Treatment of latent infection with Mycobacterium tuberculosis: update 2010

THE LATENT INFECTION STATE

M. tuberculosis is able to persist within the human host for long periods without causing clinically manifest disease. Chronic persistence within the human host involves alterations of the bacillary metabolic processes with apparent arrest or slowing of multiplicative activities 24–27. Whether “dormancy” is the correct term for this state of altered metabolism has been a matter of debate. The factors and mechanisms for such altered state of metabolism are only beginning to be elucidated, and adaptive bacillary response to hypoxia within the granuloma could be one of the factors 28. The ability of M. tuberculosis to enter a persistent phase carries major implications on its preventive treatment, as the current drugs are not expected to act when their target metabolic processes are switched off.

Using forms of environmental stress, such as hypoxia or acid suppression, in vitro models have been developed, both to elucidate the underlying mechanisms of latent infection with M. tuberculosis and to study the action of drugs under such conditions 28–31. In one such model, the Wayne model, a sealed, standing culture became progressively more hypoxic on incubation, with a concomitant shift in M. tuberculosis physiology 29. Molecular signatures of such hypoxic stress responses are accumulating, but correlations with human infection/disease status are still unclear 28, 32–34.

Models of latent infection with M. tuberculosis have been developed in mice, guinea pigs, zebrafishes, rabbits and non-human primates 28. In the Cornell model, mice infected with M. tuberculosis are treated with isoniazid and pyrazinamide to induce a temporary “latent” state, during which no detectable bacilli can be recovered by organ culture and guinea pig inoculation 35. This and similar kinds of models have been used extensively to study the action of drugs on latent infection with M. tuberculosis 13, 14, 17, 19, 28.

DIAGNOSTIC ISSUES

In contrast to the various clinical manifestations 36 and high fatality of untreated tuberculosis 37, latent infection with M. tuberculosis is, by definition, asymptomatic, and latently infected individuals do not shed bacilli into the environment and are thus noninfectious. Intervention on such latent infection state is justified primarily because of the risk of developing disease. Among young individuals with positive tuberculin skin tests identified during tuberculosis contact tracing, ∼5% will develop active disease in the first 5 yrs and another 5% in the rest of their lifetime 38–41. The risk is, however, modulated by age at infection 42 and many other host factors (table 1) 39, 43–46. In a more recent study from Germany, six (15%) out of 41 untreated close contacts with a positive result by IGRA developed active tuberculosis within 2 yrs following contact with an infectious index case 47. Most of the tuberculosis cases in that study involved ethnic minorities, and only two of them were culture confirmed. A somewhat lower proportion was observed in another study from Turkey, in which four (7%) out of 54 untreated children with a positive IGRA result during contact examination developed active tuberculosis after 93 person-yrs of follow-up 48. IGRA did not appear to perform better than the tuberculin skin test among tuberculosis contacts in the Gambia, in which most of the tuberculosis cases were detected during the initial examination, and subsequent disease rate among subjects with a positive result for either test was much lower 49.

Table 2 summarises the characteristics of the existing diagnostic tools for latent infection with M. tuberculosis 39, 47–51. With the low bacillary load and suppressed metabolic manifestations, the diagnosis of latent infection with M. tuberculosis has, all along, depended on the detection of host response, rather than the presence/activity of the pathogen itself 50. Immunodiagnostic methods may underperform among those who fail to mount an adequate immune response 39, 50, 52, 53. More importantly, currently available immunodiagnostic tests are unable to differentiate individuals who will subsequently develop tuberculosis from those who will not 39, 47–49. They also fail to distinguish recent infection from remote infection that carries a much lower risk for the development of active disease 38, 39, 50, 51. This may pose another major obstacle in areas and/or age groups with a high background prevalence of latent infection with M. tuberculosis 6. Ongoing research is focusing on antigens expressed during different metabolic phases of M. tuberculosis 54 and/or cytokines expressed by the human host 55. It is hoped that this may help to improve the performance of immunodiagnostic tests in future.

MECHANISMS OF DRUG ACTION

From observations in both animal and human studies, it has been postulated that a patient with tuberculosis can harbour four hypothetical populations of organisms (fig. 1) 24. Three major actions are proposed for the currently available antituberculosis drugs 56: 1) bactericidal action (ability to kill actively growing bacilli rapidly), often assessed by the decrease in quantitative sputum culture bacillary count in the initial few days of treatment 57; 2) sterilising action (ability to kill persisters under acid inhibition or with spurts of metabolism), as reflected by the ability to prevent relapse or its proxy marker, such as 2-month sputum culture conversion rate 58; and 3) prevention of emergence of bacillary resistance to drugs.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1–

Hypothetical components of the bacterial population in active tuberculosis.

Of the currently available antituberculosis drugs, isoniazid has the highest early bactericidal effect against rapidly growing tubercle bacilli, while rifampicin and pyrazinamide are thought to have the greatest sterilising effects against those with spurts of metabolism and under acid inhibition respectively 56, 57.

CLINICAL TRIALS WITH ISONIAZID

After the rediscovery of isoniazid there followed exploration of treatment with isoniazid alone and in combination with para-aminosalicylic acid or streptomycin in the treatment of tuberculosis with encouraging results 59. Encouraging data also emerged in its use to prevent experimental tuberculosis in guinea pigs 60. However, there were concerns over its efficacy in preventing disease or infection as well as the possible emergence of drug resistance 10. A series of large-scale clinical trials were initiated by the United States Public Health Service (USPHS) within programme settings to address this important issue. The first of these studies began in 1955 but it only examined whether the frequency of complications of primary tuberculosis could be decreased by the use of isoniazid 61. Four subsequent randomised controlled trials were started in the 1950s and were completed in the 1960s, including a total of 7,333 villagers in Alaska 62, 27,857 household contacts 63, 64, and 24,838 patients in psychiatric institutions 65. Cluster randomisation was done by household, village, or hospital ward. Isoniazid or matching placebo was given for 1 yr, at a dose of 300 mg daily or 5 mg·kg⁻¹ for children. These studies, by their sheer sizes, helped to establish the efficacy of isoniazid in the treatment of latent infection with M. tuberculosis.

Follow-up data from the Alaskan study 62, 66 also suggested that the protective effect of isoniazid preventive chemotherapy persists for up to 19 yrs, even though the offering of open-label treatment to all participants might have affected the accurate assessment of protective effect after 10 yrs. Prolonged follow-up of a cohort of children up to 30 yrs also supports a long-lasting protective effect 67. It therefore appears that, among non-HIV infected subjects in areas without excessive risk of ongoing transmission, treatment of latent infection with M. tuberculosis produces a lasting effect.

Table 3 summarises the results of randomised controlled trials on the treatment of latent infection with M. tuberculosis with a placebo/no treatment arm in non-HIV infected persons 62–65, 68–74. In a meta-analysis involving 73,375 subjects in 11 placebo-controlled randomised trials, treatment with isoniazid for 6–12 months reduced the risk of tuberculosis by 60% (risk ratio (RR) 0.40, 95% CI 0.31–0.52) over 2 yrs or longer (grade A) 11. Preventive therapy reduced deaths from tuberculosis, but not all-cause mortality.

No significant difference was found in the risk reduction between 6 months (RR 0.44, 95% CI 0.27–0.73) and 12 months (RR 0.38, 95% CI 0.28–0.50) of isoniazid 11. In the International Union Against Tuberculosis (IUAT) trial, the only study that included direct comparison between 6 and 12 months of isoniazid, a total of 28,000 persons with fibrotic pulmonary lesions compatible with tuberculosis were followed for 5 yrs after receiving varying durations of isoniazid at 300 mg daily given in 35-day packages for self-administered treatment 72. On an intention to treat basis, 12 weeks, 24 weeks and 52 weeks of isoniazid reduced the risk of tuberculosis within 5 yrs by 21%, 65% and 75% respectively, as compared with placebo. There was no statistical difference between the effectiveness of the 24-week and 52-week regimens under the study conditions, but both of them prevented significantly more tuberculosis cases than either the 12-week regimen or placebo. Hepatitis occurred at a frequency of 0.12%, 0.25%, 0.36%, and 0.52% in the placebo, 12-week, 24-week and 52-week arms respectively. The 24-week regimen prevented more tuberculosis cases (2.6 versus 2.1 tuberculosis cases) per case of hepatitis than the 52-week regimen.

When analysis was restricted to those participants who took at least 80% of doses from each calendar package and took all calendar packages for the entire assigned duration (“completer-compliers”) in the IUAT trial, 12, 24 and 52 weeks of isoniazid reduced the risk of tuberculosis by 31%, 69% and 93% respectively (all inter-regimen p<0.05). Despite possible selection bias in secondary subgroup analyses, such results and the study among Netherlands army recruits 70 suggested considerably higher efficacies might be achievable among adherent patients than the overall effectiveness as observed among all patients. In one of the USPHS household contact studies 63, the subsequent risk of tuberculosis was reduced by 68% and 16%, respectively, among those taking ≥80% of the recommended number of pills for ≥10 months and <10 months. In the Alaskan study, the decline in the case rate became nearly horizontal at 9–10 months when the tuberculosis case rate was plotted against months of treatment taken 75, 76. Based on these observations, Comstock and co-workers 75, 76 inferred that 9–10 months of isoniazid was the optimal duration, which formed the basis for the revised current recommendation in the USA 39. Extending treatment beyond 12 months did not appear to reduce the tuberculosis risk in the Alaskan study and the Veterans Administration Cooperation Study 71.

The use of a daily isoniazid dosing schedule is well supported by randomised controlled trials (grade A) 11. Twice-weekly isoniazid dosing at 15 mg·kg⁻¹ (maximum 900 mg) is also used in the USA to facilitate direct observation in the treatment of latent infection with M. tuberculosis (grade D) 39, 77. Such dosing schedule has not been tested in randomised controlled trials among non-HIV infected persons (table 3). Its likely efficacy is extrapolated mainly from a clinical trial on active tuberculosis 78.

In children, randomised trials are available only for the 12-month regimen. Extrapolation from adult studies may be reasonable. However, a recent study found a low peak serum concentration of isoniazid among young South African children (median age 3.2 yrs) given isoniazid daily at 4–6 mg·kg⁻¹, especially among intermediate or fast acetylators 79, and higher daily doses of 8–12 mg·kg⁻¹ may be required to achieve similar isoniazid concentrations as in adults. With the high serum level to minimal inhibitory concentration ratio for isoniazid 56 and the use of isoniazid at daily dosage of 5 mg·kg⁻¹ among young children in virtually all clinical trials 11, it remains an open question whether the higher dosage is necessary.

ISONIAZID-RELATED HEPATOTOXICITY

Many of the earlier trials on isoniazid were conducted before its potential hepatotoxicity was well recognised 10. At first, the isolated reports of jaundice in the early USPHS clinical trials were not conclusively linked to isoniazid 66. However, with the subsequent widespread use of the drug in the treatment of latent infection with M. tuberculosis in the USA, serum transaminase elevations and other hepatic abnormalities were soon recognised 80. After the occurrence of 19 cases of hepatitis resulting in two deaths among 2,321 contacts treated with isoniazid in an outbreak 81, the USPHS undertook a major surveillance study among 13,838 persons in 21 participating health departments 82. In that study, the overall frequency of isoniazid-related hepatitis was 10.3 per 1,000 participants (1%), with most of them occurring within the first 3 months of treatment. The hepatitis risk increased sharply with age, with 0%, 0.3% 1.2% and 2.3% among those aged below 20, 20–34, 35–49 and 50–64 yrs, respectively. Daily alcohol consumption was also an important risk factor. Hospitalisation frequency was up to 5.0 per 1,000 treatment initiations. There were eight fatalities with a mortality of 0.6 per 1,000 persons. It is noteworthy that of the eight deaths in the 20-city study, seven occurred in the Baltimore area, which was later shown to have had an excessive increase in cirrhosis of the liver when compared with the 2 yrs earlier and later 83.

In a meta-analysis involving 38,257 subjects treated with isoniazid in six earlier studies, clinical hepatitis ranged from 0.0–2.9% 84. The combined frequency was 0.6% in absence of age adjustment. More recently, the frequency of symptomatic hepatitis has been estimated to be one to three per 1,000 persons, and much lower hospitalisation (0.1 to 0.2 per 1,000 persons) and mortality (0.0 to 0.3 per 1,000 persons) have been reported 85, 86. Such a decline might reflect more careful patient selection and active monitoring for early signs of adverse effects during treatment 39.

OTHER ISONIAZID-RELATED ADVERSE DRUG EVENTS

The adverse drug events from isoniazid are summarised with those of other antituberculosis drugs in table 4 87–91. Mild and transient headache, nausea and dizziness were reported in clinical trials of isoniazid among non-HIV infected subjects 11. Peripheral neuropathy, related to a dose-dependent inhibitory effect of isoniazid on the function of pyridoxine metabolites, is uncommon (<0.2%) in healthy individuals. It is more frequently encountered in the chronically alcohol-dependent, malnourished persons, pregnant females and HIV-infected subjects 87–89, but it can be prevented as well as treated by pyridoxine. It is not generally held that pyridoxine should be given routinely 92, but it should be prescribed to patients at risk of such complication, including those with underlying nutritional deficiency or higher requirements, and in persons in whom peripheral neuropathy may develop as a result of the underlying condition, thus leading to confusion over the cause. Central nervous system reactions, such as convulsions, encephalopathy, optic neuritis, memory impairment and psychosis, are rare at normal doses. Antinuclear antibody is more common than the actual lupus-like syndrome.

HIV-INFECTED INDIVIDUALS

In a study by Selwyn et al. 93 in the USA, the risk of tuberculosis among tuberculin skin test-positive and HIV-infected subjects was at least 7.9 cases per 100 person-yrs. The risk of developing tuberculosis increases with the degree of immunosuppression 94, and remains elevated (though at a reduced level) even after the initiation of antiretroviral therapy 95. Tuberculin skin testing remains a useful primary screening test for the diagnosis of latent infection with M. tuberculosis among HIV-infected persons despite possible limitations in sensitivity 96, with 24-fold difference in risk between test-positive and test-negative subjects in the above quoted study 93. Data on the disease-predicting values of IGRA among HIV-infected persons are still scanty 97.

Table 5 summarises the results of randomised controlled trials on the treatment of latent infection with M. tuberculosis with a placebo/no treatment arm among HIV-infected persons 98–105. In a randomised controlled trial among HIV-infected subjects in Haiti 98, a 12-month course of isoniazid reduced the incidence rate of tuberculosis from 7.5 to 2.2 per 100 person-yrs (RR 0.29, 95% CI 0.09–0.91) among HIV-infected subjects. In subgroup analysis, there was significant reduction (RR 0.22, 95% CI 0.05–1.00) among tuberculin skin test-positive subjects (cut-off ≥5 mm) but not among tuberculin skin test-negative subjects (RR 0.70, 95% CI 0.15–3.28). In this particular study, isoniazid treatment also delayed progression to HIV-associated morbidity including AIDS and death. In a Ugandan trial 99, 6 months of isoniazid significantly reduced tuberculosis risk among tuberculin skin test-positive subjects (RR 0.33, 95% CI 0.14–0.77) but not subjects with anergy (RR 0.83, 95% CI 0.34–2.04). Survival did not differ between the study arms, but anergic subjects had a higher fatality than those with a positive tuberculin skin test. In a follow-up analysis, benefit in tuberculin skin test-positive persons was lost after the first year 106. No significant protective efficacy was observed for 6 months of isoniazid among HIV-infected subjects after a median follow-up of 1.83 yrs in Kenya, but the small number of tuberculin skin test-positive subjects (67–69 per arm) might not have provided adequate power to pick up any difference 101.

In a randomised controlled trial among HIV-infected subjects with anergy in the USA, tuberculosis was diagnosed in only six of 257 patients in the placebo group and three of 260 patients in the isoniazid group after a mean follow-up of 33 months 100. As the incidence of tuberculosis was low, it does not support the use of isoniazid treatment in anergic HIV-infected subjects in absence of recent exposure to a case with bacteriologically confirmed tuberculosis of the respiratory tract. There were also no significant differences between the two groups with regard to death, death or the progression of HIV disease, or adverse events.

Only a single randomised controlled trial in Zambia tested 6 months of isoniazid twice a week against placebo among HIV-infected persons 102. However, only combined analysis of treatment with either 6 months of isoniazid or 3 months of rifampicin plus pyrazinamide against placebo reached statistical significance (RR 0.60, 95% CI 0.40–0.89). The effect of preventive therapy also declined after the first year of the study. By 18 months the rates of tuberculosis in the treated groups were similar to that in the placebo group, even though the cumulative risk of tuberculosis remained significantly lower within the first 2.5 yrs 107.

Among 4,316 HIV-infected subjects in seven trials 98–104, 106, 107 included in a recent meta-analysis 12, treatment with either 6 or 12 months of isoniazid was associated with a significantly lower incidence of tuberculosis (RR 0.67, 95% CI 0.51–0.87). This benefit was more pronounced in individuals with a positive tuberculin skin test (RR 0.36, 95% CI 0.22–0.61) (grade A), but failed to achieve statistical significance among those who had a negative tuberculin skin test (RR 0.86, 95% CI 0.59–1.26) (insufficient evidence). Overall, adverse events leading to stopping treatment occurred in 2.8% of those receiving isoniazid, compared to 1.8% of the placebo arm (RR 1.66, 95% CI 1.09–2.51). The reduction in mortality just reached statistical significance among individuals with a positive tuberculin skin test (RR 0.74, 95% CI 0.55–1.00). Despite the delay in HIV progression that was observed with isoniazid treatment in the Haiti trial 98, confirmation of such an effect awaits further trials. It is not possible to determine whether the effects of treatment are influenced by the HIV/AIDS progression stages. The optimal duration of isoniazid for the prevention of tuberculosis in HIV-seropositive individuals is also unclear. Both a 6-month regimen 99 and a 12-month regimen 98 are effective (grade A), but none of the trials directly compared isoniazid regimens of different durations. Given the uncertainty, the American Thoracic Society (ATS) discourages the use of a 6-month isoniazid regimen for patients with HIV infection 39.

From the only two available studies with long-term follow-up data 106, 107, the protection of isoniazid treatment in HIV-infected persons appears to be short lasting (1–2.5 yrs). As both of these studies were conducted in areas with a high incidence of tuberculosis, reinfection after completion of isoniazid preventive therapy could have been an important factor, taking into account the heightened risk of rapid progression to disease among HIV-infected subjects.

ALTERNATIVE SHORT-COURSE REGIMENS

The long duration required for the treatment of latent infection with M. tuberculosis triggered a series of clinical trials in search for a “short-course” regimen. Attention was naturally directed to those drugs with putative sterilising capacity against persisters 13, 14, 56, especially rifampicin and pyrazinamide, the pivotal agents that successfully shortened the treatment duration for tuberculosis in human trials 15, 16.

In an animal experiment, a nonreplicating bacillary population of limited size was developed in mice vaccinated with bacille Calmette–Guérin (BCG) vaccine to mimic latent infection with M. tuberculosis 13. After treatment with 6 months of isoniazid, 3 months of rifampicin, 2 months of rifampicin plus pyrazinamide, and 2 months of rifampicin plus pyrazinamide plus isoniazid, the proportions of mice with positive spleen cultures for M. tuberculosis were 100%, 20%, 0% and 80% respectively. The 2-month regimen of rifampicin plus pyrazinamide appeared to show the best treatment-shortening potential, and addition of isoniazid showed an antagonistic effect. These results were very encouraging, even though there could be pitfalls in extrapolating experimental observations from a murine model to man 28, 108.

In the 1990s, effective regimens were already fully established for the treatment of both latent infection with M. tuberculosis 62–74 and tuberculosis 15, 16. In such a situation, there would be ethical and logistic difficulties in assigning a huge number of subjects to experimental preventive regimens in a manner similar to that which occurred in the 1950s 62–65. Most of the explorative trials on alternative regimens were, therefore, carried out among groups with a very high risk for tuberculosis, such as silicosis patients 73 or HIV-infected subjects, in order to reduce the sample size and follow-up time. Only some of them contained a placebo/no treatment arm (tables 3 and 5), as it would be acceptable to compare a new regimen with the standard regimen of isoniazid for 6 or 12 months 109, 110. Table 6 summarises the comparison of efficacy between isoniazid and the alternative regimens, while tables 7 and 8 summarise the adverse events leading to termination of treatment in those randomised controlled trials for latent infection with M. tuberculosis conducted in or after the 1990s.

RIFAMPICIN PLUS PYRAZINAMIDE

In 2000, a multinational trial reported that 2 months of rifampicin and pyrazinamide was as effective as 12 months of isoniazid in reducing tuberculosis in HIV-infected individuals with latent infection with M. tuberculosis (table 6) 113. Drug discontinuation was significantly higher in the rifampicin plus pyrazinamide arm than the isoniazid arm (table 8). However, a lower number of patients in the rifampicin plus pyrazinamide arm had abnormal liver function test results that were grade 4 or resulted in study drug discontinuation (11/791 versus 28/792; p = 0.02). Treatment completion was also higher for rifampicin plus pyrazinamide arm than the isoniazid arm (80% versus 69%; p<0.001).

In a meta-analysis of different regimens for latent infection with M. tuberculosis among HIV-infected subjects 12, combined analysis of 428 subjects on rifampicin plus pyrazinamide for 2–3 months and 428 subjects on placebo in two trials 102, 104 the combined relative risk of tuberculosis was 0.54 (95% CI 0.34–0.86) for rifampicin plus pyrazinamide versus placebo, even though treatment was more likely to be stopped because of adverse drug events (RR 7.84, 95% CI 2.60–23.67). Equivalent efficacy (RR 1.03, 95% CI 0.75–1.40) was found between isoniazid (6 or 12 months) and rifampicin plus pyrazinamide among a total of 3,409 subjects in five randomised clinical trials in the same meta-analysis (table 6) 102, 104, 111, 114. However, the isoniazid regimen was less likely to be stopped because of adverse events (RR 0.63, 95% CI 0.48–0.84). Partly because of this, the proportion of patients completing treatment was higher for the shorter rifampicin plus pyrazinamide regimen in only two studies 113, 114. Conversely, hepatotoxicity did not appear to be a prominent feature for the use of rifampicin plus pyrazinamide among HIV-infected subjects 12. Re-analysis of the data in the trial published by Gordin et al. 113 in 2000 also confirmed the absence of excess hepatotoxicity risk with the combination of rifampicin plus pyrazinamide in HIV-infected individuals 122.

In another trial among HIV-infected subjects, 3 months of isoniazid plus rifampicin and pyrazinamide was equivalent to 6 months of isoniazid monotherapy and significantly reduced the risk of tuberculosis as compared to placebo 99. However, treatment was much more likely to be stopped because of adverse events in comparison with either placebo or isoniazid (tables 7 and 8).

With the reported efficacy of a 2-month course of rifampicin and pyrazinamide in the treatment of latent infection with M. tuberculosis among the HIV-infected individuals 113, the ATS and Centers for Disease Control and Prevention (CDC) issued a joint statement in 2000 entitled “Targeted Tuberculin Testing and Treatment of Latent Tuberculosis Infection” 39. Emphasis was put on directing tuberculin testing to populations at risk of developing tuberculosis, as well as more treatment options including short-course rifampicin-based regimens. As the drugs involved had been used extensively in the treatment of tuberculosis, there was apparently not too much concern over adverse effects. In fact, to promote better acceptance, there was call for simplified monitoring that emphasises clinical evaluation more than laboratory examination. Soon after such recommendations, reports of severe drug-induced liver injury, including deaths, followed widespread use of rifampicin plus pyrazinamide 123, 124. These reports prompted revision of the guidelines in August 2001 to reduce the dose of pyrazinamide (to ≤20 mg·kg⁻¹) and ensure closer monitoring 124. In a national survey of 8,087 patients given rifampicin plus pyrazinamide in the USA, the frequencies of asymptomatic elevation of aspartate aminotransferase more than five times upper limit of normal and clinical hepatitis were 25.6 (95% CI 22.3–29.3) and 18.7 (95% CI 15.9–21.9) per 1,000 therapy initiations, respectively, with 23 (2.8 per 1,000) hospitalisations and seven (0.9 per 1,000) fatalities 125. Increasing age, an abnormal baseline aspartate aminotransferase level (often associated with alcohol use), and unemployment within the past 24 months were found to be independent risk factors for hepatotoxicity in a cohort study on the use of 2 months of rifampicin plus pyrazinamide among jail inmates and homeless persons 126. Such events could occur at a low pyrazinamide dose, despite biochemical monitoring, and at the end of therapy 127. Many of these required hospitalisation and some required liver transplantation 127, 128. Fatality was associated with higher age or use of other medications 127.

In three clinical trials involving randomised or systemic allocation of non-HIV infected subjects, 7.7%, 35%, and 10% of those in the 2-month rifampicin plus pyrazinamide arms developed significant hepatotoxicity with aspartate or alanine aminotransferase in excess of five times the upper normal limit 115–117. All of these proportions were significantly higher than those observed in the corresponding isoniazid arms. For unknown reasons, the frequency of hepatotoxicity in the use of rifampicin plus pyrazinamide in the treatment of latent infection with M. tuberculosis among non-HIV infected subjects appeared to be higher than those reported in the historical short-course trials (involving concomitant use of isoniazid, rifampicin and pyrazinamide) in the treatment of active tuberculosis 15, 16.

A high frequency of hepatitis of 11% to 14% was initially reported in early studies using high dosages of pyrazinamide (40–70 mg·kg⁻¹ daily) in combination with isoniazid 129. With the adoption of lower drug dosages, 2–5% of patients developed hepatitis attributable to pyrazinamide while they were put on the standard short-course regimens for treatment of tuberculosis 15, 16, 130, 131. In the Hong Kong study 116, the reported hepatotoxicity frequency of 35% with 2 months of rifampicin plus pyrazinamide (daily dosage largely below 20 mg·kg⁻¹) was much higher than that observed among patients with silico-tuberculosis put on a 8-month regimen of streptomycin, isoniazid, rifampicin and pyrazinamide (daily dosage 30 mg·kg⁻¹) 132. In a retrospective cohort study in the Netherlands, preventive treatment with rifampicin plus pyrazinamide caused severe hepatotoxicity more often than preventive treatment with isoniazid (OR 2.61, 95% CI 1.26–5.39) and triple- or quadruple-drug regimens for tuberculosis (OR 2.61, 95% CI 1.21–5.59) 133. As the substitution of isoniazid by moxifloxcin has not been reported to increase the frequency of hepatotoxicity in a tuberculosis treatment trial 134, intrinsic differences of host status between latent infection with M. tuberculosis and tuberculosis might account partly for the difference in hepatotoxicity, in line with the apparent absence of excess hepatotoxicity of rifampicin plus pyrazinamide among HIV-infected subjects 12, 113. In this regard, treatment for latent infection with M. tuberculosis was also terminated because of hepatotoxicity in six out of 12 consecutive African contacts of multidrug-resistant tuberculosis given pyrazinamide and ethambutol 135.

None of the randomised controlled trials among non-HIV infected are adequately powered to provide a conclusive answer on efficacy 115–117. However, drug-induced liver injury is much less acceptable for preventive therapy among persons who run a 10% lifetime risk of developing tuberculosis than among persons with a potentially fatal disease if left untreated. The revised ATS/CDC recommendations now state that rifampicin plus pyrazinamide should generally not be offered to persons with latent infection with M. tuberculosis 86.

RIFAMPICIN MONOTHERAPY

Rifampicin at a dose of 10 mg·kg⁻¹ daily (maximum 600 mg) for 4 months is currently an acceptable alternative regimen for treatment of latent infection with M. tuberculosis (grade B) 39. Only a single randomised clinical trial assessed the efficacy of rifampicin monotherapy 73. In this study, isoniazid for 6 months, rifampicin for 3 months and isoniazid plus rifampicin for 3 months were compared against placebo. Pulmonary tuberculosis was observed in 13% of the three treatment groups combined as compared to 27% of the placebo group within 5 yrs (p<0.01). In the 389 patients who were followed up to 5 yrs, the 3-month rifampicin regimen significantly reduced the risk of tuberculosis as compared to placebo (table 3). No significant difference in efficacy was observed between the isoniazid and the rifampicin arm (table 6) and acquired drug resistance did not appear to be a problem with either regimen. The rifampicin-alone therapy was very well tolerated, with serious adverse effects occurring with similar frequency as in the placebo arm (table 7). None of the 172 silicosis patients in the rifampicin arm developed hepatitis. Low frequencies of serious adverse events and high proportions of treatment completion were similarly demonstrated in later cohort studies 136–140. In two recent randomised controlled trials among predominantly non-HIV infected individuals 120, 121, a significantly higher proportion completed treatment with 4 months of rifampicin than with 9 months of isoniazid. Adverse events leading to termination of treatment also appeared to be fewer in the rifampicin arm in both studies (table 8). In the larger study, grade 3 or 4 hepatotoxicity occurred in 16 of 422 (3.8%) isoniazid recipients as compared to three of 418 (0.7%) rifampicin recipients 121. No randomised controlled trial has been published on the use of rifampicin alone in predominantly HIV-infected persons, perhaps because of concern over the possibility and consequence of acquired rifampicin resistance. In a meta-analysis involving a total of 3,586 subjects 141, noncompletion was 8.6% to 28.4% among patients who received 4-month rifampicin therapy and 24.1% to 47.4% among patients who received 9-month isoniazid therapy (RR 0.53, 95% CI 0.44–0.63). Grade 3 or 4 hepatotoxicity leading to drug discontinuation was also significantly lower (0% to 0.7% versus 1.4% to 5.2%; RR 0.12, 95% CI 0.05–0.30). A large scale multicentre trial is now underway to assess the effectiveness of 4-month rifampicin therapy 142. The optimal duration of rifampicin remains uncertain. Although 3 months of rifampicin only reduced the risk of tuberculosis by 52% among silicosis patients 73, tuberculosis incidence was actually the lowest in that arm. The high residual risk could well reflect the underlying host status, rather than treatment efficacy.

ISONIAZID PLUS RIFAMPICIN

A 3-month regimen of isoniazid plus rifampicin has also been recommended as an alternative for the treatment of latent infection with M. tuberculosis by the British Thoracic Society (grade A) 143. Its efficacy and adverse events leading to treatment termination were not significantly different from those of 6 months of isoniazid (tables 6 and 7). Hepatic adverse reactions (mainly elevated transaminases) also occurred at similar frequency (three cases in both arms). Additional hepatotoxicity was however shown by a meta-analysis of patients put on the combined regimen from different kinds of studies 84. No hepatitis case was reported among 556 HIV-infected subjects in the 3-month isoniazid plus rifampicin arm in the Uganda trial 99. In the recent Cochrane analysis among HIV-infected subjects 12, isoniazid plus rifampicin significantly reduced the risk of tuberculosis (RR 0.41, 95% CI 0.21–0.81) and death (RR 0.69, 95% CI 0.50–0.95) as compared to placebo among 1,179 subjects in two trials, but there was a higher incidence of adverse events leading to termination of treatment (RR 16.72, 95% CI 3.29–84.89). Equivalent efficacy (RR 0.97, 95% CI 0.52–1.83) was observed between isoniazid (for 6–12 months) and isoniazid plus rifampicin (for 3 months) among 1,601 subjects in four trials. Adverse events leading to treatment termination were somewhat less frequent in the arm receiving isoniazid monotherapy (RR 0.79, 95% CI 0.50–1.23), but the difference was not statistically significant. Very similar conclusions on relative efficacy and safety of the two regimens were made in another meta-analysis of all the five trials from both HIV-infected and non-HIV infected subjects 144. In a study among children, no clinical disease or adverse event leading to treatment termination was observed in 232 patients on 9 months of isoniazid and 238 patients on 4 months of isoniazid plus rifampicin, but patients who received isoniazid monotherapy were significantly less compliant 118. Martinez et al. 112 also reported better adherence with 3 months of isoniazid plus rifampicin as compared to 12 months of isoniazid. However, in two other studies, the proportions completing were similar between 3 months of isoniazid plus rifampicin and 6 months of isoniazid 99, 114. An observational study among First Nations Canadians in Saskatchewan reported much higher completion (80% versus 19%) for 6 months of twice weekly isoniazid plus rifampicin than for 12 months of daily, self-administered isoniazid 145. The use of direct treatment observation in a special population could be a contributing factor.

ISONIAZID PLUS RIFAPENTINE

Rifapentine, a long-acting rifamycin, has been shown to have good activity in treatment of latent infection with M. tuberculosis in the mouse model, when used together with isoniazid in highly intermittent regimens 146, 147. The combination of rifapentine 900 mg and isoniazid 900 mg once weekly for 12 weeks was found to be well tolerated in a recent human trial, with two out of 206 (1%) treated contacts showing grade 3 or 4 hepatotoxicity 148. No conclusion can be drawn on the efficacy as the comparison was made with daily rifampicin plus pyrazinamide and the number of tuberculosis cases was very low in either arm. Close to 9,000 subjects have been recruited in a large scale trial comparing 3 months of weekly isoniazid plus rifapentine with 9 months of isoniazid 149, and the results might help to establish this highly intermittent regimen for use under direct observation in the coming 2 yrs.

DRUG INTERACTIONS

Most of the clinically significant interactions involving drugs metabolised by the cytochrome P450 (CYP450) system are pharmacokinetic in nature 150, 151. Rifampicin is a powerful enzyme inducer of CYP450 and may, therefore, reduce serum concentration of oral contraceptives, corticosteroids, anticoagulants, anticonvulsants, anti-infectives (including antiretrovirals), cardiovasculo-therapeutics, immunosuppressants, psychotropics, sulphonylureas, theophylline and other drugs metabolised by the same pathway 152. As rifampicin features prominently in most of the alternative treatment options of tuberculosis preventive therapy, extra caution is called for in HIV-infected individuals 153 or among older people at higher risk of drug–drug interactions 150. Other rifamycins, like rifapentine, might also share similar enzyme-inducing activity to varying degrees. Rifabutin, a rifamycin compound with less enzyme-inducing activity, is sometimes used in the treatment of tuberculosis in HIV-infected patients together with selected protease inhibitors or non-nucleoside reverse transcriptase inhibitors. However, dose modifications may be necessary 154.

RISK OF ACQUIRING DRUG RESISTANCE

Mutants resistant to antituberculosis drugs are known to emerge spontaneously through alterations of chromosomal genes 155. For isoniazid, such mutants are thought to arise in one in 10⁷ to 10⁹ cell divisions, resulting in one effectively resistant M. tuberculosis bacillus in 10⁶ of the bacillary population. For rifampicin, the corresponding figures are one in 10¹⁰ cell divisions, and one in 10⁸ of the bacillary populations. With a bacillary load of 10⁸ to 10⁹ in a tuberculosis patient with cavitary lung lesions, combination therapy is required to prevent the selection of resistant mutants. However, while such calculations might be algebraically sound, clinical practice shows that resistance emerges against isoniazid in 0.5%, 2.0%, and 4.0% if the companion drug is rifampicin, streptomycin or ethambutol respectively, substantially higher than those suggested by simple product rule 156.

The propensity for emergence of resistance might be less with the much lower bacillary load as is the case with latent infection with M. tuberculosis. In a systematic review 157 of 13 studies published since 1951, a total of 31 isoniazid-resistant isolates were obtained from the isoniazid groups and 24 from placebo/no treatment groups, giving a summary relative risk of 1.45 (95% CI 0.85–2.47) in the emergence of bacillary resistance during isoniazid preventive therapy. Results were similar when studies of non-HIV infected and HIV-infected persons were considered separately. However, with the small numbers and incomplete testing of isolates, it is not possible to exclude a modest increase in risk entirely.

From a theoretical perspective, selection of rifampicin-resistant mutants may be less likely to occur than selection of isoniazid-resistant mutants because of the one to three order lower spontaneous mutation rate 155. However, rifamycin mono-resistance had emerged in four of five patients with a relapse among 31 HIV-infected patients treated once weekly with isoniazid plus rifapentine in the continuation phase of treatment against tuberculosis 158. It therefore appears that the putative low bacillary load after the end of the intensive phase of treatment is not a sufficient condition, at least not among HIV-infected patients, to prevent the emergence of resistance to a rifamycin. There is also some evidence that a higher serum level of isoniazid did appear to play a role in the prevention of rifamycin mono-resistance in the same study 159. A number of trials have confirmed the efficacy of 3 months of isoniazid plus rifampicin (tables 3, 4 and 6), even though adding isoniazid to rifampicin is likely to increase the risk of adverse drug events, particularly drug-induced liver injury (table 8). With the concern over emergence of rifampicin resistance, there could be a case for using isoniazid plus rifampicin, especially in HIV-infected patients. However, in the only trial that directly compared 3 months of rifampicin with 3 months of isoniazid plus rifampicin 73, a smaller number of tuberculosis cases occurred in the rifampicin arm, even though the difference fell short of statistical significance. In the mouse study by Lecoeur et al. 13, isoniazid also appeared to antagonise the sterilising activity of rifampicin plus pyrazinamide. Resolution of this dilemma will have to await further studies.

CONTACTS OF PATIENTS WITH DRUG-RESISTANT TUBERCULOSIS

In an observational study among 2,795 tuberculin-positive Southeast Asian refugees prescribed isoniazid preventive therapy at the time of their resettlement in the USA 160, 19 cases of tuberculosis were detected during follow-up. 15 of them were culture positive, with eight of these isolates being isoniazid susceptible and seven isoniazid resistant. A case–control analysis showed that taking isoniazid for 3 months or less was associated with a six-fold increase in risk (as compared to longer treatment) for subsequent isoniazid-susceptible tuberculosis, but there was no excess risk for subsequent isoniazid-resistant disease. Conversely, taking preventive treatment (with isoniazid) as a whole has been identified as a risk factor for subsequent development of isoniazid-monoresistant tuberculosis, possibly because of prevention of disease due to isonazid-senstive strains 161. In another observational study among 204 tuberculin skin test converters in an outbreak of isoniazid- and streptomycin-resistant tuberculosis among Boston's homeless population 136, six of 71 (8.6%) individuals who received no preventive chemotherapy, three of 38 (7.9%) in the isoniazid group, and none of the 86 in the rifampicin or rifampicin plus isoniazid groups developed tuberculosis. These observational studies provide evidence for the obvious assumption that isoniazid monotherapy is ineffective in reducing the risk of tuberculosis due to isoniazid-resistant M. tuberculosis.

In a systematic review of treatment for latent infection with M. tuberculosis in persons at risk for multidrug-resistant tuberculosis, only two cohort studies met the inclusion criteria 162. The earlier retrospective cohort study found isoniazid not to be effective (RR 0.46, 95% CI 0.07–2.32) 163, while a later prospective cohort study found individualised tailored treatment with high-dose isoniazid (15–20 mg·kg⁻¹·day⁻¹), pyrazinamide, ethionamide and/or ethambutol and/or ofloxacin to be effective (RR 0.20, 95% CI 0.04–0.94) 164.

In absence of sufficient data, existing recommendations are subjective and are necessarily based largely on expert opinions, rather than on clinical evidence. Rifampicin for 4 months is a rational regimen choice for the treatment of contacts infected after exposure to a source case with isoniazid-resistant tuberculosis (grade D) 39. Opinions differ on the management of contacts of patients with multidrug-resistant tuberculosis. On the basis of the currently available evidence, the WHO does not recommend second-line drugs for preventive therapy among contacts putatively infected with multidrug-resistant bacilli 165. Careful clinical follow-up for at least 2 yrs is recommended instead. If multidrug-resistant tuberculosis is diagnosed, treatment should be started promptly with an appropriate treatment regimen. In contrast to the WHO, CDC recommends 6–12 months of preventive therapy with at least two antituberculosis drugs (e.g. ethambutol and pyrazinamide or pyrazinamide and a fluoroquinolone, depending on the drug susceptibility profile of the source case's isolate) for persons likely to be infected with multidrug-resistant bacilli, especially for those thought or known to be at a high risk of progression to tuberculosis (grade D) 166. However, high frequencies of treatment termination because of adverse events were reported for pyrazinamide-containing regimens (pyrazinamide plus ethambutol or pyrazinamide plus levofloxacin/ofloxacin) 135, 167–169. Hepatotoxicity was a prominent feature in some of these studies 135, 169. Fluoroquinolones have been found to be generally well tolerated in the treatment of tuberculosis or multidrug-resistant tuberculosis 134, 170. A question naturally arises as to the role of fluoroquinolone monotherapy in such situations. However, fluoroquinolones, in combination with second-line injectable drugs, currently play an important role in the treatment of multidrug-resistant tuberculosis 170, 171. In a recent study, exposure to fluoroquinolones for >10 days, particularly >60 days before tuberculosis diagnosis, was associated with a high risk of fluoroquinolone-resistant tuberculosis 172. Caution must, therefore, be exercised towards the use of fluoroquinolone monotherapy, and active surveillance remains a possible option among contacts of cases with multidrug-resistant tuberculosis, especially those at lower risk of developing clinical disease.

DEVELOPMENT OF NEW DRUGS

Clinical trials on new drugs for treatment of latent infection with M. tuberculosis generally lag behind those for treatment of active tuberculosis, mainly because of the logistical difficulties associated with studies in the treatment of latent infection with M. tuberculosis 173. As mentioned in a previous section, weekly rifapentine (plus isoniazid) is being explored in the treatment of latent infection with M. tuberculosis 149 after two randomised controlled trials on the treatment of tuberculosis 174, 175.

Adenosine triphosphate (ATP) synthesis is essential in the metabolism of mycobacteria 176, 177. TMC207 is a novel diarylquinoline with unique activity on the mycobacterial ATP-synthase 178. In a murine model of tuberculosis, the bactericidal effect of TMC207 was modest during the first week of treatment but increased in the following 3 weeks 179. TMC207 also showed good sterilising activity in mice. In combination with pyrazinamide, TMC207 led to bacillary sterilisation in 2 months 180. In another mouse study on tuberculosis, TMC207 plus rifapentine plus pyrazinamide given once weekly was more active than the current standard regimen of rifampicin plus isoniazid plus pyrazinamide given five times per week 181. In an early bactericidal activity study, significant bactericidal activity of TMC207 (400 mg) was observed from day 4 onward in a magnitude similar to that of isoniazid and rifampicin 182. In a phase II clinical trial, the addition of TMC207 to standard therapy for multidrug-resistant tuberculosis reduced the time to culture conversion and greatly increased the proportion with culture conversion (48% versus 9%) at 2 months 183.

As diarylquinolines are claimed to also have bactericidal activity against nonreplicating mycobacteria 176, 177, TMC207 could be a potential candidate drug for the treatment of individuals with latent infection with M. tuberculosis, including contacts of patients with multidrug-resistant tuberculosis.

ISSUES BEYOND EFFICACY

As latent infection with M. tuberculosis is both asymptomatic and noninfectious, the primary target of intervention is to reduce the risk of progression to tuberculosis. However, on average, only a minority of latently infected individuals will develop disease in their lifetime 38–42, 47–49. This necessarily sets a ceiling to the maximum effectiveness of treatment for latent infection with M. tuberculosis in terms of the number of patients to be treated to prevent an active case of tuberculosis. Figure 2 plots the number needed to treat to prevent one case of tuberculosis against the incidence of disease among the target group at different assumptions of treatment efficacies. Under all scenarios, the incidence of disease is the prime determinant of this measure of effectiveness. This would justify targeting only those at a high risk of developing tuberculosis, even at the expense of decreasing population coverage and limiting the overall impact.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2–

Impact of disease incidence on the number needed to treat to prevent one tuberculosis case within 5 yrs. RR: risk ratio of treatment versus no treatment.

Actual field practices vary widely across the world. In North America, a fairly wide range of target groups are included for screening and treatment of latent infection with M. tuberculosis, e.g. recent contacts, institutional clients, healthcare workers, immigrants from high incidence areas, persons with currently inactive fibrotic lesions, HIV-infected persons and subjects with other immunocompromised states, including those on immunosuppressive treatment 39, 184. The WHO mainly focused on HIV-infected persons and young household contacts of patients with infectious pulmonary tuberculosis 165, 185. 9 months of isoniazid is now recommended in the USA 39 and Canada 185, while 6 months of isoniazid is still more frequently used in other parts of the world 165, 186. Rifampicin for 4 months is an acceptable alternative regimen in the USA and Canada 39, 184, while a combination of isoniazid and rifampicin may be used more often in the UK 187. Many factors could have entered into the strategic formulations in addition to the simple question of efficacy. With the wide variations in socioeconomic and epidemiological conditions globally, it would be necessary for each locality to set its own priority after taking into account of available scientific data and local circumstances.

SAFETY AND MONITORING

The adverse effects of current treatment regimens also constitute a major hurdle both among HIV-infected and non-HIV infected individuals (tables 7 and 8). For preventive therapy, every individual put on treatment will be subject to the potential risk of drug toxicity, but only those who would otherwise develop disease benefit from such treatment. Figure 3 shows the impact of hepatitis and disease incidence on benefit versus risk ratio in terms of number of tuberculosis cases prevented per case of hepatitis, which readily falls below one when the frequency of drug-induced liver injury is high and/or the tuberculosis incidence is low. This explains why there must generally be a much lower threshold of accepting adverse drug events during the treatment of latent infection with M. tuberculosis than is the case with treatment of tuberculosis.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3–

Impact of hepatitis rate and disease incidence on benefit versus risk ratios.

The presence of a substantial risk in latent infection with M. tuberculosis treatment also implies a need for careful screening and close monitoring, which may add further costs and barriers of access to care. Besides patient education and clinical monitoring, baseline and monthly (or biweekly) laboratory testing of liver enzymes is recommended for chronic alcohol users, HIV-infected persons, females during pregnancy and within 3 months after delivery, and those with chronic liver disease, or taking concomitant medications that can be hepatotoxic 39, 86. Transient transaminase elevations are common and may reflect the process of hepatic adaptation 188. However, isoniazid and/or rifampicin should be withheld as recommended if the serum transaminase level is higher than three times the upper limit of normal in a symptomatic patient or five times the upper limit of normal in the absence of symptoms 39, 189. With the lower degree of tolerance for risk in the treatment of latent infection with M. tuberculosis, reintroduction of drug is seldom attempted after significant hepatotoxicity 86.

ACCEPTANCE

Even in North America, there appears to be suboptimal acceptance of preventive therapy regimens among both clinicians and patients. Treatment of latent infection with M. tuberculosis was not recommended by the attending doctors in 20–30% of patients who appeared otherwise eligible in some studies 190, 191. Physicians' reluctance to prescribe is especially noticeable among the older individuals, possibly related to the higher incidence of drug-induced hepatitis in the elderly 82. Even if treatment is offered, it might be refused by considerable proportions of persons 190, 192, 193. In a retrospective survey of public and private clinics in the USA and Canada, 123 (17.1%) of 720 subjects tuberculin skin-tested and offered treatment in the same clinics declined treatment. Interestingly, there was a higher likelihood for healthcare workers than for other tuberculosis contacts to decline 192. In another study among healthcare workers at an urban teaching hospital in the USA, only 69% of eligible persons accepted treatment against latent infection with M. tuberculosis 193. These results suggested that lack of knowledge about the treatment might not be a major factor for poor acceptance.

ADHERENCE

In a systematic review of 78 studies in the USA and Canada on adherence to treatment of latent infection with M. tuberculosis, treatment completion varied widely (from 19% to 96%) but was mostly suboptimal across high-risk groups, regardless of regimen 194. Lesser variations were observed in some of the large-scale studies in public programmes involving clients with similar characteristics. For example, 64% of treatment completion was reported in two large-scale retrospective reviews of medical records (involving 9,018 and 3,048 contacts) of individuals who were being treated with isoniazid in contact investigation programmes 190, 195 and in a 7-yr prospective survey of hepatotoxicity associated with isoniazid preventive therapy (involving 11,141 patients) in a public health tuberculosis clinic 85. A similar proportion of treatment completion (61%) was reported in a retrospective medical record review involving 19,582 treated inmates in 49 correctional facilities in the USA 196.

Similarly, treatment completion was found to be 53% in 68 public and private clinics included in a recent retrospective survey in the USA and Canada 192. In that particular survey, a 9-month isoniazid regimen, residence in a congregate setting (nursing home, shelter or jail), injection drug use, age above 15 yrs and employment at a healthcare facility were significantly associated with failure to complete treatment. However, associations between adherence and patient factors, clinic facilities or treatment characteristics were inconsistent across other studies 194. Adherence is a composite behavioural end-point. Heterogeneity in social context, provider arrangement and client profile could well have accounted for these variations. While there are suggestions that shorter durations of treatment may improve treatment completion, the higher frequency of adverse events leading to treatment termination for some of the short-course regimens might nullify such effects, especially for rifampicin plus pyrazinamide, and possibly isoniazid plus rifampicin 12.

COST-EFFECTIVENESS

A number of economic analyses have been conducted on the effectiveness of the treatment of latent infection with M. tuberculosis 197–206. Most of them were performed under a number of epidemiological, healthcare utilisation and costing assumptions and might not be applicable outside the specific economic realities of industrialised countries for which they were modelled. Under such assumptions, the use of isoniazid in the treatment of latent infection with M. tuberculosis has been found consistently to be cost-effective and often cost-saving in populations that are younger, and/or at greater risk to progress from disease. However, not all of them have taken a full account of the direct or indirect screening costs 197, 198. Most of them based the analysis on highly defined target groups, especially recent tuberculin skin test converters 197, 199 or tuberculin skin test-positive close contacts in low tuberculosis incidence settings 197, 198. While these defined target groups might be more easily accessible and have a higher screening yield, high disease incidence, fewer adverse effects and lower monitoring cost, they might account only for a relatively small proportion of all the tuberculosis cases within a community. Among older tuberculin skin test reactors, the conclusions were either small positive health effects at a cost considered acceptable in developed economies 197, or in an opposite direction 200. As it is not always easy to delineate science clearly from assumptions or value judgements in some of these modelling exercises 207, the findings might have to be taken with a grain of salt.

In a more recent cost–benefit analysis that included both screening and treatment for latent infection with M. tuberculosis in Germany 201, cost of screening and treatment based on a positive result in a QuantiFERON® TB Gold In-Tube (QFT) test alone amounted to 215.79 euro per close contact, less than that of dual step-testing (227.89 euro) or using the tuberculin skin test alone (232.58 euro). The cost-effectiveness of QFT-based procedures were sensitive to low treatment completion or increasing price, but the relationship between the strategies remained robust when the disease-predicting power of QFT was lowered to that for the tuberculin skin test in a sensitivity analysis. It therefore appears that the potentially higher specificity of QFT may help to improve cost-effectiveness of targeted screening of latent infection with M. tuberculosis in a low-incidence setting.

In a cost-effective analysis 202 comparing 6 months and 12 months of isoniazid in the treatment of patients with fibrotic lung lesions in the IUAT trial 72, the cost per case prevented was US$7,112 for the 6-month regimen, compared with US$16,024 for the 12-month regimen, and each additional case prevented by the 12-month regimen would cost US$80,807. Therefore, the 6-month regimen appeared to be more cost-effective, at least under the level of treatment completion and effectiveness as observed in that trial and the costs then prevailing in the USA. The lack of difference between the 6-month and the 12-month arm was based on all patients, and not on patients actually taking the drugs as prescribed. It might be argued that a cost-effectiveness study of a treatment regimen should be based on efficacy under optimised situations, rather than effectiveness before it is being well accepted 76, 208. However, in situations where acceptance and adherence to such preventive treatment are likely to remain suboptimal in absence of major breakthroughs in the current diagnostic and treatment tools, there could be an equally valid challenge to this counter-argument from a pragmatic perspective. Like efficacy, no study has compared the cost-effectiveness of 6 months and 9 months of isoniazid.

Alternative regimens were assessed in other cost-effectiveness analyses. Jasmer et al. 203 used a Markov model to conduct a cost-effectiveness analysis based on frequency of adverse events and completion of the two treatment regimens in a recent clinical trial. Although 2 months of rifampicin plus pyrazinamide and 9 months of isoniazid both increased life expectancy by 1.2 yrs as compared to no treatment, the short-course regimen cost US$273 more per patient over a range of completion frequencies. In another modelling study focusing on tuberculin skin test converters after recent exposure to an infectious index case, 4 months of rifampicin was cost-saving compared with 9-month therapy of self-administered isoniazid, and directly observed isoniazid plus rifapentine once weekly for 3 months is cost-saving for extremely high-risk patients and cost-effective for lower-risk patients 199.

In a decision and cost-effectiveness analysis on hypothetical HIV-infected patients with CD4 counts of 200 cells·mm⁻³ or less and positive results on tuberculin skin tests 206, isoniazid for 6 months or 12 months, isoniazid and rifampicin for 3 months, and rifampicin and pyrazinamide for 2 months were all cost-saving, but a 3-month regimen of isoniazid, rifampicin and pyrazinamide was not. Short-course preventive therapy appears to be a reasonable alternative to the 12-month isoniazid regimen.

CLINICAL PERSPECTIVES

The purpose of this section is to summarise the clinically relevant aspects of what is known on the subject, so as to facilitate clinical decision by the practising physicians.

APPENDIX

The grading of evidence for recommendations, according to the Scottish Intercollegiate Guidelines Network, can be found in table 9.