American Journal of Respiratory and Critical Care Medicine

Isoniazid is an efficacious treatment for latent tuberculosis. Concerns remain, however, regarding hepatotoxicity associated with this medication. In addition, adherence may be suboptimal because at least 6 months of treatment is required. We extracted information from our latent tuberculosis treatment database to determine adverse effects and treatment completion rates associated with the use of isoniazid at a county tuberculosis clinic. Outcomes were available for 3,788 patients started on isoniazid between 1999 and 2002. Six hundred seventy-two patients (18%) experienced one or more adverse effects, including 10 (0.3%) determined to have isoniazid-associated liver injury. No hospitalizations or deaths occurred in patients experiencing an adverse effect. A higher incidence of adverse effects was associated with increasing age. Sixty-four percent of patients completed at least 6 months of isoniazid. Higher completion rates were associated with younger age, Hispanic ethnicity, and non-U.S. country of birth. Lower completion rates were associated with being homelessness, using excess alcohol, and having experienced an adverse effect. In summary, we conclude that in our clinic population isoniazid is a safe therapy for latent tuberculosis, but its effectiveness is limited by modest completion rates.

Multiple, large, controlled trials have demonstrated the efficacy of isoniazid for the treatment of latent tuberculosis (TB) infection (LTBI) (1, 2). Two factors, however, potentially limit isoniazid therapy's effectiveness as a public health intervention. The first is toxicity, especially hepatotoxicity. The second is the need for adherence to a prolonged course of therapy: a minimum of 6 months, but preferably 9 months (3).

Initial studies with isoniazid did not suggest that it caused hepatotoxicity (4). Later, however, deaths secondary to isoniazid-induced liver injury were reported (5). Subsequent studies revealed that isoniazid-associated liver injury occurred in approximately 1% of patients taking the medication. The largest such study conducted by the United States Public Health Service found 92 cases of hepatotoxicity, including eight deaths, among nearly 14,000 patients taking isoniazid (6). A meta-analysis of six studies performed during the 1960s through the 1980s showed a cumulative rate of isoniazid-associated hepatotoxicity of 0.6% (7). Recently, however, the Seattle public health clinic reported only 11 cases of hepatotoxicity among over 11,000 patients treated with isoniazid over a 7-year period (8).

A second obstacle to successful LTBI treatment with isoniazid is patient adherence. Completion rates for treatment with 6 months of isoniazid under operational conditions vary but are generally in the range of 50–60% (810). It was believed that a short-course therapy consisting of rifampin and pyrazinamide for 2 months might result in improved completion (11, 12). This regimen, however, has been associated with significant hepatotoxicity, including multiple instances of death and hospitalization (10, 1315). In addition, two recent studies found completion rates for short-course (2 to 4 months) multidrug regimens to be no different from those for longer course (6 to 12 months) isoniazid regimens (10, 16).

To evaluate the safety and effectiveness of our LTBI treatment program, we reviewed our recent experience with isoniazid treatment. In particular, we chose to examine rates of adverse effects of treatment, including hepatotoxicity, and to determine completion rates and factors that may affect completion. The ultimate goal is to use this analysis to improve local TB control through improving LTBI treatment. Some of the results of this study have been previously reported in the form of an abstract (17).

The County of San Diego's TB Control Program has maintained a computer database since July 1999 for all patients starting LTBI treatment. To evaluate the safety and effectiveness of the LTBI treatment program, an analysis of this database was performed. Patients were included in this analysis if they were started on isoniazid between July 1, 1999, and November 30, 2002, and they were not continuing therapy at the time the analysis was performed. Medical record review was used to obtain missing information critical to the analysis.

At the San Diego TB Clinic, patients were treated for LTBI according to the 1994 American Thoracic Society/Centers for Disease Control and Prevention treatment guidelines until July 2000 (18). At that time, the clinic instituted changes consistent with the updated guidelines published in May 2000 (9). One significant local modification of these guidelines was that patients greater than 49 years old were not offered LTBI treatment unless they had additional risks for reactivation (e.g., human immunodeficiency virus [HIV] infection, diabetes).

Patients were monitored for hepatotoxicity and other adverse effects according to American Thoracic Society/Centers for Disease Control and Prevention guidelines. Monthly clinical monitoring, performed by a nurse, consisted of a symptom review using a list of standard questions. To receive a prescription refill, patients had to complete the face-to-face interview with the nurse. Before July 2000, all patients older than 34 years old or with other risks (HIV infection, history of excess alcohol use, history of hepatitis or chronic liver disease, pregnancy) for isoniazid-associated hepatotoxicity had baseline and monthly transaminase levels measured. After July 2000, only patients with risk factors (HIV infection, history of excess alcohol use, history of hepatitis or chronic liver disease, pregnancy) for hepatotoxicity as delineated in the American Thoracic Society/Centers for Disease Control and Prevention guidelines had transaminase levels measured routinely, regardless of age.

An adverse effect was considered to be isoniazid related if (1) the adverse effect had been previously reported as attributable to isoniazid and (2) the signs and symptoms occurred after starting isoniazid and (3) no alternative cause for the adverse effect was found. For isoniazid-associated liver injury, one of two additional criteria had to be met: (1) elevation of transaminases greater than three times normal in the presence of symptoms compatible with liver injury or (2) greater than five times normal in the absence of symptoms. Completion of therapy was defined as completion of at least 6 months of isoniazid (even if 9 months of therapy was prescribed), as measured by monthly patient interview and prescription refill.

Data items extracted included sex, age, race, ethnicity, country of birth, self-reported excess alcohol use, self-reported intravenous drug use, HIV test results, reason for starting treatment, completion of treatment, reason for not completing treatment, and adverse effects associated with treatment. Consumption of more than six beers, six glasses of wine, or six mixed drinks per week defined excess alcohol use. Race/ethnicity was categorized as white, Hispanic; white non-Hispanic; Black, non-Hispanic; Asian-Pacific Islander; Native American; other; or unknown.

Rates of adverse effects and completion were calculated and compared in a univariate analysis using the chi-square and Fisher's exact tests. Multivariate logistic regression was performed to adjust for the effects of multiple explanatory variables on the two outcomes of interest: presence of at least one adverse effect and completion of treatment. Explanatory variables were included in the model if they were associated with the outcome in the univariate analysis (p < 0.05) or if an association between the explanatory variable and outcome had been reported previously. For explanatory variables with more than two categories, the category with the lowest rate of adverse effects or the lowest completion rate was chosen as the reference category. Goodness of fit was adequate according to the Hosmer–Lemeshow test (p > 0.05 for both outcome models). There was no evidence of collinearity as measured by variance inflation factor analysis (variance inflation factor of less than five for all independent variables). A separate univariate analysis was performed to examine the association between possible explanatory variables and the occurrence of hepatotoxicity. Because the number of hepatotoxic events was so small, a multivariate analysis for this outcome was not performed. Associations between independent and dependent variables were considered significant at a level of p < 0.05. Statistical analyses were performed using Statistical Package for the Social Sciences (version 11.0; SPSS, Inc., Chicago, IL) software.

This project was reviewed by the Office of the Associate Director of Science for the National Center for Sexually Transmitted Diseases, HIV, and TB Prevention and considered not to be research involving human subjects. It was, therefore, not submitted to an institutional review board. To protect patient confidentiality, the database is password protected. In addition, after data were exported to SPSS for analysis, all specific identifiers (e.g., name, medical record number) were removed.

Outcomes were available for 3,788 patients. Demographic characteristics of these individuals are shown in Table 1

TABLE 1. Demographic characteristics


Number (%)
Male1,552 (41)
Female2,229 (58)
Unknown7 (0.2)
0 to 14 years1,277 (34)
15 to 34 years1,939 (51)
35 to 49 years426 (11)
50 to 64 years95 (2.5)
65+ years50 (1.3)
White, Hispanic3,025 (80)
White, non-Hispanic170 (4.4)
Black, non-Hispanic117 (3.1)
Asian-Pacific Islander335 (9)
Native American4 (0.1)
Other13 (0.3)
Unknown124 (3.3)
Country of birth
United States782 (21)
Mexico2,101 (56)
Philippines178 (4.7)
Vietnam62 (1.6)
Other258 (7)
407 (11)
. Of 158 patients tested for HIV, 2 (1.3%) were positive. A total of 15 patients (0.4%) reported using excess alcohol, 9 (0.2%) reported using intravenous drugs, 47 (1.2%) were homeless, and 74 (2.0%) reported having been in a correctional facility. With regard to reasons for starting LTBI treatment, 3,049 patients (81%) were tuberculin skin test reactors, 394 (10%) were recent tuberculin skin test converters, 257 (7%) were close contacts to an active TB patient, 76 (2.0%) had old, healed TB (American Thoracic Society Class IV), and 12 (0.3%) began treatment for unknown reasons.

The number of patients experiencing adverse effects, the number of events per 1,000 patients starting and completing therapy, and the number of patients in whom the adverse effect was the primary reason for stopping treatment are shown in Table 2

TABLE 2. Adverse effects

Number in Whom Adverse Effect
 Was Primary Reason for Stopping Treatment
Adverse Effect
Number (%)
Events per
 1,000 Patients
 Completing Treatment
(% of total persons who
 experienced adverse effect)
Hepatotoxicity10 (0.3)45 (50)
Rash130 (3.4)546 (4.6)
Itching, no rash117 (3.1)482 (1.7)
Nausea or vomiting, not hepatotoxicity131 (3.5)5412 (9)
Abdominal pain, not hepatotoxicity176 (4.6)739 (5)
Headache338 (9)1408 (2.4)
Paresthesias177 (4.7)734 (2.2)
17 (0.4)
6 (35)
. Six hundred seventy-two patients (18%) experienced one or more adverse effects. The multivariate analysis revealed that the occurrence of at least one adverse effect was associated with female sex, increasing age, homelessness, and having spent time in a correctional facility (Table 3)

TABLE 3. Multivariate analysis of factors associated with occurrence of at least one adverse effect


Number with at Least
 One Adverse Effect (%)

OR (95% CI)

p Value
Male217 (14)Reference
Female453 (20)1.6 (1.4–2.0)< 0.01
0 to 14 years177 (14)Reference
15 to 34 years360 (19)1.3 (1.0–1.6) 0.04
35 to 49 years102 (24)1.8 (1.3–2.5)< 0.01
50 to 64 years25 (26)2.2 (1.3–3.8)< 0.01
65+ years8 (16)1.5 (0.6–3.2) 0.38
Asian-Pacific Islander57 (17)Reference
White, non-Hispanic42 (24)1.6 (0.9–2.6) 0.07
White, Hispanic530 (18)1.3 (0.9–1.8) 0.19
Black, non-Hispanic23 (20)1.2 (0.7–2.2) 0.49
Country of birth
United States138 (18)Reference
Other486 (19)1.1 (0.8–1.4) 0.58
Excess alcohol
No670 (18)Reference
Yes2 (13)0.6 (0.1–2.8) 0.52
Intravenous drug use
No670 (18)Reference
Yes2 (22)1.3 (0.3–7.3) 0.73
No654 (18)Reference
Yes18 (38)2.2 (1.2–4.2) 0.02
Correctional facility
No645 (17)Reference
27 (37)
2.6 (1.5–4.5)
< 0.01

Definition of abbreviations: CI = confidence interval; OR = odds ratio.


The overall incidence of hepatotoxicity was 3 per 1,000 (0.3%). There were no deaths or hospitalizations among patients with liver injury. The incidence of hepatotoxicity did not change significantly after monitoring practices, and the age criteria for initiating treatment were revised based on the new national LTBI treatment guidelines. The incidence before July 1, 2000 (n = 1,467), was 2 per 1,000; the incidence after July 1, 2000 (n = 2,321), was 3 per 1,000 (p = 0.74). The frequency of liver injury tended to increase with age (0 to 14 years, 0%; 15 to 34 years, 0.3%; 35 to 49 years, 0.9%; 50 to 64 years, 0%; 65+ years, 0%; chi-square for trend p = 0.04). The youngest patient with liver injury was 20, and there were no cases of liver injury in patients who were older than 49. The occurrence of hepatotoxicity was also associated with self-reported intravenous drug use (11%, p = 0.02). There was no association between hepatotoxicity and sex, race/ethnicity, or self-reported excess alcohol use. Patients with hepatotoxicity were more likely to have a rash than those without liver injury (20% vs. 3.4%, respectively, p = 0.04).

Outcomes of LTBI treatment are shown in Table 4

TABLE 4. Outcomes of latent tuberculosis infection treatment


Number (%)
Treatment completed2,414 (64)
Stopped early: adverse effect52 (1.4)
Stopped early: provider decision41 (1.1)
Stopped early: patient decision371 (10)
Lost732 (19)
Moved148 (3.9)
Stopped early: became pregnant24 (0.6)
6 (0.2)
. Of the 1,371 patients not completing treatment, 1,103 (80%) were either lost or decided not to continue treatment. Ninety-three (7%) stopped treatment primarily because of an adverse effect or a medical provider's decision.

Higher completion rates were associated with female sex, younger age groups (0 to 14 and 15 to 34), white/Hispanic race/ethnicity, and non-U.S. country of birth (Table 5)

TABLE 5. Multivariate analysis of factors associated with completion


Number Completing (%)

OR (95% CI)

p Value
Male961 (62)Reference
Female1,450 (65)1.2 (1.0–1.4) 0.03
0 to 14 years943 (74)4.1 (2.2–7.8)< 0.01
15 to 34 years1,173 (61)2.1 (1.1–3.9) 0.02
35 to 49 years223 (52)1.8 (0.9–3.4) 0.07
50 to 64 years54 (57)1.9 (0.9–4.1) 0.07
65+ years21 (42)Reference
Asian-Pacific Islander202 (60)1.4 (0.9–2.3) 0.12
White, non-Hispanic90 (53)1.5 (0.9–2.5) 0.10
White, Hispanic1,988 (66)1.5 (1.0–2.3) 0.04
Black, non-Hispanic 49 (42)Reference
Country of birth
United States471 (60)Reference
Other1,679 (65)1.4 (1.1–1.7)< 0.01
Excess alcohol
No2,412 (64)Reference
Yes2 (13)0.1 (0.0–0.6)< 0.01
Intravenous drug use
No2,412 (64)Reference
Yes 2 (22)0.5 (0.1–2.9) 0.47
No2,403 (64)Reference
Yes 11 (23)0.2 (0.1–0.5)< 0.01
Correctional facility
No2,389 (64)Reference
Yes 25 (34)0.6 (0.4–1.1) 0.09
No2,411 (64)Reference
Yes 3 (30)0.4 (0.1–1.8) 0.24
Any other adverse effect
No2,027 (65)Reference
387 (59)
0.8 (0.7–0.9)

Definition of abbreviations: CI = confidence interval; OR = odds ratio.

. Lower completion rates were associated with self-reported excess alcohol use, homelessness, and occurrence of at least one adverse effect other than hepatotoxicity.

Although there is abundant evidence demonstrating isoniazid's efficacy for the treatment of LTBI, concerns about potential toxicity have remained. The recent experience reported from Seattle suggests that these concerns can be addressed with a system for educating patients and monthly monitoring (8). With appropriate patient selection and clinical monitoring, only 0.1% of patients starting isoniazid developed hepatotoxicity. Of note, there were no deaths and only one hospitalization in the 11 patients with liver injury. The authors felt that the low incidence of hepatotoxicity was due to primarily two factors. First, most patients treated with isoniazid were less than 35 years old. Prior investigations determined that increasing age was a risk factor for hepatotoxicity (5, 6). Second, the public health clinic relied on clinical evaluation, which included systematic patient education, rather than routine biochemical testing as the method for monitoring for liver injury. Thus, patients were taught to recognize signs and symptoms of liver injury and stop the medication before serious consequences occurred. It is believed that without routine serial blood tests, transient, clinically insignificant serum transaminase elevations were not detected.

Our recent experience, as described in this report, is similar. We found an incidence of liver injury of 0.3% among patients starting isoniazid at our clinic. There were no deaths or hospitalizations among our patients with liver injury. The slight difference in incidence of hepatotoxicity between our clinic and the Seattle public health clinic can be attributed largely to a difference in criteria used for defining isoniazid-associated hepatotoxicity. Had we used the Seattle investigators' definition, which excluded asymptomatic transaminase elevation, only six patients would have met the criteria for hepatotoxicity, resulting in an incidence of less than 0.2%. We included patients with asymptomatic transaminase elevation, however, because we felt that this was compatible with the current American Thoracic Society/Centers for Disease Control treatment guidelines that state, “Some experts recommend that isoniazid be withheld if a patient's transaminase level exceeds 3 times the upper limit of normal if associated with symptoms and five times the upper limit of normal if the patient is asymptomatic” (9).

As in other studies, we found a trend toward an increasing incidence of liver injury with increasing age (5, 6, 8). Interestingly, there were no cases of hepatotoxicity in patients greater than 49 years old. This may have been due to selection bias, as patients in this age group with LTBI were not routinely started on isoniazid. Less than 4% of our patients started on isoniazid were greater than 49 years old. Medical providers in our clinic are likely to have been more selective in prescribing isoniazid for these patients, taking particular care to exclude any with additional risk factors for developing hepatotoxicity (e.g., history of excess alcohol use, history of viral hepatitis).

Other adverse effects occurred more frequently than hepatoxicity. Most of these were minor in terms of the likelihood of causing long-term morbidity. The more common side effects, those occurring in more than 3% of patients starting therapy, were rarely the primary cause of discontinuing treatment. For example, of 338 patients with headache and 130 patients with rash, only 8 (2.4%) and 6 (4.6%) stopped therapy because of the headaches or the rash, respectively. Conversely, hepatotoxicity was rare (occurred in 0.3% of persons starting treatment), but 50% of patients with this adverse effect discontinued treatment primarily as a result of the liver injury. Even if not the main cause for ending therapy, the presence of at least one adverse effect was associated with a lower completion rate. It is possible that minor adverse effects, which did not pose a danger in the clinician's view, nevertheless resulted in discomfort and contributed to a patient's decision not to continue treatment.

Because we report our program experience with isoniazid, rather than a clinical trial with a placebo-controlled arm, there are some potential limitations in attributing adverse effects to isoniazid therapy. This is more of an issue with some of the minor, commonly reported adverse effects such as headache. Although these symptoms may have occurred in association with isoniazid treatment, they may not have been caused by isoniazid. In addition, patients may have become more sensitized to possible symptoms because they were starting a new medication.

Another limitation in determining actual rates of adverse effects is the potential impact of low rates of adherence and treatment completion. For both the Seattle and San Diego programs, the low completion rates are likely to have led to lower estimates of the incidence of adverse effects, as patients who do not take a full course of treatment are less likely to suffer adverse effects (8).

Perhaps the greatest obstacle to successful use of isoniazid for LTBI treatment is the relatively low completion rate for patients starting the drug. In our clinic, 64% of patients starting isoniazid completed therapy. This completion rate is based on a 6-month regimen. Had 9 months of therapy been required, as recommended in the new guidelines, the rate would have been lower. The completion rate is comparable to that reported in previous articles and essentially the same as that reported by the Seattle group (64%) (810). In our population, we found higher completions rates among Hispanics. The reasons for this are not clear. One possible reason may be attention to cultural sensitivity. Our clinic staff includes many bilingual, bicultural employees, particularly Hispanic-American employees. This may foster a greater sense of trust among our patients. Conversely, several factors were associated with poor completion rates, including homelessness and substance abuse. It is not surprising that individuals in these groups, which largely overlap, have difficulty adhering to a prolonged public health intervention.

Possible limitations with regard to completion rates should also be noted. Observed completion rates may have been lower if more rigorous adherence monitoring had been used. Reported completion rates were based on prescription refills. Adherence monitoring methods, such as directly observed therapy and urine-metabolite testing, were not used. It is possible that some patients refilled medication and did not actually take all of it. Some risk factors for low completion rates, such as substance abuse, were based on patient self-reporting. Underreporting of these factors may have occurred.

In summary, we conclude that in our clinic isoniazid is a safe therapy for LTBI, with a low incidence of hepatotoxicity. Based on the findings described in our report, we plan to continue treating LTBI patients who are 35–49 years old, even in the absence of additional risk factors for progression to active TB. We will also continue to use clinical evaluation as the primary monitoring method for most patients.

Isoniazid's effectiveness in our population was limited, however, by modest completion rates. More intensive patient education with regard to the benefits of LTBI treatment, more active follow-up of patients who do not return for prescription refills, and the use of incentives and enablers will be considered to improve completion rates in the future.

P.A.L. has no declared conflict of interest. K.S.M. has no declared conflict of interest.

1. Ferebee SH. Controlled chemoprophylaxis trials in tuberculosis: a general review. Adv Tuberc Res 1970;17:28–106.
2. Efficacy of various durations of isoniazid preventive therapy for tuberculosis: five years of follow-up in the IUATL trial: International Union Against Tuberculosis Committee on Prophylaxis. Bull World Health Organ 1982;60:555–564.
3. Comstock GW. How much isoniazid is needed for the prevention of tuberculosis among immunocompetent adults? Int J Tuberc Lung Dis 1999;3:847–850.
4. Ferebee SH, Mount FW. Tuberculosis morbidity in a controlled trial of the prophylactic use of isoniazid among household contacts. Am Rev Respir Dis 1962;85:490–521.
5. Garibaldi RA, Drusin RE, Ferebee SH, Gregg MB. Isoniazid-associated hepatotoxicity: report of an outbreak. Am Rev Respir Dis 1972;106:357–365.
6. Kopanoff DE, Snider DE, Caras GJ. Isoniazid-related hepatotoxicity: a US Public Health Service cooperative surveillance study. Am Rev Respir Dis 1978;117:991–1001.
7. Steele MA, Burk RF, DesPrez RM. Toxic hepatotoxicity with isoniazid and rifampin: a meta-analysis. Chest 1991;99:465–471.
8. Nolan CM, Goldberg SV, Buskin SE. Hepatotoxicity associated with isoniazid preventive therapy: a seven year survey from a public health clinic. JAMA 1999;281:1014–1018.
9. Targeted tuberculin testing and treatment of latent tuberculosis infection: American Thoracic Society and Centers for Disease Control and Prevention. Am J Respir Crit Care Med 2000;161:S221–S247.
10. Jasmer RM, Saukonnen JJ, Blumberg HM, Daley CL, Bernardo J, Vittinghoff EV, King MD, Kawamura LM, Hopewell PC. Short-course rifampin and pyrazinamide compared with isoniazid for latent tuberculosis infection: a multicenter clinical trial. Ann Intern Med 2002;137:640–647.
11. Halsey NA, Coberly JS, Desormeaux J, Losikoff P, Atkinson J, Moulton LH, Contave M, Johnson M, Davis H, Geiter L, et al. Randomised trial of isoniazid versus rifampicin and pyrazinamide for prevention of tuberculosis in HIV-1 infection. Lancet 1998;351:786–792.
12. Gordin FM, Chaisson RE, Matts JP, Miller C, Garcia M, Hafner R, Valdespino JL, Coberly J, Schechter M, Klukowicz AJ, et al. Rifampin and pyrazinamide vs. isoniazid for prevention of tuberculosis in HIV-infected persons: an international randomized trial. JAMA 2000;283:1445–1450.
13. Lambert L. Update: fatal and severe liver injuries associated with rifampin and pyrazinamide treatment for latent tuberculosis infection. Morb Mortal Wkly Rep 2002;51:998–999.
14. Stout JE, Engemann JJ, Cheng AC, Fortenberry ER, Hamilton CD. Safety of 2 months of rifampin and pyrazinamide for treatment of latent tuberculosis. Am J Respir Crit Care Med 2003;167:824–827.
15. Burman WJ, Reves RR. Hepatotoxicity from rifampin and pyrazinamide: lessons for policy makers and messages for care providers. Am J Respir Crit Care Med 2001;164:1112–1113.
16. Jasmer RM, Snyder DC, Chin DP, Hopewell PC, Cuthbert SS, Antonio Paz E, Daley CL. Twelve months of isoniazid compared with four months of isoniazid and rifampin for persons with radiographic evidence of previous tuberculosis: an outcome and cost-effectiveness analysis. Am J Respir Crit Care Med 2000;162:1648–1652.
17. LoBue PA, Moser K. Factors associated with completion of latent tuberculosis treatment [abstract]. Am J Respir Crit Care Med 2002;163:A713.
18. Treatment of tuberculosis and tuberculosis infection in adults and children: American Thoracic Society. Am J Respir Crit Care Med 1994;149:1359–1374.
Correspondence and requests for reprints should be addressed to Philip A. LoBue, M.D., Centers for Disease Control and Prevention, National Center for HIV, STD, and TB Prevention, Division of Tuberculosis Elimination, Field Services Branch, P.O. Box 85222, Mail Stop P511D, San Diego, CA 92186–5222. E-mail:


No related items
American Journal of Respiratory and Critical Care Medicine

Click to see any corrections or updates and to confirm this is the authentic version of record