Effect of HIV on bedaquiline and delamanid pharmacokinetics in patients with multidrug-resistant TB

SUMMARY

BACKGROUND:

Bedaquiline and delamanid are key drugs in the treatment of multidrug-resistant TB. It is unclear whether HIV affects the pharmacokinetics of bedaquiline and delamanid.

METHODS:

Participants with multidrug-resistant TB were randomised to treatment with bedaquiline, delamanid, or both, plus a standard-of-care treatment. Intensive pharmacokinetic sampling was performed after 8 weeks of therapy. We performed non-compartment analysis to describe bedaquiline and delamanid pharmacokinetics, comparing exposures between HIV-positive and HIV-negative participants using geometric mean ratios (GMRs), and explored covariates associated with bedaquiline and delamanid AUC, using multivariable regression modelling.

RESULTS:

Twenty-six participants were assigned to bedaquiline, 25 received delamanid, and 24 received both drugs. Fifty and 49 participants were treated with bedaquiline and delamanid, respectively. The GMR (90% confidence interval [CI]) of bedaquiline AUC_0–22 and C_max in HIV-positive compared with HIV-negative participants was 0.76 (0.61–0.94) and 0.94 (0.78–1.06), respectively. The GMR (90% CI) of delamanid AUC_0–23 and C_max in HIV-positive compared with HIV-negative participants was 0.93 (0.79–1.09) and 0.79 (0.49–1.27), respectively.

CONCLUSION:

HIV infection was associated with a 24% reduction in bedaquiline AUC_0–22, the significance of which requires further exploration. Delamanid exposure was unaffected by HIV status.

There are contrasting data describing the effect of HIV on drug exposure in the treatment of TB, as some studies have shown that HIV reduces the exposure of first-line anti-TB drugs, while others have shown no effect.¹ Bedaquiline and delamanid are now included in WHO treatment guidelines for multidrug-resistant TB (MDR-TB), classified as Group A and Group C drugs, respectively.² Low exposure to TB drugs has been associated with poor clinical outcomes, including acquisition of drug resistance.^3–5 It is therefore important to explore factors, including HIV, which could affect TB drug exposure, particularly for bedaquiline, which has been shown to significantly reduce mortality when included in MDR-TB treatment regimens.^6,7 While early studies suggest that HIV does not have an effect on bedaquiline and delamanid drug exposures, the high risk associated with sub-therapeutic drug exposures merits further investigation.^8,9 The management of MDR-TB has evolved considerably in the last decade with treatment efficacy associated with bedaquiline-based regimens exceeding 90%.^6,10–13 In South Africa, HIV prevalence estimates among patients with MDR-TB are as high as 75%; this is of concern given that TB and HIV co-infection is associated with poorer treatment outcomes than TB alone.^14–16 Dolutegravir, a key antiretroviral therapy (ART) drug, has considerable advantages over other ART drugs, including a low adverse effect profile and a high genetic barrier to resistance; a once daily dosing schedule also makes dolutegravir an ideal candidate for inclusion in fixed-dose combination ART.¹⁵ Early reports have shown that treating MDR-TB patients co-infected with HIV using bedaquiline, delamanid, and dolutegravir is safe,² although it is currently unknown whether dolutegravir affects the pharmacokinetics of either of these anti-TB drugs.

Bedaquiline is primarily metabolised by cytochrome P450 3A4 (CYP3A4) to its main metabolite M2.¹⁷ Bedaquiline is predominantly protein-bound and has a half-life of 164 days.¹⁸ The key metabolic step in the metabolism of delamanid to its main metabolite DM6705 is by albumin.^19,20 Delamanid is also predominantly protein-bound and has a half-life of 15 h.²⁰ Dolutegravir is predominantly metabolised by uridine-5-diphospho-glucuronosyltransferase of the 1A1 family (UGT1A), and to a lesser extent by CYP3A4.²¹ There is some overlap between the metabolic pathways of bedaquiline and dolutegravir; therefore, possibility of a drug–drug interaction exists. There is no significant interaction between delamanid and dolutegravir.

Nested in a randomised control trial exploring the safety of MDR-TB regimens, we evaluated whether HIV co-infection affects the pharmacokinetics of bedaquiline and delamanid. Considering the high prevalence of HIV-TB co-infection and the rising incidence of TB drug resistance, the dearth of data describing the effect of HIV and dolutegravir on the exposure of key MDR-TB drugs requires urgent attention.

METHODS

Our study is a sub-analysis of the DELamanid BEdaquiline for ResistAnt TubErculosis (DELIBERATE) study, a randomised, controlled trial aimed at evaluating the safety of adding bedaquiline and/or delamanid to standard-of-care MDR-TB regimens (ClinicalTrials.gov Identifier: NCT02583048). MDR-TB adult patients were recruited to DELIBERATE from three sites between August 2016 and July 2018: TASK in Cape Town, South Africa; the South African Tuberculosis Vaccine Initiative cohort in Worcester, South Africa; and Maria Auxiliadora Hospital in Peru.²

Inclusion criteria for DELIBERATE included MDR-TB or rifampicin monoresistance, age ≥18 years, and treatment with a country-specific standard-of-care regimen for <60 days. At the time of recruitment, both South Africa and Peru were using the World Health Organization 2016 MDR-TB treatment guidelines.²² The treatment regimen included a modified 4–6-month intensive phase of high-dose isoniazid, prothionamide/ethionamide, kanamycin/amikacin, levofloxacin, ethambutol, and pyrazinamide followed by a 5–7-month continuation phase including levofloxacin, pyrazinamide, and ethambutol. By randomisation, the background MDR-TB regimens were balanced across the study treatment arms. Eligible participants living with HIV were required to have a CD4 count greater than 100 cells/mm³ and were all treated with a dolutegravir-based ART regimen prior to enrolment. Given that the M2 metabolite of bedaquiline and the DM-6705 metabolite of delamanid both cause QT-prolongation,²³ participants with a QTcF > 450 ms were also excluded.

Participants were randomly assigned (ratio: 1:1:1) to one of the following treatment arms plus standard of care: bedaquiline 400 mg daily for 14 days followed by 200 mg thrice weekly for 22 weeks, delamanid 100 mg twice daily for 24 weeks, or both bedaquiline and delamanid with the same drug-specific dosing schedules. Participants were followed post-treatment initiation for 128 weeks. A participant was regarded to have a poor TB treatment outcome if there was no sputum culture conversion after 24 weeks of study treatment. Intensive pharmacokinetic sampling was scheduled at weeks 2, 8, and 24 post-treatment initiation at the following timepoints: pre-dose, 4, 6, 8, 11, and 23 h. Bedaquiline was dosed 1 h after delamanid administration to avoid a potential drug–drug interaction^19,21; therefore, the sampling timepoints to measure bedaquiline were effectively: pre-dose, 3, 5, 7, 10, and 22 h.

Drug assays

The measurement of bedaquiline and delamanid concentrations was performed using high-performance liquid chromatography-tandem mass spectrometry (LCMS/MS) at the Division of Clinical Pharmacology laboratory, University of Cape Town. The plasma assays for bedaquiline and delamanid have previously been described.^9,24

Pharmacokinetic and statistical analysis

We used non-compartmental pharmacokinetic analysis to determine key pharmacokinetic measures of bedaquiline and delamanid using Stata (Version 17, Stata-Corp, Texas), including area under the concentration time curve to 23 h (AUC_0–23) for delamanid and AUC_0–22 for bedaquiline, peak concentration (C_max), half-life, and AUC extrapolated to infinity (AUC_∞). We selected the 8-week intensive pharmacokinetic samples for inclusion in our analysis, which is beyond the 2-week loading dose of bedaquiline, and likely prior to any significant effect of attrition on sample size. All of the HIV-positive participants in our cohort were treated with a dolutegravir-based ART regimen. We used geometric mean ratios (with 90% confidence intervals [CIs]) to compare key bedaquiline and delamanid pharmacokinetic measures between HIV-positive and HIV-negative participants. We used linear regression modelling to explore covariates associated with a unit change in natural log of bedaquiline and delamanid AUC. As a starting point, we selected the following variables a priori for inclusion in our model: HIV status, age, sex, creatinine clearance, serum albumin, weight, and body mass index (BMI). We tested these variables using backward stepwise regression with likelihood ratios, using a significance level of 0.05 to explore the effect of the selected covariates in the model.

Ethical statement

Participants provided written informed consent prior to study participation. The trial was conducted in accordance with the Declaration of Helsinki and was approved by the respective ethics committee associated with each site.²⁵

RESULTS

The DELIBERATE study enrolled 84 patients with confirmed pulmonary MDR-TB. Randomisation was to one of three arms: Of the 84 participants randomised, 75 had intensive pharmacokinetic sampling at week 8. Of the 75 participants who had pharmacokinetic sampling, 26 were assigned to bedaquiline, 25 to delamanid, and 24 to both bedaquiline and delamanid. Eighteen (36%) of the 50 participants treated with bedaquiline were HIV-positive, 19 (39%) of the 49 participants treated with delamanid were HIV-positive, and 9 (38%) of the 24 participants treated with both bedaquiline and delamanid were HIV-positive. All the participants who had HIV co-infection were on dolutegravir-based ART. Baseline participant characteristics are summarised in Table 1.

Table 1.

Participant characteristics of 75 MDR-TB patients randomised (1:1:1) to treatment with bedaquiline (n = 26), delamanid (n = 25), or both bedaquiline and delamanid (n = 24), plus standard of care.

Variables	Bedaquiline		Delamanid
	HIV-negative, n = 32	HIV-positive, n = 18	HIV-negative, n = 30	HIV-positive, n = 19
Males (%)	28 (80)	14 (70)	24 (73)	16 (76)
Age, years	34 (23–45)	36.5 (28–42.5)	30 (20–41)	39 (30–43)
Race
Black African	10	15	7	14
Mixed race	20	3	20	5
Hispanic	2	0	3	0
Albumin, g/L	35 (29–37)	32 (25.5–37)	37 (32–40)	33 (28–34)
Creatinine clearance, μmol/L	95.3 (75.8–111.6)	85.8 (75.4–97.0)	84.8 (73.6–108.9)	87.5 (65.7–110.2)
Weight, kg	53.9 (50–61.2)	56 (51.5–59.9)	52.5 (50–58.8)	54 (52.4–62)
BMI, kg/m2	19.2 (17.8–21.6)	20.2 (19.0–21.9)	19.1 (18.0–21.1)	21.0 (18.8–22.3)
CD4 count cell/mm3	—	294 (167–572)	—	256 (187–562)
Poor TB treatment outcomes, n	2	1	2	1

Median with interquartile range is shown in brackets, where appropriate.

BMI = body mass index; MDR-TB = multidrug-resistant TB.

The key pharmacokinetic parameters of bedaquiline and delamanid after 8 weeks of therapy stratified by HIV status are shown in Tables 2 and 3, respectively. Since delamanid was administered twice daily with plasma concentrations subsequently still rising in some participants at the last scheduled sampling timepoint, we were unable to generate an accurate half-life for delamanid. The geometric mean ratio (90% CI) of bedaquiline AUC_0–22 and C_max in HIV-positive compared with HIV-negative participants was 0.76 (0.61–0.94) and 0.94 (0.78–1.06), respectively. The geometric mean ratio (90% CI) of delamanid AUC_0–23 and C_max in HIV-positive compared with HIV-negative participants was 0.93 (0.79–1.09) and 0.79 (0.49–1.27), respectively.

Table 2.

Pharmacokinetic measurements of bedaquiline in 50 participants treated with bedaquiline for MDR-TB.

	HIV-negative, n = 32 median (IQR)	HIV-positive, n = 18 median (IQR)	Geometric mean ratio (90% CI)
AUC0–22, mg·h/L	21.0 (16.1–30.7)	16.4 (12.9–20.0)	0.76 (0.61–0.94)
Cmax, mg/L	1.32 (1.00–2.14)	1.30 (0.97–1.46)	1.06 (0.94–1.20)
Tmax, h	6 (6–6)	6 (6–6)	—
T1/2, h	16.4 (12.5–24.5); n = 31	14.5 (11.9–19.6); n = 18	—

Median with interquartile range is shown in brackets, where appropriate.

CI = confidence interval; IQR = interquartile range; MDR-TB = multidrug-resistant TB.

Table 3.

Pharmacokinetic measurements of delamanid after 8 weeks of treatment with delamanid in 49 participants treated for MDR-TB.

	HIV-negative, n = 30 median (IQR)	HIV-positive, n = 19 median (IQR)	Geometric mean ratio (90% CI)
AUC0–23, mg·h/L	5.68 (4.90–6.85)	4.89 (3.86–6.40)	1.08 (0.92–1.27)
Cmax, mg/L	0.35 (0.26–0.44)	0.27 (0.22–0.33)	1.27 (0.79–2.06)
Tmax, h	4 (0–23)	4 (0–4)	—

Median with interquartile range is shown in brackets, where appropriate.

CI = confidence interval; IQR = interquartile range; MDR-TB = multidrug-resistant TB.

The univariate and multivariable linear regression analyses of factors associated with a percentage change in log bedaquiline AUC_0–22 and in log delamanid AUC_0–23 are shown in Tables 4 and 5, respectively. On univariate regression, HIV infection and creatinine clearance (per μmol/L increase) were associated with a reduction in bedaquiline AUC_0–22. HIV co-infection was also associated with lower bedaquiline exposure after adjusting for other covariates in the multivariable model, as shown in Table 4. There were no statistically significant associations between the selected covariates and log delamanid AUC_0–23 (see Table 5).

Table 4.

Mean percentage change in log bedaquiline AUC_0–22 (95% CI), associated with clinical covariates after 8 weeks of treatment with bedaquiline.

Covariate	Univariate		Multivariable
		P value		P value
HIV	−24.7 (−43.1 to −0.3)	0.048	−26.2 (−45.7 to −1.5)	0.040
Age	0.7 (−0.4 to 2.0)	0.305	0.7 (−1.2 to 2.6)	0.458
Sex (referent: male)	33.6 (−2.7 to 83.5)	0.073	12.9 (−23.6 to 66.8)	0.533
Creatinine clearance, μmol/L	−0.5 (−1.1 to −0.1)	0.048	−0.4 (−1.0 to 0.3)	0.251
Albumin, g/L	1.3 (−0.5 to 3.7)	0.247	1.6 (−1.4 to 4.5)	0.282
BMI, kg/m2	−4.4 (−9.3 to 0.9)	0.098	−3.6 (−9.1 to 2.1)	0.201

The change in AUC is based on estimates from the linear regression model of the covariates on the log-transformed AUC and expressed as percentage change in AUC for each unit change in covariate.

BMI = body mass index; CI = confidence interval.

Table 5.

Mean percentage change in log delamanid AUC_0–23 (95% CI) associated with clinical covariates after 8 weeks of treatment with delamanid.

Covariate	Univariate		Multivariable
		P value		P value
HIV	−0.075 (−0.291 to 0.141)	0.487	−0.770 (−0.340 to 0.190)	0.562
Age, years	−0.003 (−0.012 to 0.006)	0.783	−0.002 (−0.017 to 0.0134)	0.783
Sex (referent: male)	0.007 (−0.242 to 0.257)	0.953	0.045 (−0.256 to 0.346)	0.766
Creatinine clearance, μmol/L	0.001 (−0.003 to 0.005)	0.543	0.001 (−0.006 to 0.007)	0.876
Albumin, g/L	0.004 (−0.015 to 0.022)	0.689	−0.002 (−0.027 to 0.024)	0.890
BMI, kg/m2	0.004 (−0.033 to 0.041)	0.851	0.005 (−0.049 to 0.060)	0.851

The change in AUC is based on estimates from the linear regression model of the covariates on the log-transformed AUC, and expressed as percentage change in AUC for each unit change in covariate.

BMI = body mass index; CI = confidence interval.

DISCUSSION

We found that HIV co-infection was associated with a 24.7% reduction in bedaquiline AUC_0–22 in patients treated for MDR-TB. Considering that low exposure of some TB drugs has been associated with poor clinical outcomes, including the development of drug resistance, evidence indicating reduced bedaquiline exposure in HIV-infected patients is concerning.

Bedaquiline has been shown to reduce mortality when combined with standard-of-care treatment regimens for drug-resistant TB,^3–5 and is therefore now recommended as a priority inclusion (Category A) in WHO MDR-TB treatment regimens.²⁶ To date, the effect of HIV on the exposure of key TB drugs, including bedaquiline, has been unclear.¹ Factors previously shown to affect bedaquiline exposure include age, BMI, race, serum albumin concentration, and the presence of the rs776746 cytochrome P450 polymorphism.^27,28 Svensson et al.⁸ showed an association between clinical recovery and bedaquiline clearance, but this was thought to be driven by an increase in weight and serum albumin which occur during clinical recovery.²⁹ We did not observe a significant association between weight, BMI, or serum albumin concentrations and bedaquiline pharmacokinetics, although we may have been underpowered to detect the effect of some of these variables, considering their limited range in our study population (see Table 1). A limited age range may also have affected our ability to detect an age-related association with bedaquiline clearance, which has previously been reported to decrease by approximately 0.9% per year of age increase.²⁵ Similarly, the participants in our study were predominantly black African (45%) or mixed race (52%) – we were therefore unable to adequately explore the effect of race in our analysis. Haas et al.²⁷ reported a 30% decrease in bedaquiline clearance in individuals homozygous for the rs776746 cytochrome P450 polymorphism. The prevalence of the rs77674 polymorphism varies by ethnicity; small pharmacogenetic studies have shown a low prevalence among people of black African ancestry,²⁷ which may explain the lower bedaquiline exposures observed in black African populations.⁸ Only HIV status affected bedaquiline exposure in our cohort although the significance of the approximate 25% reduction we observed in HIV-positive participants is unclear and requires further exploration prior to considering a recommendation for a dose adjustment.

Using population pharmacokinetics, Svensson et al.⁸ previously showed that bedaquiline AUC was unaffected by HIV status. Considering Svensson et al.⁸ used data from two separate cohorts of which only 7% of participants were HIV-positive,^30,31 the investigators may have been underpowered to detect an effect of HIV on bedaquiline exposure. Dolutegravir and bedaquiline have a partially shared metabolic pathway²¹; we therefore cannot exclude the possibility that co-treatment with dolutegravir contributed to the reduced bedaquiline exposure we observed in HIV-positive participants. However, neither drug is known to induce nor inhibit the shared metabolising enzyme CYP3A4, reducing the likelihood of a drug–drug interaction. Svensson et al.³², using population PK modelling, previously showed a 52% reduction in bedaquiline AUC_0–24 in HIV co-infected patients treated with efavirenz, but this reduction in exposure was likely due to a drug–drug interaction between bedaquiline and efavirenz (considering their shared CYP3A4 metabolic pathway).

Apart from HIV, none of the covariates we explored affected bedaquiline exposure in our multivariable model. Delamanid AUC_0–23 was not affected by any of our selected covariates, including HIV, which is consistent with previous reports that factors such as sex, age, and weight do not affect delamanid exposure.^9,19 Our finding that HIV infection was not associated with delamanid exposure was also reported in a population pharmacokinetic model from the DELIBERATE study.⁹

Our study has limitations. First, all of the HIV-positive participants in our cohort were treated with dolutegravir, we therefore cannot be certain that the effect on bedaquiline exposure, which we observed, was due to HIV co-infection or a drug–drug interaction with dolutegravir. However, the latter is extremely unlikely as dolutegravir is neither an inhibitor nor an inducer of CYP3A4, which is the enzyme that metabolises bedaquiline. Second, we used non-compartment analysis to calculate the pharmacokinetic parameters of bedaquiline 8 weeks after treatment initiation, which is not the optimal technique for long half-life drugs like bedaquiline, which accumulate slowly and are best described using population pharmacokinetic modelling. Further to the limitations of NCA, bedaquiline has a half-life of 164 days, and BDQ AUC_0–22 at 8 weeks of therapy is therefore unlikely to completely charactarise the exposure of bedaquiline.²⁹ Third, DELIBERATE only recruited participants with pulmonary TB – our findings are therefore not necessarily applicable to patients with extrapulmonary TB. Fourth, our study did not capture the following clinical data: viral load, duration of ART, and presence of comorbidities including diarrhoea which could affect bedaquiline absorption. However, at the time of pharmacokinetic sampling, participants had been on ART for a minimum of 8 weeks, at which point it is reasonable to assume that most participants would be stable on ART. Fifth, only three participants treated with bedaquiline had poor TB treatment outcomes of which only one was HIV-positive, and our study was therefore underpowered to evaluate whether the reduction in bedaquiline exposure, which we observed in HIV-positive participants, was significant enough to affect treatment efficacy.

CONCLUSION

We found that HIV reduces the exposure of bedaquiline, the significance of which is currently uncertain and further research is necessary to confirm this finding. Delamanid exposure was unaffected by HIV status.