We enrolled 85 (77%) of 111 persons with previously treated HGE (patients) and 102 (45%) of 225 controls. Of the 111 patients, we were unable to contact 13 (12%); 8 (7%) refused participation; 4 (4%) had since died of causes unrelated to HE; and 1 (1%) did not complete the survey. Among the 225 controls, we were unable to contact 56 (25%); 61 (27%) declined to participate; and 6 (3%) surveys were incomplete.

There were 73 (86%) confirmed and 12 (14%) probable cases of HGE. Among confirmed cases, 32 (44%) were confirmed by serology alone; 13 (18%) by serology and polymerase chain reaction (PCR); 10 (14%) by PCR alone; 9 (12%) by serology, PCR, and presence of intracytoplasmic morulae; 6 (8%) by serology and presence of morulae; and 3 (4%) by PCR and morulae. Of the 12 probable cases, 9 (75%) had a single IFA serologic titer ≥64, and 3 (25%) had intracytoplasmic morulae identified in blood. The median age of patients was 58 years (range 6-88 years), and the median age of controls was 57 years (range 6-88 years); 66% of patients and 73% of controls were male. Data on coexisting conditions were available for all 102 controls and 39 (53%) of 73 confirmed cases. Comparison of the 39 cases and 102 controls demonstrated no significant difference in the prevalence of cancer (p=0.36), stroke (p=0.49), heart disease (p=0.89), or diabetes (p=0.65).

For HGE patients, the median interval from illness onset to telephone interview was 24 months (range 10-40 months). Twelve patients had onset of HGE in 1996, 35 in 1997, and 38 in 1998. The illness was severe enough in 32 (38%) patients to require hospitalization. All 85 enrolled patients received some form of antibiotic treatment for acute illness. Of the 59 patients for whom we have sufficient information, 53 (90%) were prescribed doxycycline for >7 days. The median interval from illness onset to initiation of treatment was 7 days. Nineteen (22%) enrolled patients had evidence of Lyme disease during the acute illness, either the characteristic rash, erythema migrans (14 patients), or seroconversion to Borrelia burgdorferi (7 patients). Six (8%) of 76 enrolled patients tested also seroconverted to Babesia microti after the acute illness.

Compared with controls, patients were more likely to report the following constitutional symptoms either continuously or repeatedly during the previous year: fevers, chills, sweats, and fatigue (Table 1). Patients had lower SF-36 health status scores for bodily pain (p=0.03) and health compared with 1 year earlier (p=0.02), but no differences existed for physical function, role limitations due to physical health, general health, or vitality (Table 2). When probable cases were excluded from analysis, confirmed patients were also more likely to report fevers (OR 4.1, 95% CI 1.0-15.9), chills (OR 3.8, 95% CI 1.3-11.4), sweats (OR 2.8, 95% CI 1.4-5.8), and fatigue (OR 1.7, 95% CI 1.0-3.1) during the previous year. Confirmed cases also had lower SF-36 scores for relative health (p=0.02), but no significant differences existed for bodily pain, physical function, role limitations due to physical health, general health, or vitality. When asked about their health status relative to before infection, 39% of patients believed their current health was “somewhat worse” or “much worse.”

The presence of continuous or recurrent constitutional symptoms and the duration of acute illness were not correlated. Patients who were hospitalized or who started antibiotics more than 14 days after onset of illness were no more likely to experience recurrent symptoms than those who received antibiotic treatment within 14 days. Similarly, patients with 1) a preexisting chronic illness, 2) intragranulocytic morulae in an acute-phase blood smear, 3) laboratory evidence of concurrent Lyme disease or babesiosis, 4) anemia, or 5) a high acute- or convalescent-phase reciprocal HGE IFA antibody titer (>512) were no more likely than the other patients to experience one or more of the recurrent or continuous symptoms.

Serum specimens were submitted for serologic testing by 70 (82%) of 85 patients. The HGE IFA antibody titer remained elevated (>1:64) in one (1.4%) of 70 specimens tested. This patient had a very high titer after acute infection (1:2,048), which remained elevated (1:256) 1 year later. He experienced continuous or recurrent fatigue, vomiting, and headaches. Two (2.9%) of 69 patients had elevated (>100 U/L) serum AST levels; one complained of continuous or recurrent chills, sweats, and fatigue.

Our results demonstrate that some patients with treated HGE may experience more fevers, chills, sweats, and fatigue than controls 1-3 years after onset of illness. Some patients also experience more bodily pain and have a poorer perception of their health compared with 1 year ago than controls, but they do not have any functional disability. Except for bodily pain, these findings persisted when only confirmed cases were included in the analysis. We found no serologic evidence to suggest the occurrence of persistent ehrlichial infection. These symptoms may therefore be attributed to a postinfectious syndrome rather than persistent or recurrent infection.

Few previous studies have evaluated the long-term serologic profile of treated HGE. Results of one study of treated HGE demonstrated that antibody titers remained elevated 11 to 14 months after onset in 5 of 10 patients tested (10). In another study of HGE patients, not all of whom were treated, Bakken et al. detected E. equi antibodies in 11 (46%) of 24 patients at 12 months, 4 (44%) of 9 at 18 months, and 2 patients (denominator not reported) at 30 months (5). The same researchers found that sera from 71 (15%) of 475 asymptomatic residents of northwestern Wisconsin contained antibodies to E. equi (11). In our study, only 1 of 70 patients had a persistently elevated antibody titer; however, specimens were collected more than 1 year after acute onset in approximately 95% of patients, and the geographic distribution of patients differed from that of the Bakken study.

Animal studies and case reports have suggested the possibility of chronic or recurrent HGE. Experimental ehrlichiosis infections in dogs and horses have demonstrated the presence of E. canis and E. equi, respectively, in tissues up to 2 months after treatment with doxycycline (12,13). An HGE patient from Wisconsin was one of the first anecdotal reports to suggest this possibility in humans. In that case, serologic evidence of a chronic ehrlichial infection or reinfection was identified 3 years after treatment for Lyme disease and HGE coinfection (14). Another study from Connecticut included one patient who had a specimen positive by PCR 7 weeks after treatment, indicating a possible persistent ehrlichial infection. In that study, PCR was used to demonstrate the presence of ehrlichial DNA in the blood of some patients who were seronegative (15). Dumler and Bakken detected HGE agent DNA by PCR 2 to 30 days after illness onset in four patients; two of them had no HGE antibody detectable at the time (16). Compared with the serology testing used in this study, PCR may be a more sensitive assay in detecting late ehrlichial infection (17).

Our study used current national case definitions for HGE (6). These criteria allow for positive PCR alone as laboratory confirmation. Ten (14%) of the 73 confirmed HGE cases were confirmed by positive PCR alone. However, a recent report by the American Society of Microbiology’s Task Force on Consensus Approach for Ehrlichiosis (CAFÉ) considers positive PCR alone (without other laboratory support) to represent probable laboratory evidence of HGE (18). If we had applied CAFÉ criteria to our study, we would have had 63 (74%) confirmed and 22 (26%) probable cases of HGE.

The primary outcome measures in this study were based on self-reported symptoms. Without laboratory confirmation of persistent infection, recall bias should be considered as a possible explanation for the findings. Because of the severity of their past illness and because this was not a blinded study, HGE patients may have been more aware of their constitutional symptoms and had better recall of them than controls. Selection of controls is another potential source of bias if they were less likely to have coexisting chronic diseases compared with cases. However, we found no statistical difference in the prevalence of cancer, stroke, heart disease, or diabetes between cases and controls.

Previous studies of various infectious diseases have suggested that convalescence from illness is significantly dependent on the emotional state of the patient. In those studies, as in ours, fatigue was often a persistent symptom (19–21). In a study of recovery from influenza, Imboden noted that “delayed recovery following acute self-limited illness occurs in persons who respond to psychological tests in patterns characteristic of depression-prone patients” (19). We did not collect baseline psychological data, but the subset of HGE patients with recurrent or persistent constitutional symptoms may have been psychologically predisposed to a protracted convalescence.

In summary, we found that a subset of patients with HGE have persistent constitutional symptoms 1-3 years after treatment, without functional disability. Further research is needed to determine whether these symptoms are related to the previous ehrlichial infection or other causes. In the absence of serologic evidence of persistent infection, we believe these symptoms are most likely due to a postinfectious syndrome. Further studies that use PCR testing of convalescent-phase samples would be helpful to exclude the remote possibility of persistent Ehrlichia infection.