Human Ehrlichiosis in the United States, 1985 to 1990

Methods

Since the recognition of human ehrlichiosis in 1986, the Centers for Disease Control and Prevention (CDC) has offered serologic testing for the disease. In 1988, CDC and several state health departments stimulated surveillance by publishing reports in their bulletins and sending information to physicians asking them to collect serum samples from potential case-patients. In addition, several large private laboratories that routinely did serologic testing for Rocky Mountain spotted fever were asked to send all acute and convalescent serum pairs submitted for Rocky Mountain spotted fever to the CDC [8]. Serum samples from patients suspected of having Rocky Mountain spotted fever or other rickettsial diseases were tested for antibodies to E. canis. About 50% of the patients in our series were previously reported [1, 2, 4-8].

Initially, the indirect fluorescence antibody test for antibodies to E. canis in canine serum was adapted for human serum [2]. Ehrlichia canis was used as the antigen for routine diagnostic tests until 1990, when the putative causative agent of human ehrlichiosis, E. chaffeensis, was isolated. The techniques used by the CDC and the Oklahoma State Department of Health are described elsewhere [13]. All specimens sent to the CDC were also tested for antibodies to Rickettsia rickettsii, R. typhi, and Coxiella burnetii and those sent to the Oklahoma State Department of Health were tested for R. rickettsii. The case definition for ehrlichiosis required a minimum antibody titer of 64, in addition to a fourfold change in antibody titer to E. canis or E. chaffeensis from the acute to the convalescent phase.

Patients with ehrlichiosis were interviewed with a standardized questionnaire; information about presumptive clinical diagnosis, clinical and serial laboratory findings, and treatment was abstracted from medical records. When a response to a specific question was not available, that patient was excluded from analysis for that specific variable. Because almost all case-patients had laboratory tests done at different laboratories and normal values for many of these laboratories were not available, published normal laboratory values were used to determine abnormal values. Age- and state-specific rates for ehrlichiosis were calculated using mean population estimates for the 5-year period from 1985 to 1990.

The relative clinical effectiveness of various antibiotics was evaluated by classifying the case-patients who were hospitalized according to the type of antibiotic they received and by analyzing the time (days) from when the antibiotic was first administered to the first day of defervescence (temperature, ≤ 37 °C for at least 24 hours), the total duration of hospitalization, and the time until complete recovery (return to normal activities). In univariate analyses, only case-patients treated with tetracycline, chloramphenicol, or another antibiotic (alone) for at least 48 hours were included in the latter analysis. Patients whose eventual treatment included tetracycline or chloramphenicol and another antibiotic were classified on the basis of their initial treatment as long as they were initially treated with either tetracycline, chloramphenicol, or another antibiotic alone for at least 48 hours.

The epidemiologic and clinical characteristics of case-patients whose initial treatment consisted of tetracycline were compared with those treated with chloramphenicol and with those treated with one or more other antibiotics (without tetracycline or chloramphenicol). To study possible risk factors for illness severity, we divided case-patients into the following groups: those who did not require hospitalization, those who were hospitalized but did not develop complications, those who were hospitalized and had one or more serious complications, and those who died.

The Wilcoxon rank-sum test was used to compare group medians of nonparametrically distributed data, and the Fisher exact test was used to compare differences in proportions when variables were dichotomous. A multiple logistic regression was done using PC-SAS SAS Institute Inc., Cary, North Carolina [14]. Variables that were biologically plausible and were associated with the outcome (complications or death) that had a P value≤ 0.20 were allowed to enter into the model. These variables were as follows: the time (in 4-day intervals) between the onset of illness and the day of their illness that they initially saw a physician, age group (<50 years, 50 to 59, >60 years of age, and whether treatment with tetracycline or chloramphenicol was begun during the first 3 days of illness, between days 4 to 7, or after day 7 of illness. Because treatment with antibiotics other than tetracycline or chloramphenicol was associated with an unfavorable outcome, a variable was created (classifying patients by whether they were treated with these other antibiotics [within the first 3 days of illness, between days 4 to 7, or after day 7]) to control for confounding. A backward method of elimination was used.

Results

Epidemiologic Findings

Between 1985 and 1990, paired serum specimens from 237 patients met the serologic case definition. One case-patient was included on the basis of a serologic response to E. chaffeensis; the other 236 had a fourfold increase or decrease to E. canis. Six of these patients were subsequently tested for E. chaffeensis and all had an increasing or decreasing titer to this antigen. Diagnoses were made for a median of 41 and a mean of 46.4 case-patients each year (excluding the 4 diagnoses made retrospectively from sera collected in 1985).

Most (68.2%) case-patients developed their first symptoms in May, June, or July; 93.5% had onset in the 6-month period between April and September. Patients were exposed to ticks and resided in 139 counties in 21 states. The state with the most cases was Oklahoma (57 patients; rate, 0.36 per 100 000 population per year); other states with many cases included Missouri (47 patients; 0.18 per 100 000) and Arkansas (18 patients; 0.15 per 100 000). The county with the most cases was Cherokee County, Oklahoma (10 patients; 6.25 per 100 000). Other counties where 4 or more case-patients resided or reported tick exposure included Muskogee, Oklahoma (6 patients; 1.75 per 100 000); Crawford, Arkansas (4 patients; 1.94 per 100 000); Sebastian, Arkansas (4 patients; 0.81 per 100 000); and Glynn, Georgia (4 patients; 1.31 per 100 000). Information about location of residence was available for 175 case-patients; these residences were described as “rural” for 116 (66.3%), “suburban” for 46 (26.3%), and “urban” for 13 (7.4%).

The median age of case-patients was 44 years; most (74.5%) were men. Ehrlichiosis was particularly common in the 60-years-and-older age group (26.6% of total). Age-specific incidence rates increased steadily with increasing age and were highest for persons 60 to 69 years old (Figure 1). The overall sex-specific rate for men (27 per 100 000 000) exceeded that for women (9 per 100 000 000), and rates for men exceeded those for women in all age groups.

Tick bites or attachments were reported by 67.9% of the case-patients; another 15.5% reported being in a tick-infested area but did not recall a tick bite or attachment. A single exposure date was reported by 47.6% of the patients; among these persons, the median incubation period was 9 days (Table 1).

Clinical Findings

Medical care was first sought a median of 4 days (range, 0 to 35 days) after the onset of symptoms. Three patients sought medical care for tick bite or other conditions before the onset of symptoms. Information about the clinical diagnosis at the time of the initial visit to a physician was available for 185 of the case-patients. Although a rickettsial infection (Rocky Mountain spotted fever, ehrlichiosis, or rickettsial illness of unspecified cause) was the most common initial diagnosis, only 22.2% of patients were initially suspected of having a rickettsial illness. Other initial diagnoses included upper respiratory infection, sepsis, influenza, pharyngitis, gastroenteritis, and prostatitis.

Fever, malaise, and headache were the most common presenting symptoms but no symptom was present in more than 75% of the case-patients on the first day (Table 2). Other symptoms usually developing early in the illness were myalgia, rigor, arthralgia, and nausea. Rash ultimately was noted in 36.2% of patients, initially a median of 5 days (range, 0 to 21 days) after the onset of symptoms. The rash was usually described as macular or papular or both (62.3%) but was sometimes petechial (19.5%) or erythematous (18.2%). The areas most commonly affected were the trunk (54.5% of patients with a rash), legs (49.4%), arms (41.6%), and face (15.6%). Diarrhea, lymphadenopathy, and confusion were also noted but only toward the end of the first week of illness. Seven patients had pulmonary infiltrates consistent with an acute process.

Many (61.6%) case-patients were hospitalized. Hospitalization occurred a median of 7 days after the onset of symptoms, and the median duration of hospitalization was 5 days (Table 1). The median duration of illness (from onset of symptoms to complete recovery) was 23 days. Men were as likely to be hospitalized as women, but hospitalized patients were older (median, 50.5 years) than those who were not hospitalized (median, 37.6 years; P < 0.001).

Three (1.3%) case-patients died, and 28 others (15.9%) had 1 or more serious manifestations or clinical complications. Renal dysfunction or failure was noted in 16 patients, laboratory evidence of disseminated intravascular coagulation in 15 (4 of whom had clinical bleeding [2 from the gastrointestinal tract and 2 from multiple sites]), cardiomegaly in 8, seizures in 4, and coma in 4. Except for coma (which occurred a median of 19 days [range, 13 to 24 days] after onset of symptoms) and death (which occurred a median of 69 days [range, 33 to 82 days] after onset), complications tended to develop at the end of the first or the beginning of the second week of illness. Case-patients who died were older (median, 67 years) than those who had other serious symptoms (median, 60 years), those who were hospitalized but did not have complications (median, 44.5 years), and those who were not hospitalized (median, 40 years) (Table 3).

Antibiotic Treatment, Clinical Course, and Complications

Ninety-nine case-patients received treatment with antibiotics on an outpatient basis. Forty-nine (49.5%) received only tetracycline, 2 (2.0%) received only chloramphenicol, 38 (38.4%) received another antibiotic (without tetracycline or chloramphenicol), and 10 (10.1%) received a combination. Most of the patients who were treated with tetracycline alone received 500 mg of tetracycline hydrochloride, four times per day (25 patients); the others received doxycycline (12 patients), minocycline (1 patient), or an unspecified form of tetracycline (11 patients). Only 3 (6.1%) of the 49 patients who received only tetracycline as outpatients were hospitalized compared with 35 (92.1%) of the case-patients treated with antibiotics other than tetracycline or chloramphenicol (P < 0.001). Case-patients with a history of tick bite or exposure were no more likely to be treated with tetracycline or chloramphenicol than were other patients.

Hospitalized case-patients who were treated with antibiotics were divided into 3 groups on the basis of antibiotic treatment during the first 48 hours of hospitalization (Table 4). Of the 43 patients treated with tetracycline, 20 received tetracycline hydrochloride (14 orally and 6 intravenously) and 17 received doxycycline (9 orally and 8 intravenously); the type of tetracycline or route was not specified in the remaining 6 patients. Although hospitalized case-patients treated with chloramphenicol were statistically younger than those treated with tetracycline or those treated with other antibiotics, the interval from onset to initial physician visit and from onset to initiation of treatment with any antibiotic was similar for the 3 groups. Few differences were found in the response of patients treated with tetracycline compared with those treated with chloramphenicol, although defervescence occurred about 1 day sooner in persons in the former group (median, 2 days) than in those in the latter group (median, 3 days; P = 0.011).

Hospitalized patients treated with tetracycline or chloramphenicol defervesced much more rapidly (median, 2 days for tetracycline and 3 days for chloramphenicol) than did case-patients treated with other antibiotics (median, 7 days; P ≤ 0.001 compared with tetracycline and P = 0.009 compared with chloramphenicol). Hospitalized patients treated with tetracycline had shorter hospitalizations (P = 0.001) than did those treated with other antibiotics (except for chloramphenicol), and patients treated with either tetracycline or chloramphenicol recovered more quickly after initiation of treatment than did patients treated with other antibiotics (Table 4).

In univariate analyses, case-patients with more severe illness were older, saw a physician later in the course of illness, and were started on treatment later in the course of their illness than did those with less severe illness (see Table 3). We divided patients into four groups based on illness severity (not hospitalized, hospitalized without complications, hospitalized with one or more serious nonfatal complications, and death). The median length of hospitalization increased with increasing illness severity as follows: 5 days (range, 1 to 58 days) in patients with uncomplicated illness, 11.5 days (range, 2 to 79 days) in patients with complications, and 60 days (range, 26 to 70 days) in the 3 patients who died. In univariate analyses, severity was correlated with age, time to doctor visit, time to initiation of antibiotics, and time to initiation of tetracycline and chloramphenicol (data not shown).

To determine the independent effect estimates of these variables, a multiple logistic regression analysis was done. The outcome variable was the presence or absence of complications or death. Age, time to the first physician visit, time to the initiation of therapy with tetracycline or chloramphenicol, and time to initiation of therapy with an antibiotic other than tetracycline or chloramphenicol were entered into the model. We used 193 patients to develop the model; 91 were treated only with tetracycline or chloramphenicol, 15 with only another antibiotic, and 87 with tetracycline or chloramphenicol and another antibiotic. Age 60 years or more and delay in treatment with tetracycline or chloramphenicol until 8 or more days after the onset of illness were the only variables that were statistically associated with an increased risk for complications or death (Table 5).

Hematologic and Chemistry Laboratory Results

These laboratory results are shown in Figure 2. During the first few days of illness, most laboratory values were normal or near normal. By the third day of illness, the mean leukocyte count decreased to 46.7% of the baseline value (from 8.0 × 10⁹/L to 4.3 × 10⁹/L). Leukocyte values reached a nadir during the next 3 days and then gradually increased during the following 2 weeks of illness. Of the 196 case-patients whose leukocyte counts were determined at least once during illness, 117 (59.7%) had abnormal leukocyte counts on at least 1 occasion. Most leukocyte elements tended to decrease simultaneously, with the largest decrease occurring in the total lymphocyte count (58.6% decrease from day 1 to day 3). Similar decreases occurred in segmented neutrophils and band neutrophils during this period (Figure 2, Panel A).

A relative and absolute lymphocytosis developed at the beginning of the second week of illness, with the mean absolute lymphocyte count reaching 2.9 × 10⁹ on day 13 of illness. An atypical lymphocytosis with absolute atypical lymphocyte counts exceeding 1.0 × 10⁹/L was noted in 10 case-patients. The platelet count followed a similar time course, decreasing by 37.1% (from 204 × 10⁹/L to 128 × 10⁹/L) between day 1 and day 3 (Figure 2, panel B); 68.0% of case-patients had an abnormal platelet count at some time in their illness. The time-course of the hematocrit followed a somewhat different course than the leukocyte count and platelet count, decreasing gradually but steadily throughout the first few weeks of illness (Figure 2, panel C). Liver function tests generally increased rapidly during the first days of illness and then slowly returned to normal (Figure 2, panel D). Most case-patients tested had abnormal aspartate aminotransferase levels (128 of 149 patients, 85.9%) and alanine aminotransferase levels (82 of 102, 80.3%) at some time during their illness.

Discussion

Biases in case ascertainment influenced the make-up of the study cohort and must be taken into account when interpreting the epidemiologic and clinical findings of this study. Intensive surveillance for human ehrlichiosis in five states (Oklahoma, Missouri, Virginia, Georgia, and Texas) almost certainly increased the number of case-patients, increased the incidence rates, and increased the geographic distribution in these states [4, 6, 7]. Conversely, the paucity of case-patients in other states may reflect a lack of surveillance efforts rather than an actual paucity of patients. Case-ascertainment biases may actually have facilitated the study of patients most important to the clinician: those persons at higher risk for adverse outcomes. More seriously ill patients were probably over-represented because serologic testing, the basis for inclusion of case-patients in this study, is more likely to be done with such patients.

The clinical findings, laboratory findings, and complications reported in this study are probably most representative of those of seriously ill patients. This bias is particularly important because most human infections with E. chaffeensis are entirely asymptomatic or have symptoms that may not prompt medical consultation, treatment, or serologic testing [15, 16]. For example, active surveillance of young military recruits found that only 33% of infections were symptomatic and only a fraction of these case-patients were ill enough to seek medical care [16]. The absence of asymptomatic infections and the under-representation of mildly ill case-patients also influenced the findings about the incidence and the geographic distribution of the disease.

The lack of a nationwide surveillance program probably contributes to underestimation of the rate of infection, and the rates presented here probably reflect those of more serious illness. The apparent geographic dispersion may also be partially attributable to undetected, undiagnosed, or unreported cases. For the same reason, the high rates noted in a few small rural counties may also under-represent rates that are substantially higher and approach, if not exceed, those of Rocky Mountain spotted fever [4].

The incidence of human ehrlichiosis increased with age. Rocky Mountain spotted fever (a more common disease with similar seasonality, geographic distribution, and at least one common tick vector) has a peak incidence in children [17, 18]. Simultaneous laboratory-based surveillance for Rocky Mountain spotted fever and human ehrlichiosis supports the finding that patients with ehrlichiosis are older than those with Rocky Mountain spotted fever [4, 6]. One possible explanation for these findings is that elderly persons are more likely to be infected with E. chaffeensis than are younger persons, but a study that included asymptomatic infections does not support this interpretation [16]. An explanation more consistent with all findings is that older persons with E. chaffeensis infections are more likely to develop serious infections, to be hospitalized, have extensive diagnostic evaluation, and be detected in a laboratory-based surveillance system (such as the one that formed the basis for this study).

The severity of illness also increased with the interval between onset of illness and initiation of treatment with tetracycline or chloramphenicol. Similar findings have been reported for Rocky Mountain spotted fever and murine typhus [19-24]. However, unlike these other rickettsial diseases, patients with ehrlichiosis who died, as well as many with complications who recovered, had prolonged clinical courses, sometimes associated with superinfection, suggesting that factors other than acute infection may contribute to these adverse outcomes [1, 8]. In canine ehrlichiosis, delayed treatment is associated with a chronic infection [23]. This protracted illness may represent subacute or chronic phases, as is seen with dogs infected with E. canis[25].

A combination of epidemiologic and laboratory information may help differentiate acute infection from other tick-borne and non-tick-borne illness, but a high index of suspicion will probably remain necessary because of the nonspecific clinical presentation. A history of tick exposure should, of course, prompt consideration of a tick-borne disease. The decreasing leukocyte counts and platelet counts found in patients with ehrlichiosis help differentiate this infection from Rocky Mountain spotted fever and other infections [4, 6]. The mechanism for the pancytopenia of ehrlichiosis is not clear. A wide range of bone marrow abnormalities have been reported. Although bone marrow hypoplasia has been occasionally reported in patients with ehrlichiosis, it is uncommon and probably would not be associated with the rapid decreases seen in this study [2, 8, 9, 26, 27]. Sequestration or destruction of various blood elements (the hemophagocytic syndrome) may be a more probable mechanism of the cytopenia [27].

Differences in geographic distribution must primarily be relied on to differentiate ehrlichiosis from Colorado tick fever, another tick-borne disease characterized by an acute febrile illness and a rapidly decreasing leukocyte count [28, 29]. Although the geographic distribution of ehrlichiosis overlaps that of Lyme disease in the south central states and along the southern Atlantic seaboard, the erythema migrans rash of Lyme disease should distinguish it from ehrlichiosis; in addition, acute systemic illness, leukocytopenia, and thrombocytopenia are uncommon in Lyme disease [30]. However, our results highlight the importance of interpreting the hematologic data based on the number of days since onset of symptoms rather than days since hospitalization. Pancytopenia is an uncommon manifestation of bacterial infection and is usually found only in patients with widely disseminated infection [31].

Our findings about the effect of therapy with tetracycline and chloramphenicol on the clinical course of ehrlichial infections support the choice of one of these two drugs as initial therapy for human ehrlichiosis. The lower rate of hospitalization, rapid defervescence, and shorter duration of illness in patients treated with tetracycline (compared with those treated with other antibiotics) confirms the findings of others [4, 8]. Our findings about treatment with chloramphenicol also support the efficacy of this drug in the treatment, although only seven hospitalized patients were initially treated with chloramphenicol only and they were younger than case-patients treated with other antibiotics. Recent in vitro data [10] suggesting resistance of E. chaffeensis to chloramphenicol differs from our epidemiologic data. It is probably not necessary to distinguish ehrlichiosis from other tick-borne infections because, except for Colorado tick fever, all acute tick-borne diseases found in the United States are susceptible to tetracycline. Because nonhospitalized patients treated with tetracycline and hospitalized patients treated with tetracycline or chloramphenicol responded promptly, standard doses of tetracycline (500 mg four times daily) or doxycycline (100 mg twice daily for the first day then 100 mg daily) probably can be used to treat ehrlichiosis (orally in less severely ill patients and intravenously in more severely ill patients). Similar to other rickettsial diseases, chloramphenicol can be used when tetracycline is contraindicated.

The nonspecific clinical presentation and the varied clinical manifestations of human ehrlichiosis continue to hinder prompt diagnosis and initiation of proper treatment. Future studies of human ehrlichiosis should, therefore, also focus on identifying methods to prevent infection. Although a study in military personnel suggested that permethrin-impregnated clothing is an effective prevention measure, the practicality and utility of such measures in the civilian population must be established [16]. Although rapid diagnostic tests may increase the initiation of prompt treatment with effective antibiotics and decrease the severity of complications, the disease must first be suspected. Many case-patients in this study were not given treatment with tetracycline or chloramphenicol despite consultation and a history of tick bite. Almost 75% of the patients in this and all other studies had a history of tick bite. In a patient with an unexplained febrile illness, a careful interview with the patient or family may indicate a history of tick exposure. This information is useful in identifying persons with ehrlichiosis, and if followed by prompt initiation of treatment with tetracycline, may help decrease morbidity and prevent mortality. If sensitive rapid diagnostic tests that confirm ehrlichiosis are widely used in the future, morbidity and mortality may be further limited.