Risk factors contributing to a low darunavir plasma concentration.

Darunavir is an efficacious drug; however, pharmacokinetic variability has been reported. The objective of this study was to find predisposing factors for low darunavir plasma concentrations in patients starting the once- or twice-daily dosage. Darunavir plasma concentrations from January 2010 till December 2014 of human immunodeficiency virus-infected individuals treated in the outpatient clinic of the University Medical Center Groningen were retrospectively reviewed. The first darunavir plasma concentration of patients within 8 weeks after initiation of darunavir therapy was selected. A dichotomous logistic regression analysis was conducted to select the set of variables best predicting a darunavir concentration below median population pharmacokinetic curve. In total 113 patients were included. The variables best predicting a darunavir concentration besides food intake included age together with estimated glomerular filtration rate (Hosmer-Lemeshow test P = 0.945, Nagelkerke R2  = 0.284). Systematic evaluation of therapeutic drug monitoring results may help to identify patients at risk for low drug exposure.

What is Already Known about this Subject

Darunavir exposure is improved by concomitant food intake.

Darunavir has a high pharmacokinetic variability.

Factors such as demographics, concomitant medication and polymorphisms of cytochrome P450 3A4 isoenzymes are described as potential contributors to pharmacokinetic variability.

What this Study Adds

Age and estimated glomerular filtration rate are predictors of a darunavir plasma concentration below median population pharmacokinetic curve.

The combination of tenofovir and darunavir potentially leads to decreased estimated glomerular filtration rate and increased darunavir exposure and merits further investigation.

Systematic evaluation of therapeutic drug monitoring results may help to identify risk factors for low drug exposure.

Introduction

Darunavir (DRV) is a human immunodeficiency virus (HIV) protease inhibitor (PI) that is used for the treatment of HIV‐1 infection in antiretroviral treatment‐naïve and treatment‐experienced adults and paediatric patients aged ≥6 years 1. Once‐daily DRV 800 mg is approved for use in treatment‐naïve patients and the twice daily dosage of DRV 600 mg is approved for use in treatment‐experienced adults with DRV resistance‐associated mutations 2. As DRV is extensively metabolized by cytochrome P450 3A4 and being a substrate of P‐glycoprotein as well, it is coadministered with ritonavir (RTV) 100 mg or cobicistat 150 mg to increase its exposure 1, 3, 4, 5. Furthermore, concomitant food intake is advised to improve DRV bioavailability 6.

DRV is considered a safe and efficacious drug; however, substantial pharmacokinetic variability has been reported 1, 7. Factors such as demographics, treatment adherence, concomitant medication and polymorphisms of cytochrome P450 3A4 isoenzymes contribute to the observed variability 1, 5, 7. Pharmacokinetic variability can potentially result in suboptimal DRV plasma concentrations. Suboptimal DRV plasma concentrations are highly undesirable as they can lead to drug resistance, insufficient virological response or a virological breakthrough in patients who earlier had an undetectable viral load 8. Therapeutic drug monitoring is therefore routinely performed to assure adequate drug exposure 8.

Despite the relatively long use and experience with DRV in daily practice, little is documented concerning the potential risk factors for a relatively low DRV concentration in HIV‐infected individuals in an outpatient setting. The aim of the present study was to investigate the frequency as well as predisposing factors associated with DRV plasma concentration below the median population pharmacokinetic curve in an outpatient setting.

Methods

Study design and participants

We performed a retrospective study on demographics, measured DRV plasma concentrations and patient characteristics in HIV patients treated with DRV in the University Medical Center Groningen (UMCG), The Netherlands. All patients aged ≥18 years and using DRV were eligible for inclusion in this study. A database was created by extracting all measured DRV plasma concentrations from January 2010 to December 2014 from the UMCG electronic patient database. Due to intrapatient variability and potential dose adjustments, we considered only the first DRV level of each patient within 8 weeks after initiation of the DRV therapy for inclusion. DRV plasma concentrations were excluded if the time of ingestion relative to the collection of the blood sample was unknown. The ethical review board of the UMCG evaluated the study protocol and waived the need for written informed consent due to the retrospective nature of the study (METc 2015.010).

Data collection and management

The following data were extracted from the medical records of the participants and included in the research database: sex, age, weight (during visit of drug level measurement), height, drug abuse, documented adherence, comorbidity, serum creatinine concentration, aspartate aminotransferase, alanine aminotransferase, CD4+ cell count and viral load. The medical records of all participants were studied for medication potentially influencing the DRV concentrations.

When the possibility of nonadherence was suggested in the medical record, the patient was classified as potentially nonadherent. Data concerning drug abuse included alcohol abuse (persistent use of ≥14 units of alcohol weekly) 9 and the use of recreational drugs. Comorbidities with a known effect on pharmacokinetics were included in the database (e.g. renal, hepatic, and/or gastrointestinal morbidity). The estimated glomerular filtration rate (eGFR) was calculated using the Chronic Kidney Disease Epidemiology collaboration (CKD‐EPI) formula 10. The laboratory results (i.e. creatinine, aspartate aminotransferase, alanine aminotransferase, CD4+ cell count and viral load) included in the database corresponded with the date of the DRV plasma concentration or within a period of ±3 months. The De Ritis ratio 11 was calculated to interpret the hepatic function.

Bioanalytical procedure and pharmacokinetic analyses

The concentrations of DRV in human plasma were analysed in the Laboratory of the Department of Clinical Pharmacy and Pharmacology at the UMCG by a validated liquid chromatography–tandem mass spectrometry procedure. All analyses were performed on a Finnigan TSQ Quantum Discovery (San Jose, CA, USA) triple quadrupole liquid chromatography–tandem mass spectrometer with a Finnigan Surveyor LC pump and a Finnigan Surveyor autosampler. The calibration curves were linear within the concentration range of 0.335 to 33.5 mg l–1 for DRV and had a correlation coefficient (R2) of 0.999. The lower limit of quantification for DRV was 0.27 mg l–1. This method was precise and accurate as within‐day precision ranged between 2.2 and 3.2% for DRV, and between‐day precision ranged from 3.0 to 5.2%. The calculated accuracy ranged from 0.0% to 11.8%. Performance of the assay was within accepted limits of acceptance (accuracy and precision <10%) in our laboratory as confirmed by participation in the international proficiency testing program 12.

To estimate time‐adjusted DRV plasma concentrations, a DRV‐iterative two‐stage Bayesian population pharmacokinetic model with the software package MWPharm 3.82 (Mediware, Groningen, The Netherlands) was used 13. The model for DRV was a one‐compartment model with input and elimination from the central compartment. Parameters for this model are: a volume of distribution of the central compartment of 2 l kg–1 [standard deviation (s.d.) 0.5 l kg–1], a total body clearance of 6.3 l h–1 1.85m–2 (s.d. 1.57 l h–1 1.85m–2), first order absorption constant of 1 h–1 (s.d. 0.25 h–1) and a bioavailability of 0.8 (in combination with RTV). This model was built in‐house and derived from data provided in the literature 14. Similar to standard care, each time‐adjusted DRV plasma concentration was compared with a median population pharmacokinetic curve for the once‐daily or twice‐daily dosage of DRV 15. Consistent with daily practice the DRV plasma concentrations were dichotomized as either above or below the median curve. The median curves do not represent the minimal effective concentration, but are used in standard care as cut‐off values for follow‐up 15. A DRV trough concentration below 1.07 mg l–1 for the once‐daily dosage and 2.60 mg l–1 for the twice daily dosage is an indication for follow‐up. This follow‐up consists of repeated plasma drug concentration measurement, additional food intake advice and additional questions and guidance concerning therapy adherence.

Statistical analysis

Age, body mass index, eGFR, De Ritis ratio, comorbidity, drug abuse and documented adherence were tested for association with DRV plasma concentration above or below median population pharmacokinetic curve. We conducted a dichotomous logistic regression analysis with manual backward selection based on P‐values to identify which set of variables best predicted a DRV plasma concentration below the median population pharmacokinetic curve. Comorbidity, recreational drug abuse and documented adherence were marked as categorical covariates. We assessed the variance of the model using Nagelkerke R 2 and determined the goodness‐of‐fit by Hosmer–Lemeshow. Variables were checked for linearity and included as dummy variables (indicator variables) if necessary. All statistical analyses were performed using SPSS for Windows, version 22.0 (IBM SPSS, Chicago, Illinois).

Nomenclature of targets and ligands

Key protein targets and ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 16, and are permanently archived in the Concise Guide to PHARMACOLOGY 2015/16 17.

Results

In total, 113 initial DRV plasma concentrations from 113 patients were measured within 8 weeks after initiation of DRV therapy and were therefore included in the research database. Ninety‐four participants were using DRV/RTV 800/100 mg once‐daily and 19 participants were using DRV/RTV 600/100 mg twice‐daily. All participants were using RTV as a booster. Eighty‐eight (78%) participants were among others using tenofovir simultaneously with DRV and seven (6%) participants were using raltegravir simultaneously with DRV. No other concomitant use of nonantiretroviral medication interacting with DRV was found after assessing the medical records of the participants. Further demographic characteristics are presented in Table 1.

Table 1

Demographic characteristics of the 113 study participants at time of sampling

Characteristics	Value
Number of participants	113 (89 male)
Mean age (years)	43 (20–67)
Median BMI (kg m –2 ), IQR	23.50 (21.1–26.0)
BMI subgroups, n (%)
< 18 kg m –2	5 (4.4%)
18–25 kg m –2	64 (56.6%)
> 25 kg m –2	34 (30%)
Missing a	10 (9%)
eGFR (ml min 1.73 m –2 ), mean (SD)	103.3 (23.3)
Median AST (IQR)	29 (23–35)
Median ALT (IQR)	25 (17.5–33.5)
Median CD4+ cell count (IQR)	430 (285–605)
Undetectable viral load in 55 patients
Median Viral load in patients with detectable load ( n  = 58), copies ml –1 (IQR)	600 (242–1486)
Documented comorbidity
Number of patients with hepatic morbidity	6
Number of patients with gastrointestinal morbidity	4
Number of patients with renal morbidity	2
Drug abuse
Recreational drugs	18
Alcohol	7
Documented potentially nonadherent	9

Body mass index (BMI) of these individuals could not be calculated as only body weight was recorded. IQR, interquartile range; SD, standard deviation; eGFR, estimated glomerular filtration rate; AST, aspartate aminotransferase, ALT, alanine aminotransferase

The median (interquartile range) DRV plasma concentration was 2.8 (1.8–4.2) mg l–1 and 31 (27.4%) DRV plasma concentrations were classified below the median population pharmacokinetic curve. In Table 2, the patient characteristics and DRV drug concentrations below or above median population pharmacokinetic curve values are shown. Due to a lack of linearity with the dependant variable the variables age, eGFR and body mass index were included in the logistic regression as dummy variables divided into four quartiles. The model best predicting a DRV level below the median population pharmacokinetic curve included the variables age and eGFR (Hosmer–Lemeshow goodness‐of‐fit test P = 0.945, Nagelkerke R 2 = 0.284). The results of the dichotomous logistic regression are presented in Table 3.

Table 2

Patient characteristics and darunavir plasma concentrations below or above median population pharmacokinetic curve values

Variable	Below population median (n = 31)	Above population median (n = 82)	P‐value
Age mean (years)	41	44	0.138 (t test)
Sex (male), n (%)	21 (68%)	68 (83%)	0.078 (Pearson χ2)
BMI (median, IQR)	24.3 (21.3–27.5)	23.2 (20.8–25.7)	0.298 (MWU)
eGFR, mean (SD)	114.5 (16.9)	99.0 (24.1)	<0.001, t test
Earlier documented nonadherence, n (%)	5 (16.1%)	4 (4.9%)	0.062 (Fisher exact)
Detectable HIV viral load, n (%)	17 (54.8%)	41 (50%)	0.65 (Pearson χ2)
Recreational drugs/alcohol abuse	8 (26.8%)	17 (20.7%)	0.56 (Pearson χ2)
De Ritis, median (IQR)	1.3 (0.9–1.6)	1.1 (0.9–1.5)	0.33 (MWU)

MWU, Mann–Whitney U test; BMI, body mass index; IQR, interquartile range; SD, standard deviation; eGFR, estimated glomerular filtration rate; HIV, human immunodeficiency virus.

Table 3

Results of the dichotomous logistic regression analysis with manual backward selection

Variable	Significance	Odds ratio	95% CI for odds ratio
			Lower	Upper
Age
35 years	0.025
43 years	0.028	3.916	1.161	13.207
53 years	0.021	5.608	1.298	24.240
>53 years	0.712	0.634	0.056	7.138
eGFR
89 ml min –1	0.096
106 ml min –1	0.070	7.744	0.845	70.969
116.5 ml min –1	0.020	15.246	1.533	151.579
>116.5 ml min –1	0.016	19.458	1.721	220.025

CI, confidence interval.

Due to a lack of linearity with the dependant variable the variables age and eGFR were included in the logistic regression as dummy variables divided into four quartiles. The corresponding significance and odds ratio per quartile are shown.

Discussion

In this descriptive study we assessed the predictive value of patient‐ and disease‐related factors on DRV plasma concentrations after initiation of the DRV therapy. In a dichotomous logistic regression analysis, we found that the combination of age and eGFR were associated with a DRV plasma concentration below the median population pharmacokinetic curve.

Like other antiretroviral PIs, DRV pharmacokinetics are characterized by large interpatient variability 1, 7. The observed variability could partly be caused by inadequate food intake as DRV is advised to be taken concomitantly with food to achieve adequate exposure 6. To identify other potential contributors to a DRV plasma concentration below the median population pharmacokinetic curve, besides food intake, in this study we focused on several patient‐ and disease‐ related factors. More than a quarter (27.4%) of the DRV plasma concentrations were below the median curve triggering physicians to change their follow‐up strategy. Theoretically it would be expected that 50% of the DRV plasma concentrations would be below the median curve. The median population pharmacokinetic curves used do not have a relation with therapeutic effectiveness of DRV, but provide a cut‐off value in standard care to determine whether follow‐up is prudent 15, 18. Follow‐up consists of an additional food intake advice, additional questions and guidance concerning therapy adherence and a repeated blood sample for DRV monitoring. Therefore, in daily practice physicians are aiming for higher then median DRV plasma concentrations.

That eGFR is found to be a predictor for below median DRV plasma concentration in combination with age in the current study is a remarkable finding, since DRV is mainly eliminated by the liver. A higher than average eGFR alone would therefore presumably have a minor impact on DRV plasma concentration in daily practice. The observed inverse relation between eGFR and DRV concentrations in the current study might potentially be induced by the concomitant use of tenofovir. During the study period, 78% of the participants were using tenofovir simultaneously with DRV. A higher risk of renal impairment has been reported in patients receiving tenofovir and a RTV boosted PI, such as DRV 19. In addition, the summary of product characteristics of DRV demonstrates that coadministration of tenofovir (300 mg once daily) and DRV/RTV (300/100 mg twice daily) increased DRV AUC, Cmax and Cmin by 21%, 16% and 24%, respectively 20. The observed inverse relationship between eGFR and DRV concentration might be due to concomitant tenofovir usage, which is nephrotoxic on the one hand and inhibits DRV clearance on the other 19, 20. Earlier studies suggested that raltegravir also could lower DRV concentrations 21, 22. Of the seven participants using raltegravir and DRV simultaneously in the current study, two had a DRV concentration below median curve. Unfortunately, due to the unequally divided and small sample sizes of both tenofovir and raltegravir users and nonusers a subanalysis was not possible.

The other variable in the prediction model was age. The highest quartile of the variable age (>53 years) showed lower risk of a DRV plasma concentration below median curve in combination with eGFR. The effect of age on plasma antiretroviral drug (PI) exposure was shown earlier 23. The plasma concentrations of DRV might be higher with increasing age due to a lowered hepatic (CYP450 3A4 enzyme) activity in older people 24, 25.

An important limitation in the current study is that no data concerning concomitant food intake with DRV ingestion were available. The observed variability in DRV concentrations in the current study might potentially be confounded by inadequate concomitant food intake, although that is unlikely based on a prior study assessing food intake concomitantly with DRV 26. Another weakness is the limited number of events in the model. Nevertheless, the results of this study provide an insight into the DRV plasma concentrations after start of therapy at an HIV outpatient clinic.

Linking several patient‐ and disease‐related factors to the routinely measured DRV plasma concentrations shows that younger patients with a higher than average eGFR more frequently have a DRV plasma concentration below median curve. The impact of tenofovir has to be clarified in future studies in populations with a larger proportion of tenofovir‐free regimens to facilitate interpretation. Although therapeutic drug monitoring is not a substitute for clinical judgement, it can be a powerful tool for identifying patients with lower DRV concentrations and subsequently at risk for drug resistance.

Competing Interests

There are no competing interests to declare.

Contributors

A.D., Y.S. and J.W.C.A. were responsible for the conception and design of the study. The acquisition of laboratory and clinical data was performed by A.D., Y.S. and W.F.W.B. The data was analysed by A.D., Y.S and D.J.T. Both the drafting and the later critical revision of the article was conducted by all authors. The final approval of the manuscript was also done by all authors.