Mycobacterium tuberculosis bacteraemia: experience from a non-endemic urban centre

Authors

  • H. M. El Sahly,

    Corresponding author
    1. Departments of Molecular Virology and Microbiology and Medicine, Baylor College of Medicine, Houston, TX, USA
    • Corresponding author: H. M. El Sahly, Department of Molecular Virology and Microbiology, One Baylor Plaza, BCM-MS280, Houston, TX 77030, USA

      E-mail: hanae@bcm.edu

    Search for more papers by this author
  • L. D. Teeter,

    1. Center for Molecular and Translational Human Infectious Diseases Research, The Methodist Hospital Research Institute, Houston, TX, USA
    Search for more papers by this author
  • J. M. Musser,

    1. Center for Molecular and Translational Human Infectious Diseases Research, The Methodist Hospital Research Institute, Houston, TX, USA
    Search for more papers by this author
  • E. A. Graviss

    1. Center for Molecular and Translational Human Infectious Diseases Research, The Methodist Hospital Research Institute, Houston, TX, USA
    Search for more papers by this author

  • The data were presented at the 48th annual meeting of the Infectious Diseases Society of America, Vancouver, Canada, October 2010.

Abstract

The isolation of Mycobacterium tuberculosis from blood culture specimens has been associated with human immunodeficiency virus (HIV) co-infection with variable impact on tuberculosis (TB) mortality reported. The overwhelming majority of M. tuberculosis bacteraemia cases were described in developing countries. We present a nested case–control analysis of clinical, sociodemographic and behavioural risk factors in patients with positive M. tuberculosis blood cultures compared with patients with negative blood cultures from a 9-year population-based active TB surveillance study conducted in Houston, Texas. There were 42 patients with M. tuberculosis bacteraemia, 47 blood culture negative patients and 3573 patients for whom no mycobacterial blood culture was requested. HIV infection was more common in patients for whom a mycobacterial blood culture was requested (79.8% versus 15.1% p <0.001). Of the patients with M. tuberculosis bacteraemia, six were HIV negative or had no documentation of HIV status, including five with immunosuppressive conditions other than HIV. Patients with M. tuberculosis bacteraemia were more likely than patients with negative blood cultures to be deceased at diagnosis or to die while on TB therapy (50.0% versus 17.0%, p <0.01), to report men-who-have-sex-with-men behaviour (31.7% versus 13.0%, p 0.03), to have renal failure (28.6% versus 6.4%, p 0.01), and to have HIV RNA levels higher than 500 000 copies/mL (61.9% versus 17.2%, p 0.01). Requests for mycobacterial culture of blood specimens were more common in HIV-infected individuals, and the presence of M. tuberculosis bacteraemia was associated with a significant increase in mortality.

The isolation of Mycobacterium tuberculosis from blood samples of tuberculosis (TB) patients has been documented since the early twentieth century [1]. However, M. tuberculosis blood cultures were rarely used as a modality of diagnosing TB until the human immunodeficiency virus (HIV) epidemic occurred, when mycobacterial blood cultures became widely used to diagnose disseminated Mycobacterium avium-intracellulare complex disease. It was then noted that M. tuberculosis was increasingly identified from blood culture samples. Initially, the phenomenon was thought to be strictly associated with co-infection with HIV, but various reports pointed to M. tuberculosis isolation from blood cultures in HIV-negative patients as well, albeit at much lower frequency [2, 3]. The usefulness of blood cultures as a diagnostic tool depends on the pre-test probability of TB in the study population: Gopinath et al. [3, 4] found that 19 out of 79 (24%) patients suspected of TB had M. tuberculosis isolated from blood in a tertiary-care centre in India; while McDonald et al. [3, 4] found that only 10% of all febrile patients admitted to a hospital in Malawi or Thailand had mycobacteraemia in 1997. In a more recent study by Crump et al. [5], 5.7% of febrile patients admitted to a hospital in Tanzania had M. tuberculosis bacteraemia. Mycobacterial blood culture can be a useful diagnostic adjunct, as documented in a study in HIV-infected individuals in Spain where blood was the only source of a positive culture in 15% of 33 patients who were documented as having TB [6].

Predictors of M. tuberculosis bacteraemia include advanced immunosuppression (as documented by a low CD4 count), HIV infection, prolonged cough, prolonged fever, weight loss, lymphadenopathy, left shift in the white blood cell count and anaemia [2, 5-8]. The overall mortality associated with M. tuberculosis bacteraemia is generally high: 25–55%, probably a reflection of the advanced immunosuppression observed in these patients [2, 5-7]. However, in one prospective study in Malawi, the presence of M. tuberculosis bacteraemia in HIV/TB co-infected individuals was not associated with a statistically significant increase in the 1-month mortality: 21% in M. tuberculosis bacteraemic versus 32% in not M. tuberculosis-bacteraemic patients [9]. In addition, co-infection with HIV does not seem to predict an additional mortality risk. In a retrospective medical record review study from Spain, all-cause mortality in patients with M. tuberculosis bacteraemia was comparable between the HIV-negative and HIV-positive individuals (25% and 29%, respectively) [2].

With a few exceptions, the evaluations of the clinical characteristics of and mortality associated with M. tuberculosis bacteraemia have been described and performed in developing countries. We present data on the clinical characteristics and outcomes of patients with M. tuberculosis bacteraemia, identified from a 9-year population-based active surveillance study conducted in Houston, Texas.

Methods

Study population

We conducted a prospective, population-based, active surveillance study of all TB cases reported in Harris County, Texas (hereafter referred to as Houston) between 1 October 1995 and 30 September 2004. Written informed consent was obtained from all participants. Patients who consented to the study were interviewed to collect pertinent sociodemographic and behavioural variables. The patients’ next of kin, guardians or proxies were interviewed when patients were deceased or unable to be interviewed. Clinical variables were obtained using public health and medical record review. All enrolled TB patients were used as the underlying study population and those who had a mycobacterial blood culture requested by the clinician were included in the nested case–control analysis. The study was approved by the Institutional Review Boards of the participating institutions.

Mycobacterium tuberculosis isolates

Mycobacterium tuberculosis isolates were collected and molecularly characterized using three genotyping methods: principal genetic group designation, spoligotyping and restriction fragment length polymorphism (RFLP) of the IS6110 insertion element [9, 10]. Based on the genotypic findings, isolates were grouped into ‘clusters’: strains with at least five IS6110 copies were considered to be part of the same cluster if they had matching IS6110 RFLP patterns; strains with fewer than five IS6110 copies were considered clustered if they had matching IS6110 RFLP patterns, spoligotypes and principal genetic group designations [9, 10].

Statistical analysis

Data from study instruments were entered into a Microsoft Access database (Redmond, WA, USA) for storage and analysis. Characteristics of M. tuberculosis blood-culture-positive subjects (M. tuberculosis bacteraemic) were compared with those of M. tuberculosis blood culture negative subjects (not M. tuberculosis bacteraemic) by chi-squared analysis (or Fisher's exact testing when at least one cell size was ≤5); and odds ratios (OR) presented are estimates for the relative risk associated with M. tuberculosis bacteraemic patients. Similarly, characteristics of patients with blood specimens evaluated by culture were compared with patients without blood specimens evaluated (but statistical measures are not shown in the tables). Mortality and days to treatment completion were visualized using Kaplan–Meier methods with significance assessed by Cox Proportional Hazard analysis. Values of p <0.05 were considered statistically significant. Statistical analyses were conducted with STATA version 10 SE (StataCorp, College Station, TX, USA) and SAS 9.3 (SAS Institute Inc., Cary, NC, USA).

Results

During the 9 years of the study, there were 4312 patients with confirmed TB reported in Houston of whom 3662 (84.9%) were enrolled and 85.6% of the 3063 culture-positive cases were enrolled. We identified 42 patients with TB for whom at least one M. tuberculosis isolate was recovered from a blood culture sample and 47 patients with TB for whom a mycobacterial blood culture sample was requested and was M. tuberculosis culture negative. Characteristics of patients with TB with no blood specimen evaluated (= 3573), M. tuberculosis bacteraemic patients and the blood culture negative group are presented in Table 1. Overall, there were no significant differences in the age, race/ethnicity or gender distribution between the M. tuberculosis bacteraemic and blood culture negative patients. Patients with M. tuberculosis bacteraemia were more likely than blood culture negative patients to be deceased at diagnosis or to die while on TB therapy (50% versus 17.0%, p <0.01), and were more likely to report men who-have-sex-with-men behaviour (31.7% versus 13.0%, p 0.03), to have renal failure (28.6% versus 6.4%, p 0.01), and to have HIV RNA levels (viral load) higher than 500 000 copies/mL (61.9% versus 17.2%, p 0.01, Fig. 1).

Table 1. Sociodemographic, behavioural and clinical characteristics of patients with tuberculosis not evaluated by blood culture and in patients with and without Mycobacterium tuberculosis bacteraemia
 TB Cases with no blood specimen evaluated = 3573Mtb blood culture positive = 42Mtb blood culture negative = 47ORp value*
  1. HIV, human immunodeficiency virus; Mtb, M. tuberculosis; TB, tuberculosis

  2. *p values and OR refer to the comparison between M. tuberculosis blood culture positive and M. tuberculosis blood culture negative patients.

  3. a

    Denominator limited to subjects with data on M. tuberculosis drug susceptibility or given genotyping method.

  4. b

    Denominator limited to HIV-infected subjects with CD4 or HIV RNA data within 90 days of diagnosis.

Age median (mean)42 (41.7)41 (40.4)40 (42.0)0.980.43
Male2402 (67.2%)34 (81.0%)32 (68.1%)1.990.16
Race/Ethnicity    0.47
Black1339 (37.6%)29 (69.1%)31 (66.0%)1.150.76
White555 (15.6%)4 (9.5%)6 (12.8%)0.720.74
Hispanic1207 (33.9%)8 (19.1%)5 (10.6%)1.980.37
Asian458 (12.9%)1 (2.4%)4 (8.5%)0.260.37
Foreign born1333 (37.3%)7 (16.7%)9 (19.2%)0.840.76
Current smoker1470 (41.2%)15 (36.6%)27 (57.5%)0.430.05
Near daily alcohol use1186 (33.3%)19 (45.2%)18 (38.3%)1.330.51
Current homelessness239 (6.7%)3 (7.1%)4 (8.5%)0.831.00
Men-who-have-sex-with-men behaviour241 (6.8%)13 (31.7%)6 (13.0%)3.100.03
Injection drug use306 (8.6%)5 (12.5%)8 (17.0%)0.700.76
Fever2076 (58.6%)34 (81.0%)41 (87.2%)0.620.42
Cough2582 (72.9%)35 (83.3%)35 (74.5%)1.710.31
Any drug resistancea289 (9.7%)3 (7.1%)3 (8.3%)0.851.00
Clustered Mtb isolatea1718 (62.6%)27 (77.1%)20 (71.4%)1.350.61
Beijing family isolatea583 (22.6%)10 (29.4%)4 (17.4%)1.980.36
Principal genetic groupa    0.05
Group 1856 (32.8%)11 (31.4%)5 (21.7%)1.650.55
Group 21310 (50.2%)21 (60.0%)10 (43.5%)1.950.22
Group 3446 (17.1%)3 (8.6%)8 (34.8%)0.180.02
Disease location    0.33
Pulmonary2602 (72.8%)15 (35.7%)20 (42.6%)0.750.51
Extrapulmonary681 (19.1%)8 (19.1%)11 (23.4%)0.770.62
Both290 (8.1%)16 (38.1%)16 (34.0%)1.190.69
Blood only 3 (7.1%)0 (0%) n/a
Death at diagnosis or during therapy363 (10.2%)21 (50.0%)8 (17%)4.88<0.01
Completed TB therapy2874 (80.4%)19 (45.2%)30 (63.8%)0.320.01
Diabetes mellitus522 (14.6%)3 (7.1%)2 (4.3%)1.730.66
Renal failure135 (3.8%)12 (28.6%)3 (6.4%)5.870.01
Cancer128 (3.6%)1 (2.4%)2 (4.3%)0.940.91
Immunosuppressive therapy30 (0.8%)3 (7.1%)2 (4.3%)7.03<0.01
HIV diagnosis541 (15.1%)36 (85.7%)35 (74.5%)2.060.18
CD4 <200 cells/µLb295 (67.5%)30 (96.8%)30 (90.9%)3.000.61
HIV RNA >500 000 copies/mLb58 (21.4%)13 (61.9%)5 (17.2%)7.80<0.01
Figure 1.

Distribution of human immunodeficiency virus RNA levels in Mycobacterium tuberculosis (Mtb) bacteraemic and blood culture negative patients with tuberculosis (p <0.01).

Human immunodeficiency virus co-infection was highly prevalent (79.8%) among patients for whom a mycobacterial blood culture was ordered (85.7% in M. tuberculosis bacteraemic group and 74.5% in the not M. tuberculosis bacteraemic group), while HIV co-infection was present in 15.1% of the patients for whom no mycobacterial blood culture specimen was requested (p <0.001). Other characteristics that were significantly more common in the group in whom acid-fast bacillus blood culture specimens was obtained than the group in whom a blood culture specimen was not obtained were: Black race, birth in the USA, men who-have-sex-with-men behaviour, fever, disease location involving both extrapulmonary and pulmonary sites, death at diagnosis or during therapy, renal failure, cancer diagnosis, immunosuppressive therapy, a CD4 count less than 200 cells/μL and an HIV RNA level of >500 000 copies/mL (p <0.05). In addition, they were more likely to report sweats, chills, diarrhoea, loss of appetite, dyspnoea and weight loss (data not shown). However, patients in whom a blood culture specimen was obtained were less likely to be of Hispanic ethnicity, Asian race or to complete TB therapy than patients in whom no blood culture specimen was obtained. (p < 0.05).

Among M. tuberculosis bacteraemic patients, HIV co-infection was present in 36 (85.7%), two patients were HIV seronegative, and four had undocumented HIV serostatus. Mycobacterium tuberculosis bacteraemic patients with HIV co-infection had a median CD4 count of 40 cells/μL (mean 56.3). All but one of the six M. tuberculosis bacteraemic patients with negative/undocumented HIV serology had concomitant significant immunosuppressive conditions: two patients had systemic lupus erythematosus with end-stage renal disease on haemodialysis, one patient with rheumatoid arthritis receiving methotrexate with a history of cytomegalovirus colitis, one patient with diabetes mellitus, agranulocytosis and polymyositis, and one patient on cyclosporine for renal transplantation. A comparison of the characteristics of HIV-negative and HIV-positive M. tuberculosis bacteraemic patients is presented in Table 2.

Table 2. Sociodemographic and clinical characteristics of patients with Mycobacterium tuberculosis bacteraemia by HIV status
 HIV-positive = 36HIV-negative/undocumented = 6
  1. HIV, human immunodeficiency virus; TB, tuberculosis

  2. a

    In addition to isolation of M. tuberculosis from blood cultures, the patients had clinical evidence of TB involving the lungs (pulmonary), organs other than the lungs (extrapulmonary) or both.

  3. b

    Denominator limited to subjects with data on M. tuberculosis drug susceptibility or spoligotyping.

  4. c

    We found documentation of TB skin test results for five HIV-negative/undocumented and 22 HIV-positive patients.

Male31 (86.1%)3 (50%)
Race
Black26 (72.2%)3 (50%)
White4 (11.1%)0
Hispanic6 (16.7%)2 (33.3%)
Asian01 (16.7%)
Foreign born6 (16.7%)1 (16.7%)
Age median (mean)40.5 (40.0)44.0 (42.7)
Smoker15 (42.9%)0
Near daily alcohol use19 (52.8%)0
Disease sitea
Pulmonary12 (33.3%)3 (50%)
Extrapulmonary8 (22.2%)0
Both13 (36.1%)3 (50%)
Only blood3 (8.3%)0
Cavities on chest X-ray4 (11.1%)3 (50%)
Any drug resistanceb3 (8.3%)0
Beijing spoligotypeb7 (25%)3 (50%)
Positive TB skin testc6 (27.3%)2 (40%)

Clinical presentation

Overall, there were no significant differences in the symptoms on presentation between M. tuberculosis bacteraemic and blood culture negative patients (Table 1). However, patients evaluated by blood culture were clearly sicker than those not evaluated by blood culture, as assessed by symptoms, co-morbidities, clinical outcomes and, among HIV-co-infected, CD4 and HIV viral load values (see above).

Mycobacterium tuberculosis isolates

Blood specimens were the only source from which M. tuberculosis was isolated in four patients (9.8% of all patients with M. tuberculosis bacteraemia), including one with biopsy-confirmed cervical TB lymphadenitis. Drug resistance was noted in three isolates from three HIV-positive patients (one isolate was resistant to rifampin, one isolate was resistant to streptomycin and one isolate was resistant to rifampin and ethambutol). We genotyped six M. tuberculosis strains from the HIV-negative/undocumented M. tuberculosis bacteraemic patients and 29 M. tuberculosis strains from the HIV-positive, M. tuberculosis bacteraemic patients: 4 (66.7%) and 23 (79.3%) of the M. tuberculosis strains belonged to a genotypic cluster, respectively. The frequency of clustering observed in the M. tuberculosis bacteraemic patients (77.1%) was comparable to the frequency of clustering observed in clinical M. tuberculosis strains in Houston (62.9%) during the same time period (p 0.36) [11]. Of note, M. tuberculosis bacteraemic patients were less likely than blood culture negative patients to be infected with a principal genetic group 3 organism (OR = 0.18, p 0.02) [9].

TB therapy completion

Mycobacterium tuberculosis bacteraemic patients were less likely than blood culture negative patients to complete TB therapy (OR = 0.32, p 0.01). Among all TB patients who completed TB therapy, M. tuberculosis bacteraemic patients took longer to finish treatment (= 19, median 349 days) than blood culture negative patients (= 30, median 270 days, p 0.08) and patients for whom blood cultures specimens were not evaluated (median 245 days, p <0.01) (Fig. 2a).

Figure 2.

Kaplan–Meier curves for Mycobacterium tuberculosis (Mtb) bacteraemic, blood culture negative and tuberculosis (TB) cases for whom blood specimens were not evaluated: (a) Estimates of days to TB therapy completion for patients that completed treatment (p 0.08 and p <0.01 comparing blood culture negative and not-evaluated groups to the Mtb bacteraemic group, respectively). (b) Probability of survival, where death occurred either at TB diagnosis or while on therapy (age-adjusted p <0.01 and p 0.04 comparing Mtb bacteraemic and blood culture negative groups to the not-evaluated group, respectively).

Mortality

There were 21 deaths (50%) among the M. tuberculosis bacteraemic patients and eight deaths (17.0%) among blood culture negative patients (p <0.01). Fig. 2b shows the Kaplan–Meier survival curve for patients who were M. tuberculosis bacteraemic, blood culture negative and for whom blood specimens were not evaluated. Compared with patients in whom blood specimens were not evaluated, mortality was more likely for both M. tuberculosis bacteraemic and blood culture negative cases: age-adjusted hazard ratios 7.77 (95% CI 4.97–12.13, p <0.01) and 2.10 (95% CI 1.04–4.25, p <0.04), respectively.

Discussion

We present the clinical characteristics and outcomes from the largest case series of patients with M. tuberculosis bacteraemia in an industrialized country. In our series, M. tuberculosis bacteraemia was predominantly found in HIV-positive patients and the mortality associated with M. tuberculosis bacteraemia was significantly higher than that seen in TB patients without documented M. tuberculosis bacteraemia. In 9.8% of the cases of M. tuberculosis bacteraemia, blood was the only positive source for M. tuberculosis isolation.

Our case series and other case series have documented the association between HIV and M. tuberculosis bacteraemia. Importantly, we found that other relatively common immunosuppressing conditions may be associated with isolation of M. tuberculosis from the blood. In fact, we may be underestimating the importance of M. tuberculosis bacteraemia in HIV-negative patients because clinicians were more likely to order mycobacterial blood culture in HIV-infected patients: HIV infection was present in 79.8% of patients for whom a mycobacterial blood culture was obtained, whereas HIV infection was present in only 15.1% of the TB patients in Houston for whom no mycobacterial blood cultures were obtained. Also, patients with more overt symptoms (fever, chills, sweats, weight loss) were more likely to have acid-fast bacilli blood culture specimens obtained; which also denotes that we may be underestimating the prevalence of M. tuberculosis bacteraemia in TB patients with fewer overt symptoms. Given that the number of patients on immunosuppressive medications or of individuals living with immunosuppressing conditions is growing, mycobacterial blood cultures should be considered in such patients with unexplained fever as an adjunctive diagnostic modality, especially in the presence of other TB risk factors (immigration, minority status, exposure history). Notably, we found that for patients who completed therapy, the time from therapy initiation to completion was longer than the typical 180 days of therapy expected for most patients. We did not examine the records specifically for factors contributing to the longer time to therapy completion, but potential factors contributing include treatment interruptions due to drug reactions, episodes of loss to follow up (most of our patients received directly observed therapy) and hospital admission(s). Identifying these factors is important, as it may result in changes that would improve the cost efficiency of TB treatment.

In our study, the mortality of patients with M. tuberculosis bacteraemia was significantly higher than that found in both blood culture negative cases and TB patients not evaluated by blood culture. Mortality was not always assessed in previous studies, and when evaluated, mortality was not always found to be increased in patients with M. tuberculosis bacteraemia [2, 6-8]. Most other studies have not evaluated mortality systematically through therapy completion; and this could be the reason for the difference, although one can see that the survival curves start diverging early in our TB population (Fig. 2b). Increased mortality in M. tuberculosis bacteraemic patients can be partially explained by the advanced immunosuppression in this group of patients: all but one of the HIV-positive patients had CD4 counts below 200 cells/μL, and all but one HIV-negative/undocumented had significant immunosuppressing co-morbidities. Recently, Abdool Karim et al. [12] demonstrated that initiation of antiretrovirals within 2 months of initiating TB therapy improved survival, when compared with delayed initiation of therapy. This points to reversal of immunosuppression (at least partially) as an important factor in improving the outcomes of patients with M. tuberculosis bacteraemia. In our study, HIV-infected patients who were M. tuberculosis bacteraemic had significantly higher HIV viral loads than blood culture negative patients (Fig. 1). Other investigators have documented that an elevated HIV RNA level is an independent predictor of the incidence of opportunistic infections, including extrapulmonary TB [13, 14].

Our study has its limitations, the most evident of which is that mycobacterial blood cultures were not systematically collected on all patients. This probably resulted in an underestimation of the frequency of M. tuberculosis bacteraemia, which was 1.1% in our study population. Depending on the epidemiological setting and the criteria used to obtain the blood cultures, other studies have shown the prevalence of M. tuberculosis bacteraemia of 5 to >30% [4, 15, 16]. Another limitation is the possibility that mycobacterial blood culture results might have been missed in the abstraction of medical records, especially if the result was negative. However, such omissions are not expected to affect the observed increased mortality in patients with M. tuberculosis bacteraemia.

In summary, mycobacterial blood cultures are potentially useful diagnostic adjuncts. M. tuberculosis bacteraemia identified in a large metropolis in the USA is associated with advanced immunosuppression (HIV or non-HIV-related immunosuppression) and with significantly increased mortality.

Funding

The project has been supported with federal funds through the National Institute of Allergy and Infectious Diseases, the National Institutes of Health, under contract N01-AO-02738 and the National Institute on Drug Abuse, grant DA-09238.

Transparency Declaration

The authors do not have commercial or other associations that might pose a conflict of interest.

Ancillary