INTRODUCTION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Persons with a positive tuberculin skin test who have radiographic evidence of prior tuberculosis (American Thoracic Society/Centers for Disease Control [ATS/CDC] Class 4) and who have not received prior treatment are at increased risk for the subsequent development of tuberculosis (1). Case rates in one study (4) were approximately two and one-half times those for persons infected with Mycobacterium tuberculosis without chest radiographic abnormalities. Because disease activity cannot be determined from a single chest radiograph, unless previous radiographs showing that the abnormality has not changed are available, it is recommended that sputum examination be performed to assess the possibility of active tuberculosis. Once active tuberculosis has been excluded by sputum examination, these persons are high-priority candidates for treatment of latent tuberculosis infection.

Previous ATS/CDC recommendations offered two options for treatment of latent tuberculosis infection in this setting: isoniazid taken daily for 12 mo or isoniazid and rifampin taken daily for 4 mo (5). Current ATS/CDC guidelines recommend 9 rather than 12 mo of isoniazid (6). In this study, we compared the completion rates, frequency of adverse events, and cost effectiveness of isoniazid for 12 mo compared with isoniazid and rifampin for 4 mo in patients treated in the San Francisco Tuberculosis Clinic from 1993 through 1996.

	METHODS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Study Population and Treatment

The study population consisted of all patients who were evaluated at the San Francisco Tuberculosis Clinic from January 1, 1993 through December 31, 1996, who had positive (> 5 mm) tuberculin skin tests together with radiographic evidence of prior tuberculosis (TB-4) and who had not received adequate treatment. The radiographic findings that constituted evidence of prior tuberculosis were limited to apical fibronodular infiltrations, often with volume loss. Sputum examination was obtained to exclude the presence of active tuberculosis unless previous radiographs were available and showed that the infiltrates were stable. Patients were not classified as having radiographic evidence of prior tuberculosis if another diagnosis was found to account for the radiographic findings. Patients were given either isoniazid 300 mg daily for 12 mo or isoniazid 300 mg and rifampin 600 mg in combined formulation (Rifamate; Hoechst Marion Roussel, Kansas City, MO) daily for 4 mo. The study was not randomized, and patients treated in 1993-1994 always received isoniazid whereas those treated in 1995-1996, after the ATS/CDC guidelines were published, almost always received isoniazid and rifampin (4% of patients treated in 1995-1996 were begun on isoniazid). Both regimens were self-administered.

Evaluation of Treatment Outcome

Patients were evaluated monthly for adherence to therapy and for possible side effects. Completion of therapy was said to have occurred if patients returned for all scheduled monthly clinic visits and were deemed to have been adherent with therapy by nurses experienced in tuberculosis control. A standard protocol was used to question patients regarding medication ingestion and symptoms of possible adverse effects. The following data were collected from medical records: place of birth, age, type and duration of treatment in months, completion of treatment, reason for noncompletion of treatment, and the nature and effect of adverse effects, including whether any adverse event led to discontinuation of treatment. Hepatitis was defined by the presence of both symptoms and any abnormal serum transaminase level.

Evaluation of Cost Effectiveness

For evaluation of cost effectiveness, we developed a Markov model to track hypothetical cohorts of 1,000 patients with abnormal chest radiographs consistent with prior tuberculosis from age 52 (the mean age of our study population) to 99 yr who received one of three treatment strategies: (1) isoniazid daily for 12 mo, (2) isoniazid and rifampin daily for 4 mo, and (3) no treatment. Expected completion rates for treatment of latent tuberculosis infection and adverse reactions to treatment were determined from data obtained in our study. In the Markov model, all patients were at risk for developing active tuberculosis and for dying of tuberculosis or other causes. We assumed that no patients had infection with M. tuberculosis strains resistant to isoniazid and that all tuberculosis-related deaths occurred within the same year as development of active disease.

Baseline Model Parameters

Outcome of treatment of latent tuberculosis infection. Based on data obtained from our study, we used the following therapy outcomes in the baseline analysis:

1. Isoniazid strategy: completion of 12 mo of treatment, 79.8%; 6-11 mo of treatment, 3.4%; less than 6 mo of treatment, 16.7%; adverse reaction to treatment, 4.9%.

2. Isoniazid and rifampin strategy: completion of 4 mo of treatment, 83.6%; 2-3 mo of treatment, 7.1%; less than 2 mo of treatment, 9.2%; adverse reaction to treatment, 6.1%.

Risk of tuberculosis, death from tuberculosis, and death from other causes. We estimated from the literature that the annual risk of developing active tuberculosis among patients with radiographic evidence of prior tuberculosis was 1% (4). We estimated that for patients who developed tuberculosis between ages 52 and 64 yr, 7.4% would die from tuberculosis (7). For patients who developed tuberculosis at age 65 yr and older, 16.8% would die from tuberculosis (7). Using the Declining Exponential Approximation of Life Expectancy (DEAL) method (8) and U.S. data on life expectancy (9), we estimated the following annual all cause risk of death: age 52-59 yr: 3.61%; age 60-69 yr: 4.72%; age 70-79 yr: 7.09%; age 80 yr and older: 11.9%.

Efficacy of treatment of latent tuberculosis infection. Based on a previous study of 8,428 patients, completion of 12 mo of treatment of latent tuberculosis infection with isoniazid reduced the annual risk of tuberculosis by 89% (4). Six to eleven months of isoniazid had a preventive efficacy of 67%. Less than 6 mo was assumed to confer no prevention benefit. For patients who received isoniazid and rifampin, completion of 4 mo of therapy was assumed to reduce the annual tuberculosis reactivation risk by 89%. We assumed that at least 2 mo of treatment but less than 4 mo had an efficacy equivalent to that of at least 6 mo of isoniazid, 67% (the rate for those in the International Union Against Tuberculosis and Lung Disease study who completed at least 50% of a 12-mo regimen). Completion of less than 2 mo treatment was assumed to confer no benefit.

Outcomes and cost. We determined expected tuberculosis cases, tuberculosis-related deaths, and the difference in life expectancy between each cohort. We calculated the net incremental cost (savings) per 1,000 patients treated for each treatment strategy relative to the other.

Cost analysis. Costs were analyzed from the perspective of the health care system. All costs are in 1999 U.S. dollars. Costs were adjusted to their 1999 values using the medical care component of the U.S. Consumer Price Index (10). We discounted tuberculosis cases and costs prevented by 3% annually.

Cost of initial treatment. Costs were determined from the San Francisco Department of Public Health. We included only variable costs. Expected usage of treatment, chest radiographs, sputum examinations, monitoring for adverse reactions to therapy, and medical examinations were determined from the study described above. Initial treatment costs are presented in Table 1. The weighted average cost per patient treated was obtained by determining the cost for each outcome that we observed in our patients and then calculating the average cost based on the frequency of these outcomes. For example, 26 patients for whom isoniazid was prescribed for 12 mo took it for a cumulative total of 57 mo (an average of 2.2 mo per patient), required additional physician evaluations costing $598 ($23 per patient), and underwent liver function testing costing $4,446 (an average of 1.6 liver function test panels). The average cost for each of the 26 patients was $570, and this was factored into the total average cost per patient for the isoniazid group.

Averted cost. We assumed that 99% of active tuberculosis developing in the study population would be drug sensitive. We estimated the average cost of treating an active tuberculosis case, including hospitalization, as $13,530 (11).

The average cost of contact tracing and treatment was $2,396 (11). This includes the cost of contact investigation, tuberculin skin testing, medical evaluation for active tuberculosis, treatment of latent tuberculosis infection in infected contacts, and treatment of active tuberculosis. Costs were adjusted to their 1999 values using the medical care component of the U.S. Consumer Price Index (10).

Threshold and sensitivity analysis. Several model parameters were varied to determine the point at which the isoniazid strategy becomes cost saving relative to the isoniazid and rifampin strategy. We also varied the percentage of isoniazid resistance in the treated population to determine the effect on the net incremental savings per 1,000 treated for the isoniazid and rifampin strategy relative to the isoniazid strategy. We determined the net incremental cost of using 9 mo of isoniazid compared with 4 mo of isoniazid and rifampin within our model. This was done because current guidelines for the treatment of latent tuberculosis infection recommend 9 mo of daily isoniazid rather than 12 mo as was previously recommended in 1994 (5, 6).

	RESULTS

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

Treatment Outcome and Expected Tuberculosis Cases

Demographic and outcome data. A total of 1,022 patients were evaluated at the San Francisco Tuberculosis Clinic from January 1, 1993 through December 31, 1996 and begun on treatment for radiographic evidence of prior tuberculosis with either 12 mo of daily isoniazid (545 patients) or 4 mo of daily isoniazid and rifampin (477 patients). Sputum cultures were collected in 1,015 (99%) patients to exclude active tuberculosis. There were no significant demographic differences between the groups. The average age of the entire study population was 52.3 ± 16.3 (standard deviation) yr. The majority of patients (579) were male (57%). Ninety-three percent of patients were born outside of the United States, and over 75% of patients overall were born in countries in Asia or the Pacific Islands, The Republic of Philippines (33% of all patients), China (32%), and Vietnam (8.7%) being the three most common.

The only significant difference between the two cohorts was the time period in which treatment was begun. Ninety-six percent of those given isoniazid alone started treatment in 1993-1994 and 100% of those given isoniazid and rifampin began treatment in 1995-1996.

There was no significant difference in the percentage of patients who completed treatment for latent tuberculosis infection between the two groups, and overall 81.6% completed treatment (Table 2). Among patients who received isoniazid, 3.4% self-stopped the medication after taking it for 6-11 mo, and 16.7% self-stopped after less than 6 mo. Among patients who received isoniazid and rifampin, 7.1% self-stopped after taking 2-3 mo of medication, and 9.2% did so after less than 2 mo.

The overall rate of adverse effects was low (5.4%), and there was no significant difference in the percentage of patients who experienced adverse effects between the two groups. The most common adverse effects observed were rashes and nausea/vomiting. Hepatitis was very uncommon, occurring in only 1.1% of patients who received isoniazid and in 0.2% of patients who received isoniazid and rifampin.

There were six cases of tuberculosis among the 545 patients treated with 12 mo of isoniazid during a 5.5-yr follow-up period, compared with 3 of 477 treated with 4 mo of isoniazid and rifampin during a 3.5-yr follow-up period (p > 0.05 by the comparison of incidence rates test). Eight of these nine tuberculosis cases had disease due to drug-susceptible M. tuberculosis, whereas one patient (who received isoniazid for 12 mo) had disease due to isoniazid-resistant M. tuberculosis.

Expected tuberculosis cases and tuberculosis-related deaths. Without treatment of latent tuberculosis infection, 157.5 (116.9 discounted) tuberculosis cases would be expected in the cohort over the remainder of its lifetime. With isoniazid, 43.7 (32.1 discounted) cases would be expected, and with isoniazid and rifampin, the expected number would be 34.7 (25.5 discounted). Without treatment of latent infection, 17.7 tuberculosis-related deaths would be expected in the cohort over the remainder of its lifetime. With isoniazid, 5.0 tuberculosis-related deaths would be expected, and with isoniazid and rifampin, 4.0 tuberculosis-related deaths would be expected. Isoniazid would increase life expectancy by 1.42 yr relative to no treatment; isoniazid and rifampin would increase life expectancy by 1.53 yr.

Costs and Cost Effectiveness

Initial treatment. For a cohort of size 1,000, the isoniazid strategy would cost $593,757 ($594 per patient), whereas the isoniazid and rifampin strategy would cost $592,107 ($592 per patient) (Table 1). As shown in Table 1, the difference in the cost of medications favoring isoniazid alone is offset by the cost of monthly follow-up visits favoring isoniazid and rifampin.

Expected cost of tuberculosis treatment and contact tracing and treatment. Without treatment of latent tuberculosis infection, the discounted expected cost of treatment of active disease and contact tracing and treatment is $1,861 per patient. With isoniazid, the discounted expected cost is $539 per patient. With isoniazid and rifampin, the discounted expected cost is $405 per patient.

Incremental cost-effectiveness. Relative to no treatment of latent infection, both the isoniazid and isoniazid and rifampin strategies are more effective and less costly. The isoniazid strategy would produce net incremental savings of $728 per patient treated compared with no treatment, whereas the isoniazid and rifampin strategy would produce net incremental savings of $864 per patient treated. Relative to the isoniazid strategy, isoniazid and rifampin is cost saving, producing net incremental savings of $135 per patient treated.

Threshold analysis. We determined conditions in which treatment with 12 mo of isoniazid would be cost saving relative to 4 mo of isoniazid and rifampin. If the cost of 4 mo of isoniazid and rifampin were more than doubled to $286 (i.e., for the medications only), then the 12 mo of isoniazid regimen would become cost saving. In addition, if the efficacy of 4 mo of isoniazid and rifampin were decreased by 8.6% to 81%, then the 12 mo of isoniazid strategy would become cost-saving.

Sensitivity analysis. We varied the rate of isoniazid resistance to determine how this would affect the results of our cost-effectiveness analysis. A small increase in the rate of isoniazid resistance to 10% (close to the true rate of isoniazid resistance among incident cases for the past 5 yr) would result in substantial incremental savings for patients in the 4 mo of isoniazid and rifampin group, increasing from $135 per patient at baseline (assuming no significant isoniazid resistance) to $288 per patient treated assuming no loss of effectiveness for the isoniazid and rifampin regimen. Further increases in the rate of isoniazid resistance would lead to even greater incremental savings in the 4 mo of isoniazid and rifampin group.

We also determined the cost effectiveness of 9 mo of daily isoniazid, rather than 12 mo, compared with 4 mo of daily isoniazid and rifampin. Decreasing the duration of isoniazid to 9 mo did not change the overall results, although the net incremental cost effectiveness of 4 mo of isoniazid and rifampin was decreased to $92 per patient.

	DISCUSSION

TOP ABSTRACT INTRODUCTION METHODS RESULTS DISCUSSION REFERENCES

This study shows that for patients with a positive tuberculin skin test indicative of infection with M. tuberculosis and who have radiographic evidence of prior tuberculosis (inactive pulmonary tuberculosis), both 12 mo of isoniazid and 4 mo of isoniazid and rifampin have reasonably high rates of completion and few adverse effects. Furthermore, both treatments are cost saving and improve survival compared with no treatment, with 4 mo of isoniazid and rifampin saving $135 per patient treated compared with 12 mo of isoniazid. The cost advantage of using 4 mo of isoniazid and rifampin is maintained even when it is compared with 9 mo of daily isoniazid, the new recommendation for duration of treatment in patients with radiographic evidence of prior tuberculosis (6).

The data utilized for the cost-effectiveness analysis were the observed completion rates and adverse effects among the 1,022 patients in our study population. The study population largely (94%) consisted of foreign-born persons from countries having a high prevalence of tuberculosis infection. Thus, the prior probability was high that an upper lobe fibrotic-appearing lesion on chest radiograph was tuberculous in origin. Nevertheless, patients were carefully evaluated for the presence of both active tuberculosis and other diseases before treatment of latent infection was begun. Treating patients with radiographic evidence of prior tuberculosis should be undertaken only after such an evaluation. An advantage of isoniazid and rifampin treatment is that if a mistake in classification is made and the patient actually has active tuberculosis, the treatment will not lead to drug resistance and nothing is lost, assuming that the mistake is found and appropriate treatment is continued. Results of our analysis of completion rates and adverse effects were then incorporated into a Markov model to assess impact on cost effectiveness of treatment.

From the perspective of a tuberculosis control program, the initial cost of treatment is almost the same for 12 mo of isoniazid or 4 mo of isoniazid and rifampin (Table 1). That is, the same investment has to be made regardless of the regimen used. However, the 4-mo treatment is more cost saving because of more averted cases and the cost of eight additional monthly follow-up visits for patients treated with isoniazid.

An important point in the evaluation of the results of our cost-effectiveness analysis is that we made the conservative assumption that no patients had infection with M. tuberculosis strains resistant to isoniazid. This is likely not accurate given that most patients in our study population were from parts of the world where resistance to isoniazid is relatively common. For example, in San Francisco for the past 5 yr, 8.0% of all incident tuberculosis cases were infected with strains resistant to isoniazid. Only 1.2% of the incident cases in San Francisco had strains that were resistant to isoniazid and rifampin for the past 5 yr. If a 10% rate of isoniazid resistance is present, the incremental cost benefit of 4 mo of isoniazid and rifampin becomes even greater, increasing to $288 per patient treated, compared with the isoniazid alone strategy. This result is based on the assumption that treatment with rifampin alone is efficacious. This assumption is supported by a study of tuberculosis-infected persons with silicosis conducted in Hong Kong, in which rifampin alone for 3 mo was found to have an efficacy similar to isoniazid alone for 6 mo (12). Because our baseline model assumed no infection with isoniazid-resistant organisms, the superiority of isoniazid and rifampin shows that our results are generalizable to all areas regardless of their level of isoniazid resistance.

To our knowledge, this is the first study comparing the outcomes and cost effectiveness of treatment for patients having radiographic evidence of prior tuberculosis. In one of the first studies conducted on patients having radiographic evidence of prior tuberculosis, Dutt and coworkers (13) treated 452 patients in Arkansas with 1 mo of daily isoniazid and rifampin followed by twice-weekly isoniazid and rifampin for 3 mo. They found that 414 (92%) completed therapy and that adverse effects occurred in 21 (4.7%), with hepatitis in only 4 (0.9%). We found a lower completion rate but similarly low rates of adverse effects (6.1%) and hepatitis (0.2%) among the 477 patients who were treated with daily isoniazid and rifampin for 4 mo. More recently, Goldberg and coworkers (14) compared the results of treating these patients with isoniazid, rifampin, ethambutol, and pyrazinamide for 4 mo versus isoniazid for 12 mo. They found that the rate of completion of treatment was similar in the two groups (68%), but that the four-drug treatment was associated with significantly more side effects and was more costly. We did not find any significant difference in side effects comparing 4 mo of isoniazid and rifampin with 12 mo of isoniazid alone. Taken together, these findings suggest that the addition of rifampin to isoniazid does not result in any increase in side effects but that the addition of ethambutol and pyrazinamide can lead to significantly greater rates of toxicity.

The primary outcomes of this study were completion rates and rates of adverse effects of the two regimens. Passive surveillance detected nine cases of tuberculosis through 1999, six occurring among those who received 12 mo of isoniazid and three among those who received 4 mo of isoniazid and rifampin. However, our study was not adequately powered to detect differences in efficacy. In addition, the shorter follow-up period of patients who received isoniazid and rifampin (treated in 1995-1996) likely contributed to the lower number of active tuberculosis cases subsequently diagnosed in that group.

Because our study patients were not allocated randomly to one of the two regimens, we cannot be certain that the groups were not different in ways not accounted for in the analysis and that nonrandom allocation may have influenced our results. However, we did not observe any significant differences in the age, sex, or countries of origin of the patients. In addition, other than the differences in the two regimens, patients were managed according to the same clinic protocol throughout the study period, thus tending to minimize bias.

In summary, treatment of patients having radiographic evidence of prior tuberculosis in whom active tuberculosis has been excluded with either 12 mo of daily isoniazid or 4 mo of daily isoniazid and rifampin is both safe and associated with a relatively high rate of completion of therapy. Both treatments are cost saving and prolong survival compared with no treatment. The 4-mo isoniazid and rifampin treatment is cost saving compared with isoniazid alone, and the cost saving increases as the prevalence of infection with strains resistant to isoniazid increases. This advantage was maintained even when compared with 9 mo of isoniazid, the new ATS/CDC recommendation for treatment with isoniazid alone. These data support the use of the 4-mo treatment strategy with isoniazid and rifampin as the preferred option, especially in patients at risk for isoniazid-resistant tuberculosis infections.