37.0 22.2 67.3 90.4
18/52 39/52 16/52 14/52 56/82 77/83
34.6 75.0 30.8 26.9 68.3 92.8
15/31
48.4
15/52
28.8
30/83
36.1
26/31 25/31 29/31 8/31 5/31 9/31 6/31
83.9 80.6 93.5 25.8 16.1 29.0 19.4
50/52 42/52 39/52 20/52 10/52 12/52 8/52
96.2 80.8 75.0 38.5 19.2 23.1 15.4
76/83 67/83 68/83 28/83 15/83 21/83 14/83
91.6 80.7 81.9 33.7 18.1 25.3 16.9
Risk Factors for Lyme Disease
TABLE 2.
1363
Further characteristics of study subjects,* Ukiah area, California, 1988-1989 Characteristic
Age (years) Years of residence Work and leisure (hours/ week outdoors) Winter Spring Summer Fall
Total Warm weather clothing index Clothing index x no. of hours outdoors Winter Spring Summer Fall Total No. of tick bites in last 2 years
Lyme disease probable
Lyme disease unlikely (n = 52)
Total (n = 83)
36.3 ±12.2 9.7 ± 4.5
29.6 ±14.9 11.0 ±3.7
32.1 ± 14.3 10.5 ±4.0
21.3 ±19.6 31.7 ±21.3 35.8 ± 23.8 30.2 ± 32.9 30.2 ± 20.6 3.2 ± 1.1
13.8 ± 14.1 24.0 ±17.7 36.7 ± 27.4 21.1 ±17.7 23.9 ±17.4 3.1 ± 0.9
16.5 ±16.5 26.8 ±19.3 36.3 ± 26.0 24.4 ±19.4 26.1 ±18.7 3.1 ± 0.9
69.4 ± 60.8 105.9 ±80.2 125.2 ±91.5 101.5 ±79.3 102.3 ±74.4
41.5 ±46.5 72.8 ± 62.8 109.2 ±80.1 62.6 ± 58.5 71.5 ±55.8
51.1 ±53.1 84.5 ± 70.7 114.8 ±84.0 76.3 ± 68.6 82.1 ± 64.0
3.8 ± 3.3
3.0 ± 3.8
3.3 ± 3.7
" All characteristics are expressed as the mean ± the standard deviation
blouse. Of these variables, only the warm weather clothing index weighted by time spent outdoors revealed differences that approached statistical significance (p = 0.07). As these differences were generally exceeded by those of the index that weighted pants or dresses doubly, only the latter were retained for further analyses. Subjects with probable Lyme disease were somewhat older and included a higher proportion of females as compared with subjects without Lyme disease. On average, all subjects had resided at the ranch for approximately 10 years, and individuals without Lyme disease had resided there slightly longer than individuals with probable Lyme disease. All subjects reported spending an average of approximately 36 hours per week outdoors pursuing work and leisure activities in summer. They also reported spending 1012 fewer hours per week outdoors in spring and fall than in the summer and 20 fewer hours outdoors during the winter (table 2). Subjects with probable Lyme disease spent approximately 8-9 hours more per week outdoors in winter, spring, and fall than subjects without Lyme disease. The same amount of clothing was worn by subjects with and without probable Lyme
disease, as indicated by the warm weather clothing index (table 2). The product of the clothing index and hours spent outdoors may be interpreted as an index of exposure. That is, the clothing index is scaled (weighted) such that the value of the index is inversely proportional to the amount of clothing worn during warm weather. The mean values for this composite measure of exposure indicate substantial differences between the two groups (table 2). For all seasons, subjects with probable Lyme disease had higher average values for the composite indicators than subjects without Lyme disease, with the largest differences occurring in spring and fall. Subjects in both groups reported relatively few tick bites during the previous 2 years (table 2). Persons with probable Lyme disease reported an average of 3.8 bites, compared with 3.0 bites for those without Lyme disease (p > 0.05). Subjects with probable Lyme disease reported tick bites more often in fall and winter and less frequently in summer as compared with subjects without Lyme disease (table 1). Tick bites were commonly reported by both groups in spring. The incidence of reported flea bites was approximately 70 percent in both groups, and that for mosquito bites exceeded 90
1364
Laneetal.
percent for each group (table I). Subjects with probable Lyme disease reported bites by biting flies much more frequently (48.4 percent) than subjects without Lyme disease (28.8 percent). Hiking and sunbathing were reported somewhat more frequently, and woodcutting somewhat less frequently, by subjects without Lyme disease (table 1). Similar percentages of each group reported participating in gardening, bird watching, jogging, and horseback riding. Analytic epidemiology
The prevalence of seropositivity for B. burgdorferi was significantly (p = 0.017) higher in subjects with probable Lyme disease (12 of 31, 38.7 percent) than in subjects without Lyme disease (7 of 52, 13.5 percent). The overall agreement between serodiagnosis and clinical determination of probable Lyme disease was 68.7 percent (12 seropositive subjects with Lyme disease plus 45 seronegative subjects without Lyme disease divided by a total of 83 subjects), and the value for the kappa statistic was 0.27 (standard error = 0.10, Z = 2.65). Crude odds ratios and 95 percent confidence intervals for probable Lyme disease are summarized for several potential risk factors in table 3. Two factors, female versus male sex and bites by biting flies each had odds ratios of 2.3, which approached statistical significance at the 5 percent level (p = 0.068 and 0.076, respectively). Subjects who cut wood were nearly five times more likely to have probable Lyme disease than other subjects (p = 0.048). These and several other variables were evaluated in logistic regression analyses to explore their association with' probable Lyme disease. All variables specified in tables 1-3 were included in these analyses. Three factors were found to be significantly associated with probable Lyme disease among all single-predictor models. These factors were hours per week outdoors in fall multiplied by a clothing index (p = 0.033), woodcutting (p = 0.048), and age (p = 0.043) with associated logistic coefficients of 1.008, 4.83, and 1.037, respectively. Of
TABLE 3. Crude odds ratios for probable Lyme disease associated with selected characteristics of study subjects, Ukiah area, California, 1988-1989 Variable Females versus males Bitten by Biting flies (unspecified) Fleas Mosquitoes Ticks Winter Spring Summer Fall Outdoor activities (woodcutting)
95% Odds ratio confidence interval 2.3
0.94-5.81
2.3 1.1 3.2
0.92-5.83 0.43-3.00 0.36-28.7
2.2 1.1 0.5 1.6 4.8
0.70-7.22 0.31-3.90 0.16-1.79 0.48-5.67 1.01-23.10
these, only the coefficient for woodcutting can be interpreted as an odds ratio. For all hours spent outdoors multiplied by a clothing index, there is an increase in logistic risk of 0.008 units per unit increase in the indicator. Similarly, an increase of 0.037 in logistic risk was observed for each additional year of increase in age. None of these models could be improved with the inclusion of other risk factors or covariables from among all of the remaining variables. DISCUSSION
Several researchers have expressed serious reservations about the lack of standardization and reliability of the assays (enzymelinked immunosorbent, IFA) that are being used routinely for the serodiagnosis of Lyme disease (15-18). Therefore, we thought it would be prudent to compare and validate the IFAs performed in three of our laboratories before linking the serologic and entry questionnaire data. In so doing, we hoped to increase the reliability of our collective judgment about the serologic status of each subject and, consequently, to enhance the sensitivity of our epidemiologic analysis for identifying potential risk factors for Lyme disease in the study population. The interlaboratory agreement (88-93 percent) for sera obtained at entry was found to be statistically highly significant, which was attributed to the fact that the participating laboratories adhered to a uniform standard for test specificity (11).
Risk Factors for Lyme Disease
Seroprevalence
In 1988, the range of the seropositivity rate at entry was 15-20 percent among the three laboratories. One year later, blood samples were again drawn from 81 percent of the current residents who volunteered at entry. All of these were retested for antibodies to B. burgdorferi in one of our laboratories (R. S. L.), and most of them were also retested in another (E. T. L.). The seropositivity rate in both laboratories was 24 percent, and the interlaboratory agreement was significant again (96 percent, n — 76). The somewhat higher seropositivity rate I year after entry was not unexpected, since the entry cohort (n = 119 subjects) included 20 former residents, only one of whom was determined to be both seropositive and immunoblot-positive. It is not known whether the demographic characteristics and serologic status of the study subjects are truly representative of the entire community, since only 66 percent of the estimated 150 current residents volunteered at entry and even fewer (53 percent) participated 1 year after entry. Moreover, demographic data for the nonparticipants are unavailable for comparison with those who volunteered. One might expect intuitively that healthy individuals would be less likely to volunteer for such a study than individuals who were experiencing symptoms of illness or aging phenomena (e.g., arthralgia or myalgia) suggestive of Lyme disease. If this is so, and even if all of the nonparticipants are assumed to be seronegative, a highly unlikely event, then the seroprevalence rate for all 150 residents would still be as high or higher than that determined for most other high-risk communities or occupational groups (e.g., outdoor employees) that have been studied in the northeastern United States (4-9, 19). Clinical correlates
The percentage of the 16 current seropositive residents who reported clinical manifestations consistent with Lyme disease were as follows: erythema migrans, 37.5 percent; arthritis/arthralgia, 68.8 percent; cardiac ab-
1365
normalities, 6.3 percent; and neurologic involvement, 37.5 percent (11). In another small series (n = 10) of Lyme disease patients from northern California (10), 100 percent reported erythema migrans; 60 percent reported arthritis/arthralgia; 20 percent reported cardiac abnormalities; and 60 percent reported neurologic symptoms. The difference in the prevalence of erythema migrans among the subjects in these two studies may reflect chance variation due to small sample sizes or host-related factors such as the mean age of the subjects, which was 17 years less in the present study (i.e., 32 vs. 49 years). In studies done in the northeastern United States, 5-100 percent of seropositive subjects reported a history of Lyme disease (4-7, 9). Such striking differences in the ratio of apparent-to-inapparent infections between regions may be due to a lack of standardized testing or to regional differences in the pathogenicity, virulence, or both of B. burgdorferi strains. In California, considerable antigen ic heterogeneity has been observed among the 30 wild-rodent or tick-derived strains of B. burgdorferi that have been characterized so far (20-23). In New York, a tick-derived variant of B. burgdorferi was shown to be infectious but nonarthritogenic in laboratory rats and mice (24). This nonarthritogenic strain, like a number of Californian strains (22, 23), possesses somewhat heavier outer surface proteins A and B than the B3l-type strain, plus a prominent protein with a low molecular weight of about 23,500. Cumulative frequency and incidence
Twenty-four of 119 serum specimens (20 percent) tested at entry were seropositive (11); of these, 23 were from the 99 current residents (23 percent) and only one came from the group of 20 former residents (5 percent). This marked difference in seropositivity rates between current and former residents may be related to the duration of their residency (10.5 vs. approximately 7.3 years) and to the amount of potential exposure to vector ticks. A year later, one of 59 subjects (1.7 percent) who was found to be seronegative at entry had seroconverted. Thus, the
1366
Laneetal.
approximate cumulative frequency of serologic infection with B. burgdorferi in the resident population after 1 year of prospective study was at least 24 percent (i.e., 24 of 99). However, the cumulative frequency of Lyme disease may be much higher, since 19 percent fewer residents were retested a year later (i.e., 80 vs. 99), 31 of 83 residents (37 percent) who were examined physically were given a diagnosis of probable Lyme disease, and not all patients with the disease had detectable antibody titers. The cumulative frequency of Lyme disease ranged between 4 and 35 percent in the northeastern United States, with annual or seasonal incidence rates of < 1-10 percent in various studies (3-7, 25-27). The highest annual incidence of 10 percent was recorded over a 2-year period in the Ipswich area of coastal Massachusetts (7). Except for the Ipswich study, the cumulative frequency and annual incidence of Lyme disease in the Ukiah study area are comparable to or higher than the cumulative frequencies/ incidences reported from the Northeast. Risk factors
The high seropositivity rate and cumulative frequency of clinical Lyme disease in the subject population is perplexing in view of the generally low (1 -6 percent) spirochetal infection rates in Ixodes pacificus in California (28). In fact, only nine of 684 adult /. pacificus (1.3 percent) swept from vegetation in the study area between 1988 and 1990 were found to contain spirochetes (R. S. Lane and S. A. Manweiler, unpublished data). In contrast, spirochetal infection rates in Ixodes dammini average roughly 25 percent in nymphal and 50 percent in adult ticks in the Northeast (28). Furthermore, the average number of tick bites reported over a 2-year period by subjects with probable Lyme disease (3.8) versus subjects deemed free of Lyme disease (3.0) was low for both groups (table 2). Regional differences in the epidemiology of Lyme disease may explain this paradox. In the Northeast, human exposure to /. dammini ticks occurs in peridomestic environ-
ments, such as on or near well-maintained lawns in residential areas (1, 7, 29-32). Thus, it is not surprising that occupational or leisure-related tick exposure, presence of ticks near residences, time spent outdoors in various activities, and the abundance of deer have been implicated as actual or potential risk factors in this region (2-4, 6, 7, 9, 10, 27, 30-33). In the Pacific states, especially California, where most human cases have been reported, the increased age distribution (i.e., with an apparent peak of reported cases in the 25- to 44-year-old age group) may reflect the typical pattern for tick-borne diseases in the United States that is associated with outdoor occupational or recreational exposure (1). Our findings seem to confirm this observation, since the residents we studied built their homes in what is essentially a wildlands area/nature preserve (i.e., hunting is largely prohibited). Domiciles are surrounded by natural vegetation that is occupied by ticks and an abundance of wildlife, many of which serve as hosts of ticks and of B. burgdorferi (R. S. Lane & S. A. Manweiler, unpublished data). The two most abundant human-biting ticks, Dermacenlor occidentalis and /. pacificus, are infected naturally with B. burgdorferi in northern California (28). The dusky-footed wood rat, Neotomafuscipes, which lives in close proximity to human dwellings, has been implicated as a reservoir of B. burgdorferi infection (23). Nearly equal numbers of adult residents work in the community or elsewhere. Although time spent working outdoors was not identified as a risk factor, about one-half of subjects working in the community are engaged in occupational activities that may place them at risk for Lyme disease (e.g., caretaker, nurseryman, property manager, and handiperson). In contrast, over 90 percent of residents who work elsewhere are employed in occupations that would not place them at much, if any, risk of infection (e.g., artisan, carpenter, electrician, reporter, teacher, waitress). Seropositive subjects resided in the community about 2 years longer, were bitten by
Risk Factors for Lyme Disease
unspecified biting flies more frequently, and hiked less often than seronegative individuals. In Massachusetts, length of residency and location of homes were not associated with higher risk for Lyme disease on Great Island, although increased age was (5), whereas risk was correlated with proximity of residence to a nature preserve in the Ipswich area (7). The latter finding parallels and substantiates our discovery of heightened risk among residents of a natural area. The bites of mosquitoes, deer flies, and horse flies have been incriminated as potential risk factors for the acquisition of Lyme disease in Connecticut and New York (2, 26, 34, 35). Our finding of a significant association between bites by bloodsucking flies and seropositivity suggests that biting flies should be evaluated for infection with spirochetes in the Far West. However, this association may represent an indirect measure of outdoor exposure, or it may be due merely to chance as a result of the large number of statistical tests we performed. The crude (univariate) analyses disclosed that cutting wood was significantly associated with the risk for Lyme disease. Two other factors, female versus male sex and bites by biting flies, approached statistical significance. However, logistic regression analyses revealed that the only dichotomous factor that could be interpreted as a measure of risk based on its odds ratio was woodcutting. Although the season(s) during which woodcutting was usually performed is not known, the combined activity periods of /. pacificus nymphs and adults span about 9 months (approximately mid-November to mid-August) in the study region. Increased risk due to cutting wood may simply reflect the amount of time spent outdoors in favorable tick habitats and, therefore, the likelihood of tick exposure. In future studies, various property maintenance activities (e.g., cutting brush) should also be evaluated as potential risk factors for Lyme disease. We had hypothesized that time spent outdoors and pet ownership would be associated with elevated risk for infection with B. burgdorferi or contraction of Lyme disease. Although time spent outdoors was not iden-
1367
tified as a significant risk factor, subjects with probable Lyme disease spent approximately 8-9 hours more per week outdoors than subjects without Lyme disease during the peak activity periods of nymphal and adult /. pacificus (fall, winter, and spring); they also reported more tick bites than subjects who. did not have the disease (table 2). Pet ownership was also not associated with increased risk which, with one exception (2), is consistent with epidemiologic findings from the northeastern United States (4, 5, 9, 19). Like Eng et al. (19), however, we found that dogs are at much greater risk for infection with B. burgdorferi because they are exposed to considerably more vector ticks than are people (J. E. Madigan & R. S. Lane, unpublished data).
REFERENCES 1. Miller GL. Craven RB, Bailey RE, et al. The epidemiology of Lyme disease in the United States 1987-1988. Lab Med 199O;21:285-9. 2. Steere AC. Broderick TF, Malawista SE. Erythema chronicum migrans and Lyme arthritis: epidemiologic evidence for a tick vector. Am J Epidemiol 1978; 108:312-21. 3. Bowen GS, Schulze TL. Hayne C, et al. A focus of Lyme disease in Monmouth County, New Jersey. Am J Epidemiol 1984; 120:387-94. 4. Hanrahan JP, Benach JL, Coleman JL, et al. Incidence and cumulative frequency of endemic Lyme disease in a community. J Infect Dis 1984:150: 489-96. 5. Steere AC, Taylor E, Wilson ML, et al. Longitudinal assessment of the clinical and epidemiological features of Lyme disease in a defined population. J Infect Dis 1986:154:295-300. 6. Smith PF, Benach JL, White DJ, et al. Occupational risk of Lyme disease in endemic areas of New York State. Ann NY Acad Sci 1988;539: 289-301. 7. Lastavica CC, Wilson ML, Berardi VP, et al. Rapid emergence of a focal epidemic of Lyme disease in coastal Massachusetts. N Engl J Med 1989:320: 133-7. 8. Goldstein MD, Schwartz BS, Friedmann C, et al. Lyme disease in New Jersey outdoor workers: a statewide survey of seroprevalence and tick exposure. Am J Pub Health 1990:80:1225-9. 9. Schwartz BS, Goldstein MD. Lyme disease in outdoor workers: risk factors, preventive measures, and tick removal methods. Am J Epidemiol 1990; 131:877-85. 10. Lane RS, Lavoie PE. Lyme borreliosis in California: acarological, clinical, and epidemiological studies. Ann NY Acad Sci 1988;539:192-203. 11. Lane RS, Lennette ET, Madigan JE. Interlabora-
1368 Laneetal.
12.
13. 14. 15. 16. 17. 18.
19. 20.
21.
22. 23. 24.
tory and intralaboratory comparisons of indirect immunofluorescence assays for serodiagnosis of Lyme disease. J Clin Microbiol 1990:28:1774-9. Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol. I. The analysis of case-control studies. (IARC scientific publication no. 32). Lyon, France: International Agency for Research on Cancer, 1980. Zar JH. Biostatistical analysis. Englewood Cliffs, NJ: Prentice-Hall, 1984. Fleiss JL. Statistical methods for rates and proportions. New York, NY: John Wiley and Sons. 1973. Osterholm MT, Forfang JC, White K.E, et al. Lyme disease in Minnesota: epidemiologic and serologic findings. Yale J Biol Med 1984:57:677-83. Wilkinson HW, Russell H, Sampson JS. Caveats on using nonstandardized serologic tests for Lyme disease. (Letter). J Clin Microbiol 1985:21:291. Hedberg CW, Osterholm MT. MacDonald KL, et al. An interlaboratory study of antibody to Borrelia burgdorferi.i Infect Dis 1987; 155:1325-7. Schwartz BS. Goldstein MD, Ribeiro JMC. et al. Antibody testing in Lyme disease: a comparison of results in four laboratories. JAMA 1989:262: 3431-4. Eng TR, Wilson ML. Spielman A, et al. Greater risk of Borrelia burgclorferi infection in dogs than in people. J Infect Dis 1988:158:1410-11. Burgdorfer W, Lane RS, Barbour AG, et al. The western black-legged tick, lxodes pacifwus: a vector of Borrelia burgdorfen. Am J Trop Med Hyg 1985; 34:925-30. Bissett ML, Hill W. Characterization of Borrelia burgdorferi strains isolated from lxodes pacifwus ticks in California. J Clin Microbiol 1987:25: 2296-2301. Lane RS, Pascocello JA. Antigenic characteristics of Borrelia burgdorferi isolates from ixodid ticks in California. J Clin Microbiol 1989:27:2344-9. Lane RS, Brown RN. Wood rats and kangaroo rats: potential reservoirs of the Lyme disease spirochete in California. J Med Entomol 1991:28:299-302. Anderson JF, Barthold SW, Magnarelli LA. Infectious but nonpathogenic isolate of Borrelia burg-
dorferi. J Clin Microbiol 1990;28:2693-9. 25. Steere AC, Grodzicki RL, Kornblatt AN, et al. The spirochetal etiology of Lyme disease. N Engl J Med 1983;308:733-40. 26. Benach JL, Coleman JL. Clinical and geographic characteristics of Lyme disease in New York. Zentralbl Bakteriol Mikrobiol Hyg A 1986:263: 477-82. 27. Petersen LR, Sweeney AH. Checko PJ, et al. Epidemiological and clinical features of 1,149 persons with Lyme disease identified by laboratory-based surveillance in Connecticut. Yale J Biol Med 1989; 62:253-62. 28. Lane RS, Piesman J, Burgdorfer W. Lyme borreliosis: relation of its causative agent to its vectors and hosts in North America and Europe. Annu Rev Entomol 1991;36:587-609. 29. Williams CL, Curran AS, Lee AC, et al. Lyme disease: epidemiologic characteristics of an outbreak in Westchester County, NY. Am J Pub Health 1986;76:62-5. 30. Falco RC, Fish D. Prevalence of lxodes dammini near the homes of Lyme disease patients in Westchester County, New York. Am J Epidemiol 1988; 127:826-30. 31. Falco RC, Fish D. A survey of tick bites acquired in a Lyme disease endemic area in southern New York State. Ann NY Acad Sci 1988:539:456-7. 32. Maupin GO, Fish D, Zultowsky J, et al. Landscape ecology of Lyme disease in a residential area of Westchester County, New York. Am J Epidemiol 1991:133:1105-13. 33. Bowen GS, Griffin M, Hayne C. et al. Clinical manifestations and descriptive epidemiology of Lyme disease in New Jersey, 1978 to 1982. JAMA 1984:251:2236-40. 34. Magnarelli LA, Anderson JF, Barbour AG. The etiologic agent of Lyme disease in deer flies, horse flies, and mosquitoes. J Infect Dis 1986; 154: 355-8. 35. Magnarelli LA, Anderson JF. Ticks and biting insects infected with the etiologic agent of Lyme disease, Borrelia burgdorferi. J Clin Microbiol 1988:26:1482-6.
Vol. 136. No. 11 Printed in U.S A.
Risk Factors for Lyme Disease in a Small Rural Community in Northern California
Robert S. Lane,1 Stephen A. Manweiler,1-2 Harrison A. Stubbs,3 Evelyne T. Lennette,4 John E. Madigan,5 and Paul E. Lavoie6
A 1-year prospective study of risk factors for seropositivity to and contraction of Lyme disease among members of a small rural community (population, approximately 150) was conducted in northwestern California in 1988-1989. The initial rate of seropositivity for Borrelia burgdorferi for 119 current or former residents ranged from 15 to 20% among three laboratories, with statistically significant interlaboratory agreement. Questionnaires were completed by 93 current residents at entry and 80 residents a year later to evaluate the association of serologic status with 20 categorical and 47 continuous variables. Seropositive subjects had resided in the study area about 2 years longer, were bitten by unspecified biting flies more often, and were less likely to have engaged in hiking than seronegative subjects. One of 59 seronegative subjects seroconverted a year later (annual incidence = 1.7%). The cumulative frequency of seropositivity for Lyme disease in the study population was >24%. Of 83 subjects examined physically, 13 were diagnosed as having definite and 18 as having probable Lyme disease. The seropositivity rate was significantly higher (38.7%) among individuals with definite/probable Lyme disease than in asymptomatic subjects (13.5%). Subjects who were seronegative or free of Lyme disease reported nearly as many tick bites as subjects who were seropositive or had a diagnosis of the disease. Age, time spent outdoors in the fall multiplied by a clothing index, and woodcutting were significantly associated with Lyme disease in logistic regression analyses. Am J Epidemiol 1992;136:1358-68. Borrelia; Lyme disease; ticks
In the United States, autochthonous cases of Lyme disease were reported in 43 states through 1988; six northeastern states (New York, New Jersey, Connecticut, Pennsylvania, Rhode Island, Massachusetts) and Wisconsin, Minnnesota, and California collectively accounted for 95 percent of the total number of cases for 1987 and 1988 (1).
However, factors that potentially may place people at risk for contracting the disease in this country have been assessed only in Connecticut, Massachusetts, New Jersey, and New York (2-9). Although these studies have produced inconsistent findings, various leisure or work-related outdoor activities that lead to greater tick exposure have been
Received for publication October 21, 1991, and in final form July 17, 1992. Abbreviations: ACIF, anticomplement indirect immunofluorescence assay; IFA, indirect immunofluorescence assay. 1 Department of Entomological Sciences, University of California-Berkeley, Berkeley, CA. 2 Current address: Biosys, Palo Alto, CA. 3 Program in Biostatistics, Biomedical and Environmental Health Sciences, School of Public Health, University of California-Berkeley, Berkeley, CA. 4 Virolab, Inc., Berkeley, CA. 5 Department of Medicine, School of Veterinary Medi-
cine, University of California-Davis, Davis, CA. 6 California Pacific Medical Center, San Francisco, CA Reprint requests to Dr. Robert S. Lane, Department of Entomological Sciences, 201 Wellman Hall, University of California-Berkeley, Berkeley, CA 94720. This research was supported in part by grants to Dr. Lane from the National Institutes of Health (AI22501) and from the Alpha Omicron Pi Foundation. We thank Phyllis and Lynn, whose support and help not only facilitated but made possible this study, and Elfnede DeRock, Judith A Pascocello, Kenji W Takeda, Mary M. Herrmann, and Mariko I Yasuda for technical assistance
1358
Risk Factors for Lyme Disease
implicated as risk factors. Time spent outdoors and pet ownership may also be risk factors for Lyme disease in northern California (10). In 1986, one of the authors (J. E. M.), while providing veterinary care in a small rural community located near Ukiah, California, learned that a number of the human inhabitants were afflicted with Lyme disease. A preliminary serologic survey was subsequently conducted to determine the prevalence of antibodies to Borrelia burgdorferi among members of this community (L. Szucs, University of California-Davis, unpublished data, 1990). This survey, which involved fewer than half of the estimated 150 people residing there, suggested that the cumulative frequency of Lyme disease was comparable to that reported previously for other small communities in areas of the northeastern United States in which the disease is hyperendemic (4, 5, 7). Consequently, a 1-year prospective epidemiologic investigation of the community near Ukiah was undertaken to evaluate the potential risk factors for Lyme disease. We hypothesized that antibody titers to B. burgdorferi and the acquisition of Lyme disease would be associated with pet ownership (in particular, dog ownership) and time spent outdoors. Ecologic studies were performed concurrently to identify the tick vectors and vertebrate reservoirs of B. burgdorferi. Veterinary studies were also conducted to determine the seroprevalence of B. burgdorferi in dogs, cats, horses, and other domestic animals as well as clinical manifestations in these animals that might be associated with Lyme disease. Here we present our epidemiologic findings on the cumulative frequency and incidence of, and the risk factors for, Lyme disease and seropositivity to B. burgdorferi in the human population. MATERIALS AND METHODS The study area
This investigation was conducted in a small community in Mendocino County, several kilometers northwest of Ukiah, California. The residents inhabit a former cattle
1359
ranch that was subdivided into 40-acre (16.2 hectares) parcels and sold during the early 1970s. The property encompasses approximately 5,300 acres (2,145 hectares) of rolling hills interspersed with ravines and covered predominantly by woodland-grass, dense woodland, and grassland. Lesser amounts of rock outcrops, wet meadows, and chaparral are present. The range of elevation is 305781 m. The climate is mediterranean, with cool, moist winters and hot, dry summers. Snow occurs occasionally, but usually melts within a few days. There is an abundance of wildlife on the ranch, including native Columbian black-tailed deer (Odocoileus hemionus columbianus), exotic fallow deer (Cervus dama), feral pigs (Susscrofa), blacktailed jackrabbits (Lepus californicus), several species of rodents, insectivores, and carnivores, many species of birds, western fence lizards (Sceloporus occidentalis), northern alligator lizards (Gerrhonotus multicarinatus), and other reptiles. Seroepidemiology
In September 1988, a community meeting was held at the ranch to explain the purpose of the study and to draw blood from volunteers after obtaining informed consent (n = 99 current and 20 former residents, some of whom had blood drawn on subsequent dates). Ninety-three current residents also completed a questionnaire that elicited information about several demographic factors (age, sex, length of residency), 67 potential risk factors (e.g., exposure to ticks and other arthropods during the last 2 years; pet ownership; the presence of rodents, deer, and other wildlife in the vicinity of homes; the average time spent outdoors in leisure or work-related activities; the type of clothing worn in different seasons; and the use of repellents), and clinical manifestations of the disease (11). Participants were asked to respond over a time-interval that ranged from the previous 2 years (number of tick or insect bites) to the duration of their residency (other variables). In total, 20 categorical and 47 continuous variables were evaluated. One variable, the
1360
Lane et al.
warm weather clothing index, requires detailed explanation. This index ranged from 0 to 6, with larger values indicating less protective clothing usually worn in warm weather. The index for males was constructed by scoring 0, I, and 2 for long sleeves, short sleeves, and no shirt, respectively; pants were scored similarly. The index for females was scored likewise, with a blouse and dress as possible options for a shirt and pants; they also had the same opportunity as males to respond to "no shirt," etc. The score for pants or a dress was doubled, so that the index score consisted of the shirt or blouse score plus two times the pants or dress score. This index was multiplied by the average number of hours spent outdoors per week each season to derive a seasonal index of potential exposure. The sera were stored at -74°C, blind coded, and split three ways. They were tested by conventional indirect immunofluorescence assay (IFA) or by anticomplement indirect immunofiuorescence assay (ACIF) in three of our laboratories (R. S. L., E. T. L., J. E. M.) and by immunoblotting in two of them (R. S. L., E. T. L.) (II). One year later, new blood samples from 80 current residents who had participated in the initial serologic survey were tested for antibodies to B. burgdorferi in two laboratories (R. S. L., E. T. L.). Details of the IFA, ACIF, immunoblotting (Western blot analysis), and control group have been presented previously (II). In each laboratory, the cutoff titer was the lowest serum titer below which 99 percent of the group of control sera (n = 115) would be considered negative. Thus, the minimum test specificity was preset at 99 percent and standardized among the three laboratories. Using this criterion, the cutofftitersweresetat 1:128, 1:8, and 1:256 in the laboratories of R. S. L., E. T. L., and J. E. M., respectively. Of these, the cutoff titer of 1:8 was ACIF based, whereas the titers of 1:128 and 1:256 were IFA based. To be considered seropositive for B. burgdorferi infection, a subject's serum specimen had to test positive by IFA in at least two of the three laboratories and by immunoblotting. A positive immunoblot was defined as
one in which the serum contained antibodies that were reactive to polypeptides having molecular weights (Mr) of approximately 31,000/32,000, 34,000/35,000, and 40,000/ 41,000. In a few cases, however, only one of the Mr 31,000/32,000 and Mr 34,000/ 35,000 polypeptides was present along with the Mr 40,000/41,000 polypeptide. Clinical manifestations
Of the 99 current residents from whom blood was drawn initially, 83 were examined physically by one of the authors (P. E. L.) within several months of entry to the study to determine whether they met the Centers for Disease Control's 1988 standardized case definition for exposure to Lyme disease in a county in which the disease is endemic. Subjects having erythema migrans alone or, in the absence of erythema migrans, involvement of one or more organ systems (i.e., cardiac, neurologic, or arthritic complications) as well as laboratory confirmation were included as cases. Subjects were classified according to the following system: 1) definite Lyme disease (i.e., the subject met the standardized case definition and the disease had been diagnosed previously by a physician); 2) probable Lyme disease (i.e., the subject had chronic multisystemic disorders consistent with the clinical spectrum of Lyme disease or had involvement of a single system and a history of an erythema migrans type skin lesion); or 3) no Lyme disease. Each subject was evaluated clinically only once in the present study. The physical examination was retrospective, strictly investigational, and relied predominantly upon histories given by participants. The few pathologic signs that were observed at entry were limited to rare oligoarticular or subtle neurologic abnormalities (e.g., areflexia in a child). Evaluations provided by the primary care physicians of individuals who had a previous diagnosis of Lyme disease were relied on to exclude potential confounding medical diagnoses. However, because the main focus of this study was epidemiologic and not clinical, neither history since entry nor treatment were reviewed.
Risk Factors for Lyme Disease
Statistical analyses
All data analyses were performed on a personal computer using SPSS/PC+ (SPSS Inc., Chicago, Illinois) for descriptive statistics and preliminary tests of group differences and PECAN (SERC, Seattle, Washington) for logistic regression analysis (12). Univariate analyses of group differences employed Student's / test and the MannWhitney U test (13) for continuous and quantitative variables and the chi-square test (14) for categorical variables, with Yates' correction for continuity for fourfold tables. The measure of association between various risk factors and definite/probable Lyme disease was the odds ratio, which approximates the relative risk (13). Logistic regression analysis was employed to adjust for and evaluate potential confounding variables. All variables included in logistic regression analyses are specified in the results section. The kappa statistic (14) was employed to describe the agreement between clinical determination of Lyme disease and seropositivity. The serologic results for the 93 current residents who completed the entry questionnaire were linked with their responses to evaluate potential risk factors by both univariate and multivariate analyses. These analyses were not repeated with questionnaire data gathered 1 year after entry because the small sample size at entry had diminished by 14 percent, and only one of the subjects who was retested had seroconverted. Furthermore, we believe that the baseline responses of the subjects provide the most accurate representation of potential risk factors for B. burgdorferi infection with the least amount of recall bias. Of the 83 subjects who were examined physically for evidence of Lyme disease, only 13(15.7 percent) reported having had a previous diagnosis of the disease, an inadequate sample for statistical purposes. Therefore, subjects having definite and probable Lyme disease were combined (n = 31) under a single category, probable Lyme disease, and compared statistically with subjects who did
1361
not have Lyme disease (n = 52) in the ensuing analyses. RESULTS Serologic status of subjects 1 year after entry
Serum specimens obtained 1 year after entry from 80 of the 93 residents who completed the entry questionnaire were tested in the laboratory of R. S. L., and all but four of these specimens were also tested in the laboratory of E. T. L. In the former laboratory, 16 of 17 subjects whose sera were positive at entry were still positive, and three of 63 sera (4.8 percent) from subjects who were seronegative at entry were positive by the IFA. Among the apparent seroconversions, however, one specimen was positive by immunoblot at entry, and the remaining two specimens could not be confirmed by immunoblotting. Thus, none of these three subjects was considered to have seroconverted. The remaining sera that were positive by the IFA test were confirmed by immunoblotting. A representative sample (n 20) of the 61 sera that were negative by the IFA were also evaluated and confirmed as negative by immunoblotting. Overall, the concordance between the IFA and immunoblotting was 94.7 percent (36 of 38 test results agreed), and the intralaboratory agreement for the 76 sera that were tested twice was 100 percent. In the laboratory of E. T. L., all 17 subjects found to be seropositive by the ACIF at entry were also seropositive when new blood samples were tested 1 year after entry, and one of the 59 subjects (1.7 percent) who were seronegative at entry had a positive titer 1 year later that was confirmed by immunoblotting. Agreement between the ACIF and immunoblotting was 98.7 percent (n = 76). Serum specimens were tested only once, so data concerning intralaboratory agreement are unavailable. The interlaboratory agreement was 96.1 percent (73 of 76) for the IFA and 89.5 percent (34 of 38) for immunoblotting. One of the four discrepant immunoblot test re-
1362
Laneetal.
suits was positive in one laboratory and partially reactive in the other (i.e., to the genusspecific Mr 40,000/41,000 polypeptide). Descriptive epidemiology
Initial analyses of all 67 variables for their association with serologic status yielded only three variables that were significant at the 5 percent level. Seropositive individuals were more likely to have resided in the study area longer (12.2 vs. 10.0 years, p = 0.032), were more likely to have been bitten by biting flies (p - 0.048), and were less likely to have participated in hiking as an outdoor activity (p = 0.006). The latter finding is based on five of 14 seropositive subjects who reported hiking versus 62 of 64 seronegative subjects. The seropositivity rate among subjects who had been bitten by biting flies was 36.7 percent (11 of 30) versus 15.1 percent (eight of 53 subjects) among those who reported that they had not been bitten. Of the 83 study subjects who were examined physically, 31 (37 percent) had a clinical diagnosis of probable Lyme disease (table
1). Overall, 19 of the subjects (23 percent) were seropositive, 12 of whom had a diagnosis of probable Lyme disease. Among these 31 subjects, the 12 who were seropositive were compared statistically with the 19 seronegative individuals for all 67 baseline variables. Only one of these variables, years of residence, approached statistical significance. On average, seropositive subjects tended to have resided longer (11.5 ± 3.9 years (mean ± standard deviation)) than seronegative subjects (8.5 ± 4.6 years). Characteristics of the study subjects are summarized in tables 1 and 2. Many of the other variables that were examined but are not tabulated here (« = 20) involved pets (dog, cat, horse), both individually and collectively, with respect to duration of ownership, use of insecticides, and, for dogs and cats, access to the interior of the home. Other variables omitted from the tables include certain warm weather clothing scores (shirt or blouse and pants or dress), and an additional warm weather clothing index that weighted pants or dress equally with shirt or
TABLE 1. Characteristics of study subjects, Ukiah area, California, 1988-1989
Characteristic
No. of males No. of females No. seropositive No. bitten by Ticks Winter Spring Summer Fall Fleas Mosquitoes Biting flies (unspecified) No. engaging in outdoor activities Hiking Gardening Woodcutting Sunbathing Bird watching Jogging Horseback riding
Lyme disease probable (" = 31)
Lyme disease unlikely (n = 52)
Total In — AT (n — BJ,
No.
%
No.
%
No.
%
12 19 12
38.7 61.3 38.7
31 21 7
59.6 40.4 13.5
43 40 19
51.8 48.2 22.9
11/25 19/25 6/25 8/25 21/30 30/31
44.0 76.0 24.0 32.0 70.0 96.8
7/27 20/27 10/27 6/27 35/52 47/52
25.9 74.1 37.0 22.2 67.3 90.4
18/52 39/52 16/52 14/52 56/82 77/83
34.6 75.0 30.8 26.9 68.3 92.8
15/31
48.4
15/52
28.8
30/83
36.1
26/31 25/31 29/31 8/31 5/31 9/31 6/31
83.9 80.6 93.5 25.8 16.1 29.0 19.4
50/52 42/52 39/52 20/52 10/52 12/52 8/52
96.2 80.8 75.0 38.5 19.2 23.1 15.4
76/83 67/83 68/83 28/83 15/83 21/83 14/83
91.6 80.7 81.9 33.7 18.1 25.3 16.9
Risk Factors for Lyme Disease
TABLE 2.
1363
Further characteristics of study subjects,* Ukiah area, California, 1988-1989 Characteristic
Age (years) Years of residence Work and leisure (hours/ week outdoors) Winter Spring Summer Fall
Total Warm weather clothing index Clothing index x no. of hours outdoors Winter Spring Summer Fall Total No. of tick bites in last 2 years
Lyme disease probable
Lyme disease unlikely (n = 52)
Total (n = 83)
36.3 ±12.2 9.7 ± 4.5
29.6 ±14.9 11.0 ±3.7
32.1 ± 14.3 10.5 ±4.0
21.3 ±19.6 31.7 ±21.3 35.8 ± 23.8 30.2 ± 32.9 30.2 ± 20.6 3.2 ± 1.1
13.8 ± 14.1 24.0 ±17.7 36.7 ± 27.4 21.1 ±17.7 23.9 ±17.4 3.1 ± 0.9
16.5 ±16.5 26.8 ±19.3 36.3 ± 26.0 24.4 ±19.4 26.1 ±18.7 3.1 ± 0.9
69.4 ± 60.8 105.9 ±80.2 125.2 ±91.5 101.5 ±79.3 102.3 ±74.4
41.5 ±46.5 72.8 ± 62.8 109.2 ±80.1 62.6 ± 58.5 71.5 ±55.8
51.1 ±53.1 84.5 ± 70.7 114.8 ±84.0 76.3 ± 68.6 82.1 ± 64.0
3.8 ± 3.3
3.0 ± 3.8
3.3 ± 3.7
" All characteristics are expressed as the mean ± the standard deviation
blouse. Of these variables, only the warm weather clothing index weighted by time spent outdoors revealed differences that approached statistical significance (p = 0.07). As these differences were generally exceeded by those of the index that weighted pants or dresses doubly, only the latter were retained for further analyses. Subjects with probable Lyme disease were somewhat older and included a higher proportion of females as compared with subjects without Lyme disease. On average, all subjects had resided at the ranch for approximately 10 years, and individuals without Lyme disease had resided there slightly longer than individuals with probable Lyme disease. All subjects reported spending an average of approximately 36 hours per week outdoors pursuing work and leisure activities in summer. They also reported spending 1012 fewer hours per week outdoors in spring and fall than in the summer and 20 fewer hours outdoors during the winter (table 2). Subjects with probable Lyme disease spent approximately 8-9 hours more per week outdoors in winter, spring, and fall than subjects without Lyme disease. The same amount of clothing was worn by subjects with and without probable Lyme
disease, as indicated by the warm weather clothing index (table 2). The product of the clothing index and hours spent outdoors may be interpreted as an index of exposure. That is, the clothing index is scaled (weighted) such that the value of the index is inversely proportional to the amount of clothing worn during warm weather. The mean values for this composite measure of exposure indicate substantial differences between the two groups (table 2). For all seasons, subjects with probable Lyme disease had higher average values for the composite indicators than subjects without Lyme disease, with the largest differences occurring in spring and fall. Subjects in both groups reported relatively few tick bites during the previous 2 years (table 2). Persons with probable Lyme disease reported an average of 3.8 bites, compared with 3.0 bites for those without Lyme disease (p > 0.05). Subjects with probable Lyme disease reported tick bites more often in fall and winter and less frequently in summer as compared with subjects without Lyme disease (table 1). Tick bites were commonly reported by both groups in spring. The incidence of reported flea bites was approximately 70 percent in both groups, and that for mosquito bites exceeded 90
1364
Laneetal.
percent for each group (table I). Subjects with probable Lyme disease reported bites by biting flies much more frequently (48.4 percent) than subjects without Lyme disease (28.8 percent). Hiking and sunbathing were reported somewhat more frequently, and woodcutting somewhat less frequently, by subjects without Lyme disease (table 1). Similar percentages of each group reported participating in gardening, bird watching, jogging, and horseback riding. Analytic epidemiology
The prevalence of seropositivity for B. burgdorferi was significantly (p = 0.017) higher in subjects with probable Lyme disease (12 of 31, 38.7 percent) than in subjects without Lyme disease (7 of 52, 13.5 percent). The overall agreement between serodiagnosis and clinical determination of probable Lyme disease was 68.7 percent (12 seropositive subjects with Lyme disease plus 45 seronegative subjects without Lyme disease divided by a total of 83 subjects), and the value for the kappa statistic was 0.27 (standard error = 0.10, Z = 2.65). Crude odds ratios and 95 percent confidence intervals for probable Lyme disease are summarized for several potential risk factors in table 3. Two factors, female versus male sex and bites by biting flies each had odds ratios of 2.3, which approached statistical significance at the 5 percent level (p = 0.068 and 0.076, respectively). Subjects who cut wood were nearly five times more likely to have probable Lyme disease than other subjects (p = 0.048). These and several other variables were evaluated in logistic regression analyses to explore their association with' probable Lyme disease. All variables specified in tables 1-3 were included in these analyses. Three factors were found to be significantly associated with probable Lyme disease among all single-predictor models. These factors were hours per week outdoors in fall multiplied by a clothing index (p = 0.033), woodcutting (p = 0.048), and age (p = 0.043) with associated logistic coefficients of 1.008, 4.83, and 1.037, respectively. Of
TABLE 3. Crude odds ratios for probable Lyme disease associated with selected characteristics of study subjects, Ukiah area, California, 1988-1989 Variable Females versus males Bitten by Biting flies (unspecified) Fleas Mosquitoes Ticks Winter Spring Summer Fall Outdoor activities (woodcutting)
95% Odds ratio confidence interval 2.3
0.94-5.81
2.3 1.1 3.2
0.92-5.83 0.43-3.00 0.36-28.7
2.2 1.1 0.5 1.6 4.8
0.70-7.22 0.31-3.90 0.16-1.79 0.48-5.67 1.01-23.10
these, only the coefficient for woodcutting can be interpreted as an odds ratio. For all hours spent outdoors multiplied by a clothing index, there is an increase in logistic risk of 0.008 units per unit increase in the indicator. Similarly, an increase of 0.037 in logistic risk was observed for each additional year of increase in age. None of these models could be improved with the inclusion of other risk factors or covariables from among all of the remaining variables. DISCUSSION
Several researchers have expressed serious reservations about the lack of standardization and reliability of the assays (enzymelinked immunosorbent, IFA) that are being used routinely for the serodiagnosis of Lyme disease (15-18). Therefore, we thought it would be prudent to compare and validate the IFAs performed in three of our laboratories before linking the serologic and entry questionnaire data. In so doing, we hoped to increase the reliability of our collective judgment about the serologic status of each subject and, consequently, to enhance the sensitivity of our epidemiologic analysis for identifying potential risk factors for Lyme disease in the study population. The interlaboratory agreement (88-93 percent) for sera obtained at entry was found to be statistically highly significant, which was attributed to the fact that the participating laboratories adhered to a uniform standard for test specificity (11).
Risk Factors for Lyme Disease
Seroprevalence
In 1988, the range of the seropositivity rate at entry was 15-20 percent among the three laboratories. One year later, blood samples were again drawn from 81 percent of the current residents who volunteered at entry. All of these were retested for antibodies to B. burgdorferi in one of our laboratories (R. S. L.), and most of them were also retested in another (E. T. L.). The seropositivity rate in both laboratories was 24 percent, and the interlaboratory agreement was significant again (96 percent, n — 76). The somewhat higher seropositivity rate I year after entry was not unexpected, since the entry cohort (n = 119 subjects) included 20 former residents, only one of whom was determined to be both seropositive and immunoblot-positive. It is not known whether the demographic characteristics and serologic status of the study subjects are truly representative of the entire community, since only 66 percent of the estimated 150 current residents volunteered at entry and even fewer (53 percent) participated 1 year after entry. Moreover, demographic data for the nonparticipants are unavailable for comparison with those who volunteered. One might expect intuitively that healthy individuals would be less likely to volunteer for such a study than individuals who were experiencing symptoms of illness or aging phenomena (e.g., arthralgia or myalgia) suggestive of Lyme disease. If this is so, and even if all of the nonparticipants are assumed to be seronegative, a highly unlikely event, then the seroprevalence rate for all 150 residents would still be as high or higher than that determined for most other high-risk communities or occupational groups (e.g., outdoor employees) that have been studied in the northeastern United States (4-9, 19). Clinical correlates
The percentage of the 16 current seropositive residents who reported clinical manifestations consistent with Lyme disease were as follows: erythema migrans, 37.5 percent; arthritis/arthralgia, 68.8 percent; cardiac ab-
1365
normalities, 6.3 percent; and neurologic involvement, 37.5 percent (11). In another small series (n = 10) of Lyme disease patients from northern California (10), 100 percent reported erythema migrans; 60 percent reported arthritis/arthralgia; 20 percent reported cardiac abnormalities; and 60 percent reported neurologic symptoms. The difference in the prevalence of erythema migrans among the subjects in these two studies may reflect chance variation due to small sample sizes or host-related factors such as the mean age of the subjects, which was 17 years less in the present study (i.e., 32 vs. 49 years). In studies done in the northeastern United States, 5-100 percent of seropositive subjects reported a history of Lyme disease (4-7, 9). Such striking differences in the ratio of apparent-to-inapparent infections between regions may be due to a lack of standardized testing or to regional differences in the pathogenicity, virulence, or both of B. burgdorferi strains. In California, considerable antigen ic heterogeneity has been observed among the 30 wild-rodent or tick-derived strains of B. burgdorferi that have been characterized so far (20-23). In New York, a tick-derived variant of B. burgdorferi was shown to be infectious but nonarthritogenic in laboratory rats and mice (24). This nonarthritogenic strain, like a number of Californian strains (22, 23), possesses somewhat heavier outer surface proteins A and B than the B3l-type strain, plus a prominent protein with a low molecular weight of about 23,500. Cumulative frequency and incidence
Twenty-four of 119 serum specimens (20 percent) tested at entry were seropositive (11); of these, 23 were from the 99 current residents (23 percent) and only one came from the group of 20 former residents (5 percent). This marked difference in seropositivity rates between current and former residents may be related to the duration of their residency (10.5 vs. approximately 7.3 years) and to the amount of potential exposure to vector ticks. A year later, one of 59 subjects (1.7 percent) who was found to be seronegative at entry had seroconverted. Thus, the
1366
Laneetal.
approximate cumulative frequency of serologic infection with B. burgdorferi in the resident population after 1 year of prospective study was at least 24 percent (i.e., 24 of 99). However, the cumulative frequency of Lyme disease may be much higher, since 19 percent fewer residents were retested a year later (i.e., 80 vs. 99), 31 of 83 residents (37 percent) who were examined physically were given a diagnosis of probable Lyme disease, and not all patients with the disease had detectable antibody titers. The cumulative frequency of Lyme disease ranged between 4 and 35 percent in the northeastern United States, with annual or seasonal incidence rates of < 1-10 percent in various studies (3-7, 25-27). The highest annual incidence of 10 percent was recorded over a 2-year period in the Ipswich area of coastal Massachusetts (7). Except for the Ipswich study, the cumulative frequency and annual incidence of Lyme disease in the Ukiah study area are comparable to or higher than the cumulative frequencies/ incidences reported from the Northeast. Risk factors
The high seropositivity rate and cumulative frequency of clinical Lyme disease in the subject population is perplexing in view of the generally low (1 -6 percent) spirochetal infection rates in Ixodes pacificus in California (28). In fact, only nine of 684 adult /. pacificus (1.3 percent) swept from vegetation in the study area between 1988 and 1990 were found to contain spirochetes (R. S. Lane and S. A. Manweiler, unpublished data). In contrast, spirochetal infection rates in Ixodes dammini average roughly 25 percent in nymphal and 50 percent in adult ticks in the Northeast (28). Furthermore, the average number of tick bites reported over a 2-year period by subjects with probable Lyme disease (3.8) versus subjects deemed free of Lyme disease (3.0) was low for both groups (table 2). Regional differences in the epidemiology of Lyme disease may explain this paradox. In the Northeast, human exposure to /. dammini ticks occurs in peridomestic environ-
ments, such as on or near well-maintained lawns in residential areas (1, 7, 29-32). Thus, it is not surprising that occupational or leisure-related tick exposure, presence of ticks near residences, time spent outdoors in various activities, and the abundance of deer have been implicated as actual or potential risk factors in this region (2-4, 6, 7, 9, 10, 27, 30-33). In the Pacific states, especially California, where most human cases have been reported, the increased age distribution (i.e., with an apparent peak of reported cases in the 25- to 44-year-old age group) may reflect the typical pattern for tick-borne diseases in the United States that is associated with outdoor occupational or recreational exposure (1). Our findings seem to confirm this observation, since the residents we studied built their homes in what is essentially a wildlands area/nature preserve (i.e., hunting is largely prohibited). Domiciles are surrounded by natural vegetation that is occupied by ticks and an abundance of wildlife, many of which serve as hosts of ticks and of B. burgdorferi (R. S. Lane & S. A. Manweiler, unpublished data). The two most abundant human-biting ticks, Dermacenlor occidentalis and /. pacificus, are infected naturally with B. burgdorferi in northern California (28). The dusky-footed wood rat, Neotomafuscipes, which lives in close proximity to human dwellings, has been implicated as a reservoir of B. burgdorferi infection (23). Nearly equal numbers of adult residents work in the community or elsewhere. Although time spent working outdoors was not identified as a risk factor, about one-half of subjects working in the community are engaged in occupational activities that may place them at risk for Lyme disease (e.g., caretaker, nurseryman, property manager, and handiperson). In contrast, over 90 percent of residents who work elsewhere are employed in occupations that would not place them at much, if any, risk of infection (e.g., artisan, carpenter, electrician, reporter, teacher, waitress). Seropositive subjects resided in the community about 2 years longer, were bitten by
Risk Factors for Lyme Disease
unspecified biting flies more frequently, and hiked less often than seronegative individuals. In Massachusetts, length of residency and location of homes were not associated with higher risk for Lyme disease on Great Island, although increased age was (5), whereas risk was correlated with proximity of residence to a nature preserve in the Ipswich area (7). The latter finding parallels and substantiates our discovery of heightened risk among residents of a natural area. The bites of mosquitoes, deer flies, and horse flies have been incriminated as potential risk factors for the acquisition of Lyme disease in Connecticut and New York (2, 26, 34, 35). Our finding of a significant association between bites by bloodsucking flies and seropositivity suggests that biting flies should be evaluated for infection with spirochetes in the Far West. However, this association may represent an indirect measure of outdoor exposure, or it may be due merely to chance as a result of the large number of statistical tests we performed. The crude (univariate) analyses disclosed that cutting wood was significantly associated with the risk for Lyme disease. Two other factors, female versus male sex and bites by biting flies, approached statistical significance. However, logistic regression analyses revealed that the only dichotomous factor that could be interpreted as a measure of risk based on its odds ratio was woodcutting. Although the season(s) during which woodcutting was usually performed is not known, the combined activity periods of /. pacificus nymphs and adults span about 9 months (approximately mid-November to mid-August) in the study region. Increased risk due to cutting wood may simply reflect the amount of time spent outdoors in favorable tick habitats and, therefore, the likelihood of tick exposure. In future studies, various property maintenance activities (e.g., cutting brush) should also be evaluated as potential risk factors for Lyme disease. We had hypothesized that time spent outdoors and pet ownership would be associated with elevated risk for infection with B. burgdorferi or contraction of Lyme disease. Although time spent outdoors was not iden-
1367
tified as a significant risk factor, subjects with probable Lyme disease spent approximately 8-9 hours more per week outdoors than subjects without Lyme disease during the peak activity periods of nymphal and adult /. pacificus (fall, winter, and spring); they also reported more tick bites than subjects who. did not have the disease (table 2). Pet ownership was also not associated with increased risk which, with one exception (2), is consistent with epidemiologic findings from the northeastern United States (4, 5, 9, 19). Like Eng et al. (19), however, we found that dogs are at much greater risk for infection with B. burgdorferi because they are exposed to considerably more vector ticks than are people (J. E. Madigan & R. S. Lane, unpublished data).
REFERENCES 1. Miller GL. Craven RB, Bailey RE, et al. The epidemiology of Lyme disease in the United States 1987-1988. Lab Med 199O;21:285-9. 2. Steere AC. Broderick TF, Malawista SE. Erythema chronicum migrans and Lyme arthritis: epidemiologic evidence for a tick vector. Am J Epidemiol 1978; 108:312-21. 3. Bowen GS, Schulze TL. Hayne C, et al. A focus of Lyme disease in Monmouth County, New Jersey. Am J Epidemiol 1984; 120:387-94. 4. Hanrahan JP, Benach JL, Coleman JL, et al. Incidence and cumulative frequency of endemic Lyme disease in a community. J Infect Dis 1984:150: 489-96. 5. Steere AC, Taylor E, Wilson ML, et al. Longitudinal assessment of the clinical and epidemiological features of Lyme disease in a defined population. J Infect Dis 1986:154:295-300. 6. Smith PF, Benach JL, White DJ, et al. Occupational risk of Lyme disease in endemic areas of New York State. Ann NY Acad Sci 1988;539: 289-301. 7. Lastavica CC, Wilson ML, Berardi VP, et al. Rapid emergence of a focal epidemic of Lyme disease in coastal Massachusetts. N Engl J Med 1989:320: 133-7. 8. Goldstein MD, Schwartz BS, Friedmann C, et al. Lyme disease in New Jersey outdoor workers: a statewide survey of seroprevalence and tick exposure. Am J Pub Health 1990:80:1225-9. 9. Schwartz BS, Goldstein MD. Lyme disease in outdoor workers: risk factors, preventive measures, and tick removal methods. Am J Epidemiol 1990; 131:877-85. 10. Lane RS, Lavoie PE. Lyme borreliosis in California: acarological, clinical, and epidemiological studies. Ann NY Acad Sci 1988;539:192-203. 11. Lane RS, Lennette ET, Madigan JE. Interlabora-
1368 Laneetal.
12.
13. 14. 15. 16. 17. 18.
19. 20.
21.
22. 23. 24.
tory and intralaboratory comparisons of indirect immunofluorescence assays for serodiagnosis of Lyme disease. J Clin Microbiol 1990:28:1774-9. Breslow NE, Day NE, eds. Statistical methods in cancer research. Vol. I. The analysis of case-control studies. (IARC scientific publication no. 32). Lyon, France: International Agency for Research on Cancer, 1980. Zar JH. Biostatistical analysis. Englewood Cliffs, NJ: Prentice-Hall, 1984. Fleiss JL. Statistical methods for rates and proportions. New York, NY: John Wiley and Sons. 1973. Osterholm MT, Forfang JC, White K.E, et al. Lyme disease in Minnesota: epidemiologic and serologic findings. Yale J Biol Med 1984:57:677-83. Wilkinson HW, Russell H, Sampson JS. Caveats on using nonstandardized serologic tests for Lyme disease. (Letter). J Clin Microbiol 1985:21:291. Hedberg CW, Osterholm MT. MacDonald KL, et al. An interlaboratory study of antibody to Borrelia burgdorferi.i Infect Dis 1987; 155:1325-7. Schwartz BS. Goldstein MD, Ribeiro JMC. et al. Antibody testing in Lyme disease: a comparison of results in four laboratories. JAMA 1989:262: 3431-4. Eng TR, Wilson ML. Spielman A, et al. Greater risk of Borrelia burgclorferi infection in dogs than in people. J Infect Dis 1988:158:1410-11. Burgdorfer W, Lane RS, Barbour AG, et al. The western black-legged tick, lxodes pacifwus: a vector of Borrelia burgdorfen. Am J Trop Med Hyg 1985; 34:925-30. Bissett ML, Hill W. Characterization of Borrelia burgdorferi strains isolated from lxodes pacifwus ticks in California. J Clin Microbiol 1987:25: 2296-2301. Lane RS, Pascocello JA. Antigenic characteristics of Borrelia burgdorferi isolates from ixodid ticks in California. J Clin Microbiol 1989:27:2344-9. Lane RS, Brown RN. Wood rats and kangaroo rats: potential reservoirs of the Lyme disease spirochete in California. J Med Entomol 1991:28:299-302. Anderson JF, Barthold SW, Magnarelli LA. Infectious but nonpathogenic isolate of Borrelia burg-
dorferi. J Clin Microbiol 1990;28:2693-9. 25. Steere AC, Grodzicki RL, Kornblatt AN, et al. The spirochetal etiology of Lyme disease. N Engl J Med 1983;308:733-40. 26. Benach JL, Coleman JL. Clinical and geographic characteristics of Lyme disease in New York. Zentralbl Bakteriol Mikrobiol Hyg A 1986:263: 477-82. 27. Petersen LR, Sweeney AH. Checko PJ, et al. Epidemiological and clinical features of 1,149 persons with Lyme disease identified by laboratory-based surveillance in Connecticut. Yale J Biol Med 1989; 62:253-62. 28. Lane RS, Piesman J, Burgdorfer W. Lyme borreliosis: relation of its causative agent to its vectors and hosts in North America and Europe. Annu Rev Entomol 1991;36:587-609. 29. Williams CL, Curran AS, Lee AC, et al. Lyme disease: epidemiologic characteristics of an outbreak in Westchester County, NY. Am J Pub Health 1986;76:62-5. 30. Falco RC, Fish D. Prevalence of lxodes dammini near the homes of Lyme disease patients in Westchester County, New York. Am J Epidemiol 1988; 127:826-30. 31. Falco RC, Fish D. A survey of tick bites acquired in a Lyme disease endemic area in southern New York State. Ann NY Acad Sci 1988:539:456-7. 32. Maupin GO, Fish D, Zultowsky J, et al. Landscape ecology of Lyme disease in a residential area of Westchester County, New York. Am J Epidemiol 1991:133:1105-13. 33. Bowen GS, Griffin M, Hayne C. et al. Clinical manifestations and descriptive epidemiology of Lyme disease in New Jersey, 1978 to 1982. JAMA 1984:251:2236-40. 34. Magnarelli LA, Anderson JF, Barbour AG. The etiologic agent of Lyme disease in deer flies, horse flies, and mosquitoes. J Infect Dis 1986; 154: 355-8. 35. Magnarelli LA, Anderson JF. Ticks and biting insects infected with the etiologic agent of Lyme disease, Borrelia burgdorferi. J Clin Microbiol 1988:26:1482-6.
