An estimated 1.7 billion persons, one third of the world's population, are infected with Mycobacterium tuberculosis. This reservoir of infected persons results in 8 million new cases of tuberculosis and 2.9 million deaths annually [4]. Infection with human immunodeficiency virus (HIV) is the most potent risk factor for development of tuberculosis [5], and 3.1 million people are coinfected with HIV and M. tuberculosis [6]. In African countries where coinfection with these agents is most common, the number of tuberculosis cases reported annually has increased up to 100% during the past 4 to 5 years [4]. This number is projected to increase by 41% to 463% from 1989 to 2000, depending on the local prevalence of infection with M. tuberculosis and HIV [7].

Tuberculosis in the United States

In the United States, the number of annually reported patients with tuberculosis decreased from 84 304 in 1953 to 22 201 in 1984, then increased to 26 283 in 1991 [1-3]. This increased morbidity occurs largely in specific epidemiologic groups, including racial and ethnic minorities as well as foreign-born persons (Centers for Disease Control and Prevention [CDC]. Unpublished data; Table 1). In minorities, tuberculosis has increased markedly in persons who are 25 to 44 years old, probably because of the increased prevalence of HIV infection in this age group [1, 5]. Among cases in the foreign-born, 82% were reported from eight states, and 60% were diagnosed within 5 years of arrival to the United States [2, 8]. Tuberculosis is concentrated in groups characterized by disproportionately greater percentages of blacks and Hispanics, increased prevalence of HIV infection, crowded living conditions, and inadequate access to health care. Infection with M. tuberculosis is found in as many as 50% of homeless persons [9], and active tuberculosis is diagnosed in as many as 18% of homeless persons living in shelters, where HIV infection is common [10]. Prison inmates and migrant farm workers are also at increased risk for tuberculosis. From 1980 to 1990, the incidence of tuberculosis in New York state prisons increased six times to 134 per 100 000 [11], more than 12 times the national case rate. Tuberculosis case rates in U.S.-born blacks who are migrant farm workers are more than 3000 times the national rate and are higher than those in foreign-born Hispanic migrant workers, suggesting that tuberculosis in this group is an occupational problem rather than an imported one [12].

Compounding factors that increase risk for tuberculosis, such as homelessness and HIV infection, have yielded explosive epidemics, as in New York City, where the number of tuberculosis cases increased 143%, from 1514 to 3682 (1980 to 1991; CDC. Unpublished data). In central Harlem, the incidence of tuberculosis tripled from 51 patients per 100 000 in 1979 to 169 patients per 100 000 in 1989, approaching the incidence of tuberculosis in parts of sub-Saharan Africa [15].

Tuberculosis in Patients with HIV Infection

Because this topic was recently reviewed [5], only a few salient features will be noted here. Tuberculosis is often the initial manifestation of HIV infection, and serologic testing for HIV infection is recommended in all tuberculosis patients [16]. Tuberculosis in HIV-infected patients is characterized by extrapulmonary disease in as many as 70% of patients. Chest roentgenographic findings are typical of primary disease, with hilar adenopathy, pleural effusion, and miliary disease. DNA fingerprinting of M. tuberculosis isolates from tuberculosis outbreaks has shown that 37% of HIV-infected patients develop primary tuberculosis within 5 months of exposure to a source patient [17-19]. In comparison, only 2% to 4% of immunocompetent household contacts of tuberculosis patients develop tuberculosis within 12 months [20]. Furthermore, of HIV-infected patients with positive tuberculin skin tests from previous tuberculous infection, 10% of those who do not receive chemoprophylaxis develop tuberculosis each year [21]. This risk greatly exceeds the estimated 10% lifetime risk for tuberculosis in immunocompetent adults with positive tuberculin skin tests.

Multiple-drug-Resistant Tuberculosis

Ingestion of single antituberculosis agents for prolonged periods or erratic compliance with therapy fosters emergence of drug-resistant M. tuberculosis. Multidrug-resistant organisms (resistant to both isoniazid and rifampin) have caused more than 200 cases of tuberculosis in outbreaks at hospitals and prisons in New York and Florida [17, 22-25]. Ninety-six percent of the patients with multidrug-resistant tuberculosis were infected with HIV; the case fatality rate was 80%, and death occurred a median of 4 to 16 weeks after diagnosis [17]. Of 8 health care workers in these hospitals who have developed multiple-drug-resistant tuberculosis, 6 were infected with HIV, and 4 of these 6 have died [17].

Additional cases of multiple-drug-resistant tuberculosis are expected as more infected persons develop the disease. Tuberculin skin test conversion was noted in 19 (37%) of 51 health care workers exposed to multidrug-resistant tuberculosis [17], and the extent of infection in other exposed persons remains undetermined. Reactivation disease is expected to develop at a rate of 10% per year in patients infected with HIV and coinfected with multidrug-resistant M. tuberculosis, in whom primary tuberculosis has not developed.

The outbreaks of multiple-drug-resistant tuberculosis in HIV-infected patients show the potential for nosocomial transmission of tuberculosis that can be rapidly lethal. Factors that contributed to these outbreaks include delayed diagnosis of tuberculosis, delayed recognition of drug resistance with resultant prolonged ineffective therapy, failure to observe recommended isolation precautions, and lack of appropriate negative-pressure ventilation in isolation rooms [17, 23].

New Techniques for Diagnosis of Tuberculosis

Radiometric Culture Methods and DNA Probes

The acid-fast stain is the only widely available test for a rapid presumptive diagnosis of tuberculosis. Identification of M. tuberculosis by traditional culture techniques, followed by biochemical tests, requires 4 to 8 weeks. Radiometric culture methods (Bactec, Johnston Laboratories, Inc., Towson, Maryland), combined with a DNA probe for an M. tuberculosis-specific ribosomal RNA sequence (Gen-Probe, Inc., San Diego, California), allow identification of M. tuberculosis in 1 to 3 weeks [26, 27]. Several thousand copies of target RNA are present per organism, enhancing the sensitivity of the test. This DNA probe can be used in mycobacterial cultures but is not sufficiently sensitive to detect organisms in clinical samples.

Amplification of Mycobacterial DNA by Polymerase Chain Reaction

The most promising technique for rapid detection of M. tuberculosis is based on the polymerase chain reaction (PCR). This assay yields multiple copies of M. tuberculosis-specific target nucleotide sequences, including ribosomal RNA [28], the putative insertion sequence IS6110/IS986 [29-31], the DNA sequence mtp40 [32], and DNA encoding the 38-kd and 65-kd proteins [33-35], and MPB64 [36]. Most assays involve two steps: 1) mycobacterial DNA in clinical samples is amplified by PCR and subjected to gel electrophoresis; 2) PCR product is transferred to nylon membranes and hybridized to probes that bind M. tuberculosis-specific DNA. For radiolabeled probes, a positive result is radioactivity incorporation. For probes bound to an enzyme that allows a colorimetric assay, a positive result is indicated by increased optical density.

Because PCR amplification can yield million-fold increases in target DNA, strict precautions are essential to avoid false-positive results through cross-contamination with aerosolized PCR product. In addition, controls are critical to confirm adequate recovery of mycobacterial DNA from samples, efficient PCR amplification in each reaction mixture, and absence of contaminating M. tuberculosis DNA in reagents [31].

One assay that is particularly advantageous is based on detection of a target sequence within IS6110, found only in the M. tuberculosis complex. Most M. tuberculosis strains contain 6 to 15 copies of IS6110 [27, 37], enhancing sensitivity of this assay, compared with those that detect single-copy target sequences. After amplification, PCR product is electrophoresed on acrylamide gels and stained with ethidium bromide. A positive result is represented by observing a band of expected molecular weight [29, 31]. Unlike other PCR-based assays, the high sensitivity and specificity of this method obviate the need for hybridization of PCR product to DNA probes, simplifying the test for routine clinical use. This assay was 100% sensitive in 46 acid-fast, smear-positive sputum samples from tuberculosis patients and 99% specific in 68 samples from patients without tuberculosis [31]. More experience is needed to evaluate the sensitivity of this test in acid-fast, smear-negative samples.

Immunoassay for Mycobacterium tuberculosis Antigens

Mycobacterial antigens are detectable in cerebrospinal fluid of patients with tuberculous meningitis by enzyme-linked immunoabsorbent assay (ELISA) or radioimmunoassay [38, 39]. In the ELISA, cerebrospinal fluid samples are added to microtiter wells containing plastic-bound antibodies to mycobacterial antigens. Bound antigen is detected by addition of a second anti-mycobacterial antibody, conjugated to an enzyme that produces a color change after addition of appropriate substrate. The sensitivity of immunoassay varies from 39% to 79%, and the specificity is 98% to 100%, with rare false-positive results in patients with bacterial meningitis [38, 39]. Because the sensitivity of the acid-fast stain of cerebrospinal fluid is less than 30% in most series, these assays have the potential to greatly facilitate rapid diagnosis of tuberculous meningitis.

Detection of mycobacterial antigens by ELISA in sputum is difficult because most soluble antigens are discarded or denatured during the processing of sputum samples. The sensitivity of antigen detection was 76% to 89% in two small series [40, 41], but larger confirmatory studies are needed. The specificity of these tests is uncertain because patients with nontuberculous mycobacterial disease were not evaluated.

Biochemical Tests for Mycobacterium tuberculosis Components

In pulmonary specimens from tuberculosis patients, detection of tuberculostearic acid by gas chromatography and mass spectrometry is more sensitive than the acid-fast stain [42, 43]. However, specificity is a problem because tuberculostearic acid is present in nontuberculous mycobacteria and in other Actinomycetales, some of which are oropharyngeal commensals that may be present in sputum. Mycolic acids specific to M. tuberculosis can be identified in mycobacterial cultures by high-performance liquid chromatography [44], but this method has not been applied to clinical specimens. High equipment costs for these biochemical techniques limit their potential use to reference laboratories.

Immunoassay for Anti-Mycobacterial Antibodies

The ELISA can detect anti-mycobacterial antibodies in serum and other body fluids. Clinical samples are added to plastic-bound mycobacterial antigens. Bound antibodies are detected by addition of antibodies to human immunoglobulin, which is conjugated to an enzyme that permits a colorimetric assay. In one study, the sensitivity and specificity of ELISA was 69% and 88%, respectively, compared with 79% and 100% for the sputum acid-fast stain [45]. The sensitivity of antibody detection is reduced in smear-negative tuberculosis [46], presumably because lower bacillary burdens yield decreased antibody production. Sensitivity of antibody detection is also low in HIV-infected persons [47, 48], probably because immunologic dysfunction causes diminished antibody production.

In summary, widely available radiometric culture methods, combined with a DNA probe, permit identification of M. tuberculosis within 3 weeks. The other diagnostic tests outlined above are confined to research laboratories, and some of them may become valuable clinical tools. We believe that PCR-based methods are the most promising, with the greatest potential for high sensitivity and specificity.

DNA Fingerprinting

Because most M. tuberculosis strains share drug susceptibility patterns and bacteriophage types, it has been difficult to document transmission of specific strains from person to person. Identification of individual M. tuberculosis strains is now possible through "DNA fingerprinting," based on analysis of the distribution of the insertion sequence IS6110 within the M. tuberculosis genome. In different strains, copies of IS6110 vary in number and location within the chromosomes, but the nucleotide sequence of IS6110 is conserved [30, 49]. DNA fingerprinting involves digestion of M. tuberculosis DNA with a restriction enzyme such as BamHI, which cleaves IS6110 at a single site. Because the location of the BamHI site within IS6110 is constant, but the chromosomal position of IS6110 differs across strains, the distance between the IS6110 BamHI site and the next BamHI site on adjoining DNA is variable, yielding fragments of different sizes (Figure 1). The number of DNA fragments released for a given M. tuberculosis isolate equals the number of copies of IS6110 in that isolate. The number and sizes of the DNA fragments are determined using gel electrophoresis, transfer of DNA to nylon membranes, and hybridization to a radiolabeled DNA probe that binds IS6110 sequences flanking the restriction enzyme site. The banding pattern of a specific number of DNA fragments of specific sizes constitutes a "fingerprint" unique to each strain of M. tuberculosis.

View larger version (15K): [in this window] [in a new window]   Figure 1. Production of DNA fragments by cleavage of Mycobacterium tuberculosis DNA using BamHI. IS6110 and adjoining DNA sequences of two hypothetical strains of M. tuberculosis are shown. The BamHI site of flanking DNA in strain A is closer to the end of IS6110 than is the corresponding site in strain B. Digestion of chromosomal DNA with BamHI, followed by hybridization with a radiolabeled DNA probe yields a 550-base pair fragment for strain A and an 850-base-pair fragment for strain B.

Among tuberculosis patients in whom there was no epidemiologic linkage to a common source patient, each M. tuberculosis isolate had a distinct DNA fingerprint. In contrast, identical DNA fingerprints were shared by almost all isolates from tuberculosis patients within the same household or congregate living facility [18, 37, 50]. In two intrafamilial clusters of tuberculosis, the isolate from one family member had a DNA fingerprint identical to that of isolates from other family members, except that it contained one additional DNA fragment. This indicates the presence of one extra copy of IS6110, presumably due to a replicative event of IS6110 during the course of disease transmission.

Development of drug resistance by M. tuberculosis did not affect the DNA fingerprint [37]. Strains of M. tuberculosis showed no change in DNA fingerprints after culture for 2 to 6 months [37, 51], although one strain cultured for 3 years developed differences in a few DNA fragments but retained a similar overall DNA banding pattern [50]. Thus, DNA fingerprints of individual M. tuberculosis strains remain relatively stable over time and permit definitive delineation of patterns of tuberculosis transmission.

Treatment of Tuberculosis

Drug Regimens

The microbiologic principles of antituberculosis therapy have been recently reviewed [52]. In the United States, the first-line drugs are isoniazid, rifampin, pyrazinamide, and ethambutol. The usual daily dosages for adults are 300 mg of isoniazid, 600 mg of rifampin, 15 to 30 mg/kg (maximum 2 g) of pyrazinamide, and 15 to 25 mg/kg (maximum 2.5 g) of ethambutol [53-55]. Hepatotoxicity is the major adverse effect of isoniazid, rifampin, and pyrazinamide; optic neuritis can result from ethambutol at dosages of 25 mg/kg per day. Second-line antituberculosis drugs are streptomycin, kanamycin, capreomycin, ethionamide, cycloserine, ofloxacin, and ciprofloxacin.

Several treatment regimens for drug-susceptible tuberculosis yield cure rates of more than 95%. In the United States, the most widely used regimen is isoniazid, rifampin, and pyrazinamide for 2 months, then isoniazid and rifampin for 4 months [56]. In HIV-infected persons, isoniazid and rifampin are continued for 7 months [5, 16]. Isoniazid, rifampin, and pyrazinamide are given daily for the first 1 to 2 months, and the rest of the therapy is given daily or biweekly. Treatment with isoniazid, rifampin, pyrazinamide, and ethambutol three times a week for 6 months is also highly effective [57]. Administration of isoniazid and rifampin for 9 months is effective when drug resistance rates are low [58]. However, 6-month regimens with initial use of three to four drugs are preferable for most patients because the frequency of drug resistance is increasing, and the shorter duration of therapy enhances compliance [56].

For isoniazid-resistant organisms, cure rates of more than 90% are achieved using 6-month regimens of isoniazid, rifampin, pyrazinamide, and ethambutol or streptomycin [57, 59, 60] or using 12-month regimens of rifampin and ethambutol (with or without pyrazinamide) [55]. In HIV-infected persons with isoniazid-resistant organisms, 18 months of rifampin, ethambutol, and pyrazinamide is recommended [5, 16]. Resistance to both isoniazid and rifampin markedly decreases the efficacy of treatment, with failure rates of 40% to 70% [53, 57, 59, 60]. Individualized drug regimens, based on susceptibility testing, generally include at least three drugs to which the organism is likely to be susceptible. Lung resection is helpful in selected patients [61].

Ofloxacin and ciprofloxacin are mycobactericidal quinolones that are well tolerated and penetrate most tissues well [62]. Ofloxacin is active against M. tuberculosis in human macrophages [63] and has been used more extensively than ciprofloxacin in patients with drug-resistant tuberculosis [64, 65]. A daily dose of 800 mg of ofloxacin yielded promising results and was more effective than 300 mg daily. Long-term cure rates remain undetermined; controlled studies evaluating these agents are sorely needed.

Rifamate (Marion Merrell Dow, Inc., Kansas City, Missouri), a combination capsule of 150 mg of isoniazid and 300 mg of rifampin, provides clinically significant advantages over separate formulations of isoniazid and rifampin because it encourages patients to take both drugs as prescribed, thereby reducing the risk for emergence of drug resistance. A combination pill of isoniazid, rifampin, and pyrazinamide is not approved for use in the United States. Bioavailability of the individual agents is good [60, 66], and the combination pill was as effective as separate formulations of the drugs [60, 67].

Initial Therapy

Because of increasing drug resistance in many parts of the United States, drug susceptibility testing should be done on all M. tuberculosis isolates [24]. Isoniazid, rifampin, and pyrazinamide should be given to patients in whom drug-resistant tuberculosis is unlikely (for example, persons born in the United States who have not been previously treated and who are unlikely to have been exposed to drug-resistant M. tuberculosis). Patients at increased risk for drug resistance should receive isoniazid, rifampin, pyrazinamide, and ethambutol until drug susceptibility results are available [55, 68]. Immigrants from developing nations are at increased risk for tuberculosis resistant to isoniazid and streptomycin, even without previous antituberculosis therapy [68, 69]. Patients who have received previous therapy, particularly those who have taken medication erratically, may have organisms resistant to both isoniazid and rifampin [68]. In regions where multiple-drug-resistant tuberculosis is common, initial regimens should be based on local drug susceptibility patterns. Local surveillance will identify patient groups at increased risk for drug resistance and will guide the choice of initial therapy.

Extrapulmonary Tuberculosis

Despite published recommendations to the contrary [55], physicians commonly give more prolonged therapy for extrapulmonary tuberculosis than for pulmonary disease. Little microbiologic basis exists for this practice because the bacillary burden in extrapulmonary tuberculosis is lower than that in cavitary pulmonary disease and because isoniazid, rifampin, and pyrazinamide penetrate most tissues well. Like pulmonary tuberculosis, most forms of extrapulmonary tuberculosis should be treated for 6 months [55, 70]. Only in miliary, meningeal, and skeletal tuberculosis is more prolonged therapy recommended, the duration of which is determined on a case-by-case basis. In skeletal tuberculosis, early surgical debridement is important because antituberculosis drugs may not penetrate devitalized bone and sequestrum [55].

Sputum Smear-Negative Tuberculosis

Tuberculosis patients in whom acid-fast smears of sputum are negative have fewer organisms than do those with smear-positive disease. The bacillary burden in patients with negative cultures is probably even smaller. In persons with smear-negative, culture-positive tuberculosis, 4 months of isoniazid, rifampin, pyrazinamide, and streptomycin, or 6 months of isoniazid and rifampin, cured more than 97% of patients with drug-susceptible organisms [71, 72]. In patients with negative cultures, 4 months of isoniazid and rifampin treatment was successful in more than 95% of patients [73]. These studies included patients in whom at least three sputum samples were obtained for mycobacterial studies and should not be extrapolated to patients in whom fewer sputum samples are obtained, or in whom drug resistance is suspected.

Pregnancy

In pregnant patients with tuberculosis, isoniazid and rifampin are used, together with ethambutol if drug resistance is possible [55, 74]. Streptomycin is contraindicated because of fetal ototoxicity. Pyrazinamide is usually avoided because of inadequate data about teratogenicity. Exclusion of pyrazinamide precludes use of 6-month regimens and mandates 9 months of therapy for drug-susceptible tuberculosis.

Corticosteroids

A recent randomized trial showed that corticosteroids reduced the case fatality rate and frequency of neurologic sequelae in patients with positive cerebrospinal fluid cultures for M. tuberculosis [75]. However, no benefit was observed in comatose patients or in those with negative cerebrospinal fluid cultures, suggesting that corticosteroids benefit specific groups of patients. Furthermore, rifampin and pyrazinamide were not used in this study, preventing extrapolation of results to patients treated in the United States. Additional studies are needed to define the therapeutic role of corticosteroids in tuberculous meningitis.

In a prospective, double-blind trial of patients with tuberculous pericarditis, corticosteroids caused more rapid normalization of tachycardia and jugular venous pressure but did not decrease the number of deaths or the need for pericardiectomy [76]. Corticosteroids may be warranted in severely ill patients with tuberculous pericarditis. In tuberculous pleuritis, corticosteroids hasten resolution of symptoms and resorption of pleural fluid [77]. However, because clinically significant sequelae are rare when pleuritis is treated with antituberculosis drugs alone, corticosteroids are not widely used. Corticosteroids may be appropriate in patients with severe systemic toxicity, persistent pleuritic chest pain, or slow resolution of effusions.

Monitoring during Therapy

In patients receiving effective antituberculosis therapy, symptoms typically improve within the first 4 weeks, and sputum cultures become negative within 3 months [55]. In patients with delayed resolution of symptoms or persistently positive cultures, noncompliance or drug-resistant tuberculosis should be suspected. Because noncompliance is far more common, directly observed therapy is advisable for most patients, with a health care provider observing the patient ingest the medications 2 to 5 times weekly. When drug resistance is suspected, one should add at least two drugs to which the organism is likely to be susceptible. A single agent should never be added to a failing regimen because it invites development of resistance to that agent.

Optimally, sputum cultures should be obtained monthly during therapy until they are negative and also obtained after completion of therapy. At a minimum, they should be obtained after 3 months and at the end of therapy [55]. Routine chest roentgenograms are less useful than sputum cultures for evaluating the success of therapy. Nevertheless, it is reasonable to obtain a routine chest roentgenogram after 2 to 3 months of therapy and after completion of treatment. These tests provide another parameter to assess the efficacy of therapy and provide a baseline for comparison in the event of suspected relapse.

Monitoring for Drug Toxicity

Isoniazid, rifampin, and pyrazinamide are potentially hepatotoxic. Hepatotoxicity from isoniazid is more frequent in older persons and in those who use alcohol daily [55]. Hepatitis from isoniazid or pyrazinamide typically occurs after the first month of therapy, whereas rifampin is more likely to cause cholestasis during the first month of treatment [78]. Although clinical rather than laboratory monitoring for hepatotoxicity is recommended in most cases, it is prudent to measure liver function tests every 1 to 3 months during therapy in older patients and in those who use alcohol daily or have underlying liver disease [54]. Hepatic tests should be obtained in all patients with clinically suspected hepatotoxicity. If antituberculosis therapy causes the transaminase levels to increase to more than five times the upper limit of normal, isoniazid, rifampin, and pyrazinamide should be discontinued. Alternative agents such as ethambutol, streptomycin, and ofloxacin can be given until the liver function tests normalize. The potentially hepatotoxic drugs can then be reintroduced one at a time to identify the offending agent and to determine the optimal regimen needed to complete therapy.

Compliance with Therapy

In the United States, 20% of tuberculosis patients do not complete therapy [22]. In one large hospital in New York, 83% of tuberculosis patients failed to complete 3 months of therapy [15]. Potential obstacles to compliance should be discussed with each patient, and an individualized regimen tailored to his or her daily activities. Changing the timing of drug dosages may alleviate mild adverse effects such as nausea and may facilitate compliance. Some patients prefer to take pills with meals or at night, whereas others prefer taking them twice a day to minimize the number of tablets ingested at once. Assistance of the patient's family or friends should be enlisted if possible, and compliance with appointments facilitated by ensuring transportation to the clinic and by minimizing waiting time at the clinic.

Patients should be educated about the benefits of therapy throughout its course, and individualized incentives should be provided (including snacks, food vouchers, or contracts that specify rewards for compliance) [79]. Telephone calls before clinic appointments are effective reminders, combined with immediate contact if appointments are broken [80]. All patients should be questioned in a nonthreatening manner about the frequency with which they fail to take medication. Compliance should be assessed objectively by pill counts, evaluation of urine for the orange color imparted by rifampin metabolites, or detection of urinary isoniazid by dipstick or chemical tests [81]. An elevated serum uric acid concentration is a useful marker of compliance with pyrazinamide.

If compliance is uncertain, directly observed therapy should be given, with incentives provided with each dose of medication. Treatment completion rates increased from 44% to 89% when vouchers for food and shelter were given to homeless tuberculosis patients who complied with directly observed therapy [82]. Community outreach workers are invaluable because they can transport patients to the clinic, deliver medication to patients who do not come to the clinic, and provide incentives with each dose of medication. In patients who are not amenable to outpatient therapy, prolonged hospitalization or housing in residential treatment facilities may be necessary.

Tuberculin Skin Testing

The likelihood of infection with M. tuberculosis can be assessed by intradermal injection of five tuberculin units of purified protein derivative [83]. The diameter of induration is recorded 48 to 72 hours later. Reactivity to tuberculin ( 5 mm) is found if the person has been exposed to mycobacteria and has intact cell-mediated immunity. Because of cross-reactivity between mycobacterial antigens, a positive test may result from exposure to environmental nontuberculous mycobacteria.

Diagnosis of Tuberculosis

In evaluating ill patients, a reactive tuberculin skin test supports the diagnosis of tuberculosis, but it is not specific and may indicate previous infection with M. tuberculosis and an unrelated current illness. Control skin tests (such as candida and mumps) are useful to interpret a negative tuberculin skin test. Failure to react to all skin tests suggests depressed cell-mediated immunity, which can occur with tuberculosis itself or other chronic illnesses. A negative tuberculin skin test and a positive control skin test makes the diagnosis of tuberculosis unlikely.

Screening for Tuberculous Infection

A history of vaccination with bacille Calmette–Guérin should be ignored in interpreting the results of tuberculin skin testing in adults, because skin test reactivity from bacille Calmette–Guérin usually declines by adulthood [83]. If skin testing is done periodically, as in hospital employees, the boosting phenomenon must be considered. In some tuberculin reactors, sensitivity to tuberculin declines with time, and tuberculin skin tests become negative. Administration of the skin test itself boosts immunologic memory so that a second tuberculin skin test up to 2 years later will be positive. If the person is being tested annually, the positive second response and negative first response will erroneously be thought to represent recent infection requiring chemoprophylaxis (see below). To avoid this problem, persons who are initially tuberculin negative should be retested 1 to 4 weeks later to assess the boosted response. The size of the second skin test provides a baseline to compare with subsequent skin test results.

Guidelines for Chemoprophylaxis

Ten to 15 million people in the United States are infected with M. tuberculosis [84]. Chemoprophylaxis with isoniazid greatly decreases the likelihood of progression of tuberculous infection to disease. Decisions about chemoprophylaxis are based on three major factors: the likelihood of tuberculous infection, the probability that infection will progress to disease, and the risk for hepatotoxicity from isoniazid. The likelihood of infection is estimated by considering the person's history of exposure to tuberculosis and tuberculin skin test reactivity [83], as outlined below. The probability that tuberculous infection will progress to disease is increased in recently infected persons and in those with impaired cell-mediated immunity. The risk for hepatotoxicity from isoniazid increases with age and may be increased in women, particularly during pregnancy or the postpartum period [55, 74, 85].

Guidelines for use of chemoprophylaxis [55, 84, 86] are summarized below. Some recommendations are controversial, and alternative strategies have been suggested [87]. Decisions in individual patients depend on the physician's assessment of risks and benefits for chemoprophylaxis, rather than strict adherence to published guidelines. For example, for persons in whom development of tuberculosis would be a major risk to others, such as health care workers and those who work with young children, chemoprophylaxis may be given to selected tuberculin reactors at low risk for development of tuberculosis.

Before administration of chemoprophylaxis, tuberculous disease must be excluded, because inadvertent use of isoniazid alone in tuberculosis patients may induce drug resistance. If sufficient suspicion of tuberculosis exists such that sputum specimens are sent for mycobacterial culture, chemoprophylaxis should be deferred, and patients either treated for tuberculosis or observed without medication until results of cultures are available.

Highest-Risk Groups

The following groups should receive chemoprophylaxis, regardless of tuberculin skin test status.

1. Persons with HIV infection who are likely to be infected with M. tuberculosis, including close contacts of infectious tuberculosis patients, those with a previously untreated positive tuberculin skin test, and those with chest roentgenographic findings suggestive of previous untreated tuberculosis [86].

2. Young children (5 years old or younger) who are close contacts of infectious tuberculosis patients. If tuberculin skin tests are negative, prophylaxis should be continued for 3 months and skin testing repeated. If results are negative, chemoprophylaxis may be discontinued.

3. Persons with a history of inadequately treated tuberculosis.

High-Risk Groups

The following groups should receive chemoprophylaxis if the tuberculin skin test is at least 5 mm in diameter.

1. Persons with HIV infection who are not in the categories outlined above.

2. Close contacts of infectious tuberculosis patients.

3. Persons in whom chest roentgenograms show parenchymal lesions suggestive of previous untreated tuberculosis.

Moderate-Risk Groups

The following groups should receive chemoprophylaxis.

1. Persons with recent tuberculous infection, manifested by an increase in tuberculin skin test size within the previous 2 years. In persons younger than 35 years, an increase of 10 mm is clinically significant, and in persons older than 35 years, an increase of 15 mm is clinically significant.

2. Persons with a tuberculin skin test of at least 10 mm diameter, with coexistent medical conditions that predispose to development of tuberculosis (including intravenous drug use, end-stage renal disease, poorly controlled diabetes, silicosis, malnutrition, gastrectomy, immunosuppressive therapy, and reticuloendothelial malignancies).

Low-Risk Groups

Young persons with positive tuberculin skin tests of uncertain duration who do not belong to the risk groups outlined above have a small annual risk for developing tuberculosis, but their lifetime risk may be clinically significant, and the risk for isoniazid hepatotoxicity is low. Therefore, chemoprophylaxis is recommended in persons younger than 35 years with positive tuberculin skin tests. A skin test of 10 mm in diameter is considered positive in persons from groups with a high prevalence of tuberculous infection, such as foreign-born persons from high-prevalence countries; medically underserved, low-income persons; as well as residents of nursing homes, correctional facilities, and mental institutions. In persons from groups at decreased risk for tuberculosis, a tuberculin skin test of 15 mm in diameter is considered clinically significant because smaller reactions are likely to reflect exposure to nontuberculous mycobacteria.

Pregnancy

Because of concerns about use of medication during pregnancy and about the possible increased frequency of hepatotoxicity during pregnancy and the postpartum period [74, 85], chemoprophylaxis is usually deferred until after delivery. However, in tuberculin reactors who are contacts of infectious tuberculosis patients, and in those with recent tuberculin skin-test conversion, prophylaxis is begun after the first trimester.

Duration of Chemoprophylaxis

Six months of chemoprophylaxis is the most cost-effective regimen [88], although 9 to 12 months may be slightly more effective. At least 6 months of isoniazid should be provided to all patients, and an additional 3 to 6 months is optimal if no toxicity occurs. Nine months of chemoprophylaxis is recommended in children [70] and 12 months in persons with HIV infection and in those whose chest roentgenograms are suggestive of previous healed tuberculosis [55]. Guidelines for monitoring of hepatotoxicity during chemoprophylaxis are similar to those used during treatment of tuberculosis [55].

Drug-resistant Mycobacterium tuberculosis

No chemoprophylactic regimen is known to be effective in persons infected with isoniazid-resistant M. tuberculosis. One approach is to administer rifampin, with or without ethambutol, for 12 months. Another is to prescribe isoniazid because persons exposed to patients with isoniazid-resistant tuberculosis may be infected with a subpopulation of isoniazid-susceptible organisms. Alternatively, preventive therapy can be withheld if the person is not at high risk for development of tuberculosis. Studies in animals suggest that rifampin and pyrazinamide for 2 months provides effective chemoprophylaxis [89], and this regimen is being investigated in humans. Guidelines have been recently published [24] to address the complex issue of chemoprophylaxis for persons exposed to multidrug-resistant tuberculosis.

Tuberculosis Control

Guidelines have been developed to reduce tuberculosis morbidity in high-risk groups, including the homeless, the foreign-born, as well as those in health care settings and nursing homes [8, 14, 90, 91]. Groups at high risk for tuberculosis are reservoirs for spread of disease into the general population, as shown by the deadly effects of multiple-drug-resistant tuberculosis outbreaks on health care workers [17, 24]. Similarly, tuberculosis in correctional facilities can readily affect the community because more than 8 million people are discharged from local jails, and 200 000 are discharged from state and federal prisons annually [92].

Federal funds for tuberculosis control project grants, unadjusted for inflation, declined from $20 million in 1969 to $1 million in 1982 [93]. Funds have increased to $9.1 million in 1991, but they are still inadequate. Decimation of the tuberculosis control budget in New York City has been linked to rapidly increasing tuberculosis case rates [15]. Increased funding is essential for improved tuberculosis control in the United States.

Diagnosis and Treatment of Tuberculosis

The highest priority must be given to diagnosis and treatment of patients with active tuberculosis. Delayed diagnosis is common in the elderly, in those with extrapulmonary disease, and in HIV-infected persons [94, 95]. Physicians must be vigilant about tuberculosis, and training for health care professionals should emphasize recognition of this disease. High priority should be given to development of rapid diagnostic tests for tuberculosis.

Effective treatment of tuberculosis depends on enhancing compliance through measures such as directly observed therapy. Increased short-term funding for such measures should yield long-term savings by decreasing recurrent hospitalization, development of drug resistance, and disease transmission. For example, the cost to hospitalize one noncompliant patient with drug-resistant tuberculosis and nine contacts who developed tuberculosis was $950 000, more than five times the annual budget of a tuberculosis control program that treats 100 tuberculosis patients and 1000 cases of tuberculous infection each year [96].

Multiple-drug-resistant tuberculosis is not confined to New York or Florida but was reported from 11 states in 1991 [24]. To combat drug-resistant tuberculosis, we must identify new drugs that are as effective as isoniazid and rifampin. Although new agents against bacterial pathogens are constantly being developed, no new antituberculosis agents have had extensive clinical trials since rifampin was introduced 20 years ago.

Screening for Tuberculosis and Tuberculous Infection

Screening with tuberculin skin tests is recommended in all high-risk populations, including immigrants [8], the homeless, and high-risk minority populations during tuberculosis outbreaks [91, 97], as well as persons entering long-term care facilities [14]. Screening persons who have persistent cough increases the yield of screening [98]. Local screening guidelines may vary from generally recommended ones. For example, although screening hospitalized patients with chest roentgenograms is generally not cost effective, this strategy was useful in a hospital where 1 of every 160 patients admitted had tuberculosis [99]. Screening with sputum acid-fast smears and chest roentgenograms also gave high yields in selected inner-city populations [98, 100].

Persons at highest risk for tuberculin skin test conversion are contacts of tuberculosis patients, so that contact investigation of tuberculosis patients is the most cost-effective means to detect infected persons. Increased resources are needed to quicken the pace of contact investigation and to maximize its potential benefits.

Environmental Measures

Guidelines have been published to ensure adequate ventilation in health care settings where persons are at increased risk for tuberculosis [91]. High-efficiency particulate filters and disposable particulate respirators, both of which filter tubercle bacilli, are recommended in some circumstances, but high costs have limited their use. Ultraviolet light is thought to be mycobactericidal and is recommended if the risk for transmission of tuberculosis is high, such as in correctional facilities, homeless shelters, as well as selected health care facilities [90, 91]. Theoretical and practical aspects of ultraviolet light use have been recently reviewed [101].