Isoniazid Preventive Therapy for Tuberculosis Decision Analysis Considering Ethnicity and Gender1.2

THERESA J. JORDAN, EUGENE M. LEWIT, and LEE B. REICHMAN3

Introduction

Although most tuberculosis (TB) in the United States occurs in well-defined population groups and is relatively easy to prevent, case rates have risen alarmingly for the first time since reporting began in 1953 (1). Accordingly, the U.S. Department of Health and Human Services goal of the elimination ofTB in the United States by the year 2010 (2) may be difficult to achieveeven with increased utilization of isoniazid (lNH) preventive therapy (3). The American Thoracic Society-Centers for Disease Control guidelines clearly and unequivocally mandate INH preventive therapy for high-risk, adult tuberculin reactors who are characterized by positive skin tests as well as specific additional factors that place them at high risk for developing active TB (4). Whether to prescribe INH preventive therapy for low-risk tuberculin reactors without specific risk factors remains controversial, however, largely due to hepatotoxic effects of INH. Investigators have attempted to resolve this controversy through the application of decision analysis. Some concluded that the TB reduction benefit of INH preventive therapy outweighs the risks of INH-induced hepatotoxicity (5, 6). Others favored withholding INH (7), and still others showed the decision to be too close to call one way or the other (8). This disagreement has caused confusion, which has contributed to the declining use of preventive therapy in high- as well as low-risk individuals. The disparate results of previous decision analyses have hinged to a great extent on the choice of estimates of such parameters as lifetime risk of TB, TB mortality, INH effectiveness, and INH mortality. In fact, the recent set of salvos both for and against INH preventive therapy (8-12) seemed to be aimed at establishing a single correct estimate for each of the crucial parameters involved

SUMMARY The decision to prescribe or withhold isoniazid (INH) preventive therapy for low·risk tuberculin reactors has been highly controversiel, primarily due to isoniazid's possible hepatotoxic effects. Previous analyses have explored the INH decision only from the perspective of patient age, recognizing that the risks of INH-induced hepatotoxicity are age related. Decision analyses presented in this paper assess the impact of gender and ethnic group, as well as age, on the INH decision. Results for low-risk patients favor prescribing INH preventive therapy for aIl20·yr·olds, all 35-yr-olds except black women, and no 50·yr·olds, projecting life expectancy benefits that range from 3 to 19 days. A comparison set of analyses performed for hlgh·risk patients favors prescribing INH for all groups except 50·yr-old black women. These findings suggest that ethnicity, gender, and age should be considered when making the decision to prescribe or withhold INH preventive therapy. AM REV RESPIR DIS 1991; 144:1357-1360

in this decision. However, epidemiologic data suggest that the TB problem differs substantially across subgroups of the population (13, 14).Therefore, the use of averages taken across subgroups to arrive at a single estimate may be misleading. In contrast, it might be preferable to analyze the INH decision separately for subgroups of the population using parameter estimates that best fit each of the groups under consideration. The present study was designed to provide subgroup-specific decision analyses for low-risk reactors, incorporating patient age, gender, and ethnicity, insofar as published empirical data permit. The purpose of the study was to determine whether some controversy could be eliminated by conceptualizing the INH decision as a decision for each specific subgroup of low-risk patients, tied as closely as possible to the published parameter estimates that are most appropriate for the particular group. To compare the magnitude of differences found for lowrisk patients with the noncontroversial decision for high-risk patients, the decision was analyzed for both low-risk and high-risk reactors. Methods Decision Model The decision tree presented in figure 1 illustrates the outcomes resulting from withholding or prescribing INH that are considered

in these analyses. This model is similar to that used by other investigators except for the addition in this model of the contingency that all patients who develop TB will be treated and may develop hepatotoxicity from the INH-containing treatment regimen. Life expectancy is used as the primary outcome measure. Analyses were performed for 12subgroups of patients according to age, gender, and ethnic group. The ages 20, 35, and 50 yr were selected to facilitate comparison with previous analyses that reported findings at these points (8).

Probability Estimates Wherever possible, the analyses in this investigation were performed using parameter estimates specific for age, gender, and ethnici(Received in original form January 28, 1991 and in revised form June 3, 1991) 1 From the Division of Pulmonary Medicine, the Office of Primary Health Care Education, and the Informatics Section, Department of Medicine,New Jersey Medical School, University of Medicine and Dentistry of New Jersey, Newark, New Jersey. 2 Correspondence and requests for reprints should be addressed to Lee B. Reichman, M.D., M.P.H., Director, Pulmonary Division, University Hospital I 354, 150 Bergen Street, Newark, NJ 07103-2425. 3 Supported by Preventive Pulmonary Academic Carrer Award No. I-K07-HL-02095-05 and Pulmonary Complication of HIV Infection Contract No. I-HR-76032A from the Division of Lung Diseases, National Heart, Lung and Blood Institute, National Institutes of Health.

1357

JORDAN, LEWIT, AND REICHMAN

1358 Survive

TB Premature Death

..

Hepatotoxicity

C

Survive

TB

INH

Premature Death

No Hepatotoxicity Premature Death Survive

TB Premature Death

No INH

Premature Death

Fig. 1. Decision tree for outcomes that result from withholding or prescribing INH. (A) In this model TB is treated with a regimen containing INH. (8) Hepatotoxicity from treatment of TB. (C) Hepatotoxicity from preventive therapy. (0) This branch includes individuals who obtain a partial reduction in TB risk associated with early discontinuance of preventive therapy.

ty. For the three variables INH efficacy, INH efficacy after discontinuance of therapy, and TB fatality rate, specific estimates by age, gender, and ethnicity were not available in the existing literature. In these instances, middleof-the-road estimates were used. General estimates. INH efficacy in TB prevention. Rates of reduction in tuberculosis risk resulting from INH prophylaxis range from 40 to 930/0 (5, 7, 9, 15-18). Recently, Tsevat and colleagues (8) used a 50% efficacy rate in their analyses. Snider (9) suggested an efficacy rate of 68% to characterize individuals who took 80% or more of prescribed medication for at least 10 months (16). This estimate, which takes into account both efficacy and compliance, was used in the present analyses. INH efficacy in TB prevention after early discontinuance. Although INH toxicity and therefore discontinuation tend to occur in the early phases of prophylaxis, a small benefit is obtained from a partial course of therapy. This partial benefit has been estimated as a 12 to 21% reduction in TB risk (18, 19). The present analyses assumed a midrange 16% reduction, as per the Rose (5) and Tsevat groups (8). TB case fatality rate. Although TB mortality may be reduced in individuals who re-

ceive appropriate and timely treatment for their disease (20), some patients remain untreated or are treated too late to achieve a satisfactory response (9). Accordingly, a range in TB mortality of 7.8 to 9.7% was suggested (9) to reflect these varying contingencies. The present analyses used a midrange estimate of 8.75%. Subgroup-specific estimates. Lifetime risk of TB. Estimates of lifetime TB risk were de-

rived separately by ethnicity and gender from Comstock and Edwards's (6) estimates of lifetime risks for tuberculin reactors aged 25,35, and 55 yr. To facilitate comparison with more recent decision analyses, weinterpolated these estimates to derive estimates for 20- and 50yr-olds (see table 1). In contrast to the estimates for low-risk reactors, the likelihood that high-risk reactors will develop active TB has been estimated as 5.46 to 10% (6, 11, 12,21, 22). To provide maximum contrast between the low-risk and high-risk scenarios analyzed in this study, we adopted an upper estimate of lifetime TB risk of 10% in the high-risk decisions. INH toxicity and mortality. Kopanoff and coworkers (14)reported data on probable and possible cases of INH hepatitis and associated fatalities based on surveillance of 13,838 persons in 21health departments by age, gender, and ethnic groups. We used these data to calculate estimates of subgroup-specific hepatitis rates for the 12 population groups examined (see table 1). Kopanoff's group (14) reported a total of eight INH-related fatalities among the 174patients in their series with either possible or probable INH-related hepatitis. The ethnicity and gender of the patients who died were used to calculate ethnicity-and gender-specific mortality rates (table 1).

Utilities Life expectancies. For each final outcome that does not reflect a premature death from TB or INH (23), the 1985average life expectancy for each ethnicity, gender, and age group was assigned. Previous decision analyses have assumed that fatal INH hepatitis during preventive therapy occurs 12 to 18 wk after the initiation of therapy (5, 8). An intermediate value of 15 wk was used in the present analyses. For the high-risk individual whose lifetime risk of developing TB is 100/0 following exposure, Enarson (12)suggests that, based on two empirical sources (21,22), the yearly risk is 1.5% in the first year following exposure, 1.0% in the second year, 2.5% in Years 3 through 5, and 5.0% over the balance of a

TABLE 1 SUBGROUP-SPECIFIC PARAMETER ESTIMATES SUbgroup by Ethnicity, Gender, and Age BM20 BM35 BM50 BF20 BF35 BF50 WM20 WM35 WM50 WF20 WF35 WF50

Lifetime Risk ofTB

TB LE'

INH Toxieity

INH Mortality

0.049 0.033 0.024 0.048 0.037 0.026 0.048 0.036 0.025 0.039 0.031 0.025

30.9 20.4 13.7 32.5 22.7 15.2 34.0 23.2 14.5 39.6 28.4 13.0

0.002 0.009 0.019 0.002 0.Q16 0.032 0.002 0.015 0.036 0.002 0.017 0.033

0.059 0.059 0.059 0.125 0.125

• Additional yearsof life for those who die of TB. Note: Calculations based on references 6. 14, 23, and 24.

0~125 0.029 0.029 0.029 0.026 0.026 0.026

1359

ISONIAZID PREVENTIVE THERAPY FOR TUBERCULOSIS

lifetime. These statistics yield weights that were used in calculating the average lifespan of high-risk patients who eventually die of TB. On the other hand, the pool of low-risk reactors contains a mixture 0 f individuals with various actualrisks of developing TBdepending on the time at whichinitial infection occurred. To reflect this, a mean of high and lowlifeexpectancy estimates wasused as the bestestimatefor low-risk reactorswho eventually die of TB (24). These"best" estimates are shown in table 1. Results

Results in table 2 show that for low-risk reactors, INH preventive therapy is preferred for a1l20-yr-olds, all 35-yr-olds except black women, and no 50-yr-olds. The projected advantages gained by prescribing INH range from approximately 3 to 19 days of life. When the decision does not favor INH, that is, for 35-yrold black women and all 50-yr-olds, withholding it provides advantages ranging from approximately 2 to 33 days. For

high-risk reactors, INH is favored for all groups except 50-yr-old black women. The advantages gained by prescribing INH for the high-risk group range from 1 to 44 days. Withholding INH for the 50-yr-old black female group provides an advantage of about 12 days. A supplementary view of the risks and benefits of INH therapy for low-risk reactors is obtained by considering numbers of deaths and illness events rather than life expectancies. (See table 3.) When total deaths are considered (deaths from INH plus deaths from TB), INH preventive therapy is favored for all groups except 50-yr-old black women. Subgroup-specific thresholds (points at which the decision would change) for each of the parameter estimates used in the analyses for low-risk reactors were calculated through one-way sensitivity procedures. Threshold values for most of the parameters were out of the range of estimates used by other researchers in this

TABLE 2 PROJECTED EFFECTS OF PRESCRIBING INH EXPRESSED AS DAYS OF LIFE GAINED OR LOST PER INDIVIDUAL FOR LOW· AND HIGH·RISK PATIENTS

Age

Black Black White White Black Black White White Black Black White White

20

35

50

Days of Life Gained, High Risk

Days of Life Lost, Low Risk

Days of Life Gained, Low Risk

Subgroup by Ethnicity and Gender

34 44 41 43 23 10 29 29 1

15 18 19 16 3

male female male female male female male female male female male female

Days of Life Lost, High Risk

-15 6 4

-4 -33

-12

-4

14 21

-2

TABLE 3 OUTCOMES OF DECISION ANALYSIS EXPRESSED AS DEATHS AND ILLNESS EVENTS·

Subgroup by Ethnicity, Gender, and Age BM20 BM35 BM50 BF20 BF35 BF50 WM20 WM35 WM50 WF20 WF35 WF50

TB Cases

TB Deaths

INH Deaths

Total Premature Deaths

INH

NolNH

INH

No INH

INH

NolNHt

INH

No INH

1,563 1,058 782 1,553 1,192 863 1,544 1,168 848 1,242 1,018 838

4,870 3,260 2,380 4,840 3,650 2,590 4,810 3,570 2,510 3,870 3,100 2,490

137 93 68 136 104 76 135 102 74 109 89 73

426 285 208 423 319 227 421 312 220 339 271 218

13 56 111 27 206 406 6 43 106 6 44 86

1 2 3 1 7 10 0 2 3 0 1 2

150 149 179 163 310 482 141 145 180 115 133 159

427 287 211 424 326 237 421 314 223 339 272 220

• Per 100,000 individuals. are individuals who did not receive INH preventive therapy but who developed T8 and were treated with a regimen containing INH.

t These

area and therefore indicated that the decision would not change as a result of arbitrarily choosing one estimate over another. However, the TB mortality thresholds required to switch decisions favoring INH therapy to withholding INH therapy fell between 4 and 6010 for some subgroups. This means that the preferred course of action might be no INH for a patient who was very likely to obtain timely and appropriate medical care for active TB and who therefore had a risk of TB mortality lower than the general estimate of 8.75%. Discussion

The major objective ofthis investigation was not to resolve the INH "numbers controversy" in the absence of new data but, rather, to show that it may be essential to consider ethnicity and gender as well as age when making the decision to prescribe or withhold preventive therapy for TB. In contrast to the finding of Tsevatand colleagues (8) that INH should be withheld for low-risk reactors of all ages and, by implication, for high-risk reactors over age 20, our analyses by subgroup suggest that (1) for low-risk reactors, INH is preferred for all 20-yr-olds, all 35-yrolds except black women, and for no 50yr-olds; (2) for traditionally defined highrisk reactors, INH is preferred for all groups except 50-yr-old black women. The magnitude of the differences found in both the Tsevat group's analyses and ours is not large. Their decision for low-risk reactors favored withholding INH on the basis of 4 to 17 days of life expectancy. For the subgroups of lowrisk reactors for whom our decision favored prescribing INH, the benefit was 3 to 19days oflife. For subgroups oflowrisk reactors for whom our decision favored withholding INH, the benefit was 2 to 33 days. The magnitude of the differences found for high-risk reactors in our analyses was remarkably similar to the differences found for low-risk reactors: withholding INH for 50-yr-old black women is associated with a benefit of 14 days, and prescribing INH for all other high-risk reactors results in a projected benefit of 1 to 44 days. Estimates used in this investigation do not reflect the magnitude of lifetime TB risk found to characterize individuals who are infected with both TB and human immunodeficiency virus (HIV) (25, 26) or the benefits of prescribing INH in such a population (27). In addition, the benefits of prescribing INH for black

1360

women may outweigh the risks if these patients are monitored closely enough to detect the earliest signs of adverse reactions to preventive therapy. The present investigation used life expectancies as well as total deaths and illness events to arrive at decisions for subgroups of low-riskpatients. Although impact on life expectancies may provide the better, more conservative guideline for making decisions for individual patients, the findings regarding total deaths and illnesses may be useful in supporting and strengthening those decisions. In particular, the task of deciding what constitutes a "toss-up" remains a problem in decision analysis (28). Some investigators consider a benefit of only a few days to a few weeks of expected life too close to call one way or the other (8). From this perspective, the decision arrived at in this investigation to prescribe INH for all lowrisk 20-yr-olds and all low-risk 35-yr-olds except black women could be regarded as equivocal. However, if the small life expectancy benefits to the individual are considered in conjunction with the associated manyfold reductions in number of TB cases and in total deaths, the decision to prescribe INH is strengthened considerably. Furthermore, although the magnitude of life expectancy benefits found for the low-risk reactors is not large, the benefits found for high-risk reactors are comparably modest. Yetnational policy unequivocally mandates INH preventive therapy for high-risk individuals (4). The strength ofthis national policy initiative rests on concern about total cases of TB and the potential for infectious spread of the disease. It appears particularly appropriate for those charged with developing national TB policy to evaluate most carefully the impact on individuals of implementing therapeutic decisions that appear beneficial in the aggregate but may be detrimental to the individual patient. Twomajor areas of uncertainty remain problematic. First, it is unclear why black women appear to be at greater risk for INH mortality. Since there is no theoretical rationale for this phenomenon, one is led to question the soundness of the data on which it rests. Second, in the absence of new studies, the robustness of

JORDAN, LEWIT, AND REICHMAN

the data on which subgroup-specific differences must rest is open to question since it is based on few observations for each of the 12 specific subgroups. Given the widespread use of INH preventive therapy, it is surprising that better information on the rates of INH toxicity for specific subgroups is not more readily available. Since concern about this risk factor appears to be the main reason for the declining use of preventive therapy with INH (3), clarification of the true magnitude of the specific risks through properly stratified, large-sample studies would appear crucial to success of the national initiative to eliminate TB. Acknowledgment The writers thank Carroll M. Leevy, M.D., Professor and Chairman of the Department of Medicine and Scientific Director of the Sammy Davis Jr. National Liver Institute, for introducing us to the importance and value of decision analysis. We also acknowledge helpful suggestions in early stages of this investigation from George Comstock, M.D., Dr. P.H., and Donald Enarson, M.D. Finally, we appreciate the assistance of Richard L. Montgomery, D.D.S., M.P.H., and Patty MeGowan in the preparation of this document.

References 1. Centers for DiseaseControl. Tuberculosis statistics: states and cities. HHS Pub. No. (CDC) 878249, 1986. 2. Centers for Disease Control. A strategic plan for the elimination of tuberculosis in the United States. MMWR 1989; 38(suppl to 5-3:1-25). 3. Hansen C, Reichman LB. Tuberculin testing and preventive therapy. Semin Respir Infect 1989; 4:182-8. 4. Centers for DiseaseControl Advisory Committee for the Elimination of Tuberculosis. Screening for tuberculosis and tuberculosis infection in high risk populations; and the use of preventive therapy for tuberculosis infection in the United States. MMWR 1990; 39(RR8:1-12). 5. Rose DN, Schechter CB. Silver AL. The age threshold for isoniazid chemoprophylaxis: a decision analysis for low-risktuberculin reactors. JAMA 1986; 256:2709-13. 6. Comstock GW, Edwards PQ. The competing risks of tuberculosis and hepatitis for adult tuberculin reactors. Am Rev Respir Dis 1975; Ill: 573-7. 7. TaylorWC, Aronson MD, De1bancoTL. Should young adults with a positive tuberculin test take isoniazid? Ann Intern Med 1981; 94:808-13. 8. Tsevat J, Taylor WC, Wong JB, Pauker SG. Isoniazid for the tuberculin reactor: take it or leave it. Am Rev Respir Dis 1988; 137:215-20. 9. Snider DE. Decision analysis for isoniazid

preventivetherapy: take it or leaveit? Am RevRespir Dis 1988; 137:2-3. 10. Moulding IS, Redeker AG, Kanel GC. Twenty isoniazid-associated deaths in one state. Am Rev Respir Dis 1989; 140:700-5. 11. Rose DN, Silver AL, Schechter CB. Preventive treatment for tuberculosis in elderly persons. Ann Intern Med 1988; 106(6):908. 12. Enarson DA. Isoniazid for the tuberculin reactor: take it or leave it? Am Rev Respir Dis 1988; 137:1513-4. 13. Rieder HL, Cauthen GM, Kelly GD, Bloch AB, Snider DE. Tuberculosis in the United States. JAMA 1989; 262(3):385-9. 14. Kopanoff DE, Snider DE, Caras GJ.lsoniazidrelated hepatitis: a U.S. Public Health Service cooperative surveillance study. Am Rev Respir Dis 1978; 117:991-1001. 15. Koplan JP, Farer LS. Choice of preventive treatment for isoniazid-resistant tuberculosis infection: use of decision analysis and the Delphi technique. JAMA 1980; 244:2736-40. 16. Comstock GW, Baum C, Snider DE. Isoniazid prophylaxis among Alaskan Eskimos: a final report on the Bethel isoniazid studies. Am Rev Respir Dis 1979; 119:827-30. 17. Cooper JK. Decision analysis for tuberculosis preventive treatment in nursing homes. J Am Geriatr Soc 1986; 34:814-7. 18. Ferebee SH. Controlled chemoprophylaxis trials in tuberculosis: a general review.Adv Tuberc Res 1970; 17:28-105. 19. International Union Against Tuberculosis. Efficacy of various durations of isoniazid preventive therapy of tuberculosis: five years of follow-up in the IDAT trial. Bull WHO 1982; 60:555-64. 20. Davis CE, Carpenter JL, McAllister CK, et al. Tuberculosis: cause of death in antibiotic era. Chest 1985; 88(5):726-9. 21. Enarson DA. Personal communication, 4/11/88. 22. Burrill D, Enarson DA, Allen EA, Grzybowski S. TB in female nurses in British Columbia: implications for control programs. Can Med Assoc J 1985; 132:137-40. 23. U.S.Bureau of the Census, Statistical Abstract of the United States: 1988(108th ed.). Washington DC, 1987. 24. U.S. Department of Health and Human Services, 1985-1986: Tuberculosis in the United States. HHS Publication No. (CDC) 88-8322. Centers for Disease Control, 1988. 25. Snider NE, Hopewell PL, Mills J, Reichman CB. Mycobacteria and the acquired immunodeficiencysyndrome (joint position paper of the American Thoracic Society and the Centers for Disease Control). Am Rev Respir Dis 1987; 137:492-6. 26. SelwynPA, Hartel D, Lewis V, et al. A prospective study of the risk of tuberculosis among intravenous drug users with human immunodeficiency virus infection. N Engl J Med 1989; 320:545-50. 27. Block A, Rieder H, K¢lly G, Cauthen GM, Hayden C, Snider D. The epidemiology of tuberculosis in the United States. Clin Chest Med 1989; 10:297-313. 28. KassirerJ, Alan J, Moskowitz J, Lau J, Pauker S. Decision analysis: a progress report. Ann Intern Med 1987; 106:275-91.