683

Tuberculosis in Patients Infected with Human Immunodeficiency Virus: Perspective on the Past Decade Robert W. Shafer and Brian R. Edlin

From the Division of Infectious Diseases and Geographic Medicine, Stanford University Medical Center, Stanford, California; and the Division of HIVIAIDS, Centers for Disease Control and Prevention, Atlanta, Georgia

Tuberculosis (TB) is the most common opportunistic infection and the leading cause of death in persons infected with human immunodeficiency virus (HIV) worldwide. Because HIV is spreading in regions with the highest rates of Mycobacterium tuberculosis infection, HIV is responsible for an increasing proportion of the world's cases of TB. However, advances in molecular biology, clinical practice, and public health policy during the past 5 years offer reasons for hope. Molecular methods have provided insights into the epidemiology of M. tuberculosis transmission and the mechanisms of drug resistance. Rapid diagnostic tests have been developed to facilitate the diagnosis of TB. Retrospective and prospective studies have shown that TB in the HIV-infected person is highly treatable and often preventable. Moreover, directly observed therapy can decrease rates of treatment failure, relapse, drug resistance, and secondary spread. For two consecutive years, the incidence of TB in the United States has declined. Additional resources are needed, however, to achieve similar gains in the developing world.

When the first AIDS cases were diagnosed in 1981, onethird of the world's population was estimated to be infected with Mycobacterium tuberculosis [1]. Each year active disease developed in 8-10 million persons, and nearly 3 million persons died of tuberculosis (TB) [1]. As many as 7% of all deaths and 26% of preventable deaths in developing countries were caused by TB [1, 2]. Nonetheless, before the emergence of HIV, the vast majority of M. tuberculosis infections were kept in check by the host immune response and remained latent for the lifetime of the human host [3]. With host and pathogen in a standoff, the elimination ofTB from parts ofthe industrialized world was considered an achievable goal, and inroads into TB control had been made in many developing nations [1, 2, 4]. However, the worldwide spread of HIV infection has undermined human defenses against M. tuberculosis. HIV infection is the strongest risk factor for the progression of latent M tuberculosis infection to active TB [5-7]. The HIV pandemic has stalled the elimination of TB in the United States and has reversed many of the hard-won gains in TB control in the developing world [8-13]. By mid-1995, nearly 6 million persons worldwide were estimated to be coinfected with M. tuberculosis and HIV [12] (table 1); by the year 2000, an estimated

Received 8 August 1995; revised 17 November 1995. Use of trade names and commercial sources is for identification only and does not imply endorsement by the Public Health Service or the U.S. Department of Health and Human Services. Reprints: Dr. Brian R. Edlin, Division of HIV/AIDS, Centers for Disease Control and Prevention, 1600 Clifton Road, E-45, Atlanta, Georgia 30333. Correspondence: Dr. Robert W. Shafer, Division of Infectious Diseases, Room S-156, Stanford University Medical Center, Stanford, California 94305. Clinical Infectious Diseases 1996;22:683-704 © 1996 by The University of Chicago. All rights reserved. 1058-4838/96/2204-0013$02.00

14% of incident cases of active TB (about 1.4 million) will be attributable to HIV [12]. TB is the most common life-threatening HIV-related infection worldwide and is often the sentinel illness of HIV infection [11, 14]. An additional threat ofTB lies in its communicability through the air. During the past decade, much has been learned about the interaction of HIV and TB from research in molecular, clinical, and epidemiologic disciplines.

Pathogenesis M. tuberculosis Infection

M. tuberculosis is acquired by inhalation of infectious airborne particles small enough (,...., 1-5 microns) to reach the alveolar air spaces. The probability of infection depends on the intensity of exposure and probably also on the effectiveness of innate host defenses. Alveolar macrophages in some individuals may have a high degree of innate mycobacterial resistance, and in these persons the tubercle bacilli are presumably destroyed before infection is established [3, 15]. In other individuals the inhaled mycobacteria survive phagocytosis, replicate, and spread to regional lymph nodes and throughout the body. Although functional macrophage defects [16, 17] and abnormal lung surfactant [18] have been associated with HIV infection, it is not known whether HIV-infected persons are more susceptible than HIV-seronegative persons to acquisition of M. tuberculosis infection following exposure.

Active TB

The cell-mediated immune response to M. tuberculosis is characterized by complex interactions between different sub-

684

em

Shafer and Edlin

1996;22 (April)

Table 1. Estimated number of adults infected with TB and HIV in mid-1995. HIV- and TB-infected Region

HIV-infected (thousands)

TB-infected (%)

Sub-Saharan Africa South and Southeast Asia Latin America and Caribbean North America Western Europe North Africa and Middle East East Asia and Pacific Eastern Europe and Central Asia Australasia All regions

8,500 3,000 1,500+ 750+ 450 100+ 50+ 50+ 20 14-15,000

47 46 29 8 10 22 43 16 18

No. (thousands)

% of total

4,000 1,380 428+ 60+ 45 22+ 21+ 8+ 4 5,968+

67.0 23.1 7.2 1.0 0.8 0.4 0.3 0.1 0.1 100

NOTE. Data are from [13] and courtesy of the Tuberculosis Programme, World Health Organization, Geneva, Switzerland.

sets of lymphocytes and monocyte-macrophage cells [3, 19]. After ingesting mycobacteria, macrophages sensitize T lymphocytes by secreting IL-1 and presenting lymphocytes with processed mycobacterial antigens. M. tuberculosis-specific precursor lymphocytes are stimulated to proliferate and secrete 1ymphokines. These 1ymphokines in tum recruit circulating monocytes and induce their maturation into macrophages with enhanced phagocytic and microbicidal activity. In the ensuing granulomatous response, tubercle bacilli are killed by repeated cycles of phagocytosis, cytolysis, and exposure to microbicidal products. However, the immunologic response to M. tuberculosis is frequently not sterilizing, and surviving but dormant organisms often cause latent infection. Clinical disease occurs when the mycobacterial replication that follows initial infection cannot be controlled (progressive primary TB) or when latent organisms overcome immunologic control (reactivation TB). In '"" 5% of immunologically normal adults who become infected with M. tuberculosis, progressive primary TB develops within 2 years of initial infection. In another 5%, TB reactivates later in life [5]. CD4+ T lymphocytes are involved in many aspects of the immune response to M. tuberculosis, including binding to processed antigen, secreting cytokines, and killing mycobacteriainfected cells [3, 19]. HIV-induced CD4+ T-Iymphocyte depletion leads to a defective immunologic response to M tuberculosis [20-23]. HIV-infected persons with latent M. tuberculosis infection are at high risk of reactivation TB, and those with recently acquired M tuberculosis are at high risk of progressive primary TB. Active TB develops at an annual rate of 5%-12% in HIVinfected persons with previous M. tuberculosis infection [2430] (table 2). The risk ofTB is more than 25-30 times higher among HfV-infected persons than among HIV-seronegative controls [24-26]. Among HIV-infected persons, the risk ofTB is several times higher for those whose tuberculin skin tests

are positive rather than negative [24, 28-30], a finding which suggests that reactivation of latent M. tuberculosis is the most common mechanism of active TB. Rapid progression from recent M. tuberculosis infection to active TB (progressive primary TB) has been demonstrated in HIV-infected persons exposed to M. tuberculosis during institutional outbreaks [31-40]. HIV-infected persons are so vulnerable to progressive primary TB that active disease may develop within weeks of exposure to M. tuberculosis. In patients with advanced HIV infection, previous M tuberculosis infection is not always protective, and exogenous reinfection with a different strain of M. tuberculosis may occur [41]. HIV-infected patients with TB are usually less immunocompromised than HIV-infected patients with other AIDS-defining opportunistic infections, and their CD4+ T lymphocyte counts are generally in the range of 150-350/mm3 [42-52]. However, TB also may occur in HIV-infected persons with marked CD4+ T lymphocyte depletion as a result of newly acquired M tuberculosis infection.

Infectiousness

Although HIV infection may increase host susceptibility to M. tuberculosis infection and strongly increases the risk of progression to active TB, it may decrease the infectiousness of patients with TB. The presence of acid-fast bacilli (AFB) in a sputum smear and evidence of pulmonary cavitation on a chest radiograph are the best indicators of a patient's potential for transmitting M tuberculosis [9]. Most studies show that HIV-infected patients with TB have fewer AFB in their sputum and less frequent pulmonary cavitation than do HIV-seronegative patients with TB [42, 45, 53-59]. Indeed, the frequency of sputum-smear AFB-positivity and pulmonary cavitation decreases with increasing immunosuppression. In addition, tuberculin skin test reactivity rates among contacts of HIV-infected

CID 199;22 (April)

685

Tuberculosis and HIV

Table 2. Incidence of TB in HIV-seropositive (HIV+) and HIV-seronegative (HIV-) persons in various studies.

Type of study and/or participants (n)

Months of follow-up (median)

No. of TB cases

Incidence of TB (0/0 per year)

HIV -seropositive persons and HIV -seronegative controls Intravenous drug users, New York City [24] HIV +!PPD + (49) HIV+/PPD- (166) HIV -!PPD+ (62) HIV -/PPD- (236) Women of childbearing age, Zaire [25]* HIV+ (249) HIV- (310) Women of childbearing age, Rwanda [26] * HIV + (401) HIV- (917) HIV -seropositive persons, categorized according to tuberculin skin test results Intravenous drug users, New York City [24]

PPD- (166) PPD+ (49) Multicenter study, Italy [30] PPD- (849) Anergic! (1,649) PPD+ (197) Natural history study, Spain [28] PPD- (87) Anergic (235) PPD+ (87) Natural history study, Spain [29] PPD- (154) Anergic (112) PPD+ (84)

Impact on the Course of HIV Infection

Findings from several studies suggest that active TB may accelerate HIV-induced immunologic deterioration. First, active TB is associated with transient CD4+ T-Iymphocyte depression [65, 66]. Second, TB causes immune stimulation and increased production of cytokines, such as TNF [3, 67, 68], which increase HIV replication in vitro [69, 70]. Third, HIVinfected patients with TB appear to have a higher risk of opportunistic infections and death than do HIV -infected patients with similar CD4+ T cell counts but without TB [71]. Finally, in one study, preventive therapy with isoniazid for HIV-infected patients not only reduced the risk of active TB but also appeared to delay other opportunistic infections and death [27].

Epidemiology Resurgence of TB in the Uoited States

Between 1985 and 1992 the number of reported cases of TB in the United States increased by 19% [72]. During this interval

7.9 0.3

o o

32

19 1

3.1 * 0.1

24 26

20 2

0.1

21 22

7

0.3 7.9

19 15 17

6 62 15

0.5 3.0 5.4

16 16 18

0 8 11

2.6 8.4

29 17 30

19 20

30

NOTE. PPD+ = tuberculin skin test (purified protein derivative)-positive; PPD* Tuberculin skin testing was not performed in these studies. t Anergy was determined with use of multiple puncture skin tests.

patients with TB are generally lower than among contacts of HIV-seronegative patients with TB [60-64].

7 1

22 21 23 23

1

24

= tuberculin skin

2.5*

5,4 12,4 10,4

test-negative.

an estimated 52,000 more cases occurred than would have been expected had the downward trend of 1981-1984 continued [72, 73]. In 1993 and 1994 the number of reported TB cases decreased 5% and 4%, respectively, probably reflecting the effectiveness of recently introduced prevention and control measures [74, 75]. Epidemiologic evidence suggests that HIV has played an important role in the resurgence of TB in the United States. The largest increases in incidence of TB occurred in demographic groups and locations in which the prevalence of HIV was highest [76]. Between 1980 and 1992 the number ofcases ofTB increased > 150% in New York City, and between 1984 and 1990 the incidence of TB increased 400/0- 50% in California, Florida, and New Jersey [77, 78]. Among persons aged 25-44 years, the incidence of TB increased 52%; most of the increase occurred among blacks and Hispanic persons [9,76,79]. Between 1981 and 1991 at least 11,299 patients with AIDS in the United States also had TB [80]. Persons with AIDS were 59 times more likely to be found to have TB than the rest of the population, and persons with TB were 204 times more likely to be found to have AIDS than the rest of the population [80].

686

cm

Shafer and Edlin

1996;22 (April)

Table 3. Outbreaks of TB involving HIV-infected (HIV+) persons, 1988-1992. Type of TB, setting Drug-susceptible TB Hospital HIV unit, Verona, Italy [31]

Housing for HIV+ persons, San Francisco [33] Hospital HIV unit, Puerto Rico [38] Health clinic, Florida [32] Hospital, Texas [87]

Multidrug-resistant TB (MDR TB) >5 Hospitals in New York City [34-36, 40, 90]

Prison system, New York State (NYS) [91]

Hospital and clinic, Miami [39, 88] Substance-abuse treatment facility, Michigan [89] TB ward, New York City [41]

NOTE.

TST

= tuberculin

Comments

TB developed in 8 (44%) of 18 exposed HIV+ patients; CD4 cell counts were lower for exposed patients in whom TB developed than for exposed patients in whom it did not (232 vs. 562 cells/ul.; P < .01). TB developed in 11 residents within 4 months of exposure to the index case. HIV + patients sharing a room with a pulmonary TB patient were more likely to acquire TB than were other hospitalized HIV+ patients. A case-control study suggested an association with aerosolized pentamidine treatments. Thirty of 158 health care workers had TST conversions after exposure to an HIV-infected patient with pulmonary TB; the diagnosis of TB was initially obscured by simultaneous infection with P. carinii and M avium complex. Of >200 patients with TB, >80% were HIV-infected; mean incubation was 1-3.5 months in the different hospital outbreaks; multiple failures in infection control contributed to transmission. Thirty-eight of 39 inmates were HIV+; 29 inmates were infected with a strain resistant to isoniazid, rifampin, streptomycin, ethambutol, ethionamide, rifabutin, and kanamycin; inmates with MDR TB lived in 23 of the 68 NYS prisons while potentially infectious (12 were transferred through 20 prisons while ill with MDR TB); TST conversions occurred for ~30% of exposed inmates in one prison, 60 staff members in another prison, and > 50 health care workers. Sixty-two HIV-infected patients had MDR TB over a 3-year period. At least 15 and possibly as many as 31 exposed clients and staff members had TST conversions. Four HIV-infected patients hospitalized with drug-susceptible TB were reinfected with an MDR TB strain and had active TB within 2-9 months of initial hospitalization.

skin test.

Moreover, in the locations in which the greatest increases in the number of cases of TB have occurred, the prevalence of HIV among TB patients has been high. Studies of patients with TB in New York City, Miami, and San Francisco have revealed HIV prevalence rates of 30%-50% [45,46, 81-83]. Transmission of M. tuberculosis

The HIV epidemic contributed to the resurgence of TB by increasing the susceptibility ofHIV-infected individuals to both primary and reactivation TB. In addition, transmission of M. tuberculosis has probably increased. Over the past 2 decades, fiscal constraints led to cutbacks in many TB control programs [82, 84]. At the same time, the overlapping social problems of homelessness, substance abuse, and poverty have increased in populations affected by both TB and HIV and have limited the success of TB control in these groups. Studies employing restriction fragment length polymorphism (RFLP) analysis have suggested that perhaps one-third ofrecent cases of TB in New York City and San Francisco resulted from recently transmitted infections [85, 86]. In addition, TB outbreaks have occurred as persons with HIV and persons with active TB have been brought together in health care facilities and other institutional settings [31-40, 87-91] (table 3) (figure 1). In these outbreaks, HIV-related immunosuppression amplified and accelerated transmission of M. tuberculosis because

exposed HIV-infected patients often had active TB within weeks and then became additional sources of transmission. During the outbreaks, many health care workers' tuberculin skin test findings converted and in some active TB developed [37-40,87,91-95]. During the 1980s, the homeless and prison populations increased and included large numbers of HIV-infected persons [82, 86, 96-98]. TB among HIV-infected homeless persons became common [82, 86, 96, 97], and the number of TB cases increased sharply in the correctionalsystems of New York, California,New Jersey, and several other states [99-101]. The potential for M tuberculosis to spread within prisons was demonstrated by an outbreak in which a single highly drug-resistant organism was isolated from> 30 HIV-infected inmates who had been incarcerated in more than 20 different prisons [91, 102]. The recent increase in pediatric cases of TB is also evidence of increased transmission of TB. Between 1985 and 1991, the incidence of TB increased 36% among children 0-4 years old [76]; the largest increase occurred in New York City [103]. Some pediatric cases are due to transmission from HIV-infected adults, and some cases reflect the high risk of progression to active TB for children coinfected with HIV and M tuberculosis [104-108]. Drug-Resistant TB

Resistance of M. tuberculosis to drugs is caused by mutations in genes encoding the targets of anti-TB therapy [109-114].

CID 199;22 (April)

Tuberculosis and HIV

687

Figure 1. Restriction fragment length polymorphisms of M tuberculosis isolates obtained from HIV-infected patients with multidrug-resistant TB during a hospital outbreak (arrows) and from HIVinfected controls with drug-susceptible TB from the same hospital (reprinted with permission from [34]).

These mutations occur with a predictable frequency of one in 105 -1 08 organisms [115]. When anti-TB drugs are used in combination, growth of mutant M tuberculosis organisms resistant to any single drug is prevented by the other drugs in the combination. However, when organisms are exposed to only one effective drug because of incomplete or erratic antiTB therapy, drug resistance may develop. Once M tuberculosis becomes resistant to one drug, continued treatment or the addition of a single active drug to the treatment regimen may cause resistance to additional drugs [116]. In cases in which M tuberculosis becomes resistant to isoniazid and rifampin (i.e., multidrug-resistant TB [MDR TB]), treatment is often unsuccessful [117, 118]. Patients with active MDR TB may remain chronically ill and persistently infectious, and the condition is associated with high mortality. The prevalence of MDR TB in the United States increased from 0.5% during 1982-1986 to 3.5% during the first 3 months of 1991 [119, 120]. The highest rates of prevalence of MDR TB have been reported from New York City, New Jersey, and Florida [119]. In April 1991, 19% of all patients with TB in New York City whose M. tuberculosis cultures were positive had MDR TB [121]. Data regarding institutional outbreaks have demonstrated the high rate of disease progression among HIV-infected persons who become infected with M tuberculosis strains that are already multidrug-resistant (i.e., initial drug resistance). However, HIV-related immunosuppression does not appear to increase the likelihood that drug resistance will develop in a person infected with drug-susceptible M tuberculosis (i.e., acquired drug resistance). At one center in New York City, acquired drug resistance was more common among HIV-seronegative patients with MDR TB, whereas initial drug resistance

was more common among HIV-infected patients with MDR TB [122]. Indeed, RFLP analysis ofMDR TB strains from that center showed that most HIV-infected patients with MDR TB in 1990-1991 were infected with the same M. tuberculosis strains that had been isolated from HIV-seronegative patients in previous years [122].

TB and HIV Infection Outside the United States

Even before the AIDS pandemic, the countries of sub-Saharan Africa suffered disproportionately from TB. Approximately 50% of adults in sub-Saharan Africa are estimated to be infected with M tuberculosis, and the incidence of active TB may be as high as 200 per 100,000 persons [1, 10, 12]. In some urban areas 10%-30% of adults are HIV-seropositive [123-125], and 4 million Africans are estimated to be coinfected with HIV and M tuberculosis [11-13] (table 1). Nationwide notification rates and hospital-based studies suggest that the incidence of TB has more than doubled since the early 1980s in those countries in which the rates of HIV infection are highest [11, 123-129]. In some African cities, most hospital beds are occupied by HIV-infected patients, about one-half of whom have TB [14, 125, 126, 130-133]. Historically, the largest number of cases of TB have occurred in Asia, where HIV is spreading rapidly. Already, > 1.3 million adults in southeast Asia are estimated to be coinfected with HIV and TB (table 1) [13,134-136]. By the year 2000, because of the rapid spread of HIV in Thailand, India, and Myanmar (Burma), the number of cases of TB in Asia attributable to HIV may approximate the number of such cases in sub-Saharan Africa [12].

688

Shafer and Edlin

More than 400,000 adults in Latin America and the Caribbean islands are estimated to be coinfected with HIV and TB (table 1). TB has been noted in 26%-60% of Haitian persons with AIDS [137-139] and 7%-28% of Latin American persons with AIDS [13, 140]. In Spain and Italy, the incidence of TB appears to have increased as a result of TB among HIV-infected intravenous drug users [141]. The occurrence of TB in >30% of persons with AIDS in Spain [141-143] and in 2%-13% of HIV-infected persons in other western European countries has been reported [48, 49, 141, 144, 145].

Clinical Features In HIV-infected patients with TB, immunodeficiency is associated with increased dissemination of tuberculosis, increased number and severity of symptoms, and rapid progression to death unless treatment is begun [55]. Fever, weight loss, and other constitutional symptoms almost always occur. Cough, chest pain, and other respiratory symptoms are also common since most patients have some degree of pulmonary involvement. Shaking chills, hypotension, and acute respiratory distress may occur in patients with disseminated TB [55, 146148]. Localized signs and symptoms depend on the organs involved and coexisting HIV-related complications. Sites of Disease

Pulmonary TB occurs in 70%-90% of patients with TB, including most of those with extrapulmonary TB [6, 7, 49, 52, 149-151]. The frequency of extrapulmonary TB ranges from 40%-80% and increases with the severity ofimmunosuppression and the extent of diagnostic evaluation. Disseminated disease and lymphadenitis are the most common forms of extrapulmonary TB [55]. M tuberculosis bacteremia, extremely unusual in patients without HIV infection, has been noted in up to 20%-40% of HIV-infected patients with TB [152-155]. Cervical, supraclavicular, and axillary lymph nodes are the most common sites of peripheral TB lymphadenitis [21, 48, 55, 156-160]. The intrathoracic and intraabdominal lymph nodes, rare sites of TB in patients without HIV infection, are commonly involved in HIV-infected patients with advanced immunodeficiency [55, 161]. Tuberculous lymph nodes in HIV-infected patients appear to have an increased tendency to caseate, which may predispose these patients to abscesses, fistulas, and unusual sites of infection [55, 162, 163]. Tuberculous retroperitoneal lymph nodes may erode into the stomach or pancreas; mediastinal lymph nodes may erode into the esophagus, trachea, or bronchi; and mesenteric lymph nodes may erode into the lower intestine [55, 164-168]. CNS TB occurs in 5%-10% of HIV-infected patients with TB [50,55, 142, 143, 151, 169, 170]. Most have meningitis, but tuberculomas are also common [171-173]. Urine cultures are positive for most patients with disseminated TB, but local-

CID 1996;22 (April)

ized renal TB is rarely diagnosed [50, 55, 169]. Pleural disease and pericardial disease are commonly recognized forms of extrapulmonary TB in HIV-infected African patients [174-177]. TB of the skin and soft tissues may result from hematogenous seeding or contiguous organ involvement [55, 146, 162, 178-180]. Radiographic Findings

Chest radiographs of HIV-infected patients with TB and advanced immunosuppression are notable for evidence of nonapical distribution of infiltrates, infrequent cavitation, and an increased frequency of intrathoracic adenopathy, miliary infiltrates, and pleural effusions [6, 7, 53, 57, 181-183] (figure 2). Apical fibrocavitary infiltrates, the classic finding in adults with reactivation TB, occur predominantly in HIV-infected patients with TB who are not severely immunodeficient. Localized alveolar infiltrates may be confused with bacterial pneumonia, and diffuse interstitial infiltrates may mimic Pneumocystis carinii pneumonia. The occurrence ofhilar and/or mediastinal adenopathy, which is noted in about one-third of HIV-infected patients with TB, suggests the diagnosis of TB because intrathoracic adenopathy does not occur with most other HIV-related pulmonary complications. Miliary infiltrates and pleural effusions occur in > 10% of HIV-infected patients with TB and often develop during diagnostic evaluation. A normal chest radiograph does not preclude the diagnosis of pulmonary TB because radiographic findings may lag behind the rapid evolution of active TB [55, 182, 184]. In patients with intrathoracic adenopathy, CT scans usually demonstrate clusters of enlarged lymph nodes, often containing low-density centers consistent with caseous necrosis [55, 181, 185] (figure 3). In patients with disseminated TB, abdominal sonography and CT scans may demonstrate intraabdominal lymphadenopathy and focal hepatic and splenic lesions [55, 143, 163, 186-188] (figure 3). Tuberculin Skin Testing and Histopathology

The sensitivity of tuberculin skin testing in HIV-infected patients is inversely related to the degree of immunosuppression. Among HIV-infected patients with active TB, tuberculin reactions are ~ 10 mm in 40%-60% of those with otherwise asymptomatic HIV infection but in only 10%-30% of those with symptomatic HIV infection [6, 7, 42, 45, 46, 55, 83, 169, 189, 190]. The histopathologic appearance of TB in HIV-infected patients also depends on the degree of host immunity. Biopsy specimens from patients with early immunodeficiency tend to have granulomas composed of lymphocytes, epithelioid cells, and giant cells [21, 55, 191, 192], whereas those from more immunocompromised patients tend to contain necrosis, polymorphonuclear cells, and macrophages [55, 156, 193, 194] (figure 4). This histologic appearance ofTB contrasts with that

CID 199;22 (April)

Tuberculosis and HIV

689

Figure 2. Chest radiographs of patients with TB and HIV infection. A, right-middle-lobe infiltrate and widening of the mediastinum due to subcarinal, right paratracheal, and hilar lymphadenopathy; B, right-lower-lobe infiltrate and pleural effusion (courtesy of Dr. Bernard Suster, Saint Luke's Roosevelt Medical Center, New York; reprinted with permission from [183]).

of Mycobacterium avium complex, in which granulomas are either absent or small and nonnecrotizing [195]. Diagnosis Because the clinical features of HIV-infected patients with TB are often nonspecific, diagnosis can be difficult. Many HIVinfected patients with TB have died or been hospitalized for a prolonged period before TB has been diagnosed [55, 125, 149, 196-198]. Decreased tuberculin reactivity, atypical radio-

graphic presentations, and confusion with other HIV-related infections hinder the diagnosis ofTB in HIV-infected patients. However, failure to suspect TB and order the appropriate diagnostic tests is often the most common reason for diagnostic delays. TB should be considered when HIV-infected persons have unexplained fever, cough, pulmonary infiltrates, lymphadenopathy, meningitis, brain abscess, pericarditis, pleural effusions, or intraabdominal, musculoskeletal, or cutaneous abscesses. The probability of active TB is increased among patients who

Figure 3. CT scans of patients with TB and HIV infection. A, chest scan demonstrating multiple enlarged necrotic mediastinal lymph nodes; B, abdominal scan demonstrating a large necrotic periportal lymph node (arrow). (Courtesy of Dr. Bernard Suster, Saint Luke's Roosevelt Medical Center, New York; reprinted with permission from [183].)

690

Shafer and Edlin

CID 1996;22 (April)

Figure 4. Histopathologic appearance of biopsy specimens of lung (A) and liver (B) from patients with TB and HIV infection. Both specimens demonstrated focal areas of necrosis and cellular debris without lymphocytes, epithelioid cells, or giant cells, and both specimens contained many acid-fast bacilli on Fite staining (original magnification: A, X 200; B, X 100). (Courtesy of Dr. Ross Hill, State University of New York, Health Science Center at Brooklyn; reprinted with permission from [183].)

have a history of TB, whose tuberculin skin test is positive, or who have emigrated from a country or belong to a group in which the prevalence of TB is high (e.g., racial and ethnic minorities, homeless persons, intravenous drug users, alcoholics, and correctional facility inmates). In areas where nosocomial outbreaks have occurred, recent hospitalization should also be considered a risk factor for TB. The chest radiograph may show a classic reactivation pattern (apical fibrocavitary disease or miliary infiltrate), an "atypical" pattern (intrathoracic adenopathy, with or without single or multilobar infiltrates, or pleural effusion), or evidence of past infection (calcified lymph node or lung nodule, or pleural or parenchymal fibrosis). Although the sensitivity of the tuberculin skin test decreases with declining immunity, a positive test for symptomatic persons with advanced HIV infection suggests active TB [68, 199, 200]. AFB are found on microscopic examination of sputum specimens from 40%-67% of HIV-infected patients with TB; M tuberculosis is recovered from 74%-95% in culture [42, 45, 48, 49, 54-56, 59, 143, 169, 196, 201] (table 4). If adequate sputum specimens cannot be obtained, sputum should be induced with nebulized hypertonic saline [202]. Acid-fast staining and culture of gastric washings are also useful, particularly for infants and children. Fiberoptic bronchoscopy, with bronchoalveolar lavage and transbronchial biopsy, is indicated for patients with progressive unexplained pulmonary disease [203206]. During bronchoscopy, specimens from enlarged mediastinal lymph nodes may be obtained by endobronchial needle aspiration. Lymphatic, eNS, pericardial, pleural, and musculoskeletal TB may be suggested by physical examination findings. New signs often develop during diagnostic evaluation and may direct further tests or indicate new complications. Enlarged, tender, or fluctuant lymph nodes should be aspirated percutaneously.

In some series, as many as 90% of suspicious lymph nodes containAFB [50,55,143,169, 196,207,208] (table 4). Among patients with disseminated TB, biopsies of skin lesions have revealed AFB or granulomas [146, 179, 180,209]. Blood and urine should be cultured for mycobacteria; however, patients with M tuberculosis bacteremia will require treatment before these cultures become positive [152, 153]. Although AFB are often seen in the buffy coat smears of blood from patients with disseminated M avium complex infection

Table 4. Diagnostic yield of clinical specimens from HIV-infected patients with TB. Percent of patients for whom findings are positive Specimen Sputum Bronchoscopy Bronchoalveolar lavage Transbronchial biopsy Urine Blood Lymph nodes Bone marrow Liver biopsy CSF Pleural specimens Pleural fluid Pleural biopsy

Microscopy*

Culture

40-67

74-95

7-20 10-39 22

52-89 42-85 45-77 26-64 40-95 25-67 56-78

NAt

37-90 18-52 78 0-27 3-6 52-55

NAt NAt NAt

* Acid-fast bacilli seen on smears or granulomas seen on histopathologic specimens. t Not available. Blood smears have rarely been examined for HIV-infected patients with TB. The yields of cultures of CSF and pleural specimens are not shown because most reports present data regarding only M tuberculosispositive cultures of these specimens.

CID 199;22 (April)

Tuberculosis and HIV

[210], there has been only one report of a positive buffy coat smear of blood from an HIV-infected patient with M. tuberculosis infection [211]. Some studies have reported urine-culture positivity for most HIV-infected patients with extrapulmonary TB, and in several cases the diagnosis of TB has been based on the finding of AFB in smears of concentrated urine [50, 55, 143, 169, 212] (table 4). Although CSF and pleural fluid are generally abnormal in patients with involvement of these sites, stains for AFB are usually negative. In the absence of localized findings, biopsy of the bone marrow and/or liver may be effective in diagnosing disseminated TB [49, 55,213,214] (table 4). Abdominal sonography or CT scanning should be considered in difficult cases because these tests may demonstrate necrotic lymph nodes, which provide a high diagnostic yield when aspirated percutaneously [55, 163, 186]. The rapidly progressive nature of TB in HIV-infected patients requires that the diagnosis of suspected TB be pursued expeditiously. For patients whose initial evaluation is nondiagnostic, including AFB stains of sputum and other readily obtainable specimens, invasive procedures must be considered. For patients who are at high risk of TB or whose conditions are deteriorating rapidly, empirical anti-TB therapy should be started. Because sputum, urine, and blood cultures will most likely be positive for patients with fulminant TB, specimens for these cultures should be obtained before empirical therapy is begun to provide later confirmation of the diagnosis and to provide an M tuberculosis isolate for drug susceptibility testing. Microbiology The rapid progression of TB among HIV-infected individuals and the increase in prevalence of drug-resistant TB underscore the importance of rapidly identifying and determining the drug susceptibility of M tuberculosis strains. The preferred method for examining clinical specimens for AFB is fluorochrome staining, which is more rapid and slightly more sensitive than the Ziehl-Neelsen and Kinyoun stains [215]. Radiometric culture methods using liquid media (e.g., BACTEC [Becton Dickinson, Sparks, MD]) are recommended because they detect mycobacterial growth in 1-4 weeks, an average of 10 days before colonies can be seen on solid media [215, 216]. Solid culture media, however, should be used in conjunction with liquid media to detect mixed mycobacterial infections. Traditional methods for determining the species of mycobacterial isolates (e.g., on the basis of the growth rate, colonial morphology, pigmentation, and biochemical profile) are unacceptably slow for identifying M tuberculosis. A nucleic acid hybridization assay (Gen-Probe, San Diego, CA) can identify M. tuberculosis complex organisms within several hours after growth is detected [217]. Rapid hybridization assays also are available to identify M avium, Mycobacterium intracellulare, and several other nontuberculous mycobacteria. Because mixed

691

mycobacterial infections occur, a positive hybridization assay for another species, such as M avium, does not exclude the simultaneous presence of M. tuberculosis [87,218]. Mycobacterium bovis and M bovis BCG belong to the M. tuberculosis complex and have been isolated from HIV-infected persons [219- 221]. Differentiating these organisms from M. tuberculosis requires classic biochemical tests. HPLC, a popular method in some reference laboratories, can reliably identify any Mycobacterium species in < 4 hours on the basis of its mycolic acid profile [217]. HPLC is particularly useful in identifying mixed and unusual mycobacterial infections, which are common in HIV-infected patients [222- 225]. Because of the rising incidence of drug resistance, initial M. tuberculosis isolates from all patients should be submitted for antimicrobial susceptibility testing [120, 215]. Radiometric methods using liquid media can be employed to test susceptibility to isoniazid, rifampin, ethambutol, pyrazinamide, and streptomycin and will usually yield results 4- 7 days after the initial detection of mycobacterial growth [215]. Several gene amplification techniques have been developed that can detect M tuberculosis nucleic acid directly from clinical specimens within hours. Assays using peR (Roche Amplicor MTB; Roche Diagnostic Systems, Somerville, NJ) or transcription-mediated amplification (Gen-Probe amplified Mycobacterium direct test) have been evaluated extensively on clinical samples and are likely to be licensed soon [217, 226232]. In reference laboratories, both techniques are highly specific (>95%) and somewhat more sensitive than staining for AFB [230-235]. The sensitivity of these amplification techniques depends on the volume and means of specimen processing and on clinical circumstances. The usefulness of these techniques will be determined by their cost and their success in decreasing the need for invasive procedures and prolonged diagnostic evaluations.

Treatment The problems of HIV infection, drug resistance, and nonadherence with therapy have led to the publication of new guidelines for treating TB [236, 237]. The Centers for Disease Control and Prevention (CDC) and the American Thoracic Society (ATS) do not currently recommend longer treatment regimens for HIV-infected patients with TB than for HIVseronegative patients with the disease. Instead, the clinical and bacteriologic response to treatment of HIV-infected patients with TB should be followed closely, and therapy should be prolonged only for patients with a slow or suboptimal response [237]. Effectiveness of Therapy for DIV-Infected Patients

HIV-related immunosuppression does not interfere with the effectiveness of therapy for TB. Defervescence, sputum conversion, and resolution of chest radiographic abnormalities oc-

692

Shafer and Edlin

cur as rapidly in HIV-infected patients as in those without HIV infection [55, 58, 169, 197, 238-243]. Most early deaths among HIV-infected patients with TB result from undiagnosed TB or from nontuberculous HIV-related complications [55, 169, 197, 243-245]. Most treatment failures result from drug resistance or poor adherence with therapy, although anecdotal cases of treatment failure attributed to undrained tuberculous abscesses [246-248] or malabsorption of anti-TB medications [249-252] have been reported. Indeed, the rare occurrence of acquired rifampin resistance among HIV-infected patients with TB may result from malabsorption of isoniazid [252, 253]. HIV-infected patients with TB have a low risk of relapse during the first year after completing therapy. Of >500 HIVinfected patients who completed 6-9 months of standard antiTB treatment with regimens including at least isoniazid and rifampin and who were monitored for 12- 24 months, fewer than 5% relapsed, and some of those 5% adhered poorly to therapy [49, 169, 197, 240, 241, 254-256]. However, in the only completed prospective randomized study of HIV-infected patients with TB, 9% of those receiving therapy for 6 months relapsed, compared with only 2% of those receiving treatment for 12 months [243]. As HIV-infected patients are living longer because of advances in medical treatment, studies are urgently needed to assess the risk of relapse more than I - 2 years after the completion of anti-TB therapy. Studies are also needed to determine the prevalence and clinical significance of malabsorption of anti-TB drugs in HIV-infected patients. Initial Therapy (for Drug-Susceptible TB)

An initial three-drug regimen is currently recommended for use only in areas where careful surveillance has documented that drug resistance rates are lower than 4% [236]. For use in all other areas, a four-drug regimen consisting of isoniazid, rifampin, pyrazinamide, and either ethambutol or streptomycin is now recommended as initial therapy, to be administered while the results of drug susceptibility tests are pending [236] (table 5). Use of four-drug regimens reduces the likelihood that therapy for patients with drug-resistant M tuberculosis will fail and that organisms will develop resistance to additional drugs. In addition, four-drug regimens reduce infectiousness more rapidly than do three-drug regimens [236]. Therapy for Suspected or Proven Drug-Resistant TB

All patients with TB should be evaluated for possible drug resistance. Patients should be questioned thoroughly about previous preventive or curative therapy for TB and exposure to known cases of TB. The possibility of resistance to any drug the patient has received should be considered, and results of past susceptibility testing of isolates from the patient or from known contacts should be sought. Emigration from many developing countries could also be considered a risk factor for drug resistance [257].

em

1996; 22 (April)

Patients with drug-resistant TB should receive supervised therapy that is managed in consultation with clinicians who are experienced at treating such TB [236]. Resistance to either isoniazid or rifampin can usually be overcome by the substitution of other first-line drugs (table 5). The duration of therapy is usually determined by the extent of drug resistance, severity of TB, severity of immunodeficiency, and response to therapy. If resistance to both isoniazid and rifampin is suspected, the initial drug regimen should include isoniazid, rifampin, pyrazinamide, and three drugs to which local MDR TB strains are susceptible [236] (table 5). HIV-infected patients with MDR TB who are treated initially with at least two (or three) anti-TB drugs to which the causative organism is susceptible improve clinically, become noninfectious, and survive longer than patients treated with fewer effective drugs [194, 258261]. Isoniazid and rifampin should be withdrawn once resistance to these drugs is proved by antimicrobial susceptibility testing. For patients with MDR TB, it is necessary to determine susceptibilities to the second-line anti-TB drugs and to the quinolones, The optimal drugs for treating MDR TB include the other first-line anti-TB drugs (ethambutol, streptomycin, and pyrazinamide) and the quinolones (ofloxacin or ciprofloxacin) [117,118]. Administration of aminosalicylic acid, ethionamide, and cycloserine may need to be initiated in the hospital to permit observation of toxicity, intolerance, and initial response [118]. Resectional surgery should be considered for patients with extensive drug resistance, localized disease, and good cardiopulmonary reserve [262]. Although HIV-infected patients with MDR TB show improvement with appropriate therapy and become noninfectious [194, 258-261], it is not known whether discontinuing anti-TB treatment exposes such patients to a high risk of relapse. Nonadherence and Directly Observed Therapy

Although generally highly efficacious, therapy for TB requires a prolonged course of multiple medications that often have side effects. Because patients with TB often no longer feel ill after the first few weeks. of treatment, continuing antiTB therapy may become a low priority for them. Indeed, failure to complete anti-TB therapy is common in many parts of the United States [263]. Although persons leading disadvantaged and disorganized lives, such as homeless persons and substance abusers, are less likely than others to complete therapy, persons of all backgrounds have been nonadherent [264]. Adherence to TB therapy can be improved by "enablers" such as transportation and short waiting times, incentives such as meals or money, and a trusting relationship between patient and health care worker. The use of formulations with multiple drugs of demonstrated bioavailability, such as Rifater (isoniazid, rifampin, and pyrazinamide; Marion Merrell Dow, Kansas City, MO) and Rifamate (isoniazid and rifampin; Marion Merrell Dow) may enhance adherence and, by preventing discontin-

CID 199;22 (April)

Tuberculosis and HIV

693

Table 5. Treatment regimens for HIV-infected adults with TB. Clinical circumstances and treatment considerations DOT

DOT not considered necessary to ensure patient's compliance

Initial therapy*

Continuation phase of therapy

INH, Rif, PZA, and Eth or Stm daily for 2 w and then 2-3 X/w for 6 w or INH, Rif, PZA, and Stm or Eth 3 x/w for 6 mo INH, Rif, PZA, and Stm or Eth daily (pending susceptibility data) and then INH, Rif, and PZA to complete 8 w of therapy with these 3 drugs

Resistance (or intolerance) to INHt Resistance (or intolerance) to Rift Possible or confirmed resistance to both INH and Rif (cases of MDR TB)t

INH, Rif, PZA, and Eth or Stm, plus additional 2ndline drugs or a quinolone antibiotic, so that patient receives ~3 drugs to which local MDR TB strains are likely to be susceptible

INH and Rif 2-3 X/w to complete 6 mo of treatment

INH and Rif daily to complete 6 mo of treatment

Rif, Eth, and PZA X 18 mo (and for ~ 12 mo after culture conversion) INH, Eth, and PZA X 18 mo (and for ~ 12 mo after culture conversion) ~ 3 drugs to which patient's M tuberculosis strain is susceptible; appropriate duration of therapy is not known

NOTE. Data are from [118, 236, 237]. DOT = directly observed therapy; INH = isoniazid, 5 mg/kg (maximum [max], 300 mg) for daily therapy or 15 mg/ kg (max, 900 mg) for intermittent therapy; Rif = rifampin, 10 mg/kg (max, 600 mg) for daily and intermittent therapy; PZA = pyrazinamide, 15-30 mg/kg (max, 2 g) for daily therapy or 50-70 mg/kg (max, 8-9 g/w) for intermittent therapy; Eth = ethambutol, 15-25 mg/kg (max, 2.5 g) for daily therapy or 25-50 mg/kg (max, 2.5 g) for intermittent therapy; Stm = streptomycin, 15 mg/kg (max, 1 g) for daily therapy or 25-30 mg/kg (max, 1-1.5 g) for intermittent therapy. * In areas where surveillance for drug-resistant TB has documented drug resistance rates of 10% of fully compliant HIV-seronegative patients with TB [1] and is even less effective for HIV-infected patients [245, 278-280]. Furthermore, 10%-20% of HIV-infected patients

694

Shafer and Edlin

CID 1996;22 (April)

A _

2.0~

~ !~i t: :::s

•

Relapse

0.0 ........._---I~_--'- _ _...&...._ _I....__

I

__a.._ _...&....

C.

o c. o o o Q o

1.6 1.2

~

•

....a-_ _I . . . . _ _ - - ' -_ _

~

Primary resistance

c.

~

S ca

~

~

_ _..a...__

~

....CI)

I

___a._ ___'__

1.6,.....-----------------......, Acquired resistance 1.2 0.8 0.4 0.0

L...L _ _...L..-_ _1..-_-1._--:;:::::lI==-~.L....::=___..~

1980

In 4,000 CI)

~::~

0.0

B

1982

1984

1986

1988

1990

1992

In

ca

o

3,000

"-

o

o z

2,000 1,000

198319841985198619871988198919901991199219931994

Year

Figure 5. A, incidence of tuberculosis relapse, initial drug resistance, and acquired drug resistance in Tarrant County, Texas, before and after a comprehensive, directly observed therapy (DOT) program was instituted in 1986 (adapted from [268]); B, number of TB cases in New York City between 1983 and 1994 (dots) and the association of a decline from 1992 to 1994 with an increase in the number of persons receiving DOT (bars) (adapted from [269]).

receiving thiacetazone experience severe and occasionally fatal cutaneous hypersensitivity reactions [174, 281-283]. For these reasons, the World Health Organization (WHO) is attempting to obtain the resources to increase the use of supervised shortcourse chemotherapy with rifampin-including regimens in countries in which the prevalence of HIV and TB is high [14].

Prevention Preventive Therapy

Preventive therapy with isoniazid decreases the risk of active TB in HIV-infected persons latently infected with M tuberculosis (table 2) [24, 27, 28, 30, 256, 284, 285]. In a controlled trial in Haiti, for example, the incidence of TB over a 3-year period was >5-fold lower among tuberculin-reactive HIV-infected patients receiving isoniazid for 12 months than among tuberculin-reactive HIV-infected patients receiving placebo [27]. In a controlled trial in Zambia, a 6-month course of isoniazid decreased the risk of active TB, although the incidence of TB among isoniazid recipients gradually increased during the postprophylaxis period [284]. Because HIV-infected persons with latent M. tuberculosis infection have an extraordinary risk for reactivation and because preventive therapy can reduce that risk, identifying individuals dually infected with M. tuberculosis and HIV is critically important. Tuberculin skin testing is therefore recommended for all HIV-infected persons [286]. Skin testing and preventive therapy should be available to persons at high risk for dual infection; these include prison inmates, residents of homeless shelters, and clients of drug treatment programs [287].

Patients whose tuberculin skin test is positive require chest radiography and careful assessment to exclude both pulmonary and extrapulmonary TB, because administration of isoniazid alone to patients with active TB will select for isoniazid-resistant strains. After active TB is excluded, all HIV-infected persons whose tuberculin skin test is positive should receive isoniazid for 12 months unless such treatment is medically contraindicated [237]. Because TB may develop in HIV-infected persons following exogenous reinfection [41], preventive therapy after new exposure to an infectious case of TB should be considered, even for persons who have previously been treated for active TB or have received prophylaxis for latent TB. HIV-infected persons who have been significantly exposed to infectious MDR TB should receive preventive therapy with a combination of2 or 3 drugs to which the multidrugresistant organism is susceptible [237, 288]. The sensitivity of tuberculin skin testing for detecting latent M tuberculosis infection is reduced in HIV-infected persons. In several studies, HIV-infected persons have been less likely than matched HIV-seronegative controls to have positive tuberculin skin tests [97, 289-293] (table 6). To improve the sensitivity of tuberculin skin testing in HIV-infected persons, the CDC and ATS recommend that, in this population, induration of ~5 mm should be considered positive [237, 294]. Mantoux tuberculin skin testing with 5 tuberculin units of PPD should be done as early as possible in the course of HIV infection because the utility of tuberculin testing declines as immunodeficiency increases [190]. The yield of tuberculin skin testing is increased slightly if those with negative tests are retested after ~ 7 days [295]. Some experts recommend that HIV-infected persons whose tuberculin tests are negative be tested for skin-test anergy with at least two control antigens-such as mumps, tetanus toxoid,

em 199;22 (April)

695

Tuberculosis and HIV

Table 6. Results of tuberculin skin tests (TSTs) for asymptomatic HIV -seropositive (HIV +) persons and HIV -seronegative (HN - ) controls. Percent of subjects for whom TST was positive

TST cutoff* (mm) Subjects (study location and period), reference Postpartum women (Uganda, 1988-1989) [289] Adult residents (Haiti, 1990-1991) [290] Injection drug users (Baltimore, 1990) [291] Homeless adults (San Francisco, 1990-1992) [97] Intravenous drug users and homosexual men (pulmonary complications of HIV study group, United States, 1988-1990) [292]

HIV+

HIV-

HIV+

HIV-

P value

3 10 5 5 2 5

3 10 5 10 10 10

48 52 65 14 20 19

82 63 67 25 25 37