Importance of protease inhibitor plasma levels in HIV-infected patients treated with genotypic-guided therapy: pharmacological data from the Viradapt Study

Objective: In a prospective randomized study, the impact of plasma protease inhibitor (PI) trough levels on changes in HIV RNA were assessed in patients treated with genotypic-guided therapy.

Cited Here...: Patients failing combination therapy (HIV-1 RNA > 10 000 copies/ml, and at least 6 months of therapy with nucleoside analogues and 3 months with PI) were randomly assigned into two arms: control group (C) in which the treatment was modified according to the standard of care; genotypic group (G) in which the treatment was modified according to resistance mutation profiles. Serial PI plasma levels were performed in patients throughout the 12 month study. PI levels were determined by high performance liquid chromatography. `Suboptimal' concentration (SOC) was defined as at least two PI plasma levels below 2 × IC<sub>95</sub>. Others were defined as `optimal' concentration (OC). Patients were categorized into four groups: G1 (SOC/control); G2 (OC/control); G3 (SOC/genotype); G4 (OC/genotype). An intent-to-treat analysis was performed with viral load as the primary endpoint.

Cited Here...: A total of 81 patients [mean age 39.7 ± 8 years, 59 men, 52.7% Centers for Disease Control and Prevention (CDC) stage C] were included in the pharmacological substudy. The two groups according to randomization arms were comparable in terms of risk factor, age, sex, previous treatments, baseline CD4 cell count, HIV-1 RNA and mean PI plasma concentrations. Linear regression analysis showed a significant relationship between PI concentration and HIV RNA in the plasma. OC and SOC were found in 67.9% (55/81) and 32.1% (26/81) of patients, respectively. Mean changes in HIV RNA from baseline at month 6 were: -0.23 ± 0.29 log₁₀ copies/ml (G1); -0.97 ± 0.28 (G2); -0.68 ± 0.37 (G3); -1.38 ± 0.20 (G4). Multivariate analysis showed PI plasma concentrations to be an independent predictor of HIV-RNA evolution (P  = 0.017).

Cited Here...: Multiple parameters determine the response to antiretroviral therapy and causes other than the development of drug resistance should be considered in the setting of therapeutic failure. Suboptimal concentrations of PI limit the response to antiretroviral therapy. Therapeutic drug monitoring of the PI plasma concentration may therefore prove useful in optimizing antiretroviral therapy.

Introduction

Multiple parameters determine the response to antiretroviral therapy, and many causes other than the development of viral drug resistance should be considered when analysing treatment failures. Despite the benefit seen with genotypic-guided treatment, 70% of patients in the Viradapt Study ^[1] did not achieve complete viral suppression. Among other factors determining treatment efficacy, suboptimal concentrations of antiretroviral drugs may play a major role. In contrast to reverse transcriptase (RT) inhibitors, significant correlations between in-vivo antiviral activity and plasma drug concentrations have been demonstrated for HIV protease inhibitors (PI). PI plasma drug levels have been correlated with the decline in viral load in clinical studies ^[2-10]. The objective of this study was to correlate PI plasma levels with changes in HIV-RNA levels. We also attempted to determine the multiple factors contributing to the efficacy of genotypic-guided antiretroviral therapy in patients failing highly active antiretroviral therapy.

Methods

Eighty-seven of the 108 patients from both arms participating at the Nice centre were included in the pharmacological subanalysis of the Viradapt Study. Entry criteria were: a plasma HIV RNA remaining over 10 000 copies/ml despite at least 6 months of treatment with nucleoside analogues and at least 3 months with a PI, age over 18 years and a Karnofsky score above 50. The exclusion criteria were: a haemoglobin level below 6 mmol/l, an absolute neutrophil count less than 0.8 × 10⁹/l, a platelet count below 50 × 10⁹/l, a creatinine level above 200 μmol/l and liver enzyme levels over five times the upper limit of normal.

Trial design

The Viradapt Study randomly assigned heavily pretreated patients failing their current antiretroviral therapy into two arms: a control arm and a genotypic testing arm. Patients in the study arm had their therapeutic regimen changed in accordance with the resistance-associated mutations found in the RT and protease genes. Patients in the control arm had their therapy changed following standard of care using the published guidelines for antiretroviral use ^[11]. As a result of an interim analysis showing a significant advantage in the amount of viral suppression in the genotypic arm, both arms were offered a genotypic-guided treatment change during the second 6 months of the trial if needed ^[1]. Clinical status and adverse effects were assessed at each study visit. CD4 cell count, HIV RNA and genotype were assessed every 3 months. If the viral load failed to decline at least 0.5 log over a 3 month period, therapy could be changed again according to the randomization arm for the first 6 months and based on the latest available genotype in both arms for the second 6 months of the trial.

Plasma protease inhibitor drug level

Serial PI drug levels were analysed monthly during the 48 week study period. Plasma PI concentrations were measured by high performance liquid chromatography. The four PI (ritonavir, indinavir, saquinavir or nelfinavir) were extracted from plasma by liquid-liquid extraction with diethyl-ether. An external standard was used for calculation. Standards and control samples were prepared in human plasma and extracted under the same conditions as those of the patient samples. The PI were then separated on two analytical columns after a switching columns procedure was applied. Their concentration was determined using ultraviolet detection at a wavelength of 220 nm. Linearity of the assay has been validated for each drug up to 12 mg/l for each drug analysed when at least 0.5 ml of plasma was extracted. The intra- and interassay coefficient of variation was less than 7%. The limit of detection in plasma samples was 0.05 μg/ml ^[12]. Patients were instructed not to take the morning dose until after drug levels were drawn. The analysis was performed on batched frozen samples and levels were determined for the four PI utilized in the study. Data were analysed only for patients from whom at least three samples were obtained. A wide range of pharmacokinetic data are available in the literature for the various PI ^[13,14]. A consensus statement standardizing these values for the PI was not found. The values for IC₉₅, C_max, C_min, and efficacy threshold concentrations used for the various PI are presented in Table 1, based on the literature available. The threshold value of twice the IC₉₅ shown in Table 1 was used as a cutoff for optimal PI levels.

Table 1

Genotypic assays

Sequencing of the major part of the RT gene (codons 25-230) and the entire protease gene were performed on plasma HIV RNA. HIV RNA was extracted from patient plasma samples, and then amplified using reverse transcriptase-polymerase chain reaction (RT-PCR) ^[1]. To detect which mutations were present in the patient HIV-1 viral RNA, the resulting sequence for each sample was then compared with a database containing known drug resistance mutations. Classification of the mutations into primary, secondary and polymorphisms associated or not associated with decreased drug sensitivity was established according to the consensus statement on antiretroviral drug resistance testing ^[15]. HIV RNA was measured using the Roche PCR Amplicor assay (Roche, Basel, Switzerland). The limit of detection for the genotyping technique was 1000 copies/ml.

Treatment

Decisions regarding genotypic-guided therapeutic changes were made according to a resistance/mutations table ^[1]. When specific mutations were found, corresponding drugs were not considered for treatment. During the study period new drugs (nelfinavir, nevirapine, efavirenz and abacavir) became available and were incorporated in the therapeutic regimens.

Statistical methods

All patients who attended a study visit and who have been collected from for study purposes after the initial visit were included in the Viradapt analysis. The primary endpoint of the original study was the variation of HIV RNA from baseline at months 3 and 6 (log₁₀ transformed). Secondary endpoints were: the proportion of patients with an HIV RNA below the level of detection of 200 copies/ml, the absolute CD4 cell count, and the correlation between plasma PI levels and the observed change in viral load.

For the pharmacological substudy, `suboptimal' concentration (SOC) was defined as at least two PI plasma trough levels below a threshold defined as two times the published IC<sub>95</sub> of a particular PI. IC<sub>95</sub> values and other pharmacological parameters are detailed in Table 1. Patients not achieving these criteria were defined as having `optimal' PI concentration (OC). Patients were categorized into four groups: G1 (SOC/control); G2 (OC/control); G3 (SOC/genotype); G4 (OC/genotype). For the first 6 months of the randomized study, analysis was performed using the last observation carried forward statistical technique. The difference in the proportion of patients with a viral load of less than 200 copies/ml was analysed in an intent-to-treat (dropout equal failure) manner. For the second 6 months of the open study, statistics were performed using an on-treatment analysis method. The mean change in HIV RNA was compared by analysis of variance. Non-parametric tests were used to determine the P value of the pharmacokinetics data. The correlation between quantitative variables (HIV-RNA and plasma PI level) was evaluated by linear regression. Chi-squared tests or the Fischer's exact test were used for nominal variables. Multivariate analysis (logistic regression ) was used to determined the independent and predictive values of virological response (HIV RNA below the level of detection at months 3 and 6). All P values reported are two-tailed and all confidence intervals (CI) are 95%. A P value of less than 0.05 was considered statistically significant. Data were analysed using the StatviewTM version 5.0 software.

Results

Eighty-seven of the 108 patients participating in the Viradapt study were followed at the Nice centre. Six patients were excluded because they had less than three samples collected, and 81 patients were included in the pharmacological substudy. The mean age was 39.2 ± 7 years, there were 59 men, 50.6% were Centers for Disease Control and Prevention (CDC) stage C. The two groups according to randomization arms were comparable in terms of risk factor, age, sex, previous treatments, CDC stage, baseline CD4 cell count and HIV-1 RNA. One thousand and fifty-two PI plasma levels were evaluated (saquinavir 471 samples, ritonavir 408, indinavir 38, nelfinavir 135). The median plasma levels were not significantly different between the two randomization arms. Median concentrations are expressed in μg/ml and are shown in Table 2.

Table 2

Plasma drug levels and HIV RNA

Linear regression analysis showed a statistically significant correlation between plasma concentration and HIV-RNA levels for each PI at timepoints when both values were available. Higher drug concentrations were correlated with lower HIV-RNA levels for all four PI (Fig. 1). According to our efficacy thresholds, 32.1% (26/81) of the patients had SOC and 67.9% (55/81) had OC. Similar results were observed when IC₅₀ was used as the cutoff, 59 out of 81 (72.8%) patients had plasma levels above IC₅₀. The distribution of SOC and OC were similar in both randomization arms. Patients who had OC had a -1.28 ± 0.17 log reduction in viral load at week 48 compared with -0.36 ± 0.24 log copies/ml in the SOC patients (P  = 0.0048) (Fig. 2). Patients were categorized on the basis of the randomization arm and drug levels: group 1: SOC/control (n = 13); group 2: OC/control (n = 22); group 3: SOC managed with genotypic-guided treatment (n = 13); group 4: OC managed with genotypic-guided treatment (n = 33). When patients were divided into these four groups the week 24 viral load reductions from baseline were: -0.23 ± 0.29 log₁₀ copies/ml (G1); -0.97 ± 0.28 (G2); -0.68 ± 0.37 (G3); -1.38 ± 0.20 (G4). These data are graphically shown in Fig. 3. At month 3, HIV-1 RNA was below the level of detection (200 copies/ml) in 0% (0/13), 22.7% (5/22), 15.4% (2/13), and 33.3% (11/33) in groups 1, 2, 3, and 4, respectively. At month 6, HIV-1 RNA was below 200 copies/ml in 0% (0/13), 22.7% (5/22), 8.3% (1/12), and 39.4% (13/33) in groups 1, 2, 3, and 4, respectively (Fig. 4).

Fig. 1	Fig. 2	Fig. 3
Fig. 4

Treatments

After genotypic adaptation, 36 out of 81 (44.4%) patients were on one PI containing combination and 19 out of 36 (52.8%) were in SOC; 45 out of 81 patients were on dual PI combination but only seven out of 45 (15.6%) had SOC (Table 3).

Table 3

Multivariate analysis

In a multivariate analysis the following factors were found to affect the viral load response independently: drug concentration greater than IC₉₅ × 2, odds ratio (OR) -2.04 (CI -035-3.73, P  = 0.018), use of genotyping, OR 1.59 (CI 0.20-2.97, P  = 0.025), and the presence of primary PI resistance mutations, OR -1.96 (CI -0.40-3.50, P  = 0.014).

Discussion

In this study, PI plasma levels were inversely corrrelated with plasma HIV RNA levels. `Optimal' PI drug levels, following our criteria, correlated with an increased reduction in HIV-RNA viral load. Treatment change based on genotypic analysis and `optimal' PI drug levels were both independent predictors of a favourable 24 week virological response.

Despite the use of genotyping, many of our patients did not achieve complete viral suppression. Causes other than drug-resistant virus must be considered when analysing treatment failure. In many patients, resistance mutations alone could not fully explain the observed virological failure. In the absence of primary resistance-related mutations, failure is most likely to be due to other causes such as non-compliance, insufficient drug dosage, lack of absorption, poor drug intracellular metabolism, drug-drug interaction ^[16], overexpression of p-glycoprotein ^[17], release of virions from sanctuaries, or possibly clinically significant minor viral variants.

Insufficient plasma levels of antiretroviral drugs are a major cause of therapeutic failure. Effectively, recent studies have shown a correlation between plasma PI concentrations and antiviral effect. In a series of patients treated with variable doses of saquinavir hard gels ^[4], a larger and more durable improvement in plasma HIV-RNA levels and CD4 cell count was observed with higher doses of saquinavir compared with the published results of the standard saquinavir hard gel dose (1800 mg/day). In another series ^[6], linear regression analysis showed a significant correlation between the saquinavir plasma concentration and the decline in HIV-RNA levels at 12, 36 and 48 weeks. In a cohort of 65 patients receiving indinavir ^[7], in addition to high baseline viral load and previous PI use, low PI plasma levels were a major risk factor for virological failure. The 8 h area-under-the curve (AUC) for indinavir has been shown to be significantly higher in patient with viral loads below detection limit than in patients with a detectable HIV-RNA level. In the Trilège Study, most of the patients failing therapy harboured only the M184V mutation, which could not explain failure. However, indinavir plasma levels were found to be in the low ranges in the majority of these patients ^[18]. These patients were shown to be non-adherent to their therapeutic regimen. A recent study ^[19] showed that self-reported non-adherence (< 80%) predicted non-response of HIV RNA at 6 months. Higher PI plasma levels have also been demonstrated to delay the appearance of resistance mutations ^[20]. However, we did not analyse the sequential apparition of resistance mutations in the context of insufficient PI plasma levels.

Plasma drug dosages are available for the three antiretroviral classes but are of variable utility. PI seem to be the best candidate for therapeutic drug monitoring (TDM) as the AUCτ and trough levels appear to be related to virological outcome. Available data suggest a significant inter- and intra-patient pharmacokinetic variability of PI levels. In the literature, the pharmacokinetic and efficacy data described for the various PI are in a wide range ^[13,14]. Theoretically, the aim of antiretroviral therapy is to obtain plasma level above the IC₉₅ of the virus, taking into account mitigating factors that might lead to reduced free concentrations such as protein binding and intracellular diffusion. The estimation of concentrations required for in-vivo activity using in-vitro assays is dependent on the binding of many PI to serum protein, which can attenuate their antiviral activity ^[21,22]. Defining a correct threshold for use in a clinical situation is difficult. An arbitrary threshold value of twice the IC₉₅ was used as a cutoff for `optimal' PI levels in our analysis. It was found that 32.1% of our patients had `suboptimal' concentrations according to our threshold. The majority of these patients even had no detectable PI plasma concentration. One major cause of low plasma PI drug levels is non-compliance. In our study, no compliance evaluation was performed, although by measuring drug levels, we probably discriminate between patients adherent or non-adherent to their therapy. Using a different cutoff such as IC₅₀ only slightly modified the repartition of patients into `optimal' or `suboptimal' concentration groups (Table 3). Our definition of the `optimal' level can not be extrapolated to other situations. Effectively, only a minority of patients with genotypic-guided therapy and `optimal' drug concentrations exhibited complete viral suppression. The use of different cutoffs for `optimal' drug levels and the definitions for resistance-associated mutations may have influenced these results. Genotypic-guided therapy, drug concentrations and the presence of primary, resistance-related, protease mutations were all found to be factors that independently affect the response to therapy in experienced patients. Therapeutic drug monitoring of plasma PI concentration along with resistance testing may have a place separately and together in optimizing antiretroviral therapy, and might be helpful in understanding the causes of therapeutic failure.

Conclusion

Additional prospective clinical trials may be necessary to demonstrate conclusively that the monitoring of the PI plasma concentration provides virological response benefits to patients and decreases toxicity rates.