|Home | About | Journals | Submit | Contact Us | Français|
Antimicrobial pharmacokinetic-pharmacodynamic studies suggest that pyrazinamide doses higher than those currently recommended may be more efficacious. However, high pyrazinamide doses are believed to be hepatotoxic. Searches for clinical trials in MEDLINE, EBSCOHOST, and the Cochrane Controlled Trial Register were made. Studies that employed pyrazinamide dose scheduling and pharmacokinetic analysis design were examined. Population pharmacokinetic modeling methods were utilized to identify parameters associated with toxicity. At an equivalent area under the concentration-time curve, the time that concentration persisted above some thresholds was associated with overall adverse events (P = 0.032), arthralgia (P = 0.089), and an elevated serum aspartate aminotransferase level at 3 months (P = 0.067). Next, a meta-analysis was utilized to compare rates of adverse events (i) between different pyrazinamide doses, (ii) between different dosing schedules, and (iii) between pyrazinamide-containing and non-pyrazinamide-containing antituberculosis regimens. The 29 studies selected were heterogeneous (Cochrane Q statistic P value of <0.001; I2 of >95%). For the once-a-day dosing schedule, arthralgia was dose dependent (r2 = 0.996). However, arthralgia was less common with intermittent dosing, consistent with the time concentration persisted above the threshold. Arthralgia was generally clinically inconsequential. The frequencies of hepatotoxicity were 0.057 (95% confidence interval [CI], 0.021 to 0.141) for pyrazinamide monotherapy, 0.044 (CI, 0.033 to 0.059) for pyrazinamide-containing combination regimens, and 0.040 (CI, 0.023 to 0.040) for non-pyrazinamide-containing combination regimens. The frequencies of hepatotoxicity were 0.042 (CI, 0.026 to 0.067) for 30 mg/kg of body weight, 0.055 (CI, 0.031 to 0.094) at 40 mg/kg, and 0.098 (CI, 0.047 to 0.193) at 60 mg/kg of pyrazinamide. Thus, high-dose pyrazinamide did not significantly increase hepatotoxicity. This suggests that a considerable portion of hepatotoxicity rates may be idiosyncratic.
Pyrazinamide was responsible for shortening the duration of antituberculosis rifampin-containing regimens to the current 6-month standard for drug-susceptible tuberculosis (TB) (3, 18, 31). In addition, pyrazinamide is often part of the treatment for multidrug-resistant and extensively drug-resistant TB. Recent antimicrobial pharmacokinetic-pharmacodynamic studies suggested that doses higher than the dose of 15 to 30 mg/kg of body weight/day currently used would be more efficacious (19). In addition, pharmacokinetic studies have demonstrated increased pyrazinamide serum clearance and volume of distribution with weight (16, 53), so higher doses than those currently used are needed to achieve the same exposures in heavier patients than in leaner patients. However, the main limitation of high-dose pyrazinamide therapy has been a high rate of adverse reactions (8, 11). Frequently described adverse events (AE) are arthralgia and hepatotoxicity (3, 17). The latter is also common to many anti-TB drugs used in combination with pyrazinamide, so causality is not established. While guidelines for attributing AEs to drugs are now clear (12), most previous reports that detailed pyrazinamide AE (8, 11, 13) fell short of these standards. Unfortunately, such reports form the backbone of the idea that pyrazinamide causes most of the observed AEs in anti-TB regimens.
Toxicodynamics of antimicrobial drugs have of late started to benefit from a better understanding of pharmacokinetic variability and dose fractionation techniques that allow the identification of particular aspects of the concentration-time curve that best determine drug toxicity (35, 42). Using these techniques as well as a meta-analysis, we examined how often toxicity occurs with pyrazinamide monotherapy and with pyrazinamide combination therapy for active TB. Second, we wanted to identify the time to occurrence of nonidiosyncratic pyrazinamide AEs. Third, we wanted to determine the role of dose and dose schedule on the frequency of AEs.
The Council for the International Organization of Medical Sciences and the Food and Drug Administration definitions for drug AEs were used in our study (12, 50). AE refers to any unfavorable and unintended sign, symptom, disease, overdoses, and medication errors temporally associated with the use of a drug regardless of causality (12). For more recent studies, serum alanine aminotransferase (ALT) levels greater than three times the upper limit of normal with symptoms of hepatitis or greater than five times the upper limit of normal were defined as hepatotoxicity (3, 12). For older studies, standard laboratory values for the bromsulphthalein and/or the cephalin-cholesterol flocculation test used at the time were acceptable (39). Arthralgia was any joint pain reported during a study. Mortality refers to any deaths occurring during a study regardless of the cause.
Searches were performed in the MEDLINE, EBSCOHOST, and the Cochrane Controlled Trial Register (Cochrane Library 2004, issue 2) databases. A search for title and abstract was made by using the following search Medical Subject Heading (MeSH) terms: “tuberculosis or Mycobacterium tuberculosis” and “antituberculosis therapy or antituberculosis drugs” and “drug toxicity.” Searches were limited to English-language literature published between January 1950 and December 2008. We supplemented the electronic data search with a manual search of all references listed from the studies obtained in the above-described search using the same standardized abstraction form. The references from these articles were used to supplement the search for clinical trials performed prior to 1963 since some of them were not adequately captured by electronic databases. The search strategy is depicted in Fig. Fig.11.
Data were abstracted from the spreadsheets, text, and tables and entered into Microsoft Excel spreadsheets. All the AEs defined above, time to AE, daily pyrazinamide dose, duration of therapy, cumulative dose, and gender were captured. The data were then exported to different programs for analysis.
For the purpose of determining pharmacokinetic parameters associated with toxicity, studies were chosen if they used a dose-scheduling (dose fractionation) design and measured serum pyrazinamide concentrations at multiple time points. In the East African pyrazinamide study, patients were treated with a dose of either pyrazinamide at 0.5 g three times a day (n = 92), 1.5 g daily (n = 96), or 3 g every other day (n = 192) for 6 days of each week, for a cumulative weekly oral dose of 9 g (15). The patients all received streptomycin intramuscularly. Patients were treated in the hospital for 6 months, with directly observed therapy. A proportion of these patients had serum and urine concentrations of pyrazinamide and pyrazinoic acid measured (16). We used the previously reported values of the “Nairobi crossover” groups in the study to identify the compartmental pharmacokinetics of pyrazinamide using ADAPT 5 software (2), and we used the identified parameters to calculate the 0- to 168-h area under the concentration-time curve (AUC0-168) for each regimen. The peak maximum concentrations of drug in serum (Cmax) and the percentage of time that the concentration persisted above 5 mg/liter, 10 mg/liter, or 20 mg/liter utilized were those observed for the original pharmacokinetic study (16). Comparisons of the proportion of patients with all AEs (excluding vestibular toxicity, which is due to streptomycin), elevated aspartate aminotransferase (AST) levels, and arthralgia were made between each of the three dosing regimens as well as between the two daily doses as a combined group versus the intermittent group (every other day) using Pearson's chi-squared test.
Monotherapy studies that used many different doses of pyrazinamide were examined for relationships between dose and AE (37, 43, 54). Because data for some of the patients were missing, we used only all AEs, daily dose, and cumulative dose for analyses of associations with AEs in a logistic regression model. The model included a constant, with 20 maximum iterations, and P values of 0.05 for entry and 0.10 for removal at each backward step.
Prospective controlled studies and observational studies with concurrent controls that reported the presence and absence of AEs as an endpoint and reported the AEs in a standard manner met the inclusion criteria (12). Other inclusion criteria were (i) studies that clearly confirmed active TB by microscopy and/or culture; (ii) AEs reported as a study outcome with clear, standard, and clinically acceptable definitions; and (iii) the use of ethical and clinically acceptable methods for that time. Studies were excluded if they were retrospective or if they combined previously reported data already included in the search. Case reports, case series, as well as comparative studies with historical controls, case-control, and retrospective design were excluded because of the higher risk of bias associated with these designs. Since the observed frequencies of AEs associated with treatment regimens were the outcomes of interest, studies done with highly selected populations were excluded. These populations were defined as incarcerated or posttransplant patients or children or selected minorities as well as patients in phase I/phase II studies done as a proof of concept or to test new medicinal products.
Both authors reviewed selected abstracts and articles considered for inclusion. Discrepancies were resolved by consensus. The quality of each trial was graded by use of a validated score whose maximum possible score is 5 (25). Full description of evaluation methods, description of participants who withdrew from or dropped out of the trial, and randomization of participants were given 1 point for each item present. If randomization was concealed and the method of double-blind evaluation was appropriate, the study was assigned 1 additional point.
Individual-level data from controlled clinical trials that compared various established anti-TB treatment regimens were examined. Three categories of anti-TB regimens were examined: pyrazinamide monotherapy regimens (category A), combination regimens that included pyrazinamide (category B), or combination regimens that excluded pyrazinamide (category C). The outcomes of interest were (i) all AEs, (ii) hepatotoxicity, (iii) arthralgia, (iv) proportion of patients interrupting therapy, and (v) death.
The number of patients randomized to each treatment regimen was used as the denominator. Weighted mean percentages of patients with each AE were estimated by using either fixed- or random-effects models according to established methods (4). For both models, each study datum was weighted by the inverse of its variance. However, because random-effects models combine within-study data variance and between-study data variances, DerSimonian and Laird methods were used to obtain between-study data variance (4, 20). Heterogeneity across studies was compared by using the Cochrane chi-squared Q test and quantified by I2 index methods. Sensitivity analysis was performed to assess the robustness of obtained estimates of hepatotoxicity frequencies when the quality of pooled studies was considered.
Population pharmacokinetic analysis was performed by using ADAPT 5 software. GraphPad Prism (version 5) and Comprehensive Meta-Analysis software were used for statistical analysis and meta-analysis. All software was run on a personal computer.
Pharmacokinetic exposures reported in the East Africa study as well as those calculated using compartmental population pharmacokinetic analysis are shown in Table Table1.1. For elevated aspartate aminotransferase (AST) levels, only a cutoff point of >50 Karmen units was reported. This is equivalent to >54 international units/liter and was used to define hepatotoxicity for the purposes of the toxicodynamic analysis. No differences in overall AEs or arthralgia were found when each of the three dose schedules was compared to the other (P = 0.1). However, hepatotoxicity at 3 months occurred in 11/49 patients on three-times-a-day therapy, 19/51 patients with once-a-day therapy, and 18/90 patients with the every-other-day regimen (P = 0.067). When both daily therapy regimens were compared to the every-other-day regimen, all AEs (14/188 versus 5/190 patients; P = 0.032) and arthralgia (7/190 versus 2/192 patients) were more common with the daily therapy regimens (P = 0.09). Since the cumulative AUC was similar regardless of dose schedule, and the Cmax was lowest with 0.5 g three times a day but highest with 3 g every other day (Table (Table1),1), the AUC and Cmax were least associated with either all AEs or arthralgia. Since the percentage of time when the concentration exceeded 20 mg/liter was lowest with the 0.5-g three-times-a-day regimen, it too was inconsistent with the pattern of the rate of AEs. However, since the percentage of time when the concentration exceeded 5 mg/liter and, to a lesser extent, 10 mg/liter increased with more frequent dosing, these two parameters better explained the pattern of the higher rates of AE and arthralgia encountered more often with the daily therapy regimens.
Three separate pyrazinamide monotherapy studies were identified (Table (Table22 ) (37, 43, 54). Only the study by Yeager et al. (54) had results presented that were comprehensive enough, and these results were further analyzed. Severely ill TB patients received a mean pyrazinamide daily dose of 3.7 g (range, 1.2 to 7.6 g) for 3 to 75 days as salvage therapy. AEs were encountered for 22/43 patients, with the most frequent AEs being arthralgia (12/43 patients), palpitations (4/43), and pyrexia (4/43). The majority of AEs occurred after 30 days of therapy. Hepatotoxicity was reported for only 2 (5%) patients, both of whom had abnormal flocculation tests and jaundice after 52 and 87 days of treatment. In both cases, hepatotoxicity was resolved after the pyrazinamide therapy was stopped. Five patients died, two during therapy and the remainder after therapy. None of the deaths were judged to be drug related.
In multivariate logistic regression analysis, only the daily dose significantly predicted pyrazinamide-induced arthralgia, with a 2-fold increase in arthralgia with each 1-g increase in the daily dose (r = 0.580; P = 0.014). The relationship between daily dose and all AEs is shown in Fig. Fig.2.2. When all adverse events were combined and plotted against the daily pyrazinamide dose, the inflection point for increased toxicity was at 4.5 g/day (data not shown).
Twenty-nine controlled clinical trials with over 13,000 patients from Africa, Asia, Europe, and the Americas met inclusion criteria. Characteristics of included patients and the studies that met the inclusion criteria are shown in Table Table2.2. The median follow-up period ranged from 3 to 24 months but was statistically similar between the regimens tested. The majority of the trials (18/29 trials) had patients randomized to assigned treatment regimens. The agreement between the two reviewers was associated with κ values of 0.9 for the inclusion of trials and 0.9 for the rating of trials on considered methodological aspects (25). Tests for heterogeneity for all outcomes are shown in Fig. Fig.3.3. A pooled analysis of all studies showed substantial heterogeneity across studies (Cochrane Q statistic P value of <0.001; I2 of >95%), indicating that there is no certain frequency rate for the outcomes of interest around which all studies can be pooled (Fig. (Fig.3).3). Consequently, random-effects point estimates were used to infer statistical differences between groups. The funnel plot and classical fail-safe methods demonstrated a z value of 44.476, with a P value of <0.001. Thus, no significant bias was observed.
Over 3,000 of 13,000 (22%) patients reported one or more AEs, and 211 (2%) patients died. Twenty-nine percent of patients who received pyrazinamide monotherapy reported an AE. Pyrazinamide monotherapy studies were monitored for longer periods than were studies of combination regimens. Among those patients who received combinations that included pyrazinamide, 19% reported AEs, while among those who received combination therapy with no pyrazinamide, 14% reported AEs (Fig. (Fig.3).3). The mean times to AEs were 1.48 months for monotherapy and 1.21 months for combinations (P = 0.3017 by log rank test). Arthralgia was the most frequent AE and was encountered in 16% of patients receiving pyrazinamide alone, compared to 2.5% of patients receiving combinations that included pyrazinamide and 0.4% of patients receiving combinations that did not have pyrazinamide (P < 0.0001) (Fig. (Fig.3).3). However, those studies were highly heterogeneous. Pooled frequency estimates for the other AEs were statistically similar whether pyrazinamide was added to the regimen or not (Fig. (Fig.3).3). Specifically, the frequency of hepatotoxicity and jaundice was less than 6% across all regimens tested and did not differ whether or not pyrazinamide was included in a regimen (Fig. (Fig.44).
In terms of dose and dose schedule, the 15 studies (1, 6, 7, 9, 10, 21-23, 26, 29, 36, 40, 41, 44, 48) examined for these factors were homogeneous (Cochrane Q statistic of 1.681; P = 0.195). The pooled frequency estimate for a pyrazinamide dose of ≥60 mg/kg was 0.098 (confidence interval [CI], 0.047 to 0.193), that for 40 mg/kg was 0.055 (CI, 0.031 to 0.094), and that for 30 mg/kg was 0.042 (CI, 0.026 to 0.067) (Fig. (Fig.5).5). Among patients on pyrazinamide-based combination therapy, hepatotoxicity was more frequent in patients treated with daily doses (frequency, 0.054 [CI, 0.034 to 0.086]) than in those treated with intermittent doses (frequency, 0.027 [CI, 0.016 to 0.047]) (P = 0.058), even though the cumulative weekly dose with intermittent therapy was 10.5 g versus 9 g for daily therapy. Thus, hepatotoxicity occurred with more frequent dosing despite a slightly lower cumulative dose, which means that it was associated with the percentage of time that the concentration exceeded a threshold but not total exposure (AUC) or peak concentration.
To test for alternative explanations for these data, pyrazinamide and rifampin doses were correlated against peak serum AST level data. No significant correlation between the pyrazinamide dose and peak serum AST levels was observed (P = 0.5561; r = 0.0161). The rifampin dose was also not correlated with liver function (P = 0.8633; r = 0.0507). Further analysis of only high-quality trials, i.e., a quality score greater than 3, did not significantly alter conclusions compared to those obtained when all trials were included in the analysis.
For all the AEs, the presence or absence of pyrazinamide did not result in a significantly higher frequency of side effects. However, the studies were not homogeneous. The most frequent AE associated with pyrazinamide was arthralgia, which occurred late (after 1 month) and was nonsevere (Fig. (Fig.2).2). Second, the data did not demonstrate a relationship between pyrazinamide dose and hepatotoxicity. In this pooled analysis, hepatotoxicity was infrequent and also occurred late, after 30 days. These data question prior attestations that the addition of pyrazinamide to a regimen, or increasing pyrazinamide doses, significantly increases anti-TB-drug-related adverse events.
Polyarthralgia occurred more frequently in pyrazinamide-based regimens, consistent with data from previous reports (46, 47, 55). Pyrazinamide is hydrolyzed by microsomal deaminase to pyrazinoic acid (POA) and then hydroxylated to 5-hydroxypyrazinoic acid by xanthine oxidase for excretion by the kidneys. Pyrazinamide-induced arthralgia is caused by hyperuricemia due to the competitive inhibition of xanthine oxidase by POA. Serum pyrazinamide concentrations correlate closely with the POA concentration in serum and urine at a POA/pyrazinamide ratio of 1/3 (16). Our pharmacokinetic analysis results established that the percentage of time for the concentration to exceed 10 mg/liter, but most strongly exceed 5 mg/liter, best explained arthralgia. The meta-analysis revealed that the same, intermittent dosing reduced the frequency of arthralgia compared to the frequency of arthralgia for daily therapy for roughly equivalent cumulative doses, so the time above a certain threshold concentration was also the parameter driving arthralgia. The relationship between dose and arthralgia in monotherapy studies also reflects the same finding, since without a dose-scheduling design, all three parameters (time above threshold or AUC and/or peak concentration) will increase proportionately with the dose. We propose that xanthine oxidase gets saturated at relatively low POA concentrations so that the longer the concentration persists above the threshold, the longer the maximal inhibition of xanthine oxidase, which leads to prolonged urate deposition in joints (16). Once therapy stops, the concentrations fall and urate deposition stops, and the patients feel better.
On the other hand, hepatotoxicity is the most dreaded side effect attributed to anti-TB drugs (3, 8, 13, 14, 17, 31). Contrary to data from previous reports (8, 13, 14, 17, 40, 51), our meta-analysis demonstrated no significant relationship between pyrazinamide dose and hepatotoxicity. As shown in Fig. Fig.4,4, hepatotoxicity was infrequent and occurred in fewer than 6% of patients across all anti-TB regimens and was rarely encountered prior to a 30-day duration of therapy. In four previous studies (7, 41, 43, 52), pyrazinamide doses ranging from 43 mg/kg to >60 mg/kg were associated with sporadic hepatotoxicity that did not significantly differ in frequency compared to that for patients who had received lower pyrazinamide doses. On the other hand, higher rates of hepatotoxicity with the 25-mg/kg dose than with the 40-mg/kg daily dose were demonstrated in a U.S. Public Health Service (USPHS) study from 1959 (51). However, in that USPHS study, cephalin cholesterol flocculation tests and bromsulfalein retention tests were used to evaluate liver function. These tests have since been abandoned because of poor clinical predictive value, so the misclassification bias would have been high (39). In addition, pyrazinamide was given for prolonged periods compared to the current practice of 2 months. Furthermore, our reanalysis of the USPHS study data (51) using their own liver function tests demonstrated no significant difference in proportions of patients developing hepatotoxicity at 2 months between 25-mg/kg and 40-mg/kg daily doses (Fig. (Fig.6).6). Our current findings suggest that, indeed, doses of up to 60 mg/kg for patients with a mean weight of 86 kg, currently encountered in the United States (33), would probably not lead to higher rates of hepatotoxicity.
The mechanisms underlying drug-induced liver injury are not clearly understood (27, 38). While we should not generalize from the two hepatotoxic events in patients treated with pyrazinamide monotherapy, it is nevertheless intriguing that both patients had liver biopsy specimens that showed marked eosinophilia (54). Other researchers have also demonstrated features more consistent with immunoallergic liver damage (11, 32). Given the lack of dose-dependent hepatotoxicity in our analysis, we speculate that non-dose-dependent mechanisms such as immune-based damage, among others, may explain a large proportion of pyrazinamide-induced liver damage. This suggests that a portion of pyrazinamide hepatotoxicity rates may perhaps be from idiosyncratic events.
There are several limitations to our analysis. First, the selection was not exhaustive because only studies published in English were selected. Second, the identified studies were heterogeneous; however, meta-analysis is best suited for the generation of a correct hypothesis and not a proof. Third, we did not examine studies of hepatotoxicity encountered in studies of prophylactic treatment of latent TB. That analysis will be performed separately for that group of patients. In addition, only data from East Africa were available for population pharmacokinetic analyses. Despite these limitations, however, arthralgia was unequivocally induced by pyrazinamide but was clinically inconsequential. No relationship between pyrazinamide dose and hepatotoxicity was demonstrated. While Fig. Fig.55 may appear to show that higher doses were associated with greater proportions of toxicity, this is driven mainly by the large variance in patients who received >60 mg/kg. These patients were sicker patients with more advanced TB, as reflected in Table Table2.2. Given the important role of pyrazinamide in current anti-TB therapy and the potential to further optimize efficacy by increasing the dose, our data suggest that higher doses should be examined for both efficacy and safety. Nevertheless, caution is still needed in the conduct of such studies.
We are supported by the NIH Director New Innovator Award (NIGMS/NIH DP2 OD001886) and NIAID/NIH grant R01AI079497.
Published ahead of print on 3 May 2010.