Background. Severe liver injuries were attributed to the rifampin and pyrazinamide (RZ) regimen after it was recommended for treating latent tuberculosis infection. Implicating RZ as the likeliest cause required excluding alternative causes.
Methods. US health departments reported data on patients who died or were hospitalized for liver disease within 1 month after taking RZ for latent tuberculosis infection from October 1998 through March 2004. The circumstances were investigated on site for each case. Illness characteristics, reasons for RZ treatment, doses and frequency of administration of pyrazinamide, monitoring during treatment, and causes of liver injury were determined.
Results. Liver injury was attributable to RZ use for all 50 patients reported, 12 of whom died. For 47 patients, RZ was the likeliest cause of liver injury. The median patient age was 44 years (range, 17–73 years). Thirty-two patients (64%) were male. Seven (16%) of 43 patients tested had hepatitis C virus antibodies, 1 (2%) of 45 had chronic hepatitis B, 3 (14%) of 22 had positive results of HIV serologic tests, 34 (71%) of 48 had alcohol use noted, and 33 (66%) of 50 were taking additional hepatotoxic medications. Six patients, 2 of whom died, had no predictors for liver disease. Patients who died were older (median age, 52 vs. 42 years; P = .08) and took a greater number of other medications (median number of medications, 4 vs. 2; P = .05) than did those who recovered, but these 2 factors were correlated (P < .01). Thirty-one patients (62%) were monitored according to guidelines, 9 of whom died.
Conclusions. RZ was the likeliest cause of most of these liver injuries, some of which were fatal in spite of monitoring. Fatality was predicted by age or use of other medications, but none of the cofactors showed promise as a reliable clinical predictor of severe liver injury.
Three of the 4 drugs in the standard chemotherapy regimen for tuberculosis (TB) are hepatotoxic [1–3]. Rare but sometimes fatal liver injury attributed to isoniazid use raises concerns about administration of this drug for preventing TB disease [4–6]. Rifampin possibly aggravates isoniazid hepatoxicity, and pyrazinamide is associated with both dose-related hepatotoxicity and a compounding effect on liver injury with multidrug regimens [7–10]. Any drug-associated liver injury is diagnosed by exclusion, and investigating incidents on site can clarify the causes of liver injury [4, 5].
In randomized treatment trials involving HIV-infected subjects, 2 months of rifampin (600 mg) and pyrazinamide (15–20 mg/kg) in combination (RZ) proved to be as effective as 6 months of isoniazid treatment for prevention of TB disease and was well tolerated [11–13]. The briefer regimen held promise of increasing treatment-completion rates beyond those achieved with the 6- or 9-month isoniazid regimen, and it was recommended for HIV-infected patients in October 1998 . In April 2000, it was recommended for HIV-uninfected persons as an alternative to isoniazid therapy [15, 16]. Within months, severe and fatal liver injuries associated with RZ were noted [17–19]. This article summarizes investigations of 50 such incidents, including 12 that were fatal.
Definitions. Reports of hospital admissions or deaths involving patients who were treated with RZ for latent TB infection (LTBI) were pursued. Liver injury had to have been detected during RZ use or within 1 month after discontinuation, with RZ as the only anti-TB treatment at the time of onset. Attributing liver injury to RZ treatment required seeking alternative causes (environmental and/or occupational exposure to a hepatotoxin, acute viral hepatitis, acute intrinsic liver diseases, or use of another hepatotoxic drug started within 1 week of commencement of RZ therapy), and degrees of certainty were assigned as follows: for definite cases, a complete set of evidence excluded proximate causes besides RZ use; for probable cases, the evidence implicated RZ, but data about another potential cause were lacking; and for possible cases, a competing cause was supported by the evidence. Alcohol use was counted if it was in a problem list or discharge summary or if a patient reported that medical care was necessary because of it. A panel of 4 authors (K.I., J.A.J., L.A.L., and W.A.B.) reviewed the full data for each report and assigned it to a category by consensus.
The guidelines for LTBI testing and treatment provided the criteria for whether a patient was a candidate for testing or a candidate for treatment. In the initial guidelines in June 2000, active hepatitis and end-stage liver disease were listed as relative contraindications for RZ treatment . By 31 August 2001, clinicians were advised not to prescribe RZ for patients with underlying liver disease or previous isoniazid-associated liver injury .
To be monitored according to recommendations, patients had to be interviewed about liver disease at baseline and at weeks 2, 4, and 8 after commencement of treatment. Additional criteria depended on the guidelines in effect when treatment was started. Before 31 August 2001, baseline measurements of serum aminotransferase and bilirubin concentrations were recommended only for patients with specific conditions (e.g., history of chronic liver disease); follow-up tests were recommended for patients with abnormal results at baseline or with suspected liver injury . From 31 August 2001 through 8 August 2003, measurements of serum aminotransferase and bilirubin concentrations at baseline and at weeks 2, 4, and 6 of treatment were recommended for all patients taking RZ; RZ was to be dispensed in 2-week supplies . On 8 August 2003, measurements of serum aminotransferase and bilirubin concentrations at week 8 of treatment were added .
Notifications of liver injuries. From October 1998 to March 2004, the Centers for Disease Control and Prevention (CDC) requested reports at routine meetings with TB-control officials and via Morbidity and Mortality Weekly Report [17–19], letters to the 50 state and 10 city TB-control programs that receive federal funds for TB control, the American Thoracic Society newsletter, and a TB newsletter sent to health departments and specialists in pulmonary medicine and infectious disease. Reports were tracked with a data form (table 1) and a password-protected Microsoft Access database. Patient identifiers were removed. A CDC institutional ethics review determined that the intervention was an urgent public health response and not human subjects research, and patient written consent was not obtained.
For additional reports, the US Food and Drug Administration Adverse Events Reporting System was queried for this combination: treatment with RZ (and no other antituberculosis anti-TB drugs), liver injury, report date before 31 October 2003, and hospital admission or death.
On-site investigations. All reported cases were investigated on site. Medical records were reviewed. Health care providers, patients, and family members were interviewed.
Data analysis. EpiInfo, version 6.04d (CDC) , and SAS software, version 8.02 (SAS Institute) , were used. Median values were compared using the Wilcoxon rank sum test using the npar1 way function in SAS software, version 8.02 .
Data for 50 patients were reported, and all cases met the criteria for inclusion. The earliest treatment start date was April 1999, and the latest was September 2003. Forty patients were included in previous public health dispatches [17–19]. Of the 50 patients, 12 (24%) died. The reports came from 16 states and the District of Columbia. The source of the RZ prescription was a health department for 25 patients (50%), other institutions (e.g., clinics in county hospitals) for 22 (44%), and private providers for 3 (6%). Fifteen patients (30%) were inmates of correctional facilities (table 2). Twenty-five (50%) of the patients received directly observed therapy; these included 7 of 8 patients who received the twice-weekly regimen. Fifteen (30%) of the patients completed RZ treatment. HIV status was known for 22 patients (44%); 3 were infected, 2 of whom died. Five patients were referred for liver transplantation, but none received a transplant; 2 (40%) of these patients died, and 3 (60%) recovered without undergoing transplantation.
The attribution of liver injury to RZ use was “definite” for 36 patients, “probable” for 11, and “possible” for 3. The reason for all 11 “probable” designations was the lack of at least 1 result of a serologic test for hepatitis virus. For the 3 patients with a “possible” designation, 1 patient had type 1 autoimmune hepatitis diagnosed concurrently and remained ill 1 year later, another had saquinavir added to an antiretroviral regimen at the same time that RZ treatment was started, and the last had metformin treatment started at the same time as RZ treatment and also had nonalcoholic steatohepatitis previously diagnosed by biopsy.
The US Food and Drug Administration Adverse Events Reporting System had 9 compatible reports (7 from the United States, 1 from the United Kingdom, and 1 from Hong Kong, China). Five of the reports from the United States, all of which involved fatalities, already had been notified to the CDC. The other 2 reports from the United States could not be traced for investigation.
Testing for and Treatment of LTBI
At least 1 reason for skin testing was recorded for 49 patients, and the reasons were consistent with guidance for 48 patients as follows: recent contact with a person with contagious TB (13 patients), emigration from a country where TB is endemic (29 patients), occupational exposure risk as a health care worker (6 patients), connection with a congregate setting (22 patients), and an underlying medical condition (diabetes for 5 patients and HIV infection for 3 patients). The patient without an indication for testing was a US-born college student who was required to undergo a skin test for matriculation.
The decision to treat was consistent with guidelines for 48 patients. Treatment indications for the other 2 patients were ambiguous. One patient was the college student. The other had a baseline aminotransferase concentration greater than the upper limit of normal and was born overseas but had been in the United States for >5 years .
Of the 50 patients, 42 (84%) started receiving treatment before 31 August 2001. Of these, baseline measurements of aminotransferase and bilirubin concentrations were obtained for 32 (76%). Thirty of these patients were monitored with serial blood tests. Six patients with histories of alcohol use were not tested at baseline; 1 also had HIV infection.
From 31 August 2001 through 8 August 2003, a total of 7 patients (14%) started receiving treatment, and 5 underwent baseline and follow-up tests; the 2 who were not tested had histories of alcohol use. All 7 patients made follow-up appointments as recommended, but only 1 attended as scheduled.
One patient started receiving treatment after 8 August 2003, with blood tests performed at baseline. The physician in the TB clinic prescribed RZ for 12 weeks in 4-week supplies. This patient scheduled appointments for weeks 1, 3, 4, and 6 of treatment, and he attended the appointments at weeks 1 and 4. Coincidentally, at week 8, he attended an HIV clinic but was not assessed for tolerance to RZ, and he became ill 1 month later.
Onset and Detection of Liver Injury
Liver injury was diagnosed by the prescribing provider in 33 instances (66%) and by another provider in 17 instances (34%). The median interval to onset of hepatitis symptoms after commencement of RZ treatment was 38 days (range, 0–101 days), and the median time to detection of the first aminotransferase concentration greater than the upper limit of normal was 36 days (range, 3–94 days). Hepatitis symptoms were reported by 48 patients. Two patients did not report symptoms of hepatitis, but 1 had chronic dementia.
For 48 patients, RZ treatment was stopped a median of 9.5 days (range, 0–105 days) after either the onset of hepatitis symptoms (29 patients), detection of an aminotransferase concentration greater than the upper limit of normal (15 patients), or onset of symptoms and abnormal aminotransferase concentration simultaneously (4 patients). The remaining 2 patients completed the regimen 6 and 23 days before the onset of hepatitis symptoms. Eight patients had symptoms but did not seek medical attention, and their illness was discovered during regular appointments.
Potential Cofactors for Liver Injury
Alcohol and illicit drug use. Of the 48 reports that noted this information, 34 (71%) included alcohol use (any time before RZ treatment), and 5 (11%) of 45 included injection drug use. One patient intentionally inhaled paint and glue fumes.
Chemical exposures. One patient worked in an auto body paint shop where cross-flow ventilation and personal respirators were used.
Use of additional hepatotoxic medications. Nine patients initially received isoniazid, but RZ was substituted. For 6 patients, this was because of isoniazid-associated liver injury, and 3 of these patients died. For 3 patients, the regimen was switched for other reasons (1 patient preferred a briefer regimen, 1 requested directly observed therapy, and 1 had been exposed to isoniazid-resistant TB, which was discovered after commencement of isoniazid treatment).
Thirty-three patients (66%) were taking additional medications that are either metabolized in the liver or associated with liver injury (table 3) . For these 33 patients, the median number of other medications received per patient was 2 (range, 1–6 medications).
Serologic test results for viral hepatitis. Of the 50 patients, 48 (96%) were tested for hepatitis A, B, or C virus: 43 were tested for hepatitis A virus, 45 were tested for hepatitis B virus, and 43 were tested for hepatitis C virus. None had evidence of recent hepatitis A virus infection, 1 had evidence of chronic hepatitis B virus infection (first noted in 1994), and 7 had anti—hepatitis C virus antibody (2 of these patients had baseline aminotransferase concentrations that were greater than the upper limit of normal).
For 6 patients, 2 of whom died, the evidence excluded all of the aforementioned potential cofactors.
Results of Hepatic Diagnostic Tests
Liver histologic tests (with biopsy specimens or from autopsy) for all 5 patients with these data showed evidence of underlying liver disease. The 2 autopsies revealed acute massive damage superimposed on previous liver disease (previously diagnosed nonalcoholic steatohepatitis for one patient and micronodular cirrhosis with fibrosis for the other). Ultrasound examination of the abdomen was done before death for 2 other patients who died. Results were normal for one and consistent with cirrhosis (attributed to chronic hepatitis B) for the other. CT and ultrasound examination of the abdomen for a patient who recovered revealed gallstones, but there was no evidence of common bile duct obstruction or liver disease.
Seven patients underwent testing for antinuclear antibodies, with a median titer of 1:80 (range, 1:40–1:640). The patient with type 1 autoimmune hepatitis had the 1:640 titer and also had eosinophilia. The antinuclear antibody pattern was speckled for both patients who were tested for this. Results of tests for anti—double stranded DNA antibody (4 patients), anti—smooth muscle antibody (5 patients), antimitochondria antibody (3 patients), and anti—liver-kidney microsome antibody (3 patients) were negative or normal.
Comparison of Daily and Twice-Weekly Regimens
Of the 50 patients, 42 (86%) received RZ daily, and 8 (14%) received it twice weekly. Hepatitis symptoms began a median of 34 days (range, 0–85 days) after the commencement of daily treatment, compared with 60 days (range, 6–101 days) for twice-weekly treatment (P = .04). The first abnormal aminotransferase level was noted a median of 35 days (range, 13–94 days) after the initiation of daily treatment and 47 days (range, 3–83 days) after the initiation of twice-weekly treatment (P = .6). Daily treatment was stopped a median of 7 days after the onset of symptoms or an abnormal aminotransferase level, compared with 22 days for twice-weekly treatment (P = .5).
Factors Associated with Fatality
Patients who died were older than those who recovered (median age, 52 vs. 42 years; range, 32–66 vs. 17–73 years; P = .08) and were taking a greater number of other hepatotoxic medications (9 [75%] of 12 patients who died were taking a median of 4 medications [range, 1–6 medications], compared with a median of 2 medications [range, 1–4 medications] for 24 [63%] of 38 others; P = .05) (tables 2–4). The number of other medications received increased with age (P < .01). The fatality rate was similar for recipients of daily and twice-weekly regimens (9 of 42 patients vs. 3 of 8 patients; P = .3). The patients who died had symptoms later than did patients who did not die (median number of days after commencement of treatment, 51 vs. 35 days; range, 6–85 vs. 0–101 days; P = .1). For those who died, the median time to the first aminotransferase concentration greater than the upper limit of normal was 56 days (range, 6–94 days), compared with 35 days (range, 3–90 days) for those who did not die (P = .6) (figures 1 and 2). Three of 12 patients who died received isoniazid before RZ, compared with 6 of 38 patients who survived (P = .6). Nine (75%) of 12 patients who died and 22 (58%) of 38 who survived were monitored on the basis of the guidelines prevailing during the treatment period. Treatment of those who died was stopped a median of 3 days (range, 0–64 days) after the onset of symptoms or detection of an aminotransferase concentration greater than the upper limit of normal, compared with 10 days (range, 0–105 days) for those who recovered (P = .4).
RZ treatment was at least a possible cause for liver injury for each of these 50 patients who were being treated for LTBI; for 36 patients, it was the definite cause. Patients who died were older and took a greater number of other hepatotoxic medications, but the predictive value of these associations was poor: for example, 3 patients who died were not taking any other hepatotoxic medications. Six patients, including 2 who died, had no potential cofactors that could have influenced treatment and monitoring decisions.
Treatment of LTBI was indicated for almost all patients. Only two-thirds were monitored as recommended, but correct monitoring did not decrease the fatality rate. Even if more intense monitoring could improve safety, it would make RZ less practical for widespread use.
The majority of patients continued taking RZ at least 7 days after becoming ill, which may have exacerbated their illnesses. Onset was late for some—it was first noted in the seventh week or later for 19 patients (figures 1 and 2)—which raises the possibility of decreased vigilance for problems at the end of treatment. Another troubling aspect is that injury could have evolved even after treatment was stopped . This phenomenon was described in detail in separate reports of 3 patients outside the United States [24–26].
Histologic test results raised the possibility that occult liver disease predisposed the patients to RZ-associated liver injury. However, this speculation is based on data for only 5 patients, and even if the concept were confirmed, it would have limited utility for the clinician, who rarely could exclude occult liver disease. Still, this serves as a reminder that many patients who are at risk for TB also have liver disease or risk factors for it.
Pyrazinamide-associated hepatotoxicity is said to be dose dependent, in contrast to the idiosyncratic hepatotoxicity of isoniazid [9, 10, 23]. Historically, pyrazinamide caused fatal hepatitis when it was administered at a dose of >50 mg/kg for treatment of drug-resistant TB, but it was better tolerated at 20–25 mg/kg in multidrug regimens. Its dose was further reduced to 15–20 mg/kg for treatment of LTBI . Sixteen patients in our study received daily pyrazinamide doses that were >20 mg/kg but still within the range for treatment of TB disease. Pyrazinamide reportedly worsens the prognosis of liver injury during treatment of TB disease . Rifampin rarely has been implicated as a hepatocellular toxin on its own [7, 8, 23].
Two patterns of liver injury are noted for multidrug regimens for TB disease . The first involves an early increase in serum aminotransferase concentration, usually within 15 days after treatment initiation. This pattern, attributed to rifampin-augmented isoniazid hepatotoxicity, has a good prognosis. The second pattern resembles what we describe here, with a substantial increase in serum aminotransferase concentration that occurs >1 month into treatment (figure 2). This pattern, attributed to pyrazinamide treatment, has a worse prognosis .
In the 3 clinical trials and 1 study, all of which documented good tolerance to RZ, the participants were HIV infected [11–13, 28]. The initial reports of RZ-associated liver injury involved HIV-negative patients , which raised a question of whether HIV infection protected against RZ-associated liver injury . Three patients in our study were HIV infected, which argues against a protective effect. However, these patients were taking antiretroviral therapy, which was unavailable to most participants in the RZ treatment trials. Antiretroviral therapy could augment RZ hepatotoxicity, or it may restore a hypothetical immunological pathway of hepatotoxicity.
Case studies lack denominator data for estimating incidence rates and risk ratios. Our study also is limited by its retrospective design and local variances in diagnostic practices. Reporting bias could make our series nonrepresentative. By working from preselected numerator data, we might have overestimated the role of RZ when we were assigning etiology, but during our on-site investigations, we rigorously sought other causes of liver injury.
On the basis of the results of these investigations and experience using RZ [30–36], the CDC and American Thoracic Society now recommend that RZ should generally not be used for treating LTBI. Nine months of isoniazid treatment remains the preferred option. Monitoring is based on clinic visits, with face-to-face contact and with careful instructions to patients . As noted in 1989 by the US Advisory Council for the Elimination of Tuberculosis and later by the Institute of Medicine, a briefer, safer regimen for preventing TB is still needed for accelerating progress against this disease [37, 38].
We thank the National TB Controllers Association and all state and local TB controller officials for notifying the Centers for Disease Control and Prevention of the adverse effects and for collaborating in the investigations. Centers for Disease Control and Prevention personnel who did on-site investigations included McKenzie Andre, Phyllis Cruise, Beverly DeVoe, Heather Duncan, Reuben Granich, Darryl Hardge, Jane Kelly, Ventakarama R. Koppaka, Philip LoBue, Marisa Moore, Margaret Oxtoby, Farah Parvez, Renee Ridzon, Frank Romano, and Kevin Winthrop. Ameisha Sampson, Melissa Valdez, Melanie Cleary, and Elsa Villarino helped design the collection forms and set up the database, and many others provided helpful discussions during investigations. Sarah J. Singer (US Food and Drug Administration) queried the Adverse Events Reporting System. We also thank the following members of an ad-hoc expert advisory panel on RZ for their guidance: Henry Blumberg, Richard Chaisson, David Cohn, Charles Daley, Stefan Goldberg, Robert Jasmer, James McAuley, Masa Narita, Charles Nolan, Richard O'Brien, Mary Reichler, and Dixie Snider.
Financial support. Centers for Disease Control and Prevention, US Department of Health and Human Services.
Potential conflicts of interest. All authors: no conflicts.