Pharmacogenetics of plasma efavirenz exposure in HIV-infected adults and children in South Africa

Abstract

Aims

Genetic factors, notably CYP2B6 516G→T [rs3745274] and 983T→C [rs28399499], explain much of the interindividual variability in efavirenz pharmacokinetics, but data from Africa are limited. We characterized relationships between genetic polymorphisms and plasma efavirenz concentrations in HIV-infected Black South African adults and children.

Methods

Steady-state mid-dosing interval efavirenz concentrations were measured. We genotyped 241 polymorphisms in genes potentially relevant to efavirenz metabolism and transport, including ABCB1, CYP2A6, CYP2B6, CYP3A4, CYP3A5, NR1I2 and NR1I3.

Results

Among 113 participants (59 adults and 54 children), minor allele frequencies for CYP2B6 516G→T, 983T→C, and 15582C→T [rs4803419] were 0.36, 0.07, and 0.09, respectively. Based on composite CYP2B6 15582/516/983 genotype, there were 33 extensive metabolizer, 62 intermediate metabolizer and 18 slow metabolizer genotypes. Median (IQR) mid-dose efavirenz concentrations were 1.44 (1.21–1.93) µg ml^–1, 2.08 (1.68–2.94) µg ml^–1 and 7.26 (4.82–8.34) µg ml^–1 for extensive, intermediate and slow metabolizers, respectively. In univariate analyses, a model that included composite genotype best predicted efavirenz concentrations (β = 0.28, 95% CI 0.21, 0.35, P = 2.4 × 10^–11). Among individual CYP2B6 polymorphisms, 516G→T best predicted efavirenz concentrations (β = 0.22, 95% CI 0.13, 0.30, P = 1.27 × 10⁻⁶). There was also associations with 983T→C (β = 0.27, 95% CI 0.10, 0.44, P = 0.002) and 15582C→T (β = 0.11, 95% CI 0.01, 0.22, P = 0.04). Associations were consistent in adults and children. No other polymorphisms were independently associated with efavirenz concentrations.

Conclusions

Composite CYP2B6 genotype based on CYP2B6 516G→T, 983T→C, and 15582C→T best described efavirenz exposure in HIV-infected Black South African adults and children.

What is Already Known about the Subject

• The polymorphisms CYP2B6 516G→T (rs3745274) and 983T→C (rs28399499) are strongly associated with plasma efavirenz concentrations, but do not entirely explain interindividual variability.

• A recent genome-wide association study implicated CYP2B6 15582C→T (rs4803419) as an independent predictor of efavirenz trough concentrations.

• Reported pharmacokinetic associations beyond CYP2B6 are inconsistent.

What this Study Adds

• We provide evidence to show that CYP2B6 15582C→T is associated with plasma efavirenz concentrations in Black South Africans, in addition to CYP2B6 516G→T and 983T→C.

• We show that genetic associations are consistent in adults and children.

• We found no additional associations with plasma efavirenz concentrations beyond these CYP2B6 polymorphisms.

Introduction

Efavirenz is extensively prescribed for HIV-1 infection worldwide. It is metabolized primarily by cytochrome P450 (CYP) 2B6 1. Analyses with AIDS Clinical Trials Group (ACTG) protocol 5097s first showed that the non-synonymous polymorphism CYP2B6 516G→T (rs3745274) was strongly associated with increased plasma efavirenz exposure 2. Many studies have since replicated this association 3–17. The CYP2B6 516G→T polymorphism is more frequent with African ancestry than with European ancestry 18, which largely explains the greater mean plasma efavirenz concentrations reported among populations of African descent 19,20. A less frequent CYP2B6 polymorphism, 983T→C (rs28399499), also predicts increased plasma efavirenz exposure 9,13,16,21–25. The CYP2B6 983 C allele is found almost exclusively with African ancestry, where it is still much less frequent than 516G→T 18. A recent genome-wide association study of White, Black and Hispanic adults in the United States showed that a third polymorphism, CYP2B6 15582C→T (rs4803419), was also associated with estimated efavirenz trough concentrations independent of 516G→T and 983T→C 26.

Additional CYP2B6 polymorphisms suggested to affect CYP2B6 activity have been extremely infrequent 21,27,28, or have not predicted plasma efavirenz exposure 2,29,30. Polymorphisms in genes beyond CYP2B6 reported to affect efavirenz pharmacokinetics include CYP2A6 7,30, CYP3A5 2, UGT2B7 7 and CAR 31, CYP3A5 2, UGT2B7 7 and CAR 31, although results have been inconsistent. The genome-wide associations study noted above did not identify additional associations in or beyond CYP2B6 26.

Data from Africa have repeatedly replicated the association between efavirenz exposure and both CYP2B6 516G→T 7–17,24,25 and CYP2B6 983T→C 9,13,16,21,24,25, but data from Africa beyond these two polymorphisms are limited. We characterized relationships between genetic polymorphisms and plasma efavirenz concentrations among HIV-infected adults and children in South Africa.

Methods

Study Participants

We included adults and children who had been enrolled in two observational studies. In the adult study, adults from the public sector antiretroviral therapy (ART) programme were enrolled in a cross sectional study to evaluate associations between plasma efavirenz concentrations and metabolic complications. Eligible participants were on efavirenz-based therapy for at least 1 month, and were excluded for renal or hepatic disease, active opportunistic infections, known diabetes or dyslipidaemia, self-reported non-adherence, pregnancy, or concomitant drugs with the potential to interact with efavirenz.

In the paediatric study, African children (aged 3–15 years and weighing more than 10 kg) on efavirenz-based therapy with or without rifampicin-based anti-tuberculosis therapy were enrolled to evaluate the effect of rifampicin-based antituberculosis therapy on efavirenz concentrations 32. The present analyses included only data from children who were not receiving rifampicin.

All participants self-identified as Black. This study complied with the Helsinki Declaration, was approved by institutional review boards for each site, and participants or their parents gave written informed consent or assent.

Characterization of Genetic Polymorphisms

Human DNA was extracted from buffy coats using the Wizard® Genomic DNA purification kit (Promega, Madison, WI, USA). A total of 241 polymorphisms (70 in ABCB1 (also called MDR1), 22 in CYP2A6, 54 in CYP2B6, one in CYP2C19, 23 in CYP3A4, one in CYP3A5, 31 in NR1I2 (also called PXR) and 39 in NR1I3 (also called CAR)) were successfully genotyped in the Vanderbilt DNA Resources Core using MassARRAY® iPLEX Gold (Sequenom, Inc., San Diego, California, USA). Our strategy for genotyping was as follows: For CYP2B6, ABCB1 and NR1I2 we tagged each entire gene using SeattleSNPs 33 using a cosmopolitan strategy across populations (Yoruba, Asian, African-American, European-American and Hispanic) with a 5% allelic frequency cut-off, a 0.80 threshold for r², 85% data convergence for tagging polymorphisms and 70% data convergence for clustering. For CYP2B6 we included 5 kB in each 5’ and 3’ untranslated regions (UTR), and for ABCB1 and NR1I2 20 kB in each UTR. For CYP2B6, additional polymorphisms of interest (but that were not extremely infrequent) were added based on a previous report 3, as were polymorphisms with at least 5% allelic frequency in 20 kB of the 5’ UTR identified using Ensembl Genome Browser 34 and upstream polymorphisms possibly associated with CYP2B6 expression 35. We also included ABCB1 3435C→T (rs1045642) and 2677G/T/A (rs2032582), CYP2C19 681G→A (rs4244285) and CYP3A5 6986A→G (rs776746). (CYP2C19 681G→A was part of the already designed multiplex assay, but was not expected to affect efavirenz exposure). The final Sequenom assay design is available upon request. Laboratory personnel with no knowledge of clinical data performed genotyping. Ample duplicate and blank assays were included to assure validity. Polymorphisms were excluded for genotyping efficiency less than 95%.

Composite CYP2B6 15582/516/983 genotypes were assigned as follows: extensive metabolizer genotype (15582CC-516GG-983TT or 15582CT-516GG-983TT), intermediate metabolizer genotype (15582TT-516GG-983TT, 15582CC-516GT-983TT, 15582CC-516GG-983CT, 15582CT-516GT-983TT or 15582CT-516GG-983CT) and slow metabolizer genotype (15582CC-516TT-983TT, 15582CC-516GT-983CT or 15582CC-516GG-983CC) 26.

Efavirenz Concentrations

Plasma efavirenz concentrations were measured using liquid chromatography with tandem mass spectrometry (LC/MC/MS) as previously described 36. Intraday and interday precision ranged from 1.2 to 4.1% and 2.5 to 5.3%, respectively. The calibration range was linear from 0.1 to 15 µg ml^–1 and accuracy ranged from 95.2 to 104.6%. Several samples were obtained on the same day from each participant. Sampling times were not pre-specified, and time of prior dose was by self-report for adults, and was reported by caregivers for children. We excluded efavirenz data from samples obtained less than 10 h or greater than 20 h post-dose, or before 1 month of efavirenz therapy, and from children within 1 month after discontinuing rifampicin. An average of measured concentrations was used when multiple samples from the same participant were obtained. The repeated measurements of efavirenz concentrations were used in exploratory multilevel mixed effects (MLME) analyses.

Statistical Analysis

The pharmacokinetic data were not normally distributed. Therefore, a logarithmic (log₁₀) transformation was done. Genetic associations with log₁₀ transformed average efavirenz concentrations were analyzed by univariate linear regression. The tri-allelic ABCB1 rs2032582 was analyzed as A vs. not A, G vs. not G and T vs. not T. We also performed analyses adjusting for CYP2B6 516G→T. All tests used a 5% two-sided significance level. Analyses were performed with PLINK version 1.07 (http://pngu.mgh.harvard.edu/∼purcell/plink/). Bonferroni correction was used to account for multiple testing 37. For polymorphisms previously associated with efavirenz pharmacokinetics, the threshold to carry forward into subsequent linear regression models was P = 0.01.

Haplotypic blocks were defined using the D’ confidence intervals method in Haploview 38 and haplotype phases were inferred using the standard E-M algorithm in PLINK 39. Linkage disequilibrium (LD) plots and values were generated with Haploview (www.broad.mit.edu/mpg/haploview/).

The multilevel mixed effect regression model performed in Stata 11 40, which accounted for within individual correlations, evaluated associations of efavirenz concentrations with the three polymorphisms adjusted for effects of time post-dose and age group.

Results

Study Participant Characteristics

Pharmacokinetic data were available from 152 participants, of whom 113 with sufficient DNA for successful genotyping are included in the present analyses. Among 59 adults, median age was 38 years (range 20 to 59 years), median weight was 65.9 kg (range 45.2 to 104.5 kg) and 45 (76%) were female. Of participants with available plasma HIV-1 RNA data, values were <400 copies ml^–1 in 78% of adults and in 87% of children. Characteristics of study participants are presented in Table1.

Table 1.

Demographic and clinical characteristics of 113 study participants in South Africa.

	Adults	n	Children	n
Efavirenz concentration (µg ml–1)*	2.63 (0.67–29.53)	59	1.90 (0.30–8.47)	54
Time after dose (h)*	12.3 (11.1–14.5)	59	16.2 (14.0–17.8)	54
Drug dose (mg day–1)*	600 (600–600)	59	300 (200–600)	54
Age (years)*	38.0 (20.0–59.0)	59	8.2 (3.0–15.1)	54
Weight (kg)*	65.9 (45.2–104.5)	59	22.3 (13.3–46.0)	54
Female gender†	45 (76.3)	59	25 (46.3)	54
CD4 T cell count (cells mm–3)*	290 (74–904)	59	533 (264–1522)	16
HIV-1 RNA ≤ 400 copies ml–1†	14 (77.8)	18	13 (86.7)	15

^*Medians are shown (ranges in parentheses).

^†Numbers of patients are shown (percentages in parentheses).

Genetic Polymorphisms

Among the 113 participants, 241 polymorphisms were successfully genotyped, of which 18 were monomorphic (i.e. no minor alleles). Each of the remaining 223 polymorphisms was in Hardy–Weinberg equilibrium based on a Bonferroni adjusted P value threshold of 0.0002; six had unadjusted P values <0.05 (NR1I3 rs4489574, P = 0.002, NR1I2 rs2461817, P = 0.021, ABCB1 rs17149792, P = 0.024, NR1I3 rs2502815, P = 0.026, ABCB1 rs10264990, P = 0.035 and CYP3A4 rs28539499, P = 0.036). Minor allele frequencies for the 241 polymorphisms are in Supplemental Material Table S1.

Genetic Associations with Efavirenz Concentrations

The median plasma mid-dose efavirenz concentration was 2.03 µg ml^–1 (interquartile range 1.46–3.46 µg ml^–1). In univariate linear regression models, composite CYP2B6 15582/516/983 genotype was most strongly associated with efavirenz concentrations (β = 0.28, 95% CI 0.21, 0.35, P = 2.4 × 10^–11). Of the 223 polymorphisms, 15 (all in CYP2B6) were associated with efavirenz concentrations at P < 0.01. These include CYP2B6 516G→T (β = 0.22, 95% CI 0.13, 0.30, P = 1.3x10^–6) and eight polymorphisms in LD with 516G→T at r²>0.6. The minor allele frequency of CYP2B6 516G→T was 0.36, with 43 (38.1%) homozygous for GG, 58 (51.3%) heterozygous for GT and 12 (10.6%) homozygous for TT. The minor allele frequency of CYP2B6 983T→C was 0.07, with 98 (86.7%) homozygous for TT, 15 (13.3%) heterozygous for TC and none homozygous for CC. The minor allele frequency of CYP2B6 15582C→T was 0.09, with 94 (83.2%) homozygous for CC, 18 (15.9%) heterozygous for TC and one (0.9%) homozygous for TT.

To identify independent predictors of efavirenz concentrations we performed multivariable linear regression analysis adjusted for CYP2B6 516G→T. By this analysis, the only additional polymorphism associated at P < 0.01 was CYP2B6 983T→C (β = 0.37, 95% CI 0.23, 0.52, P = 2.78 × 10^-6). In an analysis that adjusted for both CYP2B6 516G→T and 983T→C, no additional polymorphism was associated with efavirenz concentrations at P < 0.01, the lowest P value being for CYP2B6 rs28723610 (β = 0.21, 95% CI 0.04, 0.38, P = 0.013). There was no apparent association with CYP2B6 15582C→T (β = 0.06, 95% CI –0.06, 0.19, P = 0.34). Final multivariable linear regression models are presented in Table2.

Table 2.

Genetic associations with efavirenz concentrations in all 113 participants.

		Unadjusted analysis		516G→T adjusted		516G→T and 983T→C adjusted
Chromosome	Polymorphism	β (95% CI)	P value	β (95% CI)	P value	β (95% CI)	P value
19	CYP2B6 516G→T	0.22 (0.13, 0.30)	1.27 × 10–6	NA	NA	NA	NA
19	rs8192719	0.22 (0.13, 0.30)	1.27 × 10–6	NA	NA	NA	NA
19	rs2279343	0.22 (0.13, 0.30)	1.79 × 10–6	NA	NA	NA	NA
19	rs10853744	0.21 (0.13, 0.29)	2.33 × 10–6	–0.13 (–0.70, 0.45)	0.66	–0.09 (–0.61, 0.43)	0.74
19	rs11083595	0.18 (0.10, 0.26)	2.30 × 10–5	0.002 (–0.17, 0.17)	0.99	0.07 (–0.08, 0.23)	0.37
19	rs2054675	0.18 (0.10, 0.26)	2.30 × 10–5	0.002 (–0.17, 0.17)	0.99	0.07 (–0.08, 0.23)	0.37
19	rs3786547	0.18 (0.10, 0.27)	2.39 × 10–5	0.0001 (–0.17, 0.17)	1.00	0.07 (–0.09, 0.23)	0.38
19	rs892216	0.18 (0.09, 0.26)	6.71 × 10–5	–0.006 (–0.16, 0.14)	0.93	0.05 (–0.09, 0.19)	0.49
19	rs7250873	0.16 (0.09, 0.25)	1.07 × 10–4	–0.004 (–0.21, 0.11)	0.58	0.02 (–0.12, 0.17)	0.74
19	CYP2B6 983T→C	0.27 (0.10, 0.44)	1.92 × 10–3	0.37 (0.22, 0.52)	2.78 × 10-6	NA	NA
19	rs1987236	–0.16 (–0.26, –0.05)	3.12 × 10–3	–0.06 (–0.17, 0.04)	0.25	–0.02 (–0.12, 0.08)	0.68
19	rs4803417	–0.16 (–0.27, –0.05)	5.14 × 10–3	–0.06 (–0.18, 0.05)	0.27	–0.03 (–0.13, 0.07)	0.58
19	rs2279345	–0.14 (–0.24, –0.03)	9.37 × 10–3	–0.04 (–0.15, 0.06)	0.39	0.004 (–0.09, 0.10)	0.93
19	rs6508966	–0.14 (–0.24, –0.03)	9.37 × 10–3	–0.04 (–0.15, 0.06)	0.39	0.004 (–0.09, 0.10)	0.93
19	rs6508965	–0.14 (–0.27, –0.05)	9.70 × 10–3	–0.04 (–0.14, 0.06)	0.43	0.01 (–0.08, 0.11)	0.82
19	CYP2B6 15582C→T*	–0.06 (–0.21, 0.08)	0.42	0.03 (–0.11, 0.17)	0.67	0.06 (–0.06, 0.19)	0.34

^*SNP of interest but did not meet criteria of P value < 0.01.

In the above analyses, two individuals had extreme outlier efavirenz values. One individual, a 41-year-old woman with a body mass index of 22 kg m^–2, had three concentrations of 28.1, 28.2 and 32.3 µg ml^–1 (mean 29.30 µg ml^–1). She was homozygous for 15582CC, 516GG and 983TT. One individual, a 7-year-old boy weighing 21 kg at a height of 114 cm, had three concentrations of 0.32, 0.29 and 0.28 µg ml^–1 (mean 0.30 µg ml^–1). He was heterozygous for 15582CT and homozygous for 516GG and 983TT. To characterize associations with CYP2B6 15582C→T further, we performed post hoc sensitivity analyses after censoring the two outliers. In such analyses, adjusting for both 516G→T and 983T→C, five more CYP2B6 polymorphisms in weak LD with CYP2B6 516G→T showed trends toward association with low efavirenz concentrations in unadjusted analyses (Table3), and CYP2B6 15582C→T was now associated with efavirenz concentrations (β = 0.11, 95% CI 0.01, 0.22, P = 0.04) (Table3). Relationships between genotypes and efavirenz concentrations in all individuals are presented in Figure1A.

Table 3.

Genetic associations with log₁₀-transformed efavirenz concentrations in all 111 participants, and excluding two outliers.

		Unadjusted analysis		516G→T adjusted		516G→T and 983T→C adjusted
Chromosome	Polymorphism	β (95% CI)	P value	β (95% CI)	P value	β (95% CI)	P value
19	CYP2B6 516G→T	0.23 (0.15, 0.30)	3.83 × 10–8	NA	NA	NA	NA
19	rs8192719	0.23 (0.15, 0.30)	3.83 × 10–8	NA	NA	NA	NA
19	rs2279343	0.22 (0.15, 0.30)	4.33 × 10–8	NA	NA	NA	NA
19	rs10853744	0.22 (0.14, 0.29)	8.64 × 10–8	–0.13 (–0.64, 0.38)	0.63	–0.08 (–0.53, 0.36)	0.71
19	rs11083595	0.17 (0.09, 0.24)	2.84 × 10–5	–0.11 (–0.26, 0.04)	0.17	–0.04 (–0.18, 0.10)	0.57
19	rs2054675	0.17 (0.09, 0.24)	2.84 × 10–5	–0.11 (–0.26, 0.04)	0.17	–0.04 (–0.18, 0.10)	0.57
19	rs3786547	0.17 (0.09, 0.24)	2.94 × 10–5	–0.11 (–0.27, 0.04)	0.16	–0.04 (–0.18, 0.10)	0.56
19	rs892216	0.16 (0.08, 0.24)	9.04 × 10–5	–0.09 (–0.22, 0.05)	0.20	–0.03 (–0.15, 0.09)	0.58
19	rs7250873	0.15 (0.08, 0.23)	1.47 × 10–4	–0.14 (–0.29, 0.0008)	0.05	–0.08 (–0.20, 0.05)	0.25
19	CYP2B6 983T→C	0.28 (0.12, 0.43)	6.10 × 10–4	0.38 (0.26, 0.51)	3.13 × 10−8	NA	NA
19	rs6508950	–0.14 (–0.23, –0.06)	1.28 × 10–3	–0.10 (–0.17, –0.02)	0.02	–0.05 (–0.12, 0.02)	0.14
19	rs1987236	–0.15 (–0.25, –0.06)	1.35 × 10–3	–0.05 (–0.14, 0.04)	0.28	–0.004 (–0.09, 0.08)	0.93
19	rs4803417	–0.16 (–0.26, –0.06)	2.43 × 10–3	–0.05 (–0.15, 0.05)	0.30	–0.01 (–0.10, 0.08)	076
19	rs10422346	–0.16 (–0.26, –0.05)	3.58 × 10–3	–0.06 (–0.16, 0.04)	0.27	–0.03 (–0.13, 0.07)	0.58
19	rs2279345	–0.13 (–0.23, –0.04)	4.98 × 10–3	–0.03 (–0.13, 0.06)	0.46	0.02 (–0.06, 0.10)	0.60
19	rs6508966	–0.13 (–0.23, –0.04)	4.98 × 10–3	–0.03 (–0.13, 0.06)	0.46	0.02 (–0.06, 0.10)	0.60
19	rs6508965	–0.13 (–0.23, –0.04)	5.20 × 10–3	–0.03 (–0.12, 0.06)	0.53	0.03 (–0.05, 0.11)	0.47
19	rs7259758	–0.14 (–0.25, –0.04)	5.47 × 10–3	–0.06 (–0.16, 0.04)	0.25	–0.02 (–0.11, 0.06)	0.64
19	rs1962261	–0.15 (–0.26, –0.04)	6.27 × 10–3	–0.06 (–0.16, 0.04)	0.23	–0.02 (–0.12, 0.02)	0.60
19	rs11671243	–0.12 (–0.22, –0.03)	9.16 × 10–3	–0.02 (–0.11, 0.07)	0.68	0.02 (–0.06, 0.10)	0.55
19	CYP2B6 15582C→T*	–0.01 (–0.15, 0.13)	0.87	0.08 (–0.04, 0.21)	0.18	0.12 (0.01, 0.22)	0.04

^*SNP of interest, did not meet criteria of P value < 0.01.

Figure 1.

Relationships between CYP2B6 polymorphisms and log₁₀ transformed efavirenz concentrations in adults and children. Efavirenz concentrations were determined as described in Methods. Relationships with CYP2B6 polymorphisms and log₁₀ efavirenz concentrations are shown in all participants (A), adults (B) and children (C). On the x-axis, CYP2B6 haplotypes represent (in order) CYP2B6 15582C→T (CC, CT, TT), 516G→T (GG, GT, TT) and 983T→C (TT, TC). Number per CYP2B6 haplotype is also displayed. On the y-axis, the log₁₀ transformed efavirenz concentrations are displayed. On each graph, each marker represents a different participant. Horizontal bars are medians and interquartile ranges. Median values are above

In the final multivariable model, each 516 T allele was associated with a 31% increase in log₁₀ efavirenz concentrations (β = 0.27, P<0.0001), each 983 C allele with a 46% increase (β = 0.38, P<0.0001), and each 15582 T allele with a 6% increase (β = 0.06, P = 0.340). These three polymorphisms explained 34% of variance in log₁₀ efavirenz concentrations. A model that included only CYP2B6 516G→T and 983T→C also explained 34% of variance in log₁₀ efavirenz concentrations. Univariate linear regression models for each individual polymorphism explained 19%, 8% and only 0.59% of variance for CYP2B6 516G→T, 983T→C and 15582C→T, respectively.

In the post hoc sensitivity analysis that censored the two outlier participants, each 516 T allele was associated with a 33% increase in log₁₀ efavirenz concentrations (β = 0.29, P < 0.0001), each 983 C allele with a 48% increase (β = 0.39, P < 0.0001) and each 15582 T allele with a 12% increase (β = 0.12, P = 0.04), which together explained 45% of variance. A model that included only CYP2B6 516G→T and 983T→C explained 43% of variance. Univariate linear regression models for each individual polymorphism explained 24%, 10% and only 0.02% of variance for CYP2B6 516G→T, 983T→C and 15582C→T, respectively.

We also performed separate analyses among children and adults. Linear regression results for children and adults are shown in Supplemental Material Table S2 and S3, respectively. In both analyses, CYP2B6 516G→T (and polymorphisms in strong LD with CYP2B6 516G→T) and 983T→C were strongly associated with log₁₀ efavirenz concentrations. In unadjusted analyses, there was no association with CYP2B6 15582C→T. In adults, unadjusted analyses revealed that NR1I2 rs9847782 had the greatest effect size (β = 0.37, 95% CI 0.11, 0.62, P = 7.3 × 10⁻³), which was no longer apparent when one outlier was excluded (a 41-year-old with very high efavirenz concentrations and homozygous for NR1I2 rs9847782TT). Interestingly, in children only, when the outlier with an exceptionally low log₁₀ efavirenz concentration was excluded (a 7-year-old heterozygous for CYP2B6 15582 CT), CYP2B6 15582C→T showed a trend towards significance when adjusted for CYP2B6 516G→T and 983T→C (β = 0.08, 95% CI 0.005, 0.30, P = 0.05). Relationships between genotypes and efavirenz concentrations in children and adults are presented in Figure1B and Figure1C, respectively.

As there were clinical differences (including age and sampling time post-dose) between adults and children in our analyses, we explored multilevel mixed effects modelling using each measured efavirenz value separately in each participant, rather than the mean of concentration values. We fitted a hierarchical model that predicted log₁₀ efavirenz concentrations as a function of 1) fixed effects of age group, time after dose, CYP2B6 516G→T, 983T→C and 15582C→T and 2) random effects for the individual to account for within individual correlations. In this model, compared with individuals in the lowest concentration stratum (i.e. 15582CC-516GG-983TT, β = 0.58), when time after dose and age group were held constant, homozygosity for CYP2B6 516TT was associated with an 0.64 increase in log₁₀ efavirenz concentrations (2.9-fold increase in measured efavirenz concentrations), heterozygosity for CYP2B6 983TC with an 0.36 increase in log₁₀ efavirenz concentrations (2.1-fold increase in measured efavirenz concentrations), heterozygosity for CYP2B6 516GT with an 0.19 increase in log₁₀ efavirenz concentrations (1.5-fold increase in measured concentrations), homozygosity for CYP2B6 15582TT with an 0.21 increase in log₁₀ efavirenz concentrations (1.4-fold increase in measured efavirenz concentrations) and heterozygosity for CYP2B6 15582CT with an 0.06 increase in log₁₀ efavirenz concentrations (1.2-fold increase in measured efavirenz concentrations).

Linkage Disequilibrium Between CYP2B6 Polymorphisms

To understand relationships between CYP2B6 polymorphisms associated with efavirenz concentrations better, we considered LD. In the first model, without adjusting for any polymorphisms, two groups of polymorphisms were associated with efavirenz concentrations (Table2), the first comprising eight polymorphisms in strong LD with CYP2B6 516G→T and with minor alleles associated with higher efavirenz concentrations (i.e. positive β values), and the second comprising five polymorphisms in weaker LD with CYP2B6 516G→T and with minor alleles trending towards an association with lower efavirenz concentrations (i.e. negative β values). In the second analyses, which adjusted for CYP2B6 516G→T, none of the above 13 polymorphisms remained associated with efavirenz concentrations, with only CYP2B6 983T→C becoming significant, with a greater effect size and even lower P value (Table2). No polymorphisms were in strong LD with CYP2B6 983T→C. In the third analyses, which adjusted for CYP2B6 516G→T and 983T→C, no other polymorphisms were associated with efavirenz concentrations (Table2). Linkage disequilibrium between the CYP2B6 polymorphisms is shown in Figure2. We only show polymorphisms with P values < 0.01 on unadjusted analyses, as well as CYP2B6 15582C→T and CYP2B6 -2320T→C (rs7251950) polymorphisms. The latter was included because it was reported to be in strong LD with CYP2B6 15582C→T in other populations 26. In our cohort, it was in weak LD with CYP2B6 15582C→T (r² =0.19).

Figure 2.

Linkage disequilibrium plot between polymorphisms in the CYP2B6 locus created in Haploview. Data from all 113 participants are included. Black denotes r² = 1, shades of grey, 0 < r² < 1, white, r² = 0. Only polymorphisms with P values <0.01 on unadjusted analyses, as well as CYP2B6 15582C→T and CYP2B6 -2320T→C (rs7251950) polymorphisms are shown. r² is displayed to show LD

Previously Reported Associations Beyond CYP2B6

Polymorphisms in genes beyond CYP2B6 reported to affect efavirenz pharmacokinetics include ABCB1 12,41, CYP2A6 7,30, CYP3A5 2, UGT2B7 7 and CAR 42,43. Univariate and multivariate analyses revealed no associations beyond CYP2B6 and the results are shown in Supplemental Material Table S4. The minor allele frequencies are also displayed. The CYP2A6 rs1801272 was monomorphic in our cohort. With only 18 participants with CYP2B6 slow metabolizer genotypes, we were unable to replicate a previously reported association with CYP2A6 -48T→G (rs28399433) in this group (P = 0.89).

Discussion

Efavirenz is one of the most extensively prescribed medications worldwide for HIV-1 infection, and multiple previous studies have associated CYP2B6 516 G→T 2–13 and CYP2B6 983 T→C 9,13,16,21–25, with increased plasma efavirenz concentrations. Our study replicated these associations in Black South Africans. In addition, we showed for the first time that CYP2B6 15582C→T was associated with plasma efavirenz concentrations in Black South Africans, which had previously only been reported for efavirenz in one study from the United States 26. In univariate analyses, a model that included composite genotype best predicted efavirenz concentrations. These associations were consistent in adults and children. An association between CYP2B6 15582C→T and slower plasma drug clearance has also been reported among Cambodians for the CYP2B6 substrate nevirapine 44, further supporting the validity of our finding. We did not find significant associations beyond CYP2B6 516G→T, 983T→C and 15582C→T. Although one analysis suggested an association with NR1I2 rs9847782, this was no longer apparent after excluding one outlier, suggesting a spurious association. Our study also provided information regarding minor allele frequencies for ABCB1, CYP2A6, CYP2B6, CYP3A4, NR1I2 and NR1I3 polymorphisms in a South African population.

We observed somewhat lower efavirenz concentrations in children, which is probably explained by the higher clearance in children compared with adults relative to their size 45. Although WHO weight-based dosing in children is intended to achieve efavirenz concentrations similar to those in adults with fixed dosing, high prevalence of low efavirenz concentrations in children has been observed in a South African population similar to ours, and has been attributed to the current weight based WHO guidelines 46. We assessed associations between polymorphisms and efavirenz concentrations separately in adults and children. In both children and adults, CYP2B6 516G→T had the strongest overall association with efavirenz concentrations, although CYP2B6 983T→C had the greatest effect size per allele. The association with CYP2B6 15582C→T could not be demonstrated in adults and children analyzed separately, possibly due to smaller samples sizes.

In the present analysis, minor allele frequencies of CYP2B6 516G→T, 983T→C, and 15882C→T were 0.36, 0.07 and 0.09, respectively. In the multivariate regression model including all three polymorphisms, CYP2B6 983T→C had the greatest magnitude of effect on log₁₀ efavirenz concentrations (β = 0.38 for 983T→C, β = 0.27 for 516G→T, β = 0.06 for 15882C→T), but due to lower frequency only explained approximately 8% of variance in univariate analysis vs. 19% for 516G→T, and negligible effect for 15882C→T. In sensitivity analyses that excluded two individuals with extreme outlier efavirenz concentrations, the variance explained by all three polymorphisms increased from 34% to 45%.

Our analyses included two participants with extreme outlier efavirenz concentrations, and in neither of the cases could we identify an explanation. Dose-to-sampling times were reported to be within the 10–20 h mid-dose interval window and there were multiple efavirenz determinations from each participant. However, non-adherence in the participant with the lowest concentrations cannot be excluded. Sensitivity analyses that excluded these two participants allowed a significant association between CYP2B6 15882C→T and log₁₀ efavirenz concentrations to be demonstrated. This same approach to censor outliers was used by Bertrand et al. 44 and highlights the critical importance of minimizing phenotype misclassification for detecting true associations.

Our primary analyses considered genetic associations with log₁₀-transformed efavirenz concentrations. Where data from multiple samples were available from the same individual, the time adjusted average was used. We also employed the hierarchical model to consider repeated measurements taking into account variation on fixed (CYP2B6 516G→T, 983T→C and 15582 C→T genotypes, age group, time after dose) and random (individual) effects. The advantage of this analysis is that it provides more information about the contribution of genotype to the increase in efavirenz concentrations when all other parameters are equal.

Previous studies from Africa have replicated the association between CYP2B6 516G→T and plasma efavirenz exposure, including studies of patients from Zimbabwe 8,9,13, South Africa 10,11,17,25, Ghana 7,24, Uganda 12,13, Tanzania 14,15, Rwanda 16 and Ethiopia 14. Multiple studies of patients from Africa have also shown the association with 983T→C 9,13,16,21,24,25. Data for association beyond CYP2B6 are limited. A study of patients in Ghana found associations with CYP2A6 and UGT2B7 polymorphisms 7, but a subsequent study of patients in Ghana did not replicate independent associations with these polymorphisms 24.

The association between CYP2B6 15582C→T and plasma efavirenz concentrations was first discovered and replicated in a genome-wide association study from the United States 26. This polymorphism was also associated with increased plasma efavirenz concentrations in Cambodians receiving concomitant antituberculosis therapy with isoniazid and rifampicin 47. In Cambodians receiving nevirapine, CYP2B6 15582C→T was also associated with decreased plasma nevirapine clearance, as was CYP2B6 -2320T→C (rs7251950), which was in strong LD with CYP2B6 15582C→T in that study 44. In contrast, in the present study CYP2B6 -2320T→C was in weak LD with 15582C→T (r² = 0.19) and was not associated with efavirenz concentrations. In addition, Lamba et al. reported that CYP2B6 15582C→T was associated with lower hepatic CYP2B6 expression in females 48. In the present study, while the association with CYP2B6 15582C→T was only significant in analyses that included both adults and children and with outliers excluded (to provide a cleaner phenotype), this association was highly likely to be valid considering the above reports in non-African populations. Additional studies on other populations will help to replicate and refine this association further.

This study increases our understanding of efavirenz pharmacogenetics and has potential clinical implications. Because many patients (especially those with intermediate and slow metabolizer CYP2B6 genotypes) have plasma efavirenz exposure in considerable excess of what is needed to control HIV-1 replication, there has been interest in alternative dosing strategies for efavirenz. The ENCORE1 study evaluated routinely initiating a lower dose of efavirenz (400 mg rather than 600 mg) in adults without genetic testing, and showed somewhat improved tolerability and no apparent loss of antiviral efficacy 49. If that approach is to be translated into clinical practice, the present study suggests that the group at greatest risk for subtherapeutic efavirenz concentrations will be patients who lack polymorphisms at all three loci (i.e. CYP2B6 516GG-983TT-15582CC). Only by genotyping for all three polymorphism can this group be defined. Consideration has also been given to using genetic testing to individualize better efavirenz dosing. This may be most practical in paediatrics, where individualized weight-based dosing is already used. In this setting, testing for polymorphisms at all three loci will allow the most precise dosing.

The present study considered combined analyses involving both adults and children, as well as subgroup analyses involving adults and children separately. A potential concern with combined analyses is the different approach to dosing efavirenz, with adults receiving a uniform dose and children receiving a weight-based dose, which might offset the advantage of a larger sample size. Several considerations provided reassurance that combined analyses were not problematic, and improved our likelihood of finding true associations. First, associations with the known loss-of-function polymorphisms, CYP2B6 516G→T and 983T→C, were far more significant in combined analyses than in subgroup analyses, suggesting that our ability to detect true associations was increased, not decreased, by combining groups. Second, the association with CYP2B6 15582C→T only achieved significance in combined analyses. Third, the rationale for a weight-based dose in children was to approximate plasma exposure in adults, which reduces concern about combining groups.

There were limitations to the present study. Sample size limited our power to identify novel genetic associations for polymorphisms that were infrequent or had small effects, even with the dataset that included both adults and children. For example, assuming a standard deviation of 0.33 for log₁₀ efavirenz concentration within each genotype group and a type II error of 0.2, we have 80% power for the following differences in pair-wise means homozygotes and heterozygotes: with 59 subjects and MAF = 0.1 a difference of 0.24 log₁₀ (∼75% absolute), with 59 subjects and MAF = 0.3 a difference 0.19 log₁₀ (∼55% absolute), with 113 subjects and MAF = 0.1 a difference of 0.17 log₁₀ (∼50% absolute), with 113 subjects and MAF = 0.3 a difference of 0.14 log₁₀ (∼40% absolute). Small sample size also limited our ability to test for interactions between CYP2B6 genotype and age in children, although we did not anticipate such interactions. Because the dose of efavirenz was not observed, we could not be certain of dose sampling times.

In summary, the present study improves the understanding of genetic determinants of efavirenz plasma exposure in an African population, including adults and children. Studies of associations between efavirenz concentrations and polymorphisms in African populations should consider CYP2B6 516G→T and 983T→C, and ideally also 15582C→T.

Author Contributions

PZS participated in the study design, DNA extraction, genotyping, acquisition of data, data analysis and interpretation and drafted the manuscript. PDL participated in the genotyping, data analysis and interpretation and critically revised the manuscript. HMM participated in the study design, data interpretation and critically revised the manuscript. PJS performed the analysis of the pharmacokinetic samples and helped to draft the manuscript. JAD participated in the study design, acquisition of data and critically revised the manuscript. NSL participated in study design and critically revised the manuscript. GM participated in study design, data interpretation and critically revised the manuscript. DWH participated in the study design, genotyping, data analysis and interpretation and drafted the manuscript. All authors read and approved the final manuscript.

Competing Interests

All authors have completed the Unified Competing Interest form at www.icmje.org/coi_disclosure.pdf (available on request from the corresponding author) and declare DWH (Principal investigator) has been a consultant to Merck. This work was supported in part by National Institute of Allergy and Infectious Diseases grants AI-077505, TR-000445 (DWH), Discovery Foundation, Welcome Trust, SA Medical Research Council and the National Health Scholar Program (PZS). The funding bodies had no role in study design, in collection, analysis and interpretation of the data, in writing of the manuscript and in the decision to submit the manuscript for publication. The rest of the authors declare no support from any organization for the submitted work and there are no other relationships or activities that could appear to have influenced the submitted work.

Supporting Information

Additional Supporting Information may be found in the online version of this article at the publisher's web-site: