Therapeutic Drug Monitoring for Slow Response to Tuberculosis Treatment in a State Control Program, Virginia, USA

TOC summary: Diabetes was associated with increased risk for slow response and low rifampin levels.

W orldwide, tuberculosis (TB) remains the leading cause of death from a curable infectious disease; ≈1.4 million deaths occurred in 2008 alone (1). Death is a consequence of delayed diagnosis and ineffective or incomplete treatment because cure rates exceed 95% with appropriate therapy (2). Slow response to therapy can lead to prolonged infectiousness, extended treatment duration, acquired drug resistance, or recurrence of TB after treatment. The reasons for slow response are diverse, but measurement of serum anti-TB drug levels, or therapeutic drug monitoring (TDM), is a potentially useful tool for uncover-ing the causes of slow response (3,4). Low serum levels can be a consequence of malabsorption, inaccurate dosing, altered metabolism, or drug-drug interactions (4), but in most instances low serum levels can be readily corrected with dose adjustment.
TDM is currently recommended in TB treatment guidelines as optional (5), and few large TB control programs have access to routine TDM. Although published reports describe patients for whom slow response was attributable to low drug levels, questions remain about how best to implement TDM on a programmatic scale (6,7). Defi nitions of slow response vary, and recommendations for which medications to prioritize for TDM are lacking. Furthermore, for general populations receiving TB therapy, TDM is unlikely to be of benefi t, given the infrequency of treatment failure or TB recurrence (8). Although it is well known that certain patients, such as those infected with HIV and thus prone to malabsorption, are at higher risk for low drug levels (9)(10)(11)(12), studies of TDM that included patients responding well to anti-TB medications found lower than expected drug levels of isoniazid and rifampin in many patients with adequate clinical response (13,14). Therefore, identifi cation of patients at risk for slow response is critical within a TB control program. In addition, TDM performed earlier in the time course of slow response may also affect other major programmatic outcomes, such as treatment duration.
In the state of Virginia it is mandatory for providers to report all cases of TB to the Virginia Department of Health. Each case is assigned to a nurse case manager, who oversees and monitors the progress of each patient until treatment is completed. Directly observed therapy is administered by the nurse case manager or a trained outreach worker. After 4 weeks of therapy, patients are screened by the nurse case manager. Medical consultation for patients with ongoing symptoms is provided by the state TB clinicians in an effort Therapeutic Drug Monitoring for Slow Response to Tuberculosis Treatment in a State Control Program, Virginia, USA to identify slow response earlier and to prevent acquired drug resistance. Clinicians defi ne slow response in a patient as after >30 days from the start of treatment the patient has >2 of the following fi ndings: sputum smear positive for acid-fast bacilli; no improvement in TB-specifi c symptoms, including fever, cough, weight loss, and/or night sweats; and no improvement in chest radiograph lesions previously identifi ed as consistent with TB. Routine TDM among patients who met criteria for slow response was instituted by March 2007.
We performed a retrospective cohort study among patients slow to respond to pulmonary TB treatment in the state of Virginia to determine the prevalence of lower than expected levels of isoniazid, rifampin, ethambutol, and pyrazinamide measured at the time of estimated peak serum concentration (C max ). Secondary aims included investigation of risk factors for levels below the expected range, evaluation of the mean change and likelihood of achieving a level within the expected range after dose adjustment, and comparison of outcomes between persons with slow responses with those with low and expected levels. The study was approved by the institutional review boards for human subjects research at the University of Virginia and the Virginia Department of Health.

Patients
Patients were identifi ed for inclusion in the study by using routine TB surveillance data recorded in the Virginia TB Registry. All patients who were >18 years of age, had confi rmed Mycobacterium tuberculosis cultures, and started TB therapy in the state of Virginia during March 1, 2007-May 1, 2009, were eligible. We included patients who had been treated for pulmonary TB or pulmonary TB and extrapulmonary TB and who began a regimen of isoniazid, rifampin, ethambutol, and pyrazinamide. All M. tuberculosis specimens were sent to the state TB laboratory, where drug-susceptibility testing was performed after secondary culture of the isolate by using the automated Bactec MGIT 960 system (Becton Dickinson, Sparks, MD, USA). Patients were excluded if their original isolate was later found to be resistant to >1 fi rst-line medication. Patients were also excluded if they had TDM performed for reasons other than slow response.
Surveillance data were retrieved from the state TB registry and included demographics (age, sex, race/ethnicity, country of origin, and homelessness), TB history (prior episodes of TB, sputum smear and culture status of current TB episode, and chest radiograph abnormalities), coexisting conditions (diabetes, HIV infection, intravenous drug use, and excessive alcohol use), and treatment outcomes (completion of TB treatment, duration of completed TB treatment, relapse of TB following treatment completion, acquisition of drug resistance in a previously susceptible TB strain, and death from any cause during TB treatment). Information about medication-related adverse events following anti-TB drug dose increase was obtained from personal communication with the state TB medical consultants.

Therapeutic Drug Monitoring
The standard procedure for TDM was for patients to be given their daily dose of TB medications in the morning while fasting and then observed for 2 hours, during which they were restricted from eating or drinking. At 2 hours after medication administration, venous blood was collected and serum was separated before transport on dry ice to the regional referral laboratory. The drug levels from blood collected 2 hours after medication administration (C 2hr ) were used as the estimated peak maximum serum concentration (C max ) as per standard practice and were determined by using high-performance liquid chromatography (for isoniazid and rifampin) or gas chromatography with mass spectrometry (for ethambutol and pyrazinamide). Expected C 2hr ranges were provided and were consistent with published norms (5). C 2hr levels were also recorded for patients with initial low levels in whom follow-up TDM was performed after dose adjustment.

Data Analysis
Demographic and clinical characteristics were compared with the χ 2 statistic or, for nonparametric data, the Mann-Whitney U test. For the determination of risk factors for C 2hr levels below the expected range, values were dichotomized into normal if the value was within or above the expected range, or low if the value fell below the expected range. Bivariate and multivariate logistic regression analyses were used to determine risk factors for either a low isoniazid or a low rifampin level. The multivariate model included any variable with p<0.1 in bivariate analysis and relevant demographic characteristics. Paired Student t tests were used to report the mean change in C 2hr levels following dose adjustment. Medications dosed >5× per week were considered daily dosed. Biweekly dosing was used for some patients for isoniazid and rifampin, with the isoniazid biweekly dose at 3× the usual daily dose. The rifampin dose was unchanged regardless of dosing frequency. The log-rank test was used to compare treatment duration and for patients who had not completed therapy at the time of analysis; data were right censored for survival analysis. All tests of significance were 2 sided. Data were analyzed with SPSS version 17.0 software (SPSS Inc., Chicago, IL, USA).

Results
During the study, 350 patients were treated with an initial regimen of isoniazid, rifampin, ethambutol, and pyrazinamide for pulmonary TB; of these patients, 45 (13%) met criteria for slow response. Thirty-seven patients were excluded from the study (34 with normal response and 3 with slow response) after drug-susceptibility testing showed resistance to >1 medication of the treatment regimen. An additional 2 patients were excluded because TDM was performed for reasons other than slow response. Thus, 311 patients were included in the study, of whom 42 (14%) met criteria for slow response (Table 1). At the time of TDM among patients meeting criteria for slow response, all had persistent TB-related symptoms. Of the 23 patients with initial smear-positive sputum specimens, 17 (74%) had specimens that remained smear positive.
The mean (SD) age for patients in the study was 46 years (20 years), and 204 (65%) were men ( Table 1). The most common ethnicity was Asian (102 [33%]) and 228 (73%) were foreign born. Among all patients, 291 (93%) had no history of TB. Most patients tested had a positive tuberculin skin test (TST) result, although 85 (27%) did not have a TST reading recorded. There were 287 patients with a sputum smear recorded at the time of diagnosis; of these smears, 193 (62%) were positive for acid-fast bacilli. Ninety-fi ve percent (295) of patients had a chest radiograph with fi ndings suggestive of TB, of which 122 (39%) were cavitary. There was no signifi cant difference in the proportion of patients with a positive TST result, a positive sputum smear, or a chest radiograph showing cavitation among the 42 with a slow response and the remaining patients with adequate response. Among patients meeting criteria for slow response, none were HIV infected and none reported using illicit drugs (either intravenous or nonintravenous). The only signifi cant predictor of slow response was diabetes (unadjusted odds ratio [OR] 6.5, 95% confi dence interval [CI] 3.2-13.5, p<0.001; adjusted OR [aOR] 6.3, 95% CI 2.8-14.0, p<0.001).

Risk Factors for Low Isoniazid or Rifampin Levels
Analyses of risk factors for low levels of isoniazid or low levels of rifampin were performed, but small sample size precluded meaningful analysis of risk factors for low ethambutol levels. Patients with diabetes were at significantly increased risk of having a low rifampin level (OR 5.8, 95% CI 1.4-23.1, p = 0.01; aOR 5.7, 95% CI 1.2-25.7, p = 0.03) ( Table 3). Patients who received isoniazid biweekly were less likely to have low isoniazid levels than those who received isoniazid daily, but this association was not statistically signifi cant in multivariate analysis (OR 0.21, 95% CI 0.05-0.91, p = 0.04; aOR 0.47, 95% CI 0.09-2.5, p = 0.37) ( Table 3).

Follow-up C 2hr Levels after Dose Adjustment
Eighteen patients with rifampin levels below the expected range had follow-up TDM after dose adjustment. Levels for all patients increased from the initial to the follow-up level with a mean (SD) change of 11.0 μg/mL (9.7 μg/mL; p<0.001); 16 (89%) had levels in the expected range after the fi rst dose adjustment (Figure 2). Fourteen patients had follow-up TDM for daily-dosed isoniazid levels below the expected range; monitoring detected increased levels in 12 patients, with a mean (SD) change of 3.4 μg/ mL (2.9 μg/mL; p = 0.001); 4 (29%) patients had levels in the expected range. Four patients had follow-up TDM for biweekly-dosed isoniazid levels below the expected range, and all had increased levels with a mean (SD) change of 11.8 μg/mL (6.1 μg/mL; p = 0.03); 3 (75%) patients had levels in the expected range.
Rifampin levels below the expected range were signifi cantly more likely to be corrected to within the expected range following the fi rst dose adjustment than were dailydosed isoniazid levels below the expected range (p = 0.01). There was no signifi cant difference in the likelihood of correction to the expected range between daily and biweekly dosed isoniazid. No follow-up levels of ethambutol or pyrazinamide were reported. There were no reported medication-related adverse events following dose increase.

Treatment Outcomes
Complete outcomes were available for 32 (76%) patients; 10 patients continued receiving treatment. Twentyseven patients successfully completed treatment, 3 patients There were no reports of relapse of infection over a median of 14.5 months (IQR 7-25 months) from the conclusion of treatment. No patient had documented acquisition of medication resistance in follow-up TB cultures while on treatment. All 3 deaths occurred shortly after TDM was performed: 1 patient had isoniazid, rifampin, ethambutol, and pyrazinamide levels within expected ranges, 1 patient had isoniazid, rifampin, and ethambutol levels within expected ranges, and 1 patient had isoniazid, rifampin, and ethambutol levels below the expected ranges.

Discussion
The major fi nding of this study is that among patients being treated for pulmonary TB in Virginia, most patients that met criteria for slow response to therapy were found to have C 2hr levels of rifampin and isoniazid below the expected range; many patients also had low levels of ethambutol. Given the high frequency of patients who were slow to respond to both key fi rst-line medications, isoniazid and rifampin, and the well-tolerated subsequent increase in levels documented after dose adjustment, TDM appears to be a useful strategy for identifying a remediable cause of slow response at a programmatic level. Furthermore, the median duration of therapy for patients with rifampin levels below the expected range was nearly 2 months shorter than that for patients with normal rifampin levels. Although it is not specifi cally known if identifi cation and correction of lower than expected levels brought about a rapid improvement in TB clinical signs and symptoms, the comparatively shorter course represents a substantial cost savings when considering personnel involved in monitoring and medication administration, as well as diagnostic tests averted in the workup of otherwise unexplained slow response.
We found that ≈90% of patients with lower than expected rifampin levels who were subsequently tested after the fi rst dose adjustment achieved target levels. Dose-titration studies of rifampin confi rm a continuously increasing response of early bactericidal activity by measurement of sputum colony counts with corresponding increase in rifampin dose (16,17). Rifampin has been tolerated at doses as high as 1,200 mg in small studies, and larger trials are ongoing to study high-dose rifampin in an effort to shorten therapeutic duration (18,19). Given increasing evidence that rifampin may be underdosed for many patients regardless of TB outcome (20), the fi ndings of this study suggest that rifampin is a prime medication to prioritize for early TDM for patients for whom TB therapy is failing.
Diabetes was signifi cantly associated with slow response in our study population, and, among persons with a slow response with diabetes, C 2hr levels of rifampin were signifi cantly more likely to be below the expected range. Patients with diabetes are at greater risk for incident TB (21,22) and are more likely to have poor TB treatment outcomes (23,24), which may partially be explained by inadequate pharmacotherapy. A growing body of evidence has 1550 Emerging Infectious Diseases • www.cdc.gov/eid • Vol. 16, No. 10, October 2010 Figure 1. Results of serum concentration 2 hours after medication administration levels (C 2hr ) of fi rst-line antituberculosis medications among patients with a slow response to tuberculosis therapy. Frequencies are reported for low, within target, and high C 2hr levels corresponding to levels below, within, or above the expected range for each medication. demonstrated reduced rifampin exposure in patients with TB and diabetes, which may be in part related to impaired absorption (25,26). Although rifampin absorption may not be blunted by delayed gastric emptying (27), hyperglycemia can decrease gastric hydrochloric acid secretion, which results in a higher gastric pH and reduced rifampin absorption (25). Markers of glycemic control were not available for analysis in this study, but it may be of further use to risk stratify patients with diabetes based on disease severity. It is suspected, though not tested, that drug-drug interactions were also playing a role in the observed lower levels of rifampin in patients with diabetes from our study population. Given preexisting knowledge of the association between diabetes and poor treatment outcome, there may have been a bias on behalf of the TB control staff to characterize a patient with diabetes as a person with a slow response based  Figure 2. Results in patients with initial serum concentrations 2 h after medication below the expected range with follow-up levels after dose adjustment for rifampin daily or biweekly (A), isoniazid daily (B), and isoniazid biweekly (C). The median initial and follow-up doses of rifampin daily or biweekly were 600 mg and 900 mg, respectively; for isoniazid daily, 300 mg and 450 mg, respectively; and for isoniazid biweekly, 900 mg and 1,200 mg, respectively. Brackets represent expected ranges for each dose of medication.
on symptom persistence. Nevertheless, our fi ndings raise the possibility of TB programs studying the benefi t of routine TDM for rifampin among all patients with diabetes at the start of TB therapy. Routine TDM among persons with slow responses at a relatively early point in the treatment course refl ects policy change within Virginia's TB control program. Given the high prevalence of low levels of key medications and the observed ease of correction after dose adjustment, we recommend that similar TB programs investigate the applicability of TDM within their own settings. Further generalization must be cautiously considered, however, because relevant factors that may adversely affect pharmacokinetics, such as the patient's weight at the time of TDM, concurrent medication use, chronic kidney disease, or cirrhosis, were not available in these surveillance data for comparison (28)(29)(30). Other limitations to the study must be taken into account. Blood was collected at 2 hours after medication administration to estimate C max . A second blood collection at 6 hours can additionally distinguish patients whose absorption may be delayed secondary to poor gastric emptying (4); however, the frequency of delayed absorption has been rare in other cohorts for which 2 and 6 hour measurements were performed (10).
Additionally, given that the prevalence of lower than expected drug levels was not known for patients with adequate response to anti-TB therapy, the overall contribution of pharmacotherapy to the cause of slow response in this cohort cannot be fully assessed. Surveillance data did not permit comparison of culture positivity in patients who met criteria for slow response at the time of TDM matched to patients with adequate response for whom TDM was not performed. Use of TDM as early as 4 weeks, as was performed in this study, may have selected for patients that might otherwise have improved after 8 weeks of therapy regardless of other interventions. Lastly, further study may fi nd, given that the cost of TDM (≈$80 US per individual drug) may be substantial for some TB control programs, that it is more economical to start therapy with increased drug dosages for patients at higher risk for slow response.
In summary, routine TDM among patients meeting criteria for slow response to TB therapy in Virginia identifi ed most of those tested to have C 2hr levels of rifampin and isoniazid below the expected range; many patients also had low levels of ethambutol. Most patients with repeat TDM following dose adjustment to rifampin and isoniazid had levels within the expected range, suggesting a clinically actionable result. Given the comparative ease of correcting low rifampin levels and the shorter duration of therapy in those with a correctable rifampin level, this medication is particularly appealing to target for programmatic intervention. Further prospective studies should evaluate the benefi t of routine TDM for rifampin early in the treatment course among patients with diabetes or of higher initial doses of rifampin among all groups at risk for slow response.