Possible Hazards of Routine Bacillus Calmette-Guérin Immunization in Human Immunodeficiency Virus-Infected Children

A Mycobacterium tuberculosis complex organism was isolated unexpectedly from blood cultures obtained from a 3-year-old child succumbing to HIV-associated cardiomyopathy at our hospital. Biochemical studies speciated the isolate as Mycobacterium bovis. Because the child had been immunized with Bacillus Calmette-Guérin (BCG) at birth, we sought to determine whether the isolate recovered from the blood before death was BCG or a pathogenic strain of M. bovis. Our report focuses on three issues: the methods of differentiation of the M. tuberculosis complex strains; the factors contributing to possible underdiagnosis of disseminated infections; and the need for additional studies to determine the true incidence of dissemination in BCG-vaccinated HIV-infected children. Case report. A Brazilian child, immunized with BCG at birth, was adopted by an American expatriate family at 1 week of age. Medical history included developmental delay, recurrent otitis media and failure to thrive. At 15 months of age he was found to be infected with HIV by serologic testing in Brazil (not repeated at our hospital). Zidovudine therapy was begun at the time of diagnosis. Approximately 1 month before arrival in the United States for parental job relocation at 3 years of age, the patient developed increasing dyspnea and signs of cardiac failure. Physical examination on admission to Vanderbilt Children's Hospital in 1993 revealed an afebrile child with a heart rate of 160 and a respiratory rate of 32, accompanied by intercostal retractions and nasal flaring. Pulmonary congestion, cardiomegaly, hepatomegaly and generalized adenopathey were present. No cutaneous lesions or splenomegaly were seen. Chest roentgenogram revealed massive cardiomegaly, severe pulmonary congestion and bilateral pleural effusions. Despite vigorous diuresis and digitalization, the child expired. Request for autopsy was denied. Postmortem Bactec® blood cultures were obtained to evaluate the possibility of disseminated Mycobacterium avium-intracellulare infection. These cultures grew a mycobacterial organism confirmed by DNA probe (Gen-Probe, San Diego, CA) to be M. tuberculosis complex. The rarity of this observation led the clinicians to pursue vigorously a more complete identification of the organism. Conventional phenotypic studies performed in the laboratories of the Tennessee State Health Department revealed the organism to be niacin- and nitrate reduction-negative, findings consistent with M. bovis or BCG. Mycolic acid characterization by high performance liquid chromatography (HPLC) at the Centers for Disease Control, Atlanta, GA, revealed a distinctive chromatogram unique to TMC 1022, a Russian BCG strain.1 Molecular characterization was also performed (by Richard Frothingham, Duke University) by a recently described method of sequence-based strain differentiation.2 The patient strain, like those of other BCG strains, belonged to the MEDIUM-C® sequevar. Strains of M. tuberculosis and non-BCG M. bovis isolates belong to other sequevars. Discussion. With improved methods to isolate mycobacteria from the blood, disseminated infections in HIV-infected patients are more commonly recognized. Although the majority of mycobacterial isolates in the United States are M. avium-intracellulare strains, M. tuberculosis complex strains are also recovered and are confirmed by gene probe technology. Because the M. tuberculosis complex contains the four species, M. tuberculosis, Mycobacterium bovis, Mycobacterium microti and Mycobacterium africanum, which are closely related genetically, strain differentiation can be difficult. Biochemical methods can be suggestive but are not definitive. Antimicrobial susceptibility testing can offer some assistance since BCG is resistant to PZA and M. tuberculosis is generally susceptible. Separation of BCG from M. tuberculosis and M. bovis by HPLC of mycolic acids has been very useful.1 Molecular markers, such as the insertion sequence, IS6110, have been extremely important in differentiating strains of M. tuberculosis for epidemiologic purposes, but have not been as useful for other members of the M. tuberculosis complex since they share common IS6110 patterns.3 Another insertion sequence, IS1081, has been used successfully to differentiate M. bovis strains of BCG and non-BCG origin and may be used more commonly in the future to distinguish between these strains.4 Finally sequence-based analysis has been helpful in distinguishing these strains.2 Biochemical, HPLC and sequence-based analysis confirmed that the strain isolated from our patient was BCG. BCG vaccines are live attenuated mycobacterial strains originating directly or indirectly at different times from the original Institut Pasteur strain which was itself derived from a virulent bovine tubercle bacillus by 230 passages on glycerol-bile-potato medium in the years between 1908 and 1921. The earliest daughter strains were the Brazilian, Russian, Romanian and Swedish strains which were all derived between 1925 and 1928. The descendants of these four strains are similar in their mycolic acid profiles, in their pattern of protein synthesis and in the possession of two copies of IS6110.5 After that time other BCG strains were derived, such as the Japanese, Danish, Pasteur and Tice strains, having different HPLC patterns, different secreted proteins and a single copy of IS6110. The HPLC pattern obtained from our patient immunized in Brazil appeared identical with that of a Russian BCG strain, suggesting a common ancestry of both strains. Attempts to more fully characterize the lineage of the BCG strain administered to our patient were unsuccessful. Live BCG vaccine is given to newborn infants throughout the developing world. Many of these areas have high rates of tuberculosis and HIV infection. It is remarkable that only 14 cases of disseminated BCG in HIV-infected patients had been reported previously in the literature,6-17 with 7 of the previously reported cases in children (Table 1).6, 7, 11, 13, 15, 16 Symptoms seen in these patients included fever, adenopathy, weight loss and organomegaly, although some of the reports provided few clinical details. Before our report definitive biochemical and molecular confirmation of the BCG strain had been made in only one patient.6 Because our patient was afebrile and had symptoms that could all be explained by HIV cardiomyopathy, we had no clinical suspicion of disseminated BCG infection. It was fortuitous that the blood culture was done. However, it is likely that many patients receiving BCG, whether infected with HIV or not, have transient bacteremia with BCG. In fact in a report of autopsy findings in 26 BCG-vaccinated non-HIV-infected individuals dying of causes not related to BCG, granulomas were seen in distant organs at least 2 to 3 years after vaccination without any clinical signs.18 These data suggest that hematogenous spread of BCG from the site of vaccination is common. It is also possible that some of the cases reported as disseminated BCG might actually be infections with other species of the M. tuberculosis complex, such as M. bovis or M. tuberculosis. In a large series of patients with M. bovis infections, Danker et al.19 described eight cases of disseminated M. bovis infections in HIV-infected patients. Isolation of M. tuberculosis from blood cultures from HIV-infected patients has also be shown in 26 to 41% of patients with clinical tuberculosis, a rate much higher than reported for non-HIV-infected patients.20, 21 This case clearly demonstrates the need for definitive identification of mycobacterial isolates obtained from HIV-infected individuals with the most sophisticated methods of HPLC, molecular markers and sequence-based analysis. In areas of the world where such techniques are unavailable, isolates should be sent to reference laboratories for this purpose. Finally larger prospective studies utilizing mycobacterial blood cultures obtained at regular intervals in BCG-immunized HIV-infected infants appear warranted. Only in this way can the true incidence of disseminated BCG infection and its impact on HIV-infected infants and children be determined. Currently the World Health Organization and UNICEF policies support the routine immunization of asymptomatic infants in countries with high rates of both HIV and tuberculosis infection.22 For children living in the United States a joint communication from two Centers for Disease Control and Prevention advisory groups, the Advisory Council for the Elimination of Tuberculosis and the Advisory Council for Immunization Practices, has recommended that BCG vaccination should not be given to HIV-seropositive or known infected infants and children, even if the risk of tuberculosis is high. While additional studies are in progress, we endorse both of these policies. Acknowledgment. Work was supported by NIH Grants AI37871 (KME) and AI-35250 (DSK). Kathryn M. Edwards, M.D.; Douglas S. Kernodle, M.D. Division of Infectious Diseases Departments of Pediatrics and Medicine Vanderbilt University Medical Center Nashville, TN