Nevirapine Plasma Exposure Affects both Durability of Viral Suppression and Selection of Nevirapine Primary Resistance Mutations in a Clinical Setting

Abstract

The relationship between nevirapine plasma concentrations and the durability of both viral suppression (VS) and selection of nevirapine primary resistance mutations (PRMs) was evaluated. A nevirapine trough concentration (C_trough) of >4,300 ng/ml was found to predict longer VS. Patients with nevirapine C_troughs ranging from 3,100 to 4,300 ng/ml had higher probabilities of developing PRMs than those with nevirapine C_troughs below and above this concentration interval.

In antiretroviral (ARV)-naive patients, nevirapine (NVP) plasma exposure was shown to affect the time to reach virological suppression (VS) and overall response at 52 weeks (5). However, no data have yet been obtained about the effects of NVP plasma concentrations on the long-term durability of VS in a heterogeneous clinical setting. Moreover, failure of NVP-containing regimens is known to be frequently, but not invariably, associated with the appearance of NVP-associated primary resistance mutations (NVP-PRMs) (2). Nevertheless, the influence of NVP plasma exposure on the selection of NVP-PRMs is unknown.

Therefore, our aims were to evaluate the relationship between NVP trough concentrations (C_trough) and the durability of VS and to assess the effect of NVP C_trough on the selection of NVP-PRMs.

Nonnucleoside reverse transcriptase inhibitor (non-NRTI)-naive patients who initiated, between 2000 and mid-2003, a regimen containing NVP (200 mg, twice a day) plus two NRTIs were retrospectively selected from our cohort in January 2004. These patients included ARV-naive subjects, ARV-experienced subjects with virological failure (VF), and patients switching from protease inhibitor-based regimens with viral loads (VLs) of <50 copies/ml. Regular follow-up (every 3 months), achievement of VS, availability of at least one detectable NVP concentration, and self-reported compliance (no more than two drug intakes missed in the last week at the time of pharmacokinetic sampling) were the main inclusion criteria. Baseline VL and CD4⁺ counts were obtained by reverse transcriptase PCR (Amplicor HIV monitor; Roche Molecular Systems, Switzerland) and flow cytometry, respectively. VF of an NVP-based regimen was defined as a VL of >50 copies/ml confirmed in two sequential determinations. In subjects with VF, the length of VS was recorded. In those maintaining VS, time was censored on 28 January 2004.

Genotypic-resistance testing (automatic sequencer ABI 377; PE Biosystems, Foster City, CA) was performed on the first plasma sample available with a VL of >1,000 copies/ml. Amino acid changes at positions L100I, K103N, V106A/M, V108I, Y181C/I, Y188L/C/H, G190A, and M230L of the pol gene were considered NVP-PRMs according to International AIDS Society-USA guidelines (www.iasusa.org).

Plasma samples for pharmacokinetic analysis were obtained before morning doses at various times after the initiation of NVP intake. NVP concentration was determined by using a reversed-phase high-performance liquid chromatography method with UV detection modified from a method of Aymard et al. (1). Our method was linear between 25 and 16,000 ng/ml and had mean interassay and intra-assay variabilities of 6.5% and 3.7%, respectively. If a sample was not obtained at the end of the dose interval, NVP concentration was normalized to the C_trough by using Mon Clin software (version 1.0.3). Mean values were considered for patients with multiple C_trough measurements over time.

Cutoffs of NVP C_trough were obtained by using the receiver operator characteristic (ROC) curve (3). Cox proportional-hazards models (forward conditional method) and logistic regression analysis were used to investigate the independent predictors of maintaining VS and developing PRMs, respectively. In both cases, NVP C_trough cutoffs, ARV-naive status, and VLs of <50 copies at baseline were analyzed as independent predictors. Cumulative probabilities of maintaining VS and of selecting PRMs were studied by using Kaplan-Meier analysis. Chi-square and Student's t tests were applied as needed. Statistical analysis was performed using NCSS software (2004 version; Kaysville, Utah).

Among the 236 subjects considered over the study period, 178 patients were eligible. Exclusions were due to a patient not having achieved VS (n = 11), no C_trough availability (n = 22), poor adherence (n = 15), and drop out (n = 10). Characteristics of patients are reported in Table 1. VF was observed in 90/178 (50.6%) patients (Table 2). VF was significantly more frequent in ARV-experienced patients and in subjects with detectable VLs at baseline (Table 2). Higher NVP C_trough were found in patients maintaining VS. The ROC curve test showed that an NVP C_trough of >4,300 ng/ml had sensitivity and specificity values of 68.2% and 63.3%, respectively, in predicting the maintenance of VS. Upon univariate Cox regression analysis, NVP C_trough of >4,300 ng/ml and ARV-naive patients, or those with undetectable VLs at baseline, independently predicted VS (odds ratio [OR] [95% confidence interval {CI}], 2.7 [1.75 to 4.16], P < 0.0001; 3.04 [1.77 to 5.23], P < 0.0001; and 7.14 [2.6 to 20], P < 0.0001, respectively). These results were confirmed by Kaplan-Meier analysis, in which patients with NVP C_trough of >4,300 ng/ml or who were ARV naive or had undetectable VLs at baseline had higher probabilities of maintaining VS for a longer time (Fig. 1). Multivariate Cox regression analysis definitively confirmed that these three variables independently predicted the maintenance of VS (OR [95% CI], 2.6 [1.69 to 4.01], P < 0.0001; 5 [2.89 to 8.26], P < 0.0001; and 12.6 [4.6 to 34.4], P < 0.0001, respectively).

TABLE 1.

Baseline characteristics of study population and pharmacokinetics results

Characteristic	Value
No. of patients	178
Demographics
No. of malesa	133 (74.7)
Age (yr)b	39 (34-44)
No. of HCV coinfected patientsa,d	68 (38.2)
Previous therapy
No. of ARV-experienced patientsa	119 (66.9)
Time on ARV therapy (days)b	2,095 (619-1,578)
No. of previous NRTIsc	2 (2-5)
Virology
HIV RNA (log no. of copies)b	4.04 (1.3-4.8)
No. of patients with HIV RNA at <50 copies/ml at baselinea	35 (19.7)
Immunologyb
No. of CD4+ T cells/mmc	347 (241-478)
Pharmacokinetics
Sampling time (h postdose)c	12.1 (6.7-12.8)
Total no. of samples	297
No. of samples/patientc	1 (1-6)
NVP Ctrough (ng/ml)b	4,471 (3,305-6,302)
Time to NVP Ctrough measurement (days)c	518 (21-1,984)

^aValue is expressed as the number of subjects (%).

^bValue is expressed as the median (IQR).

^cValue is expressed as the median (range).

^dHCV, hepatitis C virus.

TABLE 2.

Proportion of subjects with VF or maintenance of VS according to ARV-naive status, VL at baseline, and NVP C_trough

Characteristic	Value for subject with:		P value
	VF	VS
No. of subjects (%)	90 (50.56)	88 (49.43)
Time (days) with undetectable VLa	300 (146-456)	1,116 (928-1607)	<0.0001c
No. of naive subjects (%)	17 (18.8)	42 (47.7)	<0.0001b
No. of subjects with VLs of <50 copies/ml at baselinea	4 (4.4)	31 (35.2)	<0.0001b
NVP Ctrough (ng/ml)a	3,894 (2,873-5,195)	5,419 (3,869-7,248)	0.001c
No. of subjects with NVP Ctrough of >4,300 ng/ml (%)	33 (36.6)	60 (68.18)	<0.0001b

^aValues are expressed as medians IQRs.

^bBased on a chi-square test.

^cBased on Student's t test.

FIG. 1.

Kaplan-Meier curves for the time of maintenance of undetectable VL after reaching viral suppression. The probabilities represent the proportions of patients who showed virological failure. (A) Patients with NVP C_trough of <4,300 ng/ml (solid line) or ≥4,300 ng/ml (broken line). (B) Patients who were experienced (solid line) or naive (broken line) to ARVs at the initiation of the NVP-containing regimen. (C) Viral loads of >50 copies/ml (solid line) or viral loads of <50 copies/ml at baseline (broken line).

Genotypic-resistance test results could be obtained for 59/90 patients with VF. NVP-PRMs were identified in 45/59 (76.3%) subjects (L100I, 0%; K103N, 22%; V106A/M, 16.9%; V108I, 6.7%; Y181C, 28.8%; Y188L, 5.08%; G190A, 20.3%; and M230L, 0%). Subjects for whom genotypic-resistance test results could not be obtained (n = 31) had a median (interquartile range [IQR]) NVP C_trough of 4,900 ng/ml (3,689 to 6,077 ng/ml). Among subjects who showed VF, those without NVP-PRMs had lower NVP C_trough than patients with NVP-PRMs (2,853 ng/ml [1,348 to 3,932 ng/ml] versus 3,826 ng/ml [2,922 to 4,740 ng/ml]; P = 0.041). ROC testing provided an NVP C_trough cutoff (>3,100 ng/ml) that had a sensitivity and specificity of 71.1% and 71.4%, respectively, in predicting the presence of NVP-PRMs at the time of VF. Multivariate logistic regression analysis showed that having an NVP C_trough of >3,100 ng/ml was an independent predictor of developing NVP-PRMs in patients with VF (OR [95% CI], 6.15 [1.6 to 23.2]; P = 0.007), whereas no predictivity value nor association was found for either ARV-naive status or undetectable VL at baseline.

In the analysis of total population, patients with no VF were considered as lacking NVP-PRMs. Patients with VF but without genotypic-resistance test results available (n = 31) were excluded from the analysis. A total of 147 subjects were considered and categorized according to NVP C_trough in three groups according to previous cutoffs obtained for both efficacy in total population and appearance of NVP-PRM in subjects with VF: <3,100, 3,100 to 4,300, and >4,300 ng/ml. A Kaplan-Meier analysis curve showing the probability of developing NVP-PRMs over time according to NVP C_trough range is shown in Fig. 2. Associations between NVP concentration ranges and the presence of NVP-PRMs and VF are reported in Table 3.

FIG. 2.

Kaplan-Meier curves for the times of developing NVP-PRMs at virological failure. The probabilities represent the proportions of patients who developed NVP-PRMs. Curves show patients with high (>4,300 ng/ml [dashed line]), intermediate (3,100 to 4,300 ng/ml [solid line]), or low (<3,100 ng/ml [broken line]) NVP C_trough.

TABLE 3.

Proportion of subjects with NVP-PRMs according to NVP C_trough ranges by considering patients with VF and the total study population

Patient group	Valueb for subjects with NVP Ctrough of:			P valuea
	<3,100 ng/ml	3,100-4,300 ng/ml	>4,300 ng/ml
Patients with VF (n = 59)
No. of subjects	23	20	16
No. of ARV-naive subjects	9	2	0	0.004
No. of subjects with NVP-PRMs	13 (56.52)	19 (95)	13 (81.25)	0.01
Total population (n = 147)
No. of subjects	35	36	76
No. of ARV-naive subjects	17	7	29	0.03
No. of subjects with NVP-PRMs	13 (37.1)	19 (52.7)	13 (17.1)	<0.0001
No. of subjects with VF	23 (65.7)	20 (55.5)	16 (21.05)	<0.0001

^aBased on a chi-square test.

^bValues in parentheses are percentages.

Our study showed that once VS is achieved after initiation of an NVP-containing regimen, NVP plasma exposure is a contributory factor in the durability of VS over time. In fact, an NVP C_trough of >4,300 ng/ml, along with ARV-naive status and undetectability of VL at baseline, was shown to independently predict maintenance of VS for a longer time (median time, >3 years).

Moreover, NVP is known to have a low “genetic barrier.” Development of single nucleotide mutations at critical sites of the pol gene is related to a >100-fold loss of sensitivity to NVP (2). In our study, the selection of NVP-PRMs was found to be associated with NVP plasma concentrations. Natural variation of susceptibility to NVP among wild-type strains of human immunodeficiency virus type 1 (HIV-1) has been reported to be around a 4.5-fold change (4). In patients with lower NVP concentrations (<3,100 ng/ml), VF occurred in more than 65% of subjects, but NVP-PRMs were found in only 56.5% of that 65%, suggesting that NVP exposure was not enough to suppress all wild-type strains.

Similar rates of VF were seen in patients with NVP C_trough between 3,100 and 4,300 ng/ml (55.5%), but in such subjects, HIV-1 strains carrying NVP-PRMs were more frequently detected (95%). Within this concentration interval, a more complete inhibition of wild-type strains is likely to be obtained, although this concentration might still not be sufficient to suppress some natural polymorphic subpopulations of HIV-1. Prolonged maintenance of plasma concentrations at this level could allow the replication of HIV-1 strains with reduced susceptibility, thus promoting the appearance of NVP-resistant strains.

At higher NVP concentrations (>4,300 ng/ml), complete inhibition of all polymorphic natural variants is probably reached, as suggested by the lower rate of VF seen among these patients (21%). Nevertheless, 81.2% of patients who had VF with NVP C_trough of >4,300 ng/ml developed NVP-PRMs. A plausible interpretation is that irregular adherence over time could lead NVP C_trough to fall below a critical threshold, thus allowing HIV to replicate under suboptimal pharmacological pressure.

These findings prompt us toward the definition of a range of NVP concentrations in which the selection of resistant mutants is the highest (mutant selection window) (6). Based on these data, the in vivo mutant selection window could be defined as an NVP C_trough between 3,100 and 4,300 ng/ml. In fact, within this concentration range, a higher risk of NVP-PRM overgrowth has been seen than with concentrations above and below these limits. We could speculate that, although an initial virological response can be obtained at different plasma concentrations, maintenance of viral suppression for longer periods of time is reached only at higher concentrations at which replication of all wild-type subpopulations are inhibited.

Although NVP is here considered the driving force of a regimen also including two NRTIs, the contributory activity of the latter should also be taken into account. ARV-naive status was also found to be a determinant of durability of VS. It appears likely that the contributory antiretroviral effect exerted by NRTIs is reduced in ARV-experienced patients, in whom accumulation of NRTI-related mutations is frequent, and NVP is no longer efficiently protected by the coadministered drugs. Moreover, undetectability of VL at baseline has also been identified as a predictor of a more-prolonged VS.

Several limitations could influence the interpretation of this study. This was a retrospective analysis of a heterogeneous group of patients, with further limitations resulting from the exclusion of a minority of patients because of a lack of suitable plasma samples and/or genotyping. Adherence was not formally studied throughout NVP treatment, so the possibility of periods of inadequate adherence cannot be excluded. These results must be confirmed in prospective, controlled studies, and the possible value of dose adjustment must be prospectively evaluated. Nevertheless, these data provide a preliminary conceptual framework for the search of pharmacokinetic parameters that are able to predict treatment efficacy over the long term.