The Centers for Disease Control and Prevention and the American Thoracic Society recommend a 12-month regimen of isoniazid prophylaxis for HIV-infected patients [12-14]. Although prophylaxis has been proposed to reduce HIV-related tuberculosis in many countries, persistent difficulties in implementation and adherence, risk for isoniazid-associated hepatotoxicity, and unresolved questions about costs and benefits temper enthusiasm for its use [2]. After researchers found that some combination drug regimens were effective after several months in a murine model of tuberculosis and in patients with silicosis [15, 16], clinical trials were done to test short-course regimens in HIV-positive patients. The results of the trials indicate that some regimens are safe and effective [17-22]. My analysis addresses the following questions: How do the short-course prophylaxis regimens and the 12-month isoniazid regimen compare? Are the more expensive short-course regimens cost-effective?

Methods

Top Methods Results Discussion Author & Article Info References

A decision and cost-effectiveness analysis was done for hypothetical HIV-infected patients (CD4 count 200 cells/mm³) who have positive results on tuberculin skin tests. The currently recommended 12-month isoniazid regimen was compared with no prophylaxis and with six short-course prophylaxis regimens: isoniazid, 300 mg/d for 6 months; isoniazid, 600 mg or 800 mg (depending on the patient's weight) twice weekly for 6 months; isoniazid, 300 mg/d, and rifampin, 600 mg/d, with and without pyrazinamide, 2000 mg/d for 3 months; rifampin, 450 mg or 600 mg, and pyrazinamide, 1500 mg, 2000 mg, or 2500 mg twice weekly for 2 months; and rifampin, 600 mg/d, and pyrazinamide, 20 mg/kg of body weight per day for 2 months. The outcome measures were expected 5-year survival rate, lifetime incidence of tuberculosis per cohort, life expectancy, and the cost to extend life by 1 quality-adjusted life-year (QALY).

A Markov model [23] programmed on Microsoft Excel (Microsoft Corp., Redmond, Washington) was used to calculate the outcomes (Figure 1). In the first year, patients chose whether to receive prophylaxis. Those who chose prophylaxis risked a fatal adverse event with some regimens. (Nonfatal adverse events are included in the analysis and affect quality of life but not survival.) All patients risked developing tuberculosis, but this risk was decreased by prophylaxis. Regardless of whether tuberculosis occurred, patients risked death from any cause. In each subsequent year, patients who had not had tuberculosis risked active disease and all patients risked death from any cause.

View larger version (11K): [in this window] [in a new window]   Figure 1. Decision tree for HIV-infected tuberculin reactors. The square represents a decision node; circles represent chance nodes. AE = adverse event; TB = tuberculosis.

Table 1 shows the assumptions and the range of values for each variable. Transition probabilities were calculated with the exponential transformation [23]. A base-case analysis was done, followed by a sensitivity analysis that tested the full range of values for each assumption one at a time and in combination. The cost-effectiveness analysis used a U.S. societal perspective, direct medical care costs only, quality adjustments, and a 3% annual discount rate. Costs were converted to 1997 U.S. dollars by using the medical care component of the consumer price index.

Active Tuberculosis

Observational studies and clinical trials of HIV-infected tuberculin reactors show annual rates of tuberculosis (including confirmed and clinically probable cases) of 3.4% to 16.2% [11, 17-1924, 33-38]. The base-case analysis used an annual rate of 6.7%, which was the median rate of the placebo groups of the clinical trials [11, 17-19].

Effectiveness of Prophylaxis Regimens

The effectiveness of prophylaxis was taken from the results of six clinical trials [11, 17-22]. Incomplete adherence to the regimens was incorporated into this study through use of the intention-to-treat analyses of the clinical trials. Because the duration of effectiveness has not been studied, the base-case analysis used the assumption that effectiveness lasts for 3 years and diminishes to zero after 5 years.

The 12-month regimen of daily isoniazid was tested in a clinical trial in Haiti by Pape and colleagues [11]. In 63 tuberculin reactors, the tuberculosis rate was 1.7% for patients receiving isoniazid plus vitamin B₆ and 10.0% for those receiving vitamin B₆ alone. A 6-month regimen of daily isoniazid was studied in three clinical trials. Wadhawan and colleagues [17] reported the results of a trial in Zambia in which isoniazid significantly reduced the tuberculosis rate. A smaller study by Hawken and associates in Kenya [18] followed patients for less than 2 years and found a rate reduction that was not statistically significant. Whalen and colleagues [19] conducted a large trial in Uganda and reported a statistically significant reduction in the tuberculosis rate (from 3.41% in the placebo group to 1.08% in the isoniazid group; relative risk [RR], 0.33 [95% CI, 0.14 to 0.77]) [19]. The effectiveness found in this trial was used in the base-case analysis, and the range of the CI was used in the sensitivity analysis.

Whalen and colleagues [19] also studied two 3-month regimens: daily isoniazid and rifampin with and without pyrazinamide. The rate of tuberculosis in the group receiving the two-drug regimen was 1.32% (RR, 0.40 [CI, 0.18 to 0.86] compared with the placebo group). The rate in the group receiving the three-drug regimen was 1.73% (RR, 0.51 [CI, 0.24 to 1.08]; P = 0.08) compared with the placebo group.

In a trial in Haiti, Halsey and colleagues [20] compared a 2-month regimen of twice-weekly rifampin and pyrazinamide with a 6-month regimen of twice-weekly isoniazid plus vitamin B₆. In the 36 months after randomization, the rate of tuberculosis was 1.7% per year in the rifampin and pyrazinamide group and 1.8% per year in the isoniazid group, a difference that is not statistically significant. Because this trial did not include a placebo group, the base-case rate with no prophylaxis (6.7%) was used to calculate the effectiveness of these regimens compared with no prophylaxis.

The effectiveness of a 2-month regimen of daily rifampin and pyrazinamide was taken from an international trial conducted by Gordin and colleagues [21, 22]. They reported a 1.2% rate of tuberculosis. Because the study did not have a placebo group (the comparison group received 12 months of isoniazid and also had a tuberculosis rate of 1.2%), the method mentioned above was used to calculate the effectiveness of the 2-month regimen compared with no prophylaxis.

Adverse Effects of Prophylaxis

Rates of adverse reactions to prophylaxis were taken from the same clinical trials [17-22]. The methods for recording and reporting these events differed substantially among studies; this made comparisons difficult. Thus, each rate was varied through a wide range in the sensitivity analysis. Although fatal isoniazid-associated hepatotoxicity was not reported in the clinical trials, risk for this event was included in the base-case analysis of regimens that contained isoniazid because its occurrence is well documented

In the Haitian study reported by Pape and colleagues [11], isoniazid taken for 12 months caused no important adverse events. In contrast, 10.7% of the patients in the international trial done by Gordin and colleagues had adverse events [21, 22]. In other populations, clinical hepatitis occurred at rates of 0.5% to 6.9% and deaths occurred at a rate of 14 per 100 000 persons [25-28]. Because the standard protocol now includes monitoring for hepatitis, the mortality rate has decreased to 1 to 2 per 100 000 persons [13, 14, 28, 29]. The base-case analysis used the assumption that the rates of mild, moderate to severe, and fatal adverse events are 10.1%, 0.6%, and 0.002%, respectively.

Adverse event rates for the 6-month regimen of daily isoniazid and the 3-month regimens of isoniazid and rifampin with and without pyrazinamide were taken from the Ugandan trial done by Whalen and colleagues [19]. For this analysis, the rates were based on the marginal rates above those of placebo recipients. The rate of fatal isoniazid-associated hepatotoxicity for all 3- and 6-month regimens that contained isoniazid was assumed to be half that for the 12-month regimen [26]. Adverse event rates for the 6-month regimen of twice-weekly isoniazid and the 2-month regimen of twice-weekly rifampin and pyrazinamide were taken from the Haitian trial done by Halsey and colleagues [20]. Adverse event rates were not reported in detail, but diarrhea and stomach pain were more common in the isoniazid group than in the rifampin and pyrazinamide group.

Survival

Survival rates for HIV-infected patients with and those without an episode of active tuberculosis were taken from a retrospective cohort study conducted at four U.S. medical centers [9]. Patients with tuberculosis were matched with controls without tuberculosis according to CD4 cell count and study site. The two groups were similar with respect to age, sex, ethnicity, previous incidence of opportunistic infection, and use of antiretroviral therapy. The 1-year survival rate was substantially lower for patients (65%) than for controls (90%).

Costs and Quality of Life

The cost of tuberculosis diagnosis and treatment ($16 636) was determined on the basis of a national study [31]. The analysis included charges for physician services and assumed a combination of drug-susceptible and drug-resistant patients and inpatient and outpatient treatment.

The cost of prophylaxis included professional fees, radiology and laboratory tests (costs obtained from Medicare and commercial insurance fee schedules), and medication costs [32]. Patients who are infected with HIV usually visit a physician every 3 months; this cost was not included in the cost of prophylaxis. Half of the patients required chest radiography ($50 per patient) before beginning prophylaxis; the remainder had recent radiography as part of usual care. Patients receiving daily drug regimens required home visits by a nurse ($25 each) for monitoring purposes in the months without physician visits. Patients receiving twice-weekly regimens required twice-weekly home visits by an outreach worker ($20 each). Liver function testing with measurement of serum aspartate aminotransferase level ($17 each) was done monthly and was followed up by telephone. Medication costs included fixed-combination drugs, if available [32]. Eighty percent of patients adhered to the prescribed regimen.

The cost of mild adverse reactions to prophylaxis was assumed to include two physician office visits ($54 each), two liver function tests ($17 each), and one medication to treat symptoms ($20). The cost of moderate to severe or fatal adverse reactions to prophylaxis ($13 000) was assumed to include hospitalization and physician fees that were equivalent to the cost of acute liver disease (obtained from the Medicare fee schedule).

Quality-of-life adjustments for symptoms of HIV disease were made on the basis of the average results of several studies [30]. On a utility scale of 0 to 1 (equivalent to a range from death to perfect health), HIV-infected patients with CD4 counts of 200 cells/mm³ or less but no opportunistic infections had a mean utility of 0.65; patients with opportunistic infections had a mean utility of 0.62. The base-case analysis used the assumption that HIV-infected patients without tuberculosis had a utility of 0.65 and HIV-infected patients with tuberculosis had a utility of 0.62. Mild adverse reactions to prophylaxis were assumed to last for 1 week and to cause a utility of 0.5 until recovery. Severe adverse reactions were assumed to last for 3 weeks and to cause a utility of 0.25 until recovery or death.

Results

Top Methods Results Discussion Author & Article Info References

Without receiving prophylaxis, hypothetical HIV-infected patients who had CD4 counts of 200 cells/mm³ or less and positive results on tuberculin skin tests had an expected 5-year survival rate of 54.2%, an average life expectancy of 7.79 years, and an incidence of tuberculosis of 381 cases per 1000 persons in the cohort. The cost of tuberculosis diagnosis and treatment was $3709 per person, and the quality-adjusted life expectancy was 3.09 QALYs; these figures include a discount for future costs and health events of 3% per year.

Table 2 shows the results for patients who receive prophylaxis. Compared with no prophylaxis, the 12-month isoniazid regimen increased expected 5-year survival rates by 9%, life expectancy by 0.73 years (8.7 months), and quality-adjusted life expectancy by 5%; decreased lifetime incidence of tuberculosis by 27%; and reduced medical care costs by 40%. The short-course regimens also improved 5-year survival rate, life expectancy, incidence of tuberculosis, and quality-adjusted life expectancy compared with no prophylaxis. The 2-month regimen of daily rifampin and pyrazinamide and the 12-month regimen of daily isoniazid resulted in similar benefits. The other regimens resulted in somewhat fewer benefits. Short-course prophylaxis extended life expectancy from 0.42 years to 0.72 years (5.0 months to 8.6 months) and quality-adjusted life expectancy from 0.08 QALYs to 0.14 QALYs.

All short-course regimens, with the exception of the 3-month triple-drug regimen, saved medical care dollars compared with no prophylaxis. The 2-month regimen of daily rifampin and pyrazinamide resulted in savings of medical care dollars that were similar to those associated with the 12-month isoniazid regimen. The triple-drug regimen is the most expensive regimen, and it has the worst side effect profile and the lowest rate of effectiveness; however, the cost to increase life expectancy by 1 QALY with this regimen was only $1941. Figure 2 illustrates the costs and benefits of the prophylaxis regimens compared with no prophylaxis.

View larger version (9K): [in this window] [in a new window]   Figure 2. Costs and benefits of prophylaxis compared with no prophylaxis. All future costs and health events are discounted at a rate of 3% per year. The black circle represents isoniazid daily for 12 months; the white circle represents isoniazid daily for 6 months; the black square represents isoniazid twice weekly for 6 months; the white square represents isoniazid and rifampin daily for 3 months; the black triangle represents isoniazid, rifampin, and pyrazinamide daily for 3 months; the white triangle represents rifampin and pyrazinamide twice weekly for 2 months; and the diamond represents rifampin and pyrazinamide daily for 2 months.

A sensitivity analysis was done to determine whether any variation of the assumptions results in clinical outcomes for which not receiving prophylaxis is better than receiving prophylaxis in terms of 5-year survival rate, life expectancy, and incidence of tuberculosis. Even when each assumption was varied through the ranges listed in Table 1, taking any of the prophylaxis regimens is better than not taking prophylaxis. The exception is the widest variation of the effectiveness of the 3-month triple-drug regimen. In the clinical trial, the CI for the effectiveness of this regimen extended to zero. Clinical outcomes improved as long as this regimen had an effectiveness of 2% or greater. When $100 000 per QALY saved was used as the upper limit of a cost-effective medical intervention, any prophylaxis regimen remained cost-effective even when each assumption is varied through the ranges listed in Table 1.

Even after combination of the most pessimistic assumptions (the extremes of the ranges shown in Table 1), each prophylaxis regimen, with the exception of the triple-drug regimen, resulted in better clinical outcomes than not receiving prophylaxis. No prophylaxis regimen was cost-effective when the extreme assumptions were combined.

Discussion

Top Methods Results Discussion Author & Article Info References

This decision analysis shows that HIV-infected patients with CD4 counts of 200 cells/mm³ or less and positive results on tuberculin skin tests benefit greatly from taking the currently recommended 12-month regimen of isoniazid prophylaxis. Life expectancy increased by an average 8.7 months, and the lifetime risk for tuberculosis decreased by 27%. The analysis also shows that some short-course prophylaxis regimens result in benefits similar to those of the 12-month isoniazid regimen. Receiving isoniazid for 6 months, isoniazid and rifampin for 3 months, or rifampin and pyrazinamide for 2 months increased life expectancy by 6.2 to 8.6 months and decreased the lifetime risk for tuberculosis by 19% to 26%. Prophylaxis with these regimens is beneficial even when a wide range of pessimistic assumptions is made about effectiveness and risks.

The cost-effectiveness analysis shows that prophylaxis saves medical care dollars. The cost of prophylaxis was assumed to include monthly monitoring for hepatotoxicity and other adverse effects, which increased costs substantially. However, all but one regimen saved medical care dollars. The ratio of invested dollars to savings ranged from 1:1 for the 6-month regimen of twice-weekly isoniazid (prophylaxis cost, $875; savings, $876) to 1:7 for the 6-month regimen of daily isoniazid (prophylaxis cost, $190; savings, $1268). The only regimen that did not save medical care dollars was the triple-drug regimen; however, this regimen improved life expectancy and cost $1941 to save 1 QALY, a bargain compared with other preventive interventions [39].

Although the cost-effectiveness of short-course prophylaxis has not previously been studied, the cost-effectiveness of the standard isoniazid regimen is well established. Isoniazid prophylaxis reduces costs or saves medical care dollars when taken by tuberculin reactors who are at high risk for developing active tuberculosis because of HIV infection, intravenous drug use, recent conversion of the tuberculin skin test, or presence of a fibrotic scar on chest radiograph that indicates previous tuberculosis [40-42]. Isoniazid prophylaxis is also cost-effective for low-risk tuberculin reactors who have no other risk factors for active disease [41, 43, 44]. This analysis shows that even relatively expensive short-course regimens save medical care dollars because they prevent costly active tuberculosis. Most lifesaving interventions do not also save medical care dollars. Tengs and colleagues [45] reviewed 310 medical interventions and found that the median cost per life-year saved was $19 000; only 45 of the interventions saved medical care dollars.

The risk for isoniazid-associated hepatotoxicity is one of the greatest concerns inhibiting enthusiasm for isoniazid prophylaxis [46]. My analysis shows that even pessimistic assumptions about the risks for, costs of, and quality-of-life reductions caused by adverse events have little effect on the decision to choose prophylaxis. Although clinical hepatitis was not detected in two trials of HIV-infected patients that compared isoniazid prophylaxis with placebo [11, 18], this disease was reported by Whalen and colleagues [19]. Rifampin and pyrazinamide regimens eliminate concern about isoniazid-associated hepatotoxicity, and the clinical trials that compared these regimens with isoniazid prophylaxis reported similar rates of adverse events in both groups [20-22].

My study has several important limitations. First, the duration of prophylaxis effectiveness is unknown, and the assumption that protection continues for 3 years and diminishes to zero after 5 years is an estimate. Persons who do not have HIV infection receive at least 19 years of protection from a 12-month course of isoniazid [47], although one study reported a loss of benefit after 6 years [48]. No clinical trial of HIV-infected patients has lasted longer than 36 months [11, 20-22]. Nevertheless, the sensitivity analysis in this study shows that even if the benefit of prophylaxis is lost after 2 years, life expectancy and incidence of tuberculosis are better with prophylaxis than without it.

Second, although information on the effectiveness of prophylaxis comes mostly from studies of patients in developing nations, this analysis pertains to patients in the United States. It is not known whether adherence to prophylaxis regimens varies among countries; however, one study suggests that effectiveness does not. In the international trial of short-course rifampin and pyrazinamide compared with 12 months of isoniazid [21], more than two thirds of the 1583 patients were enrolled in the United States. The event rates were similar among all countries and demographic groups.

Third, I used economic data from the United States for this analysis; the results cannot be applied to patients in developing nations because the relative costs of prophylaxis regimens and tuberculosis treatment probably differ substantially.

Finally, the clinical data were derived from populations that had not received highly active antiretroviral therapy, and the effect of this therapy on the need for tuberculosis prophylaxis is not yet known. Although rates of opportunistic infection have decreased with the widespread use of combination antiretroviral therapy, even extreme viral suppression may not fully reconstitute cell-mediated immunity [49, 50]. Prophylaxis against opportunistic infections is recommended even if antiretroviral therapy substantially increases the CD4 cell count [51].

The findings of my analysis suggest that some short-course prophylaxis regimens are reasonable alternatives to a 12-month regimen of isoniazid. Because the results of some short-course regimens and the 12-month isoniazid regimen are so similar, the prevention strategy can be chosen on the basis of such considerations as individual preference, convenience, risk factors for side effects, history of non-adherence to other medication regimens, and institutional capacity for nursing or outreach worker home visits. The 2-month regimen of rifampin and pyrazinamide is especially attractive because it is safe, effective, cost-effective, and relatively convenient and eliminates concern about isoniazid-associated hepatotoxicity.