Early antituberculosis drug exposure in hospitalized patients with human immunodeficiency virus‐associated tuberculosis

Aims Patients hospitalized at the time of human immunodeficiency virus‐associated tuberculosis (HIV‐TB) diagnosis have high early mortality. We hypothesized that compared to outpatients, there would be lower anti‐TB drug exposure in hospitalized HIV‐TB patients, and amongst hospitalized patients exposure would be lower in patients who die or have high lactate (a sepsis marker). Methods We performed pharmacokinetic sampling in hospitalized HIV‐TB patients and outpatients. Plasma rifampicin, isoniazid and pyrazinamide concentrations were measured in samples collected predose and at 1, 2.5, 4, 6 and 8 hours on the third day of standard anti‐TB therapy. Twelve‐week mortality was ascertained for inpatients. Noncompartmental pharmacokinetic analysis was performed. Results Pharmacokinetic data were collected in 59 hospitalized HIV‐TB patients and 48 outpatients. Inpatient 12‐week mortality was 11/59 (19%). Rifampicin, isoniazid and pyrazinamide exposure was similar between hospitalized and outpatients (maximum concentration [Cmax]: 7.4 vs 8.3 μg mL–1, P = .223; 3.6 vs 3.5 μg mL–1, P = .569; 50.1 vs 46.8 μg mL–1, P = .081; area under the concentration–time curve from 0 to 8 hours: 41.0 vs 43.8 mg h L–1, P = 0.290; 13.5 vs 12.4 mg h L–1, P = .630; 316.5 vs 292.2 mg h L–1, P = .164, respectively) and not lower in inpatients who died. Rifampicin and isoniazid Cmax were below recommended ranges in 61% and 39% of inpatients and 44% and 35% of outpatients. Rifampicin exposure was higher in patients with lactate >2.2 mmol L–1. Conclusion Mortality in hospitalized HIV‐TB patients was high. Early anti‐TB drug exposure was similar to outpatients and not lower in inpatients who died. Rifampicin and isoniazid Cmax were suboptimal in 61% and 39% of inpatients and rifampicin exposure was higher in patients with high lactate. Treatment strategies need to be optimized to improve survival.


| INTRODUCTION
Tuberculosis (TB) is the leading cause of hospitalization and inhospital death in human immunodeficiency virus (HIV)-infected people worldwide. 1,2 In high-burden settings hospitalized patients with HIV-associated TB (HIV-TB) have case fatality rates between 11 and 32%. [3][4][5][6][7][8] The majority of these deaths occur within 2 weeks [3][4][5]8 and in postmortem series inpatient HIV-TB deaths are reported at a median of 4-5 days after admission 9,10 , with 50% of deaths occurring in patients already on anti-TB therapy. 11 Severe HIV-TB may present with clinical features of bacterial sepsis. [12][13][14] In high-burden settings Mycobacterium tuberculosis bloodstream infection is the most common diagnosis in HIV-infected patients presenting to hospital with a clinical syndrome of sepsis. [15][16][17][18] Analogous to sepsis, there are many factors in severe HIV-TB that could reduce drug exposure, such as impaired absorption of orally administered drugs due to delayed gastric emptying and decreased perfusion of the gastrointestinal tract, increased volume of distribution due to fluid shifts, and augmented renal clearance. 19,20 Other factors in advanced HIV infection such as intestinal TB, HIV-related enteropathy, and gastrointestinal opportunistic infections and macroor micronutrient deficiencies [21][22][23] could contribute to reduced drug exposure. Limited existing data suggest that anti-TB drug exposure in critically ill patients is inadequate. 24 Elevated blood lactate is used as a marker of sepsis severity 25 and is associated with mortality in hospitalized patients with HIV-TB. 5 HIV infection has a variable effect on anti-TB drug concentrations across studies, with some studies showing lower concentrations than in HIV-negative patients. [26][27][28] There are few pharmacokinetic studies in HIV-TB that assess relationships between drug exposure and clinical outcomes. [29][30][31] Rifampicin is a potent inducer of drug metabolizing liver enzymes 32 and also undergoes auto-induction. 33 The majority of rifampicin pharmacokinetic studies have been performed after administration of multiple doses when autoinduction is advanced, 27 yet mortality in hospitalized HIV-TB patients occurs early. In the parent cohort of this pharmacokinetic (PK) study 37% of deaths occurred within 7 days of enrolment. 34 Preliminary evidence suggest that higher-dose than the currently recommended 10 mg kg -1 daily may improve survival in HIV-TB patients with low CD4 counts. 35 We performed intensive PK studies on the third day of anti-TB therapy, administered at standard doses, in hospitalized patients with HIV-TB and outpatient controls and determined 12-week mortality in hospitalized patients. We compared exposure of rifampicin, isoniazid and pyrazinamide between inpatients and outpatients, between inpatients who survived and those who died within 12 weeks, and between inpatients presenting with an elevated lactate (a marker of sepsis severity) and those presenting with normal lactate. We hypothesized that exposure to rifampicin, isoniazid and pyrazinamide would be lower in inpatients than outpatients; lower in inpatients who died within 12 weeks compared to survivors, and lower in inpatients presenting with elevated venous lactate compared to those presenting with normal lactate.

| Study design and study population
We enrolled hospitalized HIV-infected adults with a CD4 count of • Deaths occur early and there is paucity of data regarding antitubercular drug exposure in hospitalized critically ill HIV-TB patients.

What this study adds
• Rifampicin, isoniazid and pyrazinamide exposure in hospitalized HIV-TB patients and outpatients on day 3 of standard treatment are described.
• Hospitalized HIV-TB patients do not have lower exposure than outpatients; however, many have suboptimal concentrations, which could play a role in mortality.
• This could inform treatment strategies in hospitalized HIV-TB patients.
18 years or older, with a suspected new diagnosis of TB were enrolled at presentation to hospital and PK studies were performed in a subgroup within the routine hospital service on the third day of anti-TB therapy. Patients who survived to the third day of TB treatment were enrolled sequentially for PK studies, provided they still required inpatient care, did not require transfer to a tertiary care facility for intensive care or investigations and there were adequate staff to fulfil the parent study's operational requirements and perform PK study.
Outpatients were enrolled at treatment initiation and returned for PK studies on the third day of therapy. Patients were enrolled regardless of antiretroviral therapy status or type. Outpatients were HIV-infected or HIV-uninfected. Clinical data and baseline blood tests were obtained at enrolment. Twelve-week vital status was ascertained for inpatients. Participants were fasted overnight and were offered a standardized breakfast after the 1-hour sample and a standardized lunch between the 4-and 6-hour samples. The study team administered the third dose of anti-TB therapy and collected samples immediately before (0 h) and at 1, 2.5, 4, 6 and 8 hours after the dose. Timing of samples were calculated from the time the dose was administered and all samples were collected within a 10-minute window (±5 min). A cold chain was maintained by placing blood samples in crushed ice immediately after collection, spinning in a cold centrifuge (8 C) and flash freezing plasma aliquots in dry ice within 30 minutes of collection.

| Anti-TB therapy and PK study methods
Plasma aliquots were transported and stored in a -80 C freezer at the end of each day.
Rifampicin, isoniazid and pyrazinamide concentrations were measured on stored plasma using high-performance liquid chromatography coupled to tandem mass spectrometry at the Division of Clinical Pharmacology Laboratory, University of Cape Town. The combined accuracy and precision statistics of the low-, medium-, and highquality control samples during analysis (n = 22) of the rifampicin assay were between 99.7% and 100.8%, and 4.7% and 7.7%, respectively.
The combined accuracy and precision statistics of the low-, mediumand high-quality control samples during analysis (n = 22) of the isoniazid assay were between 98.3% and 100.4%, and 3.0% and 5.1%, respectively. The combined accuracy and precision statistics of the low-, medium-and high-quality control samples during analysis (n = 22) of the pyrazinamide assay were between 88.1% and 92.3%, and 2.9% and 3.6%, respectively. Baseline blood tests including venous lactate measurements were performed at the National Health Laboratory Services.

| Ethical approval
The study was approved by the University of Cape Town Human Research Ethics Committee (UCT HREC reference: 057/2013) and written informed consent was obtained for the PK substudy. Eligible inpatients with a decreased level of consciousness were enrolled and followed up daily until they regained capacity to participate in the informed consent process. Permission was sought from the UCT HREC to use information of participants who died prior to providing informed consent.  (Table S5) and this group was not disaggregated for any of the other analyses. Lactate was also compared to PK variables as a continuous variable. Correlations were performed on log or square root transformed variables using Pearson's correlation test or Spearman's rank correlation where appropriate. In hospitalized patients we calculated the odds ratio for survival per doubling of lactate concentration using a logistic regression model and log2 transformed lactate concentration. We did not adjust for other clinical variables. We performed correlation tests (Pearson or Spearman's correlation) to assess relationships between PK variables, creatinine clearance and conjugated bilirubin, and pyrazinamide exposure with 2 inflammatory markers (C-reactive protein and procalcitonin). Concentrations below the lower limit of quantification (LLQ) were imputed at half the value of the LLQ. Missing concentrations were imputed using the slope of the relevant drug's log concentration curve for the patient when possible (Tables S1 and S2). The LLQ for rifampicin, isoniazid and pyrazinamide was 0.117, 0.105 and 0.203 μg mL -1 , respectively.

| Statistical analysis
Drug concentrations were log-transformed and the geometric mean was calculated by exponentiating the mean of the logtransformed values. We used published reference ranges of drug concentrations that can be expected after administration of standard doses of anti-TB therapy for comparison for comparison of our C max results (8-24 μg mL -1 for rifampicin, 3-6 μg mL -1 for isoniazid and 20-60 μg mL -1 for pyrazinamide). [31][32][33] 3 | RESULTS

| Outcomes of the parent study and baseline characteristics
The parent study enrolled 576 hospitalized patients with HIV-TB and the 12-week mortality was 124/576 (22%) at a median of 12.5 days from enrolment. 34 Intensive PK studies were performed in a subgroup of 60 inpatients and in 48 outpatients with TB. One inpatient was excluded due to a high CD4 count and an alternative diagnosis of mycetoma. We analysed data from 59 inpatients and 48 outpatients ( Figure 1). Outpatients included 19/48 (40%) HIV-uninfected patients. The median CD4 counts for inpatients and HIV-infected outpatients were 58 and 146 cells μL -1 , respectively (Table 1). Twelve-week mortality for inpatients was 11/59 (19%) with median days from PK study to death = 40 days (interquartile range = 8-60 days). One inpatient was lost to follow up after 2 months.
On baseline blood tests there were significant differences between inpatients and outpatient controls, including significantly lower CD4 count, haemoglobin, creatinine clearance and albumin, and significantly higher liver enzymes and C-reactive protein in inpatients (Table 1). Inpatients and outpatients received similar doses (mg kg -1 ) of rifampicin, isoniazid and pyrazinamide and there was a similar distribution of patients in different weight categories (Table 1). There were fewer differences between hospitalized patients who died and those who survived 12 weeks of follow up (Table S3).

| C max
Comparing hospitalized patients to outpatients, neither the median rifampicin C max (7.4 vs 8.3 μg mL -1 , P = .223), nor the median isoniazid C max (3.6 vs 3.5 μg mL -1 , P = .569) were significantly different. The median pyrazinamide C max in hospitalized patients was higher than outpatients (50.1 vs 46.8 μg mL -1 , P = .081) but this did not reach statistical significance (Table 2 and Figure 2). Rifampicin C max was below the minimum threshold of the reference range of 8 μg mL -1 in 36/59 F I G U R E 1 Study flow chart: hospitalized human immunodeficiency virus (HIV)-infected adults with a CD4 count of ≤350 cells μL -1 starting tuberculosis treatment in hospital and ambulant outpatients (HIV-infected and uninfected) were enrolled for intensive pharmacokinetic (PK) studies. Inpatients were enrolled at presentation and PK studies were performed within the routine hospital service on the third day of antituberculosis therapy, provided they still required inpatient care and did not need transfer for intensive care. Outpatients were enrolled at treatment initiation and returned for PK studies on the third day of therapy. Twelve-week mortality was ascertained for inpatients. *Exclusions are only listed if participants had consented to take part in the study and PK study could not be performed. We did not document all patients who qualified to take part in the PK study and could not be included due to logistical reasons such as early deaths, transfers to tertiary facilities and staff availability T A B L E 1 Baseline characteristics of outpatient controls with TB and hospitalized patients with HIV-TB who had intensive pharmacokinetic studies performed on the third day of anti-TB therapy Comparing hospitalized patients who survived to those who died within 12 weeks, there were no significant differences in the median C max for rifampicin (7.2 vs 7.5 μg mL -1 , P = .655), isoniazid (3.9 vs 3.2 μg mL -1 , P = .394) or pyrazinamide (48.0 vs 55.1 μg mL -1 , P = .302; Table 3 and Figure 2 Figure 2 and Table 3).

| Patients presenting with an elevated lactate concentration
Lactate concentration was positively correlated with random glucose and conjugated bilirubin concentrations (Table S4)   C max and AUC 0-8 (Table 4). These findings are contrary to our hypothesis that patients presenting with elevated lactate would have lower exposure to TB drugs.

| Associations of PK findings with selected clinical variables
Based on findings from previous studies and the physicochemical properties of the drugs we measured, we next performed an exploratory analysis to assess the relationship of selected clinical variables with our findings. In hospitalized patients we explored the correlations of PK variables with creatinine clearance and con-

| DISCUSSION
We measured concentrations of rifampicin, isoniazid and pyrazinamide on the third day of anti-TB therapy in hospitalized adults with a new diagnosis of HIV-TB and in outpatient controls. We found high 12-week mortality of 19% for inpatients and no significant difference in C max or AUC 0-8 of rifampicin, isoniazid or pyrazinamide between hospitalized patients and outpatients, or between hospitalized patients who survived and those who died. Rifampicin and isoniazid peak concentrations were below reference ranges in 61% and 39% of inpatients and 44% and 35% of outpatients. All patients attained pyrazinamide concentrations within the reference range. We found significantly higher rifampicin C max and AUC 0-8 amongst patients presenting with elevated venous lactate, taken as a marker of sepsis severity.
We observed high 12-week mortality despite treatment and patients died at a median of 40 days after the PK study. This time to death is longer than the median days to death in the main study, F I G U R E 2 Rifampicin, isoniazid and pyrazinamide peak concentrations (C max ) on the third day of antituberculosis therapy: boxplots of rifampicin, isoniazid and pyrazinamide maximum concentrations are presented in μg mL -1 for outpatients (green), hospitalized patients who survived 12-week follow up (blue) and hospitalized patients who died within 12 weeks (black). P value: Kruskal-Wallis test comparing C max values across 3 groups. Dashed horizontal lines represent the minimum threshold of the reference range: 8 μg mL -1 for rifampicin, 3 μg mL -1 for isoniazid and 20 μg mL -1 for pyrazinamide. C max : maximum (peak) concentration; Outpatients: ambulant tuberculosis patients attending outpatient clinic for treatment; Hosp.Survivors: hospitalized survivors; Hosp.Deaths: hospitalized patients who died within 12 weeks of enrolment which was 12.5 days from enrolment. 34 This PK study was performed within the routine clinical service. Critically ill patients requiring intensive care were transferred to a tertiary facility or died and stable patients were often discharged before the third day of anti-TB therapy and could thus not be included in the PK study.
A large proportion of all patients had suboptimal rifampicin and isoniazid peak concentrations. Low concentrations of anti-TB medications have been reported in other studies 29,42 and low exposure to pyrazinamide in particular have been associated with poor clinical outcomes. One study conducted intensive PK studies at 2 months on treatment and monitored 2-year outcomes in South African pulmonary TB patients. 31 They used classification and regression tree analysis, which identified pyrazinamide AUC 0-24 < 363 mg h L -1 as the highest-ranking factor associated with poor 2-year outcomes (relapse, death or therapy failure). A predominantly HIV-infected pulmonary TB cohort from Botswana had PK studies performed after at least 7 days on treatment and were followed for the duration of treatment. Lower peak concentrations of pyrazinamide (<35 μg mL -1 ) was the only PK variable associated with poor outcome and was associated with 3-fold increased risk of poor outcome. 29 In our cohort, pyrazinamide Cmax was <35 μg mL -1 in 6 patients (5 inpatients who survived and 1 outpatient) and there was a trend towards higher exposure in hospitalized patients. One potential mechanism for a trend towards higher pyrazinamide AUC 0-8 in inpatients who died is impaired renal clearance due to acute kidney injury. Pyrazinamide and its main metabolite pyrazinoic acid are excreted in the urine 43 and, although hospitalized patients and specifically inpatients who died had higher creatinine, we observed no significant correlation between pyrazinamide exposure and creatinine clearance. Pyrazinamide clearance was shown to be inversely correlated to chronic cellular immune activation in HIV-TB patients in Botswana. 44 We did not measure human leucocyte antigen-DR expression on CD8 T cells in our study and, even though hospitalized patients and specifically patients who died had higher C-reactive protein and procalcitonin, there was no significant correlation between pyrazinamide exposure and either of these markers.
In a previous study, optimal early bactericidal activity was associated with an isoniazid C max and AUC 0-∞ of >2. 19   We found significantly higher C max and AUC 0-8 for rifampicin in inpatients presenting with elevated lactate, which is contrary to our hypothesis, and a positive correlation of rifampicin C max and AUC 0-8 with conjugated bilirubin. Lactate is used as a marker of sepsis severity and probably reflects increased aerobic glycolysis on cellular level due to adrenergic stimulation 49  and but also in urine. 32 High pretreatment bilirubin levels in patients with advanced liver cirrhosis are associated with higher rifampicin exposure. 53 It is possible that the cellular metabolic changes which underly higher lactate concentrations could play a role in the higher rifampicin exposure we observed in these patients. Neither of the hydrophylic drugs (isoniazid or pyrazinamide) were correlated with creatinine clearance, but rifampicin exposure was positively correlated with creatinine clearance. The mechanism for this is unclear. We did not perform genotyping to assess patients' isoniazid acetylator status. Potential unmeasured differences in distribution of acetylator status across the comparator groups may have biased our analysis of the PK of isoniazid. The associations of PK variables with selected clinical variables could be due to other underlying mechanisms than the potential mechanisms we explored.
In conclusion, rifampicin and isoniazid peak concentrations were below reference ranges in 62% and 39% of hospitalized patients with HIV-TB, respectively. Isoniazid peak concentration and exposure were below the levels associated with optimal early bactericidal activity in The funders had no role in the study design, data collection, data analysis, data interpretation or writing of this report. The opinions, findings and conclusions expressed in this manuscript reflect those of the authors alone.

COMPETING INTERESTS
There are no competing interests to declare.  Low C max 0 (0.0) 0 (0.0) ---Lactate is used as a marker of sepsis severity and we divided patients into those presenting with high lactate (>2.2 mmol L -1 , n = 16) and those presenting with normal lactate (n = 41) concentration. One patient who survived had no lactate performed and is not included in this table.
Pharmacokinetic parameters were compared between groups using a nonparametric comparison (Wilcoxon rank sum test) for the numerical values or the Pearson's χ 2 test for categorical variables. In addition, lactate was treated as a continuous variable and correlation tests (Spearman's rank correlation (distribution not normal) or Pearson's correlation test (normal distribution)) were used to correlate lactate concentrations with PK variables. PK: pharmacokinetic a AUC: area under the time-concentration curve from 0 to 8 hours in mg h L -1 : median and interquartile range. b AUC: area under the time-concentration curve from 0to 8 hours in mg h L -1 : geometric mean and geometric standard deviation (approximate coefficient of variation). C max : maximum concentration in μg mL -1 : median and interquartile range. Low C max : number and percentage of patients with maximum concentrations below minimum threshold of reference ranges: 8 μg mL -1 for rifampicin, 3 μg mL -1 for isoniazid and 20 μg mL -1 for pyrazinamide. 31 Correlation coefficient: Spearman's ρ or Pearson's correlation coefficient. measurement of drug concentrations. C.S. curated drug concentration results and analysed data with assistance of D.B. and M.C. C.S. wrote the manuscript and all coauthors reviewed and contributed to the manuscript.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are openly available in