Mycobacterium kansasii has traditionally been considered the most virulent of the nontuberculous mycobacteria [15]; 75% to 97% of HIV-negative persons fulfill criteria for infection [1, 16]. Co-infection with HIV seems to alter the spectrum of clinical disease. Cavitary lung disease, which is diagnostic of pulmonary infection in HIV-negative patients [16-18], is documented in less than 50% of HIV-positive patients [19, 20]. Concomitant pathogens are identified in 30% to 46% of HIV-positive patients [21, 22], further obscuring the relative contribution of M. kansasii as a cause of pulmonary disease. Disparate reports documenting the frequency of colonization in the HIV population have led to confusion about the clinical significance and need for treatment of M. kansasii infection [4, 19, 23, 24].

Despite the changing epidemiology of M. kansasii infections, no updated population-based evaluation has been done in the AIDS era. Limitations of existing studies include hospital-based case ascertainment, which biases toward more seriously ill patients, and exclusion of HIV-negative patients. These limitations mask changing disease patterns in the population as a whole. Monitoring disseminated disease as part of AIDS surveillance may significantly underestimate the overall incidence of M. kansasii infection because pulmonary disease accounts for more than 65% of infections [4, 19]. Our study used population-based laboratory surveillance to find all cultures that were positive for M. kansasii in three counties in northern California. By including both HIV-positive and HIV-negative patients, we could assess the role of impaired immunity in the clinical presentation, significance of a positive respiratory culture, and risk for infection.

Methods

Top Methods Results Discussion Author & Article Info References

Surveillance Strategy and Case Mapping

Active laboratory-based surveillance and auditing of all 43 hospital and clinic laboratories and 4 health departments in Alameda County, Contra Costa County, and San Francisco County between 1 January 1992 and 31 December 1996 were used to identify persons with positive cultures for M. kansasii. The organism was isolated by using standard methods [25]. Between 1992 and 1993, a designated laboratory employee at each facility prospectively notified surveillance staff of each positive culture. Monthly summaries of laboratory isolates were reviewed to ensure complete ascertainment of cases. We accrued cases identified between 1994 and 1996 by retrospectively reviewing laboratory records using a combination of computerized database searches and manual audits of mycobacteriology laboratory records.

To abstract clinical, demographic, and microbiological data, we reviewed charts by using a standardized definition to establish the presence of pulmonary infection [26]. A diagnosis of HIV or AIDS was established by using the 1993 Centers for Disease Control and Prevention case definition [27]. Patients with negative serologic test results in the 6 months before M. kansasii culture were classified as HIV negative; those who were not tested or were seronegative more than 6 months before the index culture were classified as "HIV unknown." Serostatus was validated by cross-referencing patients with the California Office of AIDS database of patients who had a confirmed AIDS diagnosis.

California Department of Finance data [28] that projected the annual population of each county were used as rate denominators for the general population. The California Office of AIDS provided midyear estimates of the number of living persons with AIDS and HIV [29, 30]. Cumulative incidence rates were calculated for each year of the study by dividing the number of adult patients older than 20 years of age by the annual estimated adult population. Atlas GIS software (Strategic Mapping, Inc., Santa Clara, California) was used to assign patients living in San Francisco County to residential census tracts. Demographic and socioeconomic data were obtained by linkage with the 1990 U.S. Census report through the Summary Tape File 3 database. The geographic analysis was restricted to San Francisco County.

Statistical Analysis

Data were entered into an EPI-Info 6.04 database (Centers for Disease Control and Prevention, Atlanta, Georgia) and analyzed by using Stata 4.0 statistical software (Stata Corp., College Station, Texas). Dichotomous variables were assessed by using chi-square tests or the Fisher exact test. Continuous variables were compared by using the Student t-test or the two-sample Wilcoxon rank-sum test for nonparametric data. Poisson exact CIs were calculated for rates. All tests were two-sided; a P value of 0.05 or less was considered statistically significant.

Results

Top Methods Results Discussion Author & Article Info References

Mycobacterium kansasii was isolated from 283 patients. Thirteen patients were excluded because they lived outside the surveillance area (n = 7) or because their place of residence was unknown (n = 6). One hundred eighty-seven patients (69.3%) were HIV-positive, 33 (12.2%) were HIV-negative, and 50 (18.5%) had an unknown HIV status. The HIV-negative and HIV-unknown patients had homogeneous demographic characteristics (data not shown) and were combined to form a presumed HIV-negative group. Cross-reference of these patients with the California State AIDS Registry identified one (1.2%) potential case of misclassification: a patient with unknown HIV status at the time of M. kansasii isolation who subsequently received an AIDS diagnosis.

Demographic and Socioeconomic Characteristics

Patients who were HIV positive and those who were HIV negative differed significantly with respect to several characteristics (Table 1). The median age of HIV-negative patients was 22 years greater than that of HIV-positive patients, although the single infected child was an HIV-negative 2-year-old boy. Patients who were HIV positive were more likely to be male and African American and to have reported use of injection drugs. Homelessness (defined as no housing or housing in shelters) or marginal housing (defined as temporary residence in a low-rent hotel) in the year before M. kansasii culture was significantly more common among HIV-positive patients.

Incidence and Geographic Analysis

The mean annual incidence of M. kansasii infection between 1992 and 1996 (based on the 269 adult patients) was 2.4 cases per 100 000 persons (95% CI, 2.1 to 2.7 cases per 100 000); the rate of infection was higher among male patients (5.1 cases per 100 000 [CI, 4.5 to 5.9 cases per 100 000]) and African Americans (8.1 per 100 000 [CI, 6.5 to 10.0 cases per 100 000]). Mean incidence varied markedly by HIV status: The rate per 100 000 persons was 0.75 in the general population (CI, 0.6 to 0.9 cases per 100 000), 115.1 in HIV-positive persons (CI, 99.2 to 132.9 cases per 100 000), and 646.5 in persons with AIDS (CI, 553.5 to 750.7 cases per 100 000). Among persons with AIDS, the rate of isolation of M. kansasii was disproportionately elevated for African Americans (1533 cases per 100 000 [CI, 1197 to 1933 cases per 100 000]) and women (1450 cases per 100 000 [CI, 873 to 2265 cases per 100 000]). African-American women with AIDS had the highest isolation rate of M. kansasii (2137 cases per 100 000 [CI, 1196 to 3524 cases per 100 000]).

The geographic analysis was restricted to San Francisco County, which contributed 78% of HIV-positive patients and 46% of HIV-negative patients even though it had the smallest population of the three surveillance counties. Fifteen residents of San Francisco County were excluded from the geographic analysis because of homelessness (n = 14) or incomplete address (n = 1). Patients with M. kansasii infection resided in 72 (47.4%) of the 152 county census tracts, and 38 (53%) of these tracts had more than one case (Figure 1). Ten census tracts had an isolation rate of more than 1 case per 1000 residents (range, 1.08 to 7.35 cases per 1000 residents). These high-density tracts, which contained 5.6% of the San Francisco County population, accounted for 34.5% of mapped M. kansasii cases in the county.

View larger version (44K): [in this window] [in a new window]   Figure 1. Map of San Francisco, California, showing geographic distribution of patients with Mycobacterium kansasii infection by HIV status and median household income of census tract. Gridlines delineate the 152 census tracts in San Francisco County. Shaded areas indicate census tracts with a median income of less than $20 000. Black circles represent HIV-positive patients, and white circles represent HIV-negative patients.

Median income differed between the 72 census tracts that contained patients with M. kansasii infection and the 80 census tracts that did not ($32 317 compared with $38 048; P = 0.001). Median incomes were $24 112 for tracts with more than 1 case per 1000 residents and $32 514 for low-density case-containing tracts (P = 0.03). Median income of the census tracts was similar for HIV-positive and HIV-negative patients ($26 157 compared with $31 009; P > 0.2).

Clinical Characteristics

Advanced immunosuppression was common among HIV-positive patients with M. kansasii infection (Table 2). The median CD4 count for this group was 20 cells/mm³, and 75% of patients had previous or concurrent opportunistic infection. Among HIV-positive persons without an AIDS diagnosis, the median CD4 count was 290 cells/mm³. Comorbid disease was significantly more common in the HIV-negative group, but no predisposing medical condition was reported for 40.3% of this group. These persons were younger (mean age, 53 years compared with 65 years; P = 0.02) and less likely to be hospitalized at the time of index culture (37.0% compared with 67.5%; P = 0.01) but otherwise did not differ from HIV-negative patients with comorbid disease.

Clinical and radiologic evidence of infection was present in 87.5% of patients (91.5% of HIV-positive patients and 82.7% of HIV-negative patients). Infectious syndromes included pneumonia (91%), bloodstream infection (5%), bronchitis (0.9%), gastroenteritis (0.9%), lymphadenitis (0.9%), osteomyelitis (0.4%), and cellulitis (0.4%). Twenty-nine patients (11.3%), all with positive respiratory cultures, did not fulfill criteria for infection. Of note, eight of these patients had multiple positive sputum cultures for M. kansasii. Asymptomatic patients did not differ significantly with respect to age, sex, number of positive cultures, or presence of underlying illness but were less likely than symptomatic patients to be HIV positive (51.7% compared with 70.6%; P = 0.06) and to require hospitalization (44.8% compared with 74.3%; P = 0.002).

A concurrent organism was identified in 47% of patients. Because of concern that the higher proportion of symptomatic HIV-positive patients was attributable to other pathogens, a subgroup analysis was performed on the 109 patients who did not have mixed infections (the "pure-culture" subgroup). Clinically evident disease was present in 86% of HIV-positive patients and in 71% of HIV-negative patients with pure cultures (P = 0.15). Copathogens did not explain the higher rate of hospitalization in HIV-positive patients; 72.2% of HIV-positive patients and 51.8% of HIV-negative patients with pure cultures were hospitalized at the time of index culture (P = 0.009).

Microbiological Characteristics

A total of 588 specimens yielded M. kansasii, for a mean of 2.2 positive specimens per patient (range, 1 to 12). More than 93% of positive cultures were from respiratory tract specimens. More HIV-positive than HIV-negative patients had a positive acid-fast stain of sputum (42% compared with 21%; P = 0.005). Repeated mycobacterial respiratory cultures were obtained for 69.8% of patients. Patients with several cultures were divided into high-grade (>1 positive culture, 56.7% of patients) and low-grade (1 positive culture, 43.3% of patients) groups. Thirty-four percent of patients with a single culture had a pure culture of M. kansasii and radiologic and clinical symptoms that were compatible with infection.

Testing for mycobacteremia was inconsistently performed. Only one patient without a diagnosis of HIV infection had a blood culture drawn; the culture was negative. Blood cultures were obtained for 48.7% of the HIV-positive group. Mycobacteremia was detected in 9.6% of all HIV-positive patients and in 19.6% of patients from whom blood cultures were obtained. Isolated mycobacteremia accounted for 50% of cases of positive blood cultures; the remainder represented dissemination from a primary pulmonary infection.

Extrapulmonary localized disease was present in 2 (2%) patients without HIV infection: a child with cervical lymphadenitis and an adult who developed a digital abscess after lacerating his finger on a barnacle. Extrapulmonary culture sites from HIV-positive patients included stool (n = 4), urine (n = 1), bone (n = 1), lymph node (n = 1), and cerebrospinal fluid (n = 1). In four of these patients, disseminated infection was confirmed by positive blood culture.

Discussion

Top Methods Results Discussion Author & Article Info References

Mycobacterium kansasii has been considered an infrequent cause of pulmonary infection in persons residing outside of highly endemic regions in the southern and central United States [1, 2]. Our study, conducted in a nonendemic area, found a cumulative incidence of 2.4 cases per 100 000 adults, a rate almost five times higher than the national isolation rate in 1980 [2] and of a magnitude similar to that of the 8 cases of M. tuberculosis per 100 000 persons reported in 1996 [31]. The isolation rate in the HIV-negative population (0.75 cases per 100 000 persons) is similar to the 1980 California state-specific rate (0.58 cases per 100 000 persons) [2]; this suggests that culture of M. kansasii from the general population has remained relatively stable. The increase in the crude isolation rate is most likely attributable to the increased population prevalence of HIV [32, 33]. Whether the 150-fold higher incidence of M. kansasii infection among HIV-positive persons (115 cases per 100 000 persons) and the almost 900-fold increase among persons with AIDS (647 cases per 100 000 persons) in northern California is generalizable to other geographic sites is unknown. Surveillance for M. kansasii among HIV-positive persons residing in traditionally endemic areas will be important to determine whether the burden of disease caused by M. kansasii has increased proportionally in these regions.

Although M. kansasii pulmonary infection is clinically indistinguishable from tuberculosis [34, 35], person-to-person transmission has not been documented, and infection is thought to occur sporadically from environmental sources [7, 9, 11, 14, 36]. A limited number of familial clusters suggest a point-source outbreak or human transmission [12, 37, 38]. We used case mapping to reveal geographic clustering of cases in San Francisco. Median census tract income was significantly lower in tracts that contained persons with M. kansasii infection and decreased inversely with census tract incidence. Homelessness or marginal housing in the 12 months before the index culture characterized 41% of HIV-positive patients and 20% of HIV-negative patients. Poverty, urban residence, and homelessness have historically been major risk factors for contagious diseases (such as tuberculosis) because crowding and poor ventilation facilitate person-to-person transmission [39-42]. Although a previous study found isolation rates in the pre-AIDS era to be 2.5-fold higher in urban regions than in rural regions [43], our study is the first to detect an ecologic association between poverty and M. kansasii infection. Whether this association can be attributed to the existence of environmental sources for M. kansasii in low-income areas or to human transmission requires further investigation.

The ability of M. kansasii to colonize or infect the respiratory tract has resulted in considerable confusion about the clinical significance of a single positive sputum culture. Previous American Thoracic Society guidelines requiring at least two positive respiratory samples for the diagnosis of pulmonary infection [44] have been criticized as being overly rigorous in seriously ill patients with AIDS [4, 24] and have been modified in the most recent national recommendations [45]. In our study, 27.6% of patients without signs or symptoms of tissue invasion had more than one positive culture; this finding effectively excludes the possibility that contamination caused a false-positive result. Furthermore, 34% of patients with a single positive culture among many submissions were symptomatic, had abnormal chest radiographs, and had no alternative pathogen isolated; these findings suggest that low-grade culture positivity may be consistent with invasive disease. When the analysis was restricted to patients with a pure culture for M. kansasii, 85.7% of HIV-positive patients and 71.4% of HIV-negative patients had both clinical and radiologic evidence of infection. The low percentage of asymptomatic cases is all the more remarkable because the study was designed to identify patients with colonization or mild clinical illness that did not require hospitalization. Our study supports the belief that M. kansasii is a true pathogen in most patients regardless of HIV status and that clinicians should have a low threshold for initiating therapy [1, 4, 16, 19, 20, 45].

The clinical spectrum of disease varied between HIV-positive and HIV-negative patients. Most patients with HIV infection had an AIDS diagnosis at the time of positive culture [4, 19, 20, 22], and predisposing medical conditions were common among HIV-negative patients [46-48]. Not previously appreciated was the finding that 40.3% of HIV-negative patients had no identifiable underlying disease; this suggests that, similar to Mycobacterium avium complex infection, M. kansasii infection may occur in healthy, immunocompetent persons [49]. Mycobacteremia or focal extrapulmonary infection was present in 12.8% of HIV-positive patients, confirming that surveillance for disseminated infection significantly underestimates the total isolation rate. When we controlled for co-isolated organisms, we found that similar proportions of HIV-positive and HIV-negative patients had symptomatic pulmonary disease, yet hospitalization frequency, proportion of smear-positive respiratory specimens, and high-grade culture positivity were overrepresented among the HIV-positive patients; this finding suggests that this group was sicker and had a higher burden of pulmonary mycobacteria.

Our findings must be interpreted cautiously because of several limitations. Data were collected retrospectively by chart review, and such variables as homelessness and comorbidity may not have been routinely sought. Because cultures are usually obtained from patients who have signs and symptoms of pulmonary disease, the potential for detection bias and overestimation of the proportion of symptomatic patients exists. Determination of pulmonary infection compared with colonization was necessarily limited by the subjective impression of the evaluating physician. To minimize misclassification, we used previously published criteria that required both an abnormal chest radiograph and clinical signs that were compatible with chest infection [26]. Although almost one fifth of patients did not have a documented HIV test performed, diagnostic misclassification was minimal; a single unanticipated AIDS diagnosis was discovered by cross-referencing the M. kansasii database with the California AIDS registry. The relation between poverty and M. kansasii may represent ecologic bias because census tract characteristics may not adequately represent attributes of individual patients [50]. This association is less likely to be confounded by HIV infection because there is minimal overlap between the San Francisco County census tracts with the highest prevalence of AIDS and those with the highest incidence of M. kansasii [51, 52].

Mycobacterium kansasii is an important pathogen in northern California. Our findings may underrepresent trends in traditionally endemic areas, where rates of infection among patients who have AIDS and patients who do not have AIDS are five times higher than rates reported bicoastally [2, 6]. In our study, recovery of M. kansasii from respiratory tract specimens was strongly associated with symptomatic disease regardless of HIV status. The findings of a high prevalence of homelessness among patients and geographic clustering in census tracts with lower socioeconomic attributes are both unexpected and provocative. Prospective surveillance is needed to more precisely define the incidence of M. kansasii infection in diverse populations, evaluate the clinical significance of a single positive culture for M. kansasii, and understand modes of mycobacterial acquisition and transmission of infection.

From the University of California, San Francisco, San Francisco, California; Centers for Disease Control and Prevention, Atlanta, Georgia; and California Department of Health Services and the University of California, Berkeley, Berkeley, California.

Dr. Zwerling: Department of Pediatrics, University of Pennsylvania School of Medicine, Children's Hospital of Pennsylvania, 34th and Civic Center Boulevard, Philadelphia, PA 19104.

Dr. Pletcher: School of Medicine, University of California, San Francisco, 513 Parnassus Avenue, S224, San Francisco, CA 94143. Ms. Hahn: Department of Epidemiology and Biostatistics, San Francisco General Hospital, 1001 Potrero Avenue, San Francisco, CA 94110.

Dr. Gerberding: Division of Hospital Epidemiology, Centers for Disease Control and Prevention, 1600 Clifton Road, NE, Atlanta, GA 30333.

Dr. Ostroff: National Center for Infectious Diseases, Office of the Director, Centers for Disease Control and Prevention, 1600 Clifton Road, NE, Atlanta, GA 30333.

Dr. Vugia: Disease Investigation and Surveillance Branch, State of California, Department of Health Services, 2151 Berkeley Way, Berkeley, CA 94704-1011.

Dr. Reingold: Division of Public Health Biology and Epidemiology, Earl Warren Hall, University of California, Berkeley, Berkeley, CA 94720-7360.