Efavirenz concentrations in CSF exceed IC₅₀ for wild-type HIV

Abstract

Objectives

HIV-associated neurocognitive disorders remain common despite use of potent antiretroviral therapy (ART). Ongoing viral replication due to poor distribution of antivirals into the CNS may increase risk for HIV-associated neurocognitive disorders. This study's objective was to determine penetration of a commonly prescribed antiretroviral drug, efavirenz, into CSF.

Methods

CHARTER is an ongoing, North American, multicentre, observational study to determine the effects of ART on HIV-associated neurological disease. Single random plasma and CSF samples were drawn within 1 h of each other from subjects taking efavirenz between September 2003 and July 2007. Samples were assayed by HPLC or HPLC/mass spectrometry with detection limits of 39 ng/mL (plasma) and <0.1 ng/mL (CSF).

Results

Eighty participants (age 44 ± 8 years; 79 ± 15 kg; 20 females) had samples drawn 12.5 ± 5.4 h post-dose. The median efavirenz concentrations after a median of 7 months [interquartile range (IQR) 2–17] of therapy were 2145 ng/mL in plasma (IQR 1384–4423) and 13.9 ng/mL in CSF (IQR 4.1–21.2). The CSF/plasma concentration ratio from paired samples drawn within 1 h of each other was 0.005 (IQR 0.0026–0.0076; n = 69). The CSF/IC₅₀ ratio was 26 (IQR 8–41) using the published IC₅₀ for wild-type HIV (0.51 ng/mL). Two CSF samples had concentrations below the efavirenz IC₅₀ for wild-type HIV.

Conclusions

Efavirenz concentrations in the CSF are only 0.5% of plasma concentrations but exceed the wild-type IC₅₀ in nearly all individuals. Since CSF drug concentrations reflect those in brain interstitial fluids, efavirenz reaches therapeutic concentrations in brain tissue.

Introduction

Combining non-nucleoside reverse transcriptase inhibitors (NNRTIs) or protease inhibitors (PIs) with nucleoside reverse transcriptase inhibitors (NRTIs) can dramatically reduce HIV replication, preserve immune function and prolong survival. Antiretrovirals also reduce HIV replication in the CNS, contributing to observed decreases in HIV-associated neurological complications. However, the prevalence of HIV-associated neurocognitive disorders remains high for reasons that are multifactorial.¹

CSF measurements are often used as surrogates for measuring substances in the brain. Some antiretrovirals penetrate the CSF poorly, and low drug concentrations in this compartment may allow ongoing HIV replication, development of viral resistance, local tissue injury and treatment failure in spite of suppression of HIV replication in the blood.² PIs, which are substrates of the P-glycoprotein efflux pump, are highly protein bound and are large molecules, have much lower CSF than blood concentrations. In contrast, several NRTIs cross the blood–CSF barrier,³ perhaps because they are smaller, are not P-glycoprotein substrates and are not significantly protein bound. The NNRTI nevirapine is also a small molecule with less protein binding than PIs, and reaches effective CSF concentrations.³

The widely prescribed NNRTI efavirenz is one of the cornerstones of HIV treatment. Similar to nevirapine and NRTIs, efavirenz is a small molecule and is not a P-glycoprotein substrate; however, it is highly bound (>99.5%) to plasma proteins. Importantly, only unbound drug crosses the blood–CSF barrier. Two small studies have reported efavirenz CSF concentrations. Tashima et al.⁴ found mean CSF efavirenz concentrations of 11.1 ng/mL (range 2.1–18.6 ng/mL) in nine patients. The CSF-to-plasma ratio was 0.61% (0.26%–0.99%). Antinori et al.⁵ found undetectable efavirenz concentrations in the CSF of 11 patients (the assay limit of detection was not provided), with resulting CSF-to-plasma ratios of 0. Despite these data, the frequency of CNS side effects with efavirenz use clearly supports some degree of brain penetration. The objective of this study was to expand the available data on the penetration of efavirenz into the CSF.

Methods

Subjects

Subjects were enrolled in an NIH-funded, six-centre, prospective, observational, US cohort study, CNS HIV Anti-Retroviral Therapy Effects Research (CHARTER), to determine the effects of potent antiretroviral therapy on HIV-associated neurological disease. Written informed consent was obtained from all participants, and the research was conducted in accordance with the Declaration of Helsinki, national and institutional standards. Single plasma and CSF samples were collected for research purposes at biannual study visits between September 2003 and July 2007. Included in this analysis were subjects who: (i) reported taking 600 mg of efavirenz daily for at least 2weeks as part of their combination regimen; (ii) reported by validated adherence questionnaire at least 95% adherence in the 4days prior to assessment; and (iii) were sampled within 24 h of their most recent dose.

To obtain a robust plasma efavirenz pharmacokinetic model, this dataset was enriched with a complementary dataset containing 186 plasma efavirenz concentrations from 55 subjects from the CCTG578 therapeutic drug monitoring study. A detailed description of CCTG578 has already been published.⁶ In brief, these 55 subjects had three samples drawn 2 h apart ∼12 h post-dose at week 2 of therapy; no CSF efavirenz concentrations were measured in CCTG578.

Measurements

Efavirenz plasma samples were assayed by HPLC, with a lower limit of quantification of 39 ng/mL. The efavirenz CSF samples were measured by HPLC coupled to mass spectrometry (HPLC-MS). To enhance sensitivity, CSF samples were extracted with a highly non-polar solvent [methyl tert-butyl ether (MTBE)], which allowed detection levels more than one order of magnitude lower than those found without this step (detection limit of <0.1 ng/mL with extraction versus 5–10 ng/mL without extraction).

Analyses

Data were first summarized with descriptive statistics. Spearman's correlation measured the association between plasma and CSF efavirenz concentrations. Population pharmacokinetic parameters were estimated using NONMEM, version VI (Icon, Dublin, Ireland), using first-order conditional estimation with interaction. A two-compartment physiological model (ADVAN4 TRANS4) with first-order absorption and elimination provided estimates of efavirenz plasma pharmacokinetics and CSF penetration. The absorption rate (k_a) was fixed at 0.287 h⁻¹, and CSF volume was fixed at 140 mL (approximate physiological CSF volume).⁷ For plasma concentrations below detection, half of the quantification limit (19.5 ng/mL) was used. Confidence intervals were determined by bootstrapping (a resampling procedure that created 100 datasets by random sampling with replacement from the original dataset).

Results

Eighty CHARTER subjects who met eligibility criteria were selected and were taking efavirenz for a median [interquartile range (IQR)] of 7 (2–17) months. Plasma/CSF sample pairs were drawn within a median of 26 min of each other (IQR 19–38) at an average of 12.5 ± 5.4 h post-dose. Subjects were mostly male [60/80 (75%)], averaged 44 ± 8 years of age and weighed a mean of 79 ± 15 kg. The median (IQR) plasma HIV-RNA and CD4 cell counts were <50 (<50–268) copies/mL and 472 (241–633) cells/mm³, respectively. CSF HIV-RNA levels were suppressed to <50 copies/mL in 67 of 79 subjects (85%) and plasma levels were suppressed in 52 of 80 (65%). Fifty-three subjects took concomitant tenofovir, 63 were taking lamivudine or emtricitabine and 9 took a PI.

Median plasma and CSF efavirenz concentrations were 2145 and 13.9 ng/mL, respectively (Table 1). The median CSF-to-plasma ratio was 0.005 (i.e. 0.5%). Eleven plasma samples were <39 ng/mL. All CSF samples were measurable. For the 11 subjects with undetectable plasma concentrations, the median CSF concentration was 3.4 ng/mL (IQR 1.8–7.2). Plasma concentrations correlated with CSF concentrations moderately (r_s= 0.55). The median CSF-to-50% inhibitory efavirenz concentration (IC₅₀ for wild-type = 0.51 ng/mL)⁸ ratio was 26 (IQR 8–41). Two CSF concentrations (2.5%) were below the IC₅₀ for wild-type. The corresponding plasma concentrations for these two subjects were 1183 ng/mL and undetectable.

Table 1.

Efavirenz concentrations

	Plasma	CSF	CSF/plasma ratio
Number of samples	80	80	69a
Efavirenz concentration (ng/mL)
median	2145	13.9	0.005
IQR	1384–4423	4.1–21.2	0.0026–0.0076
range	<39–12 249	0.2–51.8	0.0003–0.0275

^aCSF/plasma ratios are reported only for subjects with efavirenz above the detection limit in both plasma and CSF samples.

From the population model, oral clearance and volume of distribution were 11.2 L/h and 318 L (Table 2), similar to published estimates. The median (range) efavirenz plasma concentrations from the 55 CCTG578 subjects [2480 (150–6932) ng/mL] were similar to those of the CHARTER cohort. Age (40 ± 8 years) and sex (81% male) were also similar. Significant intersubject variation for oral clearance was observed (81%). The residual variation for plasma concentration was 52%, while the residual variation in CSF concentration, 117%, was greater. The modelled estimate of efavirenz CSF penetration was 0.48%, meaning that CSF efavirenz concentrations and area under the time–concentration curves (AUCs) are <1% of the corresponding plasma concentrations and AUCs (Figure 1a and b).

Table 2.

Efavirenz population pharmacokinetic estimates

Population parameter	Estimate
n	135 plasma/80 CSF
V/F, apparent volume of distribution (L)	318 (279–357)a
CL/F, oral clearance (L/kg/h)	0.151 (0.149–0.153)a
CSF penetration (% of plasma concentration)	0.0048 (0.0047–0.0049)a
Intersubject variability of CL/F (%)	81 (79–83)a
Residual variability of plasma concentrations (%)	52 (51–53)a
Residual variability of CSF concentrations (%)	117 (109–124)a

^aExpressed as population estimate (95% confidence interval).

Figure 1.

(a) Measurable efavirenz plasma (filled circles) and CSF (open circles) concentrations on a log scale as a function of time after dose. Lines show the model-predicted plasma and CSF concentrations over time in the population. (b) The CSF/plasma efavirenz concentration ratio over time (filled circles), with a linear regression line (r²= 0.018, P = 0.289).

Discussion

In this large cohort of subjects with measurable efavirenz in CSF and plasma samples, modelled plasma pharmacokinetic parameter estimates were similar to data published previously. Efavirenz CSF concentrations were 0.5% of those in plasma, confirming the findings of Tashima et al.⁴ This is in contrast to Antinori et al.,⁵ who found no measurable CSF efavirenz. Given the very low concentrations of efavirenz in CSF, a likely explanation is that the limit of detection of their assay was too high to detect CSF efavirenz.

The estimated free plasma efavirenz concentration in our study (0.5% of 2145 ng/mL) is 10.7 ng/mL, similar to the observed median CSF concentration of 13.9 ng/mL. This suggests that unbound efavirenz is freely distributed into the CSF by passive diffusion, and supports the lack of P-glycoprotein efflux from the CSF. The correlation of CSF and plasma efavirenz concentrations suggests that increasing plasma efavirenz exposure might increase CSF exposure, although the correlation was modest overall.

Because this was a study of subjects already taking antiretrovirals, data about efavirenz CSF penetration and corresponding decreases in CSF viral load at therapy initiation were not available. Another limitation of this study was the need to include subjects from another study to develop a robust plasma pharmacokinetic model. By adding CCTG578 subjects, we established a stronger plasma pharmacokinetic model, which allowed us to develop a second compartment and estimate efavirenz CSF penetration in the model. The two cohorts were sufficiently similar in demographics and efavirenz plasma concentrations to assume that plasma pharmacokinetics would be similar. Furthermore, the main objective of this analysis was to determine efavirenz CSF penetration, and that parameter estimate was based on the 80 subjects from the CHARTER cohort who had CSF concentrations available.

While the observed efavirenz penetration of only 0.5% into the CSF is extremely low, the CSF concentrations are still 26 times the estimated efavirenz IC₅₀ for wild-type HIV. The exact IC₅₀ in the CSF has not been directly measured. Normal CSF has low amounts of binding proteins, like albumin. In a protein-free medium, the 90%–95% effective efavirenz concentration (EC_90–95) is 0.1–1.6 ng/mL.⁹ Because CSF does contain some protein, and efavirenz binds readily to protein, the true EC_90–95 in CSF is probably somewhat higher than these values. Even so, the majority (>95%) of subjects had CSF efavirenz concentrations well above these estimated effective concentrations. While CSF efavirenz concentrations may not reflect those in the brain with complete accuracy, we believe that CSF concentrations are a reasonable surrogate marker for intracerebral concentrations. Thus, efavirenz may contribute substantially to the control of wild-type HIV replication in the CNS.

Funding

This work was supported primarily by the National Institute of Mental Health (NIMH) and the National Institute of Neurological Disorders and Stroke (NINDS) at the NIH (contract number N01 MH22005). Additional support was provided by funds from the National Institute for Child Health and Human Development (5U10 HD031318 to B. M. B., E. V. C. and S. S. R.).

Transparency declarations

B. M. B. has received past research support from Abbott Laboratories and ARK Diagnostics. P. P. K. in the past has received research grants from GlaxoSmithKline, MSD and BMS. E. V. C. has consulted or served on advisory committees for Bristol-Myers Squibb, Johnson & Johnson and Trius. S. S. R. has current or past research support from Medtronic Inc. and Raptor Pharmaceuticals. He and an immediate family member own stock in Merck & Co. D. B. C. has advisory board membership with Genzyme, Genentech, Pfizer and Millennium. He has received consultancy fees from Biogen Idec, Genentech, Pfizer, Bristol-Myers Squibb and Millennium (all <$10 000 per year). He has received a speaker fee from GlaxoSmithKline and Millennium. He has received travel support from Biogen Idec. He has had research funding from Biogen Idec, Pfizer, Schering-Plough, Bavarian Nordic, NeuorgesX, Novartis and Tibotec. He receives research support from the NIMH, NINDS, National Institute of Allergy and Infectious Diseases and Fogarty Institute of NIH. A. C. C. has current or past research support from Merck & Co., Schering-Plough, Boehringer-Ingelheim, Gilead Sciences and Tibotec–Virco. She is a member of a Data, Safety and Monitoring Board for a Merck-sponsored study and has participated in Advisory Boards for GlaxoSmithKline, Merck & Co. and Pfizer. She and an immediate family member own stock in Abbott Laboratories and Bristol-Myers Squibb. R. H. receives research support from GlaxoSmithKline, Merck, Pfizer and Abbott Laboratories and is a consultant and/or receives an honorarium from Abbott, Bristol-Myers Squibb, Gilead Sciences, GlaxoSmithKline, Merck, Monogram, Pfizer, Roche, Tibotec and Virco. R. E. has participated in advisory boards and on speakers bureaus for GlaxoSmithKline and Abbott Laboratories. An immediate family member owns stock in Abbott Laboratories and Johnson & Johnson. I. G. has been a speaker for Abbott Laboratories. All other authors: none to declare.

Author contributions

All co-authors reviewed, revised for content and approved this manuscript. B. M. B., E. V. C., S. L. L. and S. S. R. participated in the conception and design of the study, with revision and approval by D. B. C., A. C. C., B. B. G., J. A. M., D. M. S., I. G. and R. E. Data were acquired by D. B. C., A. C. C., B. B. G., R. H., G. M., J. A. M., S. S. R. and D. M. S. Also, B. M. B., E. V. C., P. P. K. and S. L. L. analysed and interpreted the data, with review by D. B. C., A. C. C., R. E., B. B. G., I. G., R. H., G. M., J. A. M., S. S. R. and D. M. S.

Disclaimer

The views expressed in this article are those of the authors and do not reflect the official policy or position of the US Government.