Protein-Free Efavirenz Concentrations in Cerebrospinal Fluid and Blood Plasma Are Equivalent: Applying the Law of Mass Action To Predict Protein-Free Drug Concentration

Abstract

Efavirenz (EFV) is one of the most commonly prescribed antiretroviral drugs (ARVs) for the treatment of HIV. Highly protein-bound drugs, like EFV, have limited central nervous system (CNS) penetration when measured using total drug concentration gradients between blood plasma (BP) and cerebrospinal fluid (CSF). However, the more relevant pharmacologically active protein-free drug concentrations are rarely assessed directly in clinical studies. Using paired BP and CSF samples obtained from 13 subjects on an EFV-containing regimen, both the protein-free and total concentrations of EFV were determined. Despite a median (interquartile range [IQR]) total EFV BP/CSF concentration ratio of 134 (116 to 198), the protein-free EFV BP/CSF concentration ratio was 1.20 (0.97 to 2.12). EFV median (IQR) protein binding was 99.78% (99.74 to 99.80%) in BP and 76.19% (74.47 to 77.15%) in CSF. In addition, using the law of mass action and an in vitro-derived EFV-human serum albumin dissociation constant, we have demonstrated that the predicted median (IQR) protein-free concentration in BP, 4.59 ng/ml (4.02 to 9.44 ng/ml), compared well to that observed in BP, 4.77 ng/ml (3.68 to 6.75 ng/ml). Similar results were also observed in CSF and seminal plasma. This method provides a useful predictive tool for estimating protein binding in varied anatomic compartments. Our results of equivalent protein-free EFV concentrations in BP and CSF do not support prior concerns of the CNS as a pharmacological sanctuary from EFV. As CSF penetration of ARVs may increase our understanding of HIV-associated neurological dysfunction and antiretroviral effect, assessment of protein-free CSF concentrations of other highly protein-bound ARVs is warranted.

INTRODUCTION

Antiretroviral drugs (ARVs) are highly successful in preventing disease progression and prolonging life in HIV-infected individuals (1, 2) and preventing sexual transmission (3–6). HIV-associated neurological disease (HAND) is relatively common, resulting in significant cognitive and neurological impairment, but the etiological mechanism of HAND progression is incompletely understood (7, 8). While HIV is capable of central nervous system (CNS) infection resulting in CNS dysfunction and ARV treatment has reduced the progression to HIV-associated dementia and cognitive dysfunction, the use of ARVs has not eradicated HIV infection of the CNS. The ability of small-molecule drugs to penetrate the CNS has been demonstrated to be associated with many variables, including molecular size, lipophilicity, electric charge, plasma protein binding, and active transport (9, 10). In some cases, ARVs may contribute to HAND (9, 11, 12). Possible explanations for this include the CNS acting as a pharmacological sanctuary due to inadequate penetration of ARVs, toxicity due to ARV exposure with more-sensitive drug toxicity responses in the CNS, coinfection, or other neurological disorders, and physiological factors, including age and sex (10, 13). Several ARVs achieve lower concentrations in cerebrospinal fluid (CSF) than in blood plasma (BP), though these concentrations may still exceed the in vitro 50% inhibitory concentration (IC₅₀) for HIV inhibition (14–17). Still, ARVs that have poor CNS penetration are strongly associated with increased viral load in the CSF, and a CNS penetration effectiveness (CPE) score has been developed and refined as one among several explanatory variables of HAND (18).

None of these assessments of the ARV BP/CSF gradient, however, measured free drug concentration of the ARV in the CSF; all measured total ARV concentration (protein-free and protein-bound ARV). ARVs are predominantly bound to plasma-binding proteins, human serum albumin (HSA) and alpha-1-acid glycoprotein (AAG). Nonnucleoside reverse transcriptase inhibitors (nNRTIs) predominantly bind HSA, while protease inhibitors predominantly bind AAG (19). In addition, HSA is the most abundant binding protein in blood plasma (19, 20). Since many of the ARVs are highly bound to HSA and the concentration of HAS is far lower in the CSF than in BP (13), these reports may underestimate the relative protein-free ARV concentrations in BP and CSF. While membrane transporters also contribute to some of the previously reported BP/CSF ARV gradients, protein-binding effects are independent and will continue to lead to underestimation of relative protein-free BP/CSF concentrations where only total drug concentrations are measured. Whether the desire is to better understand ARV-related HIV inhibition or drug-related toxicity in the CNS, an assessment of protein-free ARV concentration in CSF may improve the predictive value of the CPE score for a better understanding of HAND, since protein-free, not total, drug concentration exerts pharmacological effects (20, 21).

One of the most widely used ARVs, efavirenz (EFV), is a nonnucleoside reverse transcriptase inhibitor that acts by allosterically inhibiting viral reverse transcriptase (22) and is greater than 99.75% bound to human serum albumin (23, 24). In a previous study (24), we demonstrated that BP and seminal plasma (SP) have the same protein-free EFV concentration despite a 20-fold-higher total EFV concentration in BP compared to SP. This dispelled prior concerns that the male genital tract was a pharmacological sanctuary from EFV. Leveraging our expertise from this previous study, we sought to examine EFV distribution in CSF and the potential role of protein binding. In a previous study by Best et al. (14), paired BP and CSF samples showed a median (interquartile range [IQR]) total EFV concentration of 2,145 ng/ml (1,384 to 4,423 ng/ml) in plasma and 13.9 ng/ml (4.1 to 21.2 ng/ml) in CSF. Because HSA concentrations are estimated to be far lower in CSF than in BP and SP, we hypothesized that EFV protein binding is significantly lower in CSF than in BP, which may result in a similar free EFV concentration in BP and CSF, similar to the situation in SP.

Free drug concentrations in CSF are estimated to be very low requiring highly sensitive assays, and the methods used to separate free drug from binding proteins are laborious and time-consuming. Accordingly, a method to predict free drug concentration in a peripheral anatomic compartment, like CSF, would be highly beneficial. In a previous study, we estimated the in vitro protein (albumin)-binding dissociation constant (K_d) of EFV using the law of mass action which compared favorably with the observed population mean of K_d for EFV binding to HSA (24). We wanted to test this in vitro K_d estimate in CSF for estimating protein-free EFV concentration to evaluate its generalizability to different anatomic compartments.

Therefore, the goals of this study are to (i) describe total and protein-free EFV distribution in the CNS, (ii) describe the impact of protein binding on EFV distribution, and (iii) develop and validate an application of the law of mass action for predicting protein binding within an extravascular compartment without direct assessment of free drug concentration in the compartment. This predictive tool for estimating drug distribution in a compartment would be especially useful where free drug concentrations are very low, available sample volumes are very small, or where the matrices are not readily amenable to the ultrafiltration needed for assessment of free drug concentration.

MATERIALS AND METHODS

Subjects and study design.

To establish the EFV and HSA concentrations in BP and CSF, we obtained paired blood plasma and CSF samples from 13 HIV-positive subjects on an ARV regimen containing EFV obtained both prospectively and with archived samples. Through a study approved by the Johns Hopkins Institutional Review Board (IRB), research participants provided written informed consent and were screened from among current research participants in one of two active Johns Hopkins Department of Neurology IRB-approved studies involving a lumbar puncture (LP). Research participants were recruited from both the Johns Hopkins Oxidative Stress Cohort and the Johns Hopkins National Institute of Mental Health (NIMH) Center for Novel Therapeutics of HIV-Associated Cognitive Impairment Clinical Outcomes Cohort. The subjects had to be on an EFV-containing regimen for a minimum of 4 weeks prior to participation in this study. Consenting and eligible research participants contributed a single blood sample by venipuncture and spinal fluid samples from the neurology studies. Four subjects were recruited and enrolled through this study protocol. Samples for these subjects were obtained 5, 9, 14, and 15 h following their last dose. Blood was drawn (≤30 ml) within 4 h of the lumbar puncture for each subject, providing nearly simultaneous paired BP and CSF samples given the long half-life of EFV (40 to 52 h) with less than a 3% estimated change in blood plasma concentration over this time interval at steady state.

Archived BP and CSF samples were obtained from the Northeastern AIDS Dementia Cohort (NEAD) repository (25). Samples were acquired from nine subjects with paired BP and CSF samples who were enrolled in 2007 or 2008 with the available samples stored at −80°C. Written consent was provided during the original study consenting to the use of the remaining samples for additional study. The original study was approved by the Johns Hopkins Institutional Review Board and the use of the samples in this secondary study was also approved by the board. We previously demonstrated that EFV in clinical samples is stable at this temperature over this period of time (24).

Separation of protein-free EFV from protein-bound EFV by ultrafiltration.

Ultrafiltration of BP and CSF samples was performed according to a modified version of a previously described method (24). Nonspecific binding of EFV to the apparatus is significant for the CSF matrix due to the low concentrations of HSA in the biological samples. The lower limit of HSA in a sample was previously defined as 1 mg/ml in order to prevent interference by nonspecific binding. To overcome this limitation, membrane filter plates (Millipore, Billerica, MA) were incubated at 37°C for 1 h with a blocking reagent (10% polyethylene glycol) prior to the addition of CSF samples. Addition of the blocking reagent resulted in 100% recovery of EFV in validation experiments. The plates were centrifuged at 3,000 × g for 15 min at 37°C. Remaining blocking reagent was decanted from the filter membrane and collection plate prior to sample addition. CSF samples were then processed as previously described (24), with the modification of using 5-min centrifugation intervals rather than 15-min centrifugation intervals, for a total centrifugation time of 15 min. The ultrafiltrate was extracted from the filter plate using methanol. Validation experiments were conducted using five individual lots of blank CSF (Bioreclamation, LLC., Westbury, NY), spiked with total EFV concentrations of 30 ng/ml and 150 ng/ml. The protein binding of EFV was shown to be stable for at least three freeze-thaw cycles in BP and CSF matrices. We have previously demonstrated that the concentration of EFV is also stable over time by comparing our previous results (24) with those obtained from the original study (26).

Processing of blood and CSF samples.

Approximately 1 ml of CSF was received for each subject enrolled in the study, which was subsequently stored at −80°C. Whole blood was drawn from each subject using K₂EDTA as an anticoagulant. Blood samples were centrifuged at 1,100 × g for 10 min at 20°C. The separated blood plasma was aliquoted into a clean tube and stored at −80°C until analysis. Archived samples were similarly processed and stored at −80°C until analysis.

Quantification of protein-free and protein-bound efavirenz.

Concentrations of EFV were determined using individual ultraperformance liquid chromatography tandem mass spectrometry (UPLC-MS/MS) assays performed using an AB Sciex QTRAP 5500 (AB Sciex, Foster City, CA) interfaced with an Acquity UPLC (Waters Inc., Milford, MA). EFV samples were analyzed by a modified version of a previously described method (27). Briefly, samples were extracted using a liquid-liquid extraction method of 600 μl of 50 mM ammonium formate and 900 μl of a solution of hexanes and ethyl acetate (1:1). The organic layer was decanted and evaporated to dryness. Samples were reconstituted in methanol for analysis. EFV was resolved using a reverse-phase UPLC column (2.1 by 50 mm Acquity UPLC BEH C₁₈) with a flow rate of 0.5 ml/min by gradient elution (mobile phase A of 0.1% formic acid in water; mobile phase B of 0.1% formic acid in acetonitrile). EFV was detected via negative-ion multiple reaction monitoring (MRM). The assay employed a fluorinated analog of EFV (F-EFV) as the internal standard. EFV and F-EFV were detected via the MRM transitions m/z 314.0 > 244.1 and m/z 298.0 > 227.9, respectively. The dynamic range for EFV quantitation is 0.125 to 10,000 ng/ml with an accuracy (percent deviation [%dev]) of −5.2 to 8.0% and precision (percent coefficient of variation [%CV]) of <8% (26). Standard curves were linear with coefficients of variation (r²) of >0.98.

Quantification of human serum albumin.

HSA was quantified using a colorimetric bromocresol green assay (BioAssay Systems, Hayward, CA) according to the manufacturer's instructions. Blood plasma samples were diluted 2-fold in water to obtain accurate concentrations. CSF samples were analyzed without dilution (10-μl sample volume). For those CSF samples which fell below the limit of detection, 20-μl sample volumes were analyzed.

Prediction of protein-free drug fraction by the law of mass action.

The law of mass action describes the equilibrium between protein-bound and protein-free drug:

$$\underset{P-\left(C_T-C_u\right)}{\text{Protein}}\hspace{0.17em}+\hspace{0.17em}\underset{C_u}{\text{Drug}}\stackrel{K_d}{\rightleftarrows }\underset{C_T-C_u}{\text{Protein}\times \text{Drug}}\hspace{0.17em}$$ (1)

An estimate of the protein-free unbound fraction (f_u) may be made in the CSF, or other anatomic compartment, based on total EFV concentrations (C_T), unbound EFV concentration (C_u), protein concentration (P), the dissociation constant (K_d) estimate for EFV binding to albumin, and a stoichiometry (n) estimate of HSA binding to EFV of 1 (24, 28) as follows:

$$K_d=\frac{\left(P-C_T+C_u\right)C_u}{C_T-C_u}$$ (2)

Since C_u = f_u × C_T,

$$K_d=\frac{f_u}{1-f_u}\left(nP-C_T+C_T{f}_u\right)C_T^n$$ (3)

$$f_u=\frac{\left(C_T-K_d-P\right)+\sqrt{{\left(C_T-K_d-P\right)}^2+4K_d{C}_T}}{2C_T}$$ (4)

As the direct experimental assessment of free EFV concentration described above is labor-intensive, we were interested in predicting the protein-free EFV concentration in a given anatomic compartment without its direct measurement. The predicted protein-free EFV fraction, f_u, can be calculated with equation 4 using an in vitro estimate of K_d for EFV binding to HSA, 2.05 μM (24), based on equation 3. Therefore, the only necessary clinical measurements would include total EFV and HSA concentrations, with no direct assessment of protein-free EFV concentrations. The K_d estimate was previously established (24) using a titration of HSA in phosphate-buffered saline (PBS) with different concentrations of EFV to determine the binding dynamic between HSA and EFV. To evaluate the accuracy of predicting the f_u of EFV in a biological matrix, we compared the predicted EFV f_u (calculated from equation 4) to the directly observed EFV f_u obtained by ultrafiltration and UPLC-MS/MS analysis of BP and CSF samples. To extend the method to another peripheral compartment, we also compared the predicted to observed EFV f_u for seminal plasma. For this, we used paired BP and SP samples from six subjects on an EFV-containing regimen from an earlier study (24, 26). To evaluate one source of variability in EFV f_u estimates in different compartments, we compared the in vitro K_d estimate to the median K_d estimate for each compartment using equation 3 and data from clinical samples in BP, CSF, and SP.

Statistical analysis.

A sample size of 12 was chosen to confidently exclude (80% power) differences as small as 1.1 standard deviations in outcome variables assuming a correlation factor of 0.1. Statistical analyses were performed using GraphPad Instat 3.0 (GraphPad Software, Inc., San Diego, CA). Differences between BP and CSF protein-free EFV concentrations and differences between predicted f_u and observed f_u were analyzed using a Student's paired t test. Statistical significance is defined as P ≤ 0.05.

RESULTS

Total and protein-free EFV concentrations in BP and CSF.

We examined the total and protein-free concentrations of EFV in paired BP and CSF samples from research participants to investigate how protein binding impacts the distribution of EFV and evaluate the CNS as a potential pharmacological sanctuary from EFV. The median (interquartile range [IQR]) total EFV concentration was 2,170 ng/ml (1,684 to 3,953 ng/ml) in BP and 18.8 ng/ml (9.34 to 22.74 ng/ml) in CSF. Protein-free EFV concentrations in BP and CSF were 4.8 ng/ml (3.7 to 6.7 ng/ml) and 4.1 ng/ml (2.2 to 4.8 ng/ml), respectively, and therefore were not statistically different (P = 0.075) (Fig. 1). The total (protein-free and protein-bound) EFV BP/CSF concentration ratio was 134 (116 to 198), indicating a significant reduction of the total concentration EFV penetration into the CSF. In contrast, the protein-free EFV BP/CSF ratio of paired samples was 1.20 (0.97 to 2.12) indicating similar concentrations of protein-free EFV in BP and CSF.

Fig 1.

Total (protein-free plus protein-bound) and protein-free EFV concentrations measured in blood plasma (BP) and cerebrospinal fluid (CSF). The concentrations of EFV were determined as BP total EFV (●), CSF total EFV (○), BP protein-free EFV (▼), and CSF protein-free EFV (△). Each symbol represents the value for an individual.

Protein binding and albumin concentrations.

As an explanation for the equivalent protein-free concentrations in BP and CSF despite a large total concentration gradient of EFV, we examined the protein binding and impact of HSA concentration on protein-free EFV concentrations. Protein binding of EFV in BP, 99.78% (99.74 to 99.80%), was significantly greater than protein binding in CSF, 76.19% (74.47 to 77.15%) (P = 0.0002). This result is best explained by the large observed difference in HSA concentrations in BP and CSF. The measured concentration of HSA in BP was 58.4 mg/ml (55.4 to 61.5 mg/ml), and in CSF, it was 0.27 mg/ml (0.23 to 0.30 mg/ml) with a paired BP/CSF ratio of 213 (187 to 254).

Prediction of protein-free EFV fraction.

To evaluate the use of the law of mass action for predicting the fraction of unbound EFV (f_u) in an extravascular compartment, we compared the predicted versus observed f_u values in paired BP and CSF samples and further confirmed the utility in additional samples of paired BP and SP specimens. For each biological matrix, the predicted f_u, using our in vitro estimate for K_d, 2.05 μM (1.79 to 2.39), for the binding of EFV to HSA, is compared to the observed f_u for each subject (Fig. 2). The predicted in vitro EFV f_u in BP was 0.0023 (0.0022 to 0.0025) compared to the observed EFV f_u of 0.0021 (0.0020 to 0.0026) (P = 0.120), a difference that is not statistically significant. These values correspond to predicted protein-free EFV BP concentrations of 4.97 ng/ml (4.02 to 9.44 ng/ml) and observed concentrations of 4.77 ng/ml (3.68 to 6.75 ng/ml). The in vitro predicted f_u in CSF, 0.34 (0.31 to 0.38), was 42% higher than the observed f_u of 0.24 (0.23 to 0.26) (P < 0.0001), a statistically significant difference. These values correspond to CSF predicted protein-free EFV concentrations of 5.55 ng/ml (3.01 to 8.80) and observed concentrations of 4.12 ng/ml (2.23 to 4.77). The in vitro predicted f_u in SP, 0.032 (0.030 to 0.034), was 35% lower than the observed f_u of 0.049 (0.034 to 0.060), a statistically significant difference (P = 0.016). These values correspond to SP predicted protein-free EFV concentrations of 4.33 ng/ml (3.37 to 5.64 ng/ml) and observed concentrations of 6.61 ng/ml (4.46 to 8.21 ng/ml). The calculated values for estimated in vitro K_d and matrix-specific K_d for the various compartments are listed in Table 1.

Fig 2.

Fraction of unbound EFV may be predicted using the law of mass action. Correlation between the predicted and observed fractions of unbound (f_u) EFV in CSF (squares), SP (circles), and BP (triangles), where f_u is determined using an in vitro K_d estimate of 2.05 μM. The line of unity or complete correlation is shown.

Table 1.

Comparison of K_d estimates (in vitro and clinical matrices) and predicted versus observed EFV f_u estimates (predicted f_u/observed f_u)^a in blood plasma, seminal plasma, and cerebrospinal fluid

Category, sample source, or compartment	Predictive estimate used in EFV fu calculation, median (IQR)
In vitro or clinical sample source	Kd (μM)	Predicted fu/observed fu
In vitro	2.05 (1.79–2.39)
Blood plasma	1.89 (1.65–2.13)
Seminal plasma	3.27 (1.99–4.55)
Cerebrospinal fluid	0.97 (0.84–1.11)
Compartment
Blood plasma		1.13 (0.95–1.28)
Seminal plasma		0.67 (0.58–0.73)
Cerebrospinal fluid		1.43 (1.23–1.52)

^aPredicted EFV f_u based on in vitro K_d (2.05 μM) and measured total EFV and HSA for each individual; observed EFV f_u based entirely on measured values from each individual (total EFV, free EFV, and HSA). Compartmental median (IQR) estimates were obtained from paired blood plasma and cerebrospinal fluid samples from 13 individuals and paired blood plasma and seminal plasma samples from 6 individuals.

DISCUSSION

We have shown that protein-free EFV concentrations are equivalent in BP and CSF—inconsistent with CSF acting as a pharmacological sanctuary with regard to EFV—despite the large 134-fold-higher total EFV concentration in BP compared to CSF. Prior to this study, neither protein binding or protein-free concentrations had been established in CSF for any ARVs. The equivalent free EFV concentration in CSF and BP in the face of the large total EFV gradient is best explained by the significantly decreased albumin concentrations in CSF compared to BP, resulting in lower, though still significant, protein binding in CSF. Our total EFV concentrations are consistent with previous studies (14, 29). Our protein-free EFV CSF concentrations are lower than previous estimates based on assumptions of HSA concentration and protein binding without direct assessment of protein-free EFV CSF concentrations (14). Even so, our observed protein-free EFV concentrations are well above the previously established wild-type in vitro IC₅₀ in lymphocytes for EFV of 0.51 ng/ml (30), which we would estimate to be a protein-free EFV concentration of 0.116 ng/ml assuming they used 10% serum concentration in their experiments, or 0.002 ng/ml if they used 50% serum concentration. Accordingly, our findings, consistent with those of Best et al. (14) indicate that the CNS is not a pharmacological sanctuary from EFV despite the large total concentration gradient from BP to CSF. Since many ARVs with high protein binding have also been reported to penetrate CSF poorly based on total ARV concentrations in BP and CSF, their protein-free BP/CSF concentration gradient may also be quite different. Membrane transporters, however, may also contribute to some of the previously reported gradients. EFV, and possibly other highly protein-bound ARVs, should be reevaluated with regard to their CSF penetration effectiveness (CPE) scores used in pathogenesis studies of HAND (18).

It has previously been suggested that increased penetration of HSA into CSF is indicative of neurotoxicity due, in part, to pathological alteration of the blood-brain barrier (31, 32). HSA has, therefore, been considered a potential biomarker for neurological damage; however, several studies provide conflicting results with regard to the correlation of HSA and HIV disease progression (13, 33). Elovaara et al. (13) compare CSF albumin concentrations in HIV-infected subjects to healthy subjects, with concentrations of 0.208 ± 0.068 mg/ml and 0.214 ± 0.054 mg/ml, respectively. Brew et al. (33) describe several studies which have observed increased concentrations of CSF albumin correlated with disease progression; however, they also note a correlation of increased CSF HSA to neuroasymptomatic subjects. While the etiology of elevated CSF HSA concentrations is still incompletely understood, our study results indicate that elevated HSA concentrations would decrease free drug CSF concentration for EFV and, potentially other highly HSA-bound ARVs, likely resulting in reduced pharmacological effect.

The usefulness of the law of mass action and the in vitro estimate of the EFV-HSA K_d for predicting protein-free EFV in BP, CSF, and SP was demonstrated by our ability to predict protein-free EFV concentrations within 50% of the observed concentration. We developed this predictive model using EFV with single-point sampling facilitated by the long EFV half-life. This method may be applied to estimate protein-free drug concentrations of other drugs; however, single-point sampling may not be appropriate for drugs with shorter half-lives. In BP, the predicted fraction of protein-free EFV was very similar to the observed protein-free EFV fraction. In CSF, however, the predicted fraction of protein-free EFV was greater than the observed fraction of protein-free EFV. In contrast, the predicted fraction of protein-free EFV was less than the observed fraction of protein-free EFV in SP. These results suggest compartmental differences in HSA binding or HSA itself and subsequently, protein-free EFV concentrations. To examine these possible differences, we performed a K_d analysis for the interaction of EFV with HSA with the collective clinical data for each compartment: BP, SP, and CSF. The calculated values showed distinctly different estimates for K_d in each compartment supportive of compartmental differences in the binding of EFV to HSA. One possibility for the compartmental differences in K_d values is the presence of additional binding proteins in a compartment, which may either bind EFV or affect the binding of EFV to HSA. It is also possible that compartmental binding proteins, including albumin, may be subject to certain modifications or alterations that could affect the overall binding capacity. As anatomic compartments are known to vary in pH, this may also impact the degree of protein binding, and subsequently, the K_d of EFV binding to HSA (34–36).

In summary, we demonstrated that protein-free EFV concentrations in BP and CSF are not different, despite the significantly decreased penetration of total EFV into the CSF, demonstrating that the CNS is not a pharmacological sanctuary from EFV. This raises questions about other highly protein-bound ARVs and may alter prior assessments of EFV in terms of antiviral effect or toxicity in the CNS. We also established a useful application of the law of mass action to estimate EFV protein binding with reasonable accuracy in varied anatomic compartments without direct assessment of free EFV concentration. This tool may enable compartment-specific assessment of free drug concentration, avoiding the need to develop very sensitive, laborious, and painstaking methods for quantifying free drug concentration in peripheral compartments.