Serum samples from 114 patients with acute febrile illness, including 14 with both acute- and convalescent-phase serum samples, were tested at a dilution of 1:32 by IFA with a panel of 9 Bartonella antigens. All positive samples were retested at a dilution of 1:32 and at doubling dilutions to 1:4,096. In 12 of 13 cases with titers <512 to any rodent-associated antigen, the titer to the N. albigula–associated Bartonella antigens (NA-AB antigens) were the highest measured. Therefore, only the titers to NA-AB antigens are shown in Table 1. IFA titers to NA-AB >128 were observed more often in undifferentiated febrile illness (group A, 24 of 76) than in the 3 groups with specific diagnoses (groups B–D, 4 of 38) (χ² = 4.98, p = 0.026, using Yates' correction). Among 24 patients in group A with titers >128, a total of 11 had convalescent-phase titers >512. Clinical information was sufficient to analyze for 9 of these 11 patients: 5 patients with both acute- and convalescent-phase titers (Table 2) and 4 patients with only a convalescent-phase titer (Table 3). Nine patients in group A with both acute- and convalescent-phase serum samples showed no increase in titer or a titer >64.

Of 24 patients with pneumonic disease (groups B and C), only 1 had a titer of 128 to NA-AB antigens. Of 14 patients with other diagnosed febrile illnesses (group D) not listed in Table 2 and Table 3, three had high titers to NA-AB antigens (Table 1). A 35-year-old man with aortic valve endocarditis and cultures of blood and valve positive for S. aureus had an NA-AB titer of 1,024 on admission and the following day. A 30-year-old man with fever, myalgias, headache, thrombocytopenia, and leukopenia with admission serum positive by PCR for Borrelia hermsii (tickborne relapsing fever) had an acute-phase (day 1) titer of 256 and a convalescent-phase (day 24) titer of 1,024 to NA-AB antigens. A 23-year-old woman with fever and acute hepatic injury had positive convalescent-phase (day 28) IgM phase I (512) and IgG phase II (1,024) titers for Coxiella burnetii antigens and an NA-AB antigen titer of 256 in a convalescent-phase serum sample.

Five of the 14 patients with acute- and convalescent-phase serum samples in group A showed a >4-fold increase in titer to NA-AB antigens and convalescent-phase titers >512 on days 14, 7, 7, 12, and 42, respectively, after admission (Table 2). Each of the 5 who seroconverted had a clinical syndrome characterized by fever (temperature >39°C), chills, pronounced myalgias in the back and thighs, nausea, and headache. Two who seroconverted had a sore throat and 2 had diarrhea, but none had other upper or lower respiratory symptoms, abnormal chest radiograph results, lymphadenopathy, hepatosplenomegaly, bleeding, rash, altered consciousness, or abnormal neurologic findings. Thrombocytopenia and leukopenia were common (Table 2 and Table 3), but no patients had evidence of coagulopathy, or cardiac, pulmonary, renal, or neurologic disease.

Four other patients in group A had a single titer >512 to NA-AB antigens on days 21, 7, 20, and 23, respectively, after admission (Table 3). This group had elevated levels of serum transaminase, bilirubin, and alkaline phosphatase, which is indicative of active hepatitis. These 4 patients were treated with doxycycline, and all recovered without sequelae. Of the 9 patients listed in Tables 2 and 3, one had a diagnosis of chronic alcoholism (patient 6, Table 3). All 9 were negative for hepatitis A, B, and C; Q fever; Rocky Mountain spotted fever; murine typhus; leptospirosis; granulocytic or monocytic ehrlichiosis; plague; and tularemia; they also had negative titers for HIV, hantavirus, and antinuclear antibody. Patients 6, 8, and 9 were tested for antibody to hepatitis E at the Hepatitis Branch of CDC in Atlanta, Georgia, and were negative (M. Favorov, pers. comm.). Patients 1, 4, and 6 had 6-, 3-, and 3-fold lower titers, respectively, to the known Bartonella pathogen antigens compared with the titer to NA-AB antigens (Table 2 and Table 3).

This study provides preliminary serologic evidence for a Bartonella or Bartonella cross-reactive species that is causing acute febrile illness in immunocompetent adults in the rural southwestern United States. Five patients who seroconverted to rodent-associated antigen had fever, myalgias, headache, and chills with varying degrees of leukopenia, mild hepatitis, and thrombocytopenia. Four other patients with a single elevated titer 2–5 weeks into their illness had more severe hepatic injury. In the absence of culture- or PCR-positive evidence of Bartonella infection in any of these patients, the interpretation of these serologic observations is related to the cross-reactivity between Bartonella species as well as non-Bartonella species, interpretation of the quantitative IFA titer, variations among pathogens to stimulate antibody responses, timing of serum specimen collection, and the route of exposure.

Although antigens derived from Bartonella isolated from N. albigula were used, this process does not imply that the human infection was caused by a Bartonella strain that naturally infects N. albigula. Serologic cross-reactions among Bartonella species are common (20), and the IFA is unable to distinguish between infection with B. quintana or B. henselae (21). The cross-reactivity between rodent-associated and known Bartonella pathogen–associated antigens was expected and found to some degree in nearly all cases. We did not find clear evidence for infection with Bartonella species known to cause disease in humans, including B. henselae, B. quintana, B. vinsonii, and B. elizabethae, in the sense that titers to rodent-associated, particularly NA-AB, antigens were always higher than those for known human Bartonella species. The lack of cross-reactivity in 3 patients is consistent with a rodent-associated Bartonella infection, although infection with a Bartonella associated with a nonrodent animal cannot be ruled out (22).

Identification of Bartonella infections in humans in the southwestern United States is important because cat-scratch disease is not common in this region, and cat fleas, presumed vectors for B. henselae, do not naturally exist in such arid environments (23). Cross-reactivity between Bartonella antigens and antigens of C. burnetii and Chlamydia species has been demonstrated (24,25). Except for the woman in group D who had clear evidence of acute Q fever hepatitis, significant Bartonella titers >128 were not associated with detectable antibody to phase I or II Coxiella antigens in the complement fixation test in all 8 patients tested. None of the patients had a condition associated with nonspecific immune stimulation such as HIV infection, injection drug use, or collagen vascular disease that could account for false-positive results.

The IFA was developed at CDC (21) and has been assessed most extensively in the diagnosis of B. henselae and B. quintana infection in the United States (20). At the National Referral Center of CDC, a titer of 64 is considered positive (20). When a strict case definition is used for cat-scratch disease, this titer has a sensitivity of ≈80% and a specificity of 93% to 96% (20,21,26). Other investigators have found greater specificity when titers of 128 (27), 256 (25), or 512 (28) were used to diagnose cat-scratch disease. An IFA titer of 512 to B. henselae in adults with no exposure to cats or illness compatible with cat-scratch disease was uncommon (<1%) in 1 study in Germany (27). We used a conservative threshold IFA titer of 512 to present clinical data on 9 patients based in part on this experience with cat-scratch disease, recognizing that immunogenicity to immunodominant antigens may vary among species of the same genus. The usefulness of a single titer of 1:512 to NA-AB antigens (Table 3) is unknown because IFA titers to B. henselae persist during the first year after infection (20).

The clinical syndrome associated with seroconversion to NA-AB antigens was characterized by either a brief undifferentiated febrile illness or fever accompanied by hepatic injury. Clinical evidence for inflammation in the lung, heart, kidney, and nervous system was not apparent. Infection with B. henselae, particularly in immunocompromised hosts, has been documented to involve the liver (2). Moreover, thrombocytopenia and leukopenia, which were common in our small sample of febrile patients, have also been associated with B. quintana infection (29) in immunocompetent adults and with B. henselae infection in immunocompromised adults (2). No patient had intraerythrocytic bacilli visible on Giemsa-stained blood smear (30) (F. Koster, unpub. data). A clear definition of the syndrome awaits definitive identification based on culture of the pathogenic species from patients. Thus, a concerted effort to identify acute infections with rodent-associated Bartonella should be undertaken with specific serologic assays as well as intensive PCR-based diagnostics and culture techniques specific to the fastidious Bartonella genus.