FormalPara Key Points

A population pharmacokinetics (PK) model was developed based on data from four clinical trials to describe the PK characteristics of buprenorphine (BPN) used in the treatment of opioid use disorder (OUD), and to evaluate the impact of covariates on PK parameters following administration of intravenous BPN, sublingual (SL) BPN, and subcutaneous (SC) CAM2038 formulations with different BPN doses and with weekly (Q1W) or monthly (Q4W) dosing schedules.

Simulations based on the PK model support the clinical use of CAM2038 in a variety of treatment phases and scenarios, including initiation of treatment and switching from daily SL doses of BPN. Treatment regimens using CAM2038 can be individually matched to patients’ SL BPN dose and adjusted to achieve the desired clinical effect.

Treatment with CAM2038 supports a rapid and extended opioid blockade after the initial dose, while dose titration after 3 days allows adjustment of BPN levels to patients’ individual needs. Subsequently, opioid blockade is maintained using different doses of weekly and monthly CAM2038. Compared to SL BPN, SC CAM2038 resulted in extended exposure to BPN and is administered much less frequently, and by a trained health care professional, which increases treatment compliance and reduces the potential for misuse, abuse, and accidental pediatric exposure.

1 Introduction

Opioid use disorder (OUD) is detrimental to an individual’s health [1,2,3] and has substantial public health, economic, and societal implications [1, 4]. Opioid use disorder can be treated with the µ-opioid receptor partial agonist buprenorphine (BPN), which is typically administered through daily transmucosal sublingual (SL) tablets or films, or a 6-month subdermal implant [4,5,6]. Although effective, daily BPN formulations risk misuse and diversion, and treatment noncompliance [7]; the subdermal implant also has limitations [8].

To address these limitations, CAM2038, a subcutaneous (SC) extended-release (sustained-release in the EU) BPN injectable, was developed [9, 10]. It has demonstrated clinical efficacy in treating OUD, including a significant reduction in cumulative illicit opioid use versus SL BPN [4, 11,12,13]. CAM2038 utilizes FluidCrystal® injection depot technology: a lipid-based solution transforms in situ into a liquid-crystalline gel after SC injection, then gradual and sustained release of encapsulated BPN occurs as the depot matrix biodegrades [14, 15].

CAM2038 formulations with different BPN doses and dosing intervals (weekly [Q1W]/monthly [Q4W]) allow individualized treatment during treatment initiation, switching from available SL BPN doses, and long-term maintenance therapy [15]. Subcutaneous injection of CAM2038 exhibits a rapid initial release of BPN into the systemic circulation followed by a sustained release [15]. Estimated terminal plasma half-life of BPN is 5 days for CAM2038 Q1W and 19–25 days for CAM2038 Q4W [15], versus 10–25 h for IV BPN [15, 16], and 24–43 h for SL BPN administrations [6, 15]. Thus, BPN absorption following release from the CAM2038 depot is the rate-limiting step in BPN elimination (flip-flop kinetics). Buprenorphine bioavailability after CAM2038 administration is 6- to 8-times higher than after SL BPN due to the first-pass effect of the swallowed fraction of SL BPN [15].

Together, these findings demonstrate smaller and less frequent fluctuations in BPN plasma concentrations over time for CAM2038 Q1W and Q4W versus IV BPN and SL BPN. Furthermore, the 6- to 8-fold higher bioavailability of BPN with CAM2038 reduces the total BPN dose administered to patients with OUD [6]. This, alongside less frequent administration of CAM2038 Q1W and Q4W and administration by trained health care providers, results in less BPN circulating in society and reduces the risk for diversion or misuse versus SL BPN [12, 15]. Another advantage of avoiding the first-pass effect by SC CAM2038 administration is lower exposure to the BPN active metabolite norbuprenorphine (norBPN). Given norBPN’s association with respiratory depression [17], reduced exposure with CAM2038 suggests a more predictable safety and efficacy profile than SL BPN [15]. Additionally, CAM2038 formulations enable OUD therapy to be individualized in alignment with published European guidelines [18]. Previously, overall BPN exposure was shown to increase dose-proportionally within the dose interval 8–32 mg CAM2038 Q1W (unpublished data) and within 64–192 mg CAM2038 Q4W [15], allowing therapy to be individualized. Similar BPN exposure was observed for 4 \(\times\) 16 mg CAM2038 Q1W and 1 \(\times\) 64 mg CAM2038 Q4W [15], suggesting the formulations are interchangeable on a monthly basis.

Population PK models have been presented for another SC extended-release BPN depot that utilizes a solvent-based solution of degradable polymers (BUP-XR) [5, 19]. Laffont et al. developed a population PK model for BUP-XR using data from 89 patients [20], which Jones et al. later refined using data from 570 patients [5]. Both analyses described the BPN plasma concentration-time profiles by a two-compartment model with dual absorption pathways [5, 20]. Given the different PK profile of CAM2038, the current analysis aimed to develop a population PK model describing BPN plasma concentration over time following administration of IV BPN (absent in Laffont et al. [20] and Jones et al. [5]), SL BPN, and SC CAM2038 Q1W and Q4W, and to evaluate the impact of selected covariates on PK parameters. Data from two Phase 1 and two Phase 2 studies were used to develop the model [11, 15]. The model was used to simulate BPN plasma concentration-time profiles in different treatment situations with CAM2038, including treatment initiation in individuals without previous BPN exposure, switching from SL BPN to CAM2038 Q1W or Q4W, and tapering after interrupting treatment with CAM2038.

2 Methods

2.1 Study Data

The population PK analysis was based on data from four clinical trials (Trials 1–4) that investigated SC CAM2038 Q1W and/or Q4W; IV and/or SL BPN were also administered in three of the trials (Figs. S1–4). These trials encompassed a variety of study designs including single-dose and repeat-dose regimens, and different doses and formulations of CAM2038 (see Supplementary Appendix A). Serial plasma sample collection was performed for all trial treatment groups (Table S1). Buprenorphine plasma concentration values were excluded in the data set if pre-treatment BPN dosing history was missing and there was a potential risk for carry-over effects from previous treatment. During the analyses, concentration values below the lower limit of quantification (LLOQ) were excluded. Observations with an absolute value of conditional weighted residuals (CWRES) > 5 may be regarded as outliers, but since there were very few such observations, and they comprised a small minority of the total number of samples for an individual, they were included in the analysis.

For the initial population PK model development, the data set comprised 236 participants with 10,260 BPN plasma concentration observations (Table S2). Following the generation of the initial PK model, data from a dose previously not included in the model (160 mg CAM2038 Q4W) became available. The data set for the model update for CAM2038 Q4W comprised 96 participants with 3870 BPN observations, from which data for 16 participants administered with 160 mg CAM2038 Q4W (with 398 BPN observations) had not been included in the initial model (Table S2). From the two data sets, concentrations were excluded due to uncertain dosing time, uncertain dosing history, or quantifiable pre-dose concentrations of BPN. There were no missing covariate data.

Of the 236 participants in the initial PK model, 147 (62 %) were healthy and 89 (38 %) were participants with OUD. Of the 96 participants in the model update, 62 (65 %) were healthy and 34 (35 %) were participants with OUD (baseline characteristics detailed in Table 1). Approximately 85 % of participants had normal renal function (creatinine clearance ≥ 90 mL/min) and approximately 15 % had mild renal function impairment (60–< 90 mL/min). No participants had moderate or severe renal function.

Table 1 Baseline characteristics of participants included in the population PK model
Full size table

Whole blood samples for PK assessment of BPN were collected into K2-ethylenediaminetetraacetic acid (EDTA)-coated tubes. Approximately 15 min after collection, samples were centrifuged at approximately 1500×g for 10 min at 20 °C. Plasma was separated and samples were frozen until analysis. Upon analysis, plasma samples were extracted with an organic solvent mixture and concentrations of BPN were determined using liquid chromatography-tandem mass spectrometry (LC–‍MS/MS). The method was fully validated for linearity, selectively, recovery, matrix effect, accuracy, precision, and stability before application to the sample analysis. The calibration range for BPN was 0.0250–10.0 ng/mL with a LLOQ of 0.0250 ng/mL based on analysis of 0.500 mL of plasma.

2.2 Software

The population PK analysis was performed using NONMEM version 7.3.0 (RRID:SCR_016986) [21], installed on an Intel Xeon-based server running Scientific Linux 6.3. NONMEM runs were performed using the gfortran compiler version 4.4.6. Updated simulations were performed using NONMEM version 7.5.0 installed on an Intel Xeon-based server running Scientific Linux 8.6. NONMEM runs were performed using the GCC compiler version 7.5.0 and OpenMPI 3.1.5 for the parallelization features, facilitated by Perl-speaks-NONMEM (PsN) version 5.2.6 [22, 23]. Data management and processing of NONMEM outputs were performed using R version 3.2.3 or 3.5.3 [24]. Stepwise covariate modelling (SCM) and visual predictive checks (VPCs) were run using PsN version 4.4.8 or 4.8.1 [22, 23]. Xpose version 4.5.3 [25] was used to aid model assessment. SigmaPlot version 14.0 and ggplot2 version 2.0.0 were used for plotting the simulation results.

2.3 PK Analysis

2.3.1 Structural and Statistical Model Building

Exploratory graphical analyses of BPN plasma concentration-time data and covariate distribution plots were used to describe data and identify potential outliers. Modelling was performed using log-transformed (natural logarithm) BPN plasma concentrations. Parameter estimations were performed using the first-order conditional estimation method with interaction (FOCEI) method in NONMEM. Parameter estimate standard errors were computed using the MATRIX=S option in the NONMEM $COV record. To decrease run times, the CTYPE=4 option in the NONMEM $EST record was used during and after the SCM [26].

Two- and three-compartment models with first-order elimination from the central compartment were evaluated for the disposition of BPN. Absorption was evaluated by first-order, sequential zero- and first-order, and parallel first-order absorption models, and combinations thereof. The proportion of bioavailable dose absorbed through respective pathways for the parallel absorption models was estimated. Lag time was evaluated for SL absorption. Bioavailability was set to 1 for IV BPN administration, restricted to between 0 and 1 (using a logit transformation) for SL BPN, and set to 1 for CAM2038 formulations. The influence of dose on absorption parameters for SL BPN, CAM2038 Q1W, and CAM2038 Q4W was investigated. The effect of the injection site was assessed univariately on the absorption parameters for CAM2038 Q1W.

Inter-individual variability (IIV) was evaluated on all relevant PK parameters and was generally added in an exponential form (Equation 1), where \(P\) is the typical value of the parameter, \({P}_{i}\) is the individual value of the parameter, and \({\eta }_{i}\) is a normally distributed random variable with mean 0 and standard deviation \(\omega\).

$$P_{i} = P \times e^{{\eta_{i} }}$$
(1)

For parameters restricted to between 0 and 1 (bioavailability and fractions), a logit transformation was used (Eq. 2) where \(\mathrm{logit}\left( P \right)\) is the logit-transform of the typical parameter value (\(P\); Eq. 3).

$$\mathrm{logit}\left( P \right) = \ln \left( {\frac{P}{1 - P}} \right)$$
(2)
$$P_{i} = \frac{{e^{{\left( {\mathrm{logit}\left( P \right) + \eta_{i}} \right)}} }}{{1 + e^{{\left( {\mathrm{logit}\left( P \right) + \eta_{i}} \right)}} }}$$
(3)

Additive and combined additive and proportional residual error models were explored for log-transformed BPN plasma concentrations.

2.3.2 Covariate Model Building

The base model, including effects of dose, injection site, and IIV, was used as a starting point for BPN covariate model building. All IIV values, except IIV on plasma clearance (\(CL\)) and central volume of distribution (\({V}_{\mathrm{c}}\)), were fixed to 0 before searching for covariates due to lengthy run times. The covariate model building was performed using the SCM [26] in PsN [22]. The covariate-parameter relationships in Table S3 were evaluated.

The first forward step assessed the effects of body weight (\(WT\)) on \(CL\) and \({V}_{\mathrm{c}}\). Significant effects were retained in the second forward step, where all other parameter-covariate relationships were tested. After the second forward step, all non-significant parameter-covariate relationships were excluded from the remaining forward steps. Each remaining covariate-parameter relationship was added one at a time to the model and the most statistically significant relationship retained until no more relationships were statistically significant. The model obtained in the forward step was subsequently subjected to a stepwise backward elimination of relationships. Each covariate-parameter relationship was omitted from the model one at a time and the least-significant relationship was removed from the model, given that it was not statistically significant. This was repeated until no more relationships could be removed and the SCM model was established. During the forward selection, p < 0.01; during the backward elimination, p < 0.001.

Covariate-parameter relationships (\({ParCov}_{m})\) for continuous covariates were coded as power models for untransformed parameters and linear models for logit-transformed parameters (Eq. 4), while categorical covariates were coded as a fractional difference to the most common category (Eq. 5), where \({Cov}_{ref}\) is the reference covariate value for covariate \(m\), to which the covariate model is normalized (usually median or mode). Total effect of covariates on parameter \(P\) was the product of the number of covariate terms.

$$ParCov_{m} = \left( {\frac{Cov}{{Cov_{ref} }}} \right)^{{\theta_{m} }}$$
(4)
$$ParCov_{m} = \left\{ {\begin{array}{*{20}c} 1 & {{\text{if}}} & {Cov = Cov_{ref} } \\ {1 + \theta_{m} } & {{\text{if}}} & {Cov \ne Cov_{ref} } \\ \end{array} } \right.$$
(5)

The SCM model was updated with the IIV from the base model. Parameter-covariate relationships that were estimated with poor precision (relative standard error [RSE] > 50 %) were tested for significance in the updated SCM model and, if not significant, removed from the model. Thus, the final model was established.

2.3.3 Model Evaluation

Model evaluation was based on the inspection of graphical diagnostics, including goodness-of-fit plots and VPCs, changes in the objective functional value (OFV) provided by NONMEM, and RSEs of the parameter estimates. Goodness-of-fit plots of observed concentrations versus individual predictions (IPRED) or population predictions (PRED) were evaluated for random scatter around the line of identity, while CWRES versus PRED were evaluated for random scatter around the horizontal line across the zero. Visual predictive checks were performed to evaluate the predictive performance of key models. Data were simulated 1000 times using doses and covariate data from participants in the analysis data set with the same study designs. Observed data below the LLOQ (BLQ) were included in calculations of median and percentiles and in the simulations. Prediction-corrected VPCs (pcVPCs) [27] were used since the study designs and doses differed widely among trials and trial arms. In the pcVPCs, observed and simulated dependent variables were normalized based on the typical population prediction. A 95 % confidence interval (CI) was used in the VPCs.

The discrimination between models was based mainly on inspection of graphical diagnostics and changes in OFV provided by NONMEM. A more complicated model was only retained if it provided a significant improvement (p < 0.05, − 3.84 change in OFV), plausible parameter estimates that were not associated with excessively high RSEs, improvements in the graphical diagnostics, and did not result in a high (> 1000) condition number. In the covariate analysis, p < 0.01 (− 6.63 change in OFV) was used for forward inclusion and p < 0.001 (− 10.83 change in OFV) for backward elimination of covariates. Shrinkage (SHR) [28] was taken into account when empirical Bayes estimates-based graphs and diagnostics were evaluated. The ETASTYPE=1 option in the NONMEM $EST record was used for the calculation of SHR.

2.4 Simulations

Simulations were performed using the final population PK model, based on predictions for a typical healthy individual (male participant, aged 35 years, 72.4 kg body weight), and were replicated 1000 times. It was presumed that the CAM2038 Q1W and Q4W injections were administered in the buttock. Descriptive summary statistics and 95 % prediction intervals of the individual posterior predicted PK parameters were calculated.

Simulations of BPN concentration-time profiles were performed to support treatment initiation with CAM2038 Q1W in patients not previously exposed to BPN, and to support dose conversions when switching from daily doses of SL BPN to CAM2038 Q1W or Q4W (Table 2). Buprenorphine concentration-time profiles were also simulated after interruption of CAM2038 Q1W and Q4W dosing at steady state.

Table 2 Switches from daily doses of SL BPN to SC doses of CAM2038 Q1W or doses of CAM2038 Q4W
Full size table

3 Results

3.1 Population PK Model for BPN

3.1.1 Base Model

Model development begun with the disposition of BPN after IV administration, best described by a three-compartment model with first-order elimination from the central compartment (Fig. 1). Sublingual BPN data were added, with BPN absorption into the systemic circulation best described by two parallel pathways: one with sequential zero- and first-order absorption with a lag-time, and one with first-order absorption. Estimated bioavailability for SL BPN (\({F}_{\mathrm{SL}}\)) was dose-dependent, with decreased bioavailability with increased dose. Lastly, CAM2038 Q1W and Q4W data were added in parallel. Buprenorphine absorption for CAM2038 Q1W was best described by two parallel pathways: one with sequential zero- and first-order absorption, and one with first-order absorption. For CAM2038 Q4W, BPN absorption was best described by two parallel pathways with first-order absorption (Fig. 1). Estimated bioavailability of BPN for CAM2038 Q1W and Q4W was close to the upper boundary of 1 and therefore fixed to 1.

Fig. 1
figure 1

Population PK model of BPN. BPN buprenorphine, \(CL\) plasma clearance, CMTq1w1 compartment q1w1, CMTq1w2 compartment q1w2, CMTq4w1 compartment q4w1, CMTq4w2 compartment q4w2, CMTSL1 compartment SL1, CMTSL2 compartment SL2, DSL1 duration of drug input to compartment SL1, Dq1w1 duration of drug input to compartment q1w1, FSL1 fraction of bioavailable dose going into compartment SL1, Fq1w1 fraction of dose going into compartment q1w1, Fq4w1 fraction of dose going into compartment q4w1, IV intravenous, ka,q1w1 first-order rate constant from compartment q1w1, ka,q1w2 first-order rate constant from compartment q1w2, ka,q4w1 first-order rate constant from compartment q4w1, ka,q4w2 first-order rate constant from compartment q4w2, ka,SL1 first-order rate constant from compartment SL1, ka,SL2 first-order rate constant from compartment SL2, Q1D once daily, Q1W once weekly, Q2 inter-compartmental clearance to second compartment, Q3 inter-compartmental clearance to third compartment, Q4W once monthly, SC subcutaneous, SL sublingual, tlag,SL1 lag-time to compartment SL1, V2 volume of second compartment, V3 volume of third compartment, Vc central volume of distribution

Full size image

Investigation into the influence of injection site on absorption parameters for CAM2038 Q1W found that for injections in the thigh, the entire dose was absorbed through the slower of the two absorption pathways (\({k}_{a,q1w1}\)) with the estimated dose fraction entering compartment q1w1 (\({F}_{q1w1}\)) close to 0 and subsequently fixed to 0.

Likelihood-based methods for handling BPN values BLQ had difficulties in obtaining a successful termination and gave comparable parameter estimates to when data BLQ were ignored. Data BLQ were therefore excluded in the analysis but included when producing VPCs. The residual unexplained variability (RUV) for BPN was best described by an additive error model on the log-transformed data (approximately proportional on untransformed data).

The initial population PK model was updated to include data for participants receiving SC doses of 160 mg CAM2038 Q4W. After several optimization attempts, the initial PK model with re-estimated absorption parameters for CAM2038 Q4W was established to be the best updated model, describing the data for all CAM2038 Q4W dose groups satisfactorily. The final population PK model analysis included 252 individuals and 10,658 observations.

3.1.2 Covariate Analysis

Given the strong correlation between body weight and body mass index, only body weight was included in the covariate analysis. As few trial participants were Asian, Mixed, and Other, race was only tested for Black or African American race versus all other races combined (White, Asian, Mixed, and Other).

Covariate-parameter relationships were identified in the SCM procedure for the initial base model (Table S4). The base model for BPN was updated with the significant covariate-parameter relationships. After adding the IIV back into the updated model, the effects of body weight on bioavailability after SL BPN dosing (\({F}_{\mathrm{SL}}\)) and \({V}_{c}\), population on dose fraction entering compartment q4w1 (\({F}_{q4w1}\)), and sex on \({F}_{q4w1}\), were estimated with poor precision (RSE > 50 %) and removed from the model without significant increases in OFV. Covariate analysis was not performed on the updated PK model for CAM2038 Q4W since no significant covariate-parameter relationships were observed on the absorption parameters in the covariate analysis of the initial PK model.

3.1.3 Final Model

The population PK model parameter estimates for the final model are presented in Table 3. Goodness-of-fit plots are shown in Figs. S5–6 and pcVPCs stratified by treatment are presented in Fig. 2. A selection of individual fits is presented in Fig. S7. Overall, the final population PK model predicted the observed BPN concentrations well for the different treatments. The model adequately predicted the median, 5th and 95th percentiles of the observed data, although a smaller variability than predicted was observed after IV BPN dosing. No trends in the residuals were observed. Model diagnostics for the final model indicated a good predictive performance for BPN concentrations. The RSE, which indicates precision of the parameter estimates, was < 19 % for the structural model parameters and < 37 % for the covariate model parameters. Further, the estimated IIV values had RSE < 14 %. The condition number was low (< 161) and SHR was generally low (< 20 %), except for IIV on volume of third compartment (\({V}_{3}\): 27.3 %) and first-order rate constant from compartment SL2 (\({k}_{a,SL2}\): 35.8 %).

Table 3 Parameter estimates for the final BPN population PK model
Full size table
Fig. 2
figure 2

Prediction-corrected visual predictive check of BPN concentrations for the final BPN population PK model. BPN concentrations are displayed versus time after last dose, stratified by treatment. Observations below the LLOQ were included in the calculations of median and percentiles The solid and dashed blue lines represent the median and 5th and 95th percentiles of the observations, the shaded red and blue areas represent the 95 % confidence interval of the median and the 5th and 95th percentiles predicted by the model. BPN buprenorphine, IV intravenous, LLOQ lower limit of quantification, PK pharmacokinetic, Q1W once weekly, Q4W once monthly, SC subcutaneous, SL sublingual

Full size image

Estimated BPN \(CL\) was 52.1 L/h (RSE: 1.80 %) and estimated \({V}_{\mathrm{c}}\) was 64.3 L (RSE: 10.3 %), with IIV of 20.9 % (RSE: 5.83 %) for \(CL\) and 78.6 % (RSE: 8.21 %) for \({V}_{\mathrm{c}}\). Buprenorphine bioavailability for CAM2038 was fixed to 100 %, whereas that of SL BPN decreased with increasing dose according to \({F}_{\mathrm{SL}}=0.14\times ({\frac{\mathrm{Dose}}{16})}^{-0.371}\). Thus, bioavailability for 8, 16, and 24 mg SL BPN was 18.1 %, 14.0 %, and 12.0 %, respectively. Dose-proportional BPN PK were indicated for CAM2038 Q1W and CAM2038 Q4W (Table 3).

The dose fraction entering the slowest of the two absorption rate compartments was 0.545 (= 1–‍‍0‍.455) for CAM2038 Q1W and 0.910 (= 1–0.0900) for CAM2038 Q4W, versus 0.241 (= 1–‍0.759) for SL BPN. For CAM2038 Q1W injections in the thigh, the whole dose was absorbed through the slower of the two pathways. The absorption rate constant for the slowest of the two absorption rate compartments was 0.00565 h–1 (half-life 123 h) for CAM2038 Q1W and 0.00166 h-1 (half-life 418 h) for CAM2038 Q4W, versus 0.0875 h–1 (half-life 7.9 h) for SL BPN.

The final PK model included effects of age and body weight (WT) on \(CL\), a population (healthy participants vs participants with OUD) effect on \({V}_{\mathrm{c}}\), and effects of sex and population on the dose fraction entering the faster of the two absorption pathways for CAM2038 Q1W (\({F}_{q1w1}\)). \(CL\) decreased with increasing age and increased with increasing body weight according to

\(CL={(\frac{Age}{35})}^{-0.233}\times {(\frac{WT}{72.4})}^{0.413}\). For an 18-year-old individual \(CL\) would be 60.8 L/h, versus 45.1 L/h for a 65-year-old (26 % decrease); \(CL\) for a 50 kg individual would be 44.7 L/h, versus 59.5 L/h for a 100 kg individual (33 % increase). Higher \({F}_{q1w1}\) was shown in women than in men, and in healthy individuals versus those with OUD. A higher \({F}_{q1w1}\) may result in a higher maximum concentration (\({C}_{\mathrm{max}}\)) value and a lower trough concentration (\({C}_{\mathrm{trough}}\)) value compared to a lower \({F}_{q1w1}\), but with an unaffected area under the plasma concentration-time curve (\(AUC\)) value. Additionally, estimated \({V}_{\mathrm{c}}\) was larger in individuals with OUD (237 L) than in healthy individuals (64.3 L). No additional effects of population, body weight, sex, age, race, or creatinine clearance on the model parameters were found. The RUV for the final model was 27.7 %.

3.2 Simulated BPN Plasma Concentration-Time Profiles

Simulation of BPN plasma concentrations following the recommended starting dose of 16 mg CAM2038 Q1W administered to individuals not previously exposed to BPN is shown in Fig. 3. To reach therapeutic BPN concentrations during the first treatment week, 1–2 additional 8 mg doses may be needed. Simulations for starting doses of 24 and 32 mg CAM2038 Q1W doses are presented for comparison. Both CAM2038 Q1W and Q4W exhibit rapid initial absorption of BPN into the systemic circulation after first dose, followed by a slow absorption as the depot matrix biodegrades (Fig. 4).

Fig. 3
figure 3

Simulated median BPN plasma concentration-time profiles based on the final population PK model, during treatment initiation of 16 mg CAM2038 Q1W followed by 8 mg CAM2038 after 3 and 5 days in comparison with initial treatment of 24 mg and 32 mg CAM2038 Q1W. BPN buprenorphine, Q1W once weekly

Full size image
Fig. 4
figure 4

Simulated median BPN plasma concentration-time profiles, based on the final population PK model, during the first week after administration of 24 mg CAM2038 Q1W, 96 mg CAM2038 Q4W, and 16 mg SL BPN Q1D. BPN buprenorphine, Q1D once daily, Q1W once weekly, Q4W once monthly, SL sublingual

Full size image

The model predicted that, regardless of dose, 78.9 % of the CAM2038 Q1W dose was absorbed after 7 days and 70.1 % of the CAM2038 Q4W dose was absorbed after 28 days (Fig. S8).

Simulated steady-state BPN plasma concentration-time curves illustrate switching from daily SL BPN to CAM2038 Q1W or Q4W, including bridging between daily doses of 16 mg SL BPN and 24 mg CAM2038 Q1W (Fig. 5a), daily doses of 16 mg SL BPN and 96 mg CAM2038 Q4W (Fig. 5b), and 24 mg CAM2038 Q1W and 96 mg CAM2038 Q4W (Fig. 5c). Notably, the steady-state BPN concentration range (\({C}_{\mathrm{trough},\mathrm{ss}}-{C}_{\mathrm{max},\mathrm{ss}}\); geometric mean) for 24 mg CAM2038 Q1W (1.18–‍4.97 ng/mL) and 96 mg CAM2038 Q4W (1.34–‍5.84 ng/mL) fell within the range of 16 mg SL BPN (0.852–6.09 ng/mL), but overall exposure (\({C}_{\mathrm{av},\mathrm{ss}}\)) was higher for CAM2038 versus SL BPN (Table 4, Fig. 5). Similar patterns were seen for the 8 mg/16 mg/64 mg and 24 mg/32 mg/128 mg (SL BPN/SC CAM2038 Q1W/SC CAM2038 Q4W) conversions (Table 4, Fig. 6).

Fig. 5
figure 5

Simulated steady-state BPN plasma concentration-time profiles for a daily 16 mg SL BPN and 24 mg CAM2038 Q1W, b daily 16 mg SL BPN and 96 mg CAM2038 Q4W, and c 24 mg CAM2038 Q1W and 96 mg CAM2038 Q4W. BPN buprenorphine, CI confidence interval, Q1D once daily, Q1W once weekly, Q4W once monthly, SL sublingual

Full size image
Table 4 Simulated steady-state BPN PK parameters, based on the final population PK model, following repeated dosing at steady state of SL BPN Q1D, SC CAM2038 Q1W, and SC CAM2038 Q4W (geometric mean [geometric CV %])
Full size table
Fig. 6
figure 6

Simulated steady-state BPN PK parameters (Ctrough,ss –Cav,ss–Cmax,ss), based on the final population PK model, following dosing for a) daily 8 mg SL BPN, 16 mg SC CAM2038 Q1W, and 64 mg SC CAM2038 Q4W, b) daily 16 mg SL BPN, 24 mg SC CAM2038 Q1W, and 96 mg SC CAM2038 Q4W, c) daily 24 mg SL BPN, 32 mg SC CAM2038 Q1W, and 128 mg SC CAM2038 Q4W. BPN buprenorphine, Cav,ss average plasma concentration during the dosing interval at steady state (geometric mean), Cmax,ss maximum plasma concentration at steady state, Ctrough,ss trough plasma concentration at steady state, PK pharmacokinetic, Q1D once daily, Q1W once weekly, Q4W once monthly, SC subcutaneous, SL sublingual

Full size image

Furthermore, simulations showed a slow gradual decrease of BPN plasma concentrations after discontinuation of treatment (Fig. 7). The predicted time taken for BPN concentration to fall below 0.1 ng/mL was 21–28 days for CAM2038 Q1W and 82–104 days for CAM2038 Q4W (Table 5).

Fig. 7
figure 7

Predicted BPN plasma concentrations, based on the final population PK model, after discontinuation of treatment following SC administration of 5 doses of 32 mg CAM2038 Q1W and 5 doses of 128 mg CAM2038 Q4W. BPN buprenorphine, CI confidence interval, PK pharmacokinetic, Q1W once weekly, Q4W once monthly, SC subcutaneous

Full size image
Table 5 Decay of BPN plasma concentration after interrupting treatment with CAM2038 at steady state
Full size table

4 Discussion

The current work presents a population PK model of BPN including IV BPN, SL BPN, and SC CAM2038 Q1W and Q4W treatments. Effects of selected covariates on the PK of BPN were evaluated and significant covariates were included in the model. The final PK model was used for simulations of different treatment regimens, including initiation, bridging between SL BPN and CAM2038, and tapering of treatment. The simulations showed that \({C}_{\mathrm{max}}\) and \({C}_{\mathrm{trough}}\) at steady state for CAM2038 Q1W and CAM2038 Q4W were well within steady-state plasma concentrations of daily SL BPN when dosed according to the established conversion table. Moreover, the simulations highlight the scope for individualizing treatment with CAM2038 through the flexible selection of 8, 16, 24, or 32 mg Q1W or 64, 96, 128, or 160 mg Q4W formulations, and the flexible administration of CAM2038 into the buttock, abdomen, thigh, or upper arm with similar BPN plasma exposure over time.

4.1 Population PK Model

Disposition of BPN was best described by a three-compartment PK model with first-order elimination based on data following IV BPN dosing. After SL BPN and CAM2038 Q1W and Q4W data were added, BPN absorption into systemic circulation for these treatments was best described by three separate two-parallel pathway models. Previous PK studies by Jensen et al. [29] and Yassen et al. [30] also described the disposition of BPN with a three-compartment model, whereas Laffont et al. [20] and Jones et al. [5] found that a two-compartment model best described disposition. Laffont et al. [20] and Jones et al. [5] used SL BPN data to describe the disposition of BPN, whereas the current analysis, Jensen et al. [29] and Yassen et al. [30] based the additional third compartment on data following IV BPN administration. Estimated plasma clearance (CL) was 52.1 L/h in our model-based analysis, versus 68 L/h using non-compartmental analysis (NCA) and assessments up to 48 h after a single IV dose of 0.6 mg BPN in healthy individuals (unpublished data). After administration of single increasing doses of 2, 4, 8, 12, and 16 mg IV BPN to non-dependent opioid users, estimated CL was 54.0 L/h [16].

In this model-based analysis, estimated bioavailability of BPN for CAM2038 Q1W and Q4W was close to the upper boundary of 1 and therefore fixed to 1. This is supported by estimated absolute bioavailability of BPN for CAM2038 Q1W and CAM2038 Q4W using NCA of observed data [15]. For SL BPN, our model-based analysis showed a decrease in bioavailability with increased dose. The NCA showed this by a dose-dependent decrease in dose-normalized AUC extrapolated to infinity (\({\mathrm{AUC}}_{inf}\)) between 8–24 mg SL BPN [15]. In our model-based analysis, bioavailability of BPN was estimated to be 6- to 8-times higher for CAM2038 than for SL BPN, consistent with values estimated by the NCA [15]. Initial release of BPN and absorption into systemic circulation following SC injection of CAM2038 were fast, with estimated absorption rate constants of 0.0401 h-1 for Q1W and 0.0441 h-1 for Q4W. In situ, the SC solution transforms into a liquid-crystalline gel encapsulating BPN. Sustained release of BPN occurs as the depot matrix biodegrades [11], with estimated absorption rate constants of 0.00565 h-1 (half-life 123 h; 5 days) for CAM2038 Q1W and 0.00166 h-1 (half-life 418 h; 17 days) for CAM2038 Q4W. Previous NCA estimates for terminal plasma half-life were similar: 5 days for CAM2038 Q1W and 19–25 days for CAM2038 Q4W [15]. Following IV BPN administration, the elimination half-life has been estimated to 10–25 h by NCA [15, 16]. The faster elimination rate following IV BPN administration demonstrates absorption rate-limited elimination of BPN with one fast and one slow absorption rate for CAM2038 Q1W and CAM2038 Q4W. This is consistent with release characteristics of the CAM2038 formulations. Similar to BUP-XR, BPN plasma concentration-time profiles for CAM2038 were best described by two absorption pathways: one fast and one slow [5, 20].

No differences in absorption parameters were observed after SC administration of CAM2038 Q1W in the buttock, abdomen, or upper arm. Injection in the thigh may produce a lower BPN peak concentration, as the whole dose was absorbed through the slower of the two absorption pathways. There were no overall exposure differences between injection sites. Given the similarities between CAM2038 Q1W and Q4W formulations, there are no expected injection site-related differences in overall BPN exposure.

Our model-based analysis exhibited dose-proportional PK of BPN for CAM2038 Q1W between 8–32 mg and for CAM2038 Q4W between 64–192 mg. Based on previous NCA, AUC extrapolated to infinity (\({\mathrm{AUC}}_{inf}\)) and maximum plasma concentration (\({C}_{\mathrm{max}}\)) increased proportionally to dose between 8–32 mg CAM2038 Q1W (unpublished data), while a small non-linearity in \({C}_{\mathrm{max}}\) was indicated between 192 mg and the lower doses of CAM2038 Q4W, but with dose-proportional \({\mathrm{AUC}}_{inf}\) [15]. The small non-linearity in \({C}_{\mathrm{max}}\) was not captured by the PK model.

No signs of time-dependent PK were apparent for SL BPN, CAM2038 Q1W, or CAM2038 Q4W, and steady-state conditions appear to be reached within 4 doses of CAM2038 Q1W and 4 doses of CAM2038 Q4W. If two or more doses of CAM2038 Q1W or Q4W are required to approach steady state, initial \({C}_{\mathrm{trough}}\) values may be significantly lower than steady state. Previously, the steady-state accumulation ratio was estimated to 1.4 by NCA for repeated doses of 16 mg CAM2038 Q1W [15]. In our model-based analysis, it appears that steady-state conditions are reached within 7 daily doses of SL BPN. This is supported by Compton et al, who observed no significant differences and no trend in \({C}_{\mathrm{trough}}\) values between Days 7–11 after repeated daily administration of 16 mg SL BPN to patients with opiate dependence [31].

In this analysis, the final population PK model included effects of age and body weight on \(CL\), a population (healthy participants vs participants with OUD) effect on \({V}_{\mathrm{c}}\), and effects of sex and population on the dose fraction entering the faster of the two absorption pathways for CAM2038 Q1W (\({F}_{q1w1}\)). No additional effects of population, body weight, sex, age, race, or creatinine clearance (\({CL}_{\mathrm{CR}}\)) were found.

The difference in \({V}_{c}\) in healthy participants (64.3 L) versus participants with OUD (237 L) may be because healthy participants received an IV BPN dose prior to SC CAM2038 administration, which may lead to a more precise estimate of \({V}_{\mathrm{c}}\). The predicted difference in \({V}_{\mathrm{c}}\) is of less importance for drugs with absorption rate-limited elimination, such as CAM2038, and is thereby not expected to influence overall systemic exposure of BPN. \({F}_{q1w1}\) was higher in women than men, and in healthy participants than participants with OUD, which may result in slightly higher \({C}_{\mathrm{max}}\) and lower \({C}_{\mathrm{trough}}\) values in women and healthy individuals. Overall BPN exposure (\(\mathrm{AUC}\)) was not affected. \(CL\) increased with body weight and decreased with increasing age. The differences in \(CL\) were small and are not considered clinically relevant. \({CL}_{\mathrm{CR}}\) was not identified as a predictor for \(CL\) in individuals with normal or mild renal function impairment. Overall, observed effects of age, sex, body weight, and population on the overall systemic BPN exposure (\(\mathrm{AUC}\)) are not considered clinically relevant and do not warrant dose adjustments. Therefore, patients with OUD should be individually titrated to the optimal dose of CAM2038 based on clinical effect.

Between-occasion variability was not investigated as few individuals received > 1 dose of CAM2038. Overall, the final population PK model predicted the observed BPN concentrations well for the different formulations. Model diagnostics for the final model, including goodness-of-fit plots and VPCs, indicated a satisfactory fit and predictive performance, suggesting that the model is adequate for simulation of different dosing regimens. No external evaluation of the model was performed, although it has been used in regulatory submission. The present population PK model of BPN is based on plasma concentrations in healthy individuals included in Phase 1 trials and in patients with OUD included in Phase 2 trials. Additional plasma concentration data in patients may be used to improve or validate the PK model.

In a previous evaluation of BPN PK parameters by NCA for 16 mg CAM2038 Q1W at steady state, higher \({C}_{\mathrm{max},\mathrm{ss}}\), \({C}_{\mathrm{trough},\mathrm{ ss}}\), and \({C}_{\mathrm{av},\mathrm{ss}}\) values were observed (4.30 ng/mL, 0.84 ng/mL, and 2.09 ng/mL, respectively) versus corresponding model-predicted values in our analysis (3.38 ng/mL, 0.78 ng/mL, and 1.79 ng/mL) [15]. The PK parameters observed after repeated daily administration of 8, 16, and 24 mg SL BPN at steady state were also higher than the corresponding model-predicted values in our analysis [15]. The PK parameter differences are small and may be a consequence of basing the simulations on a typical individual and not those used in the calculation of the PK parameters.

In clinical studies with CAM2038 and SL BPN, the relative norBPN to BPN exposure at steady state after CAM2038 administration was much less (5- to 7-fold lower, unpublished data) than after SL BPN administration. Given that norBPN appears to contribute to respiratory depression [17], the reduced norBPN exposure with the CAM2038 formulations may translate into a more predictable safety profile [15]. Therefore, the norBPN plasma concentrations were not included in the development of the population PK model.

4.2 CAM2038 Treatment

For individuals not exposed to BPN, the recommended treatment starting dose is 16 mg CAM2038 Q1W, with up to two additional 8 mg titration doses ≥ 1 day apart during the first treatment week. The dose in the second week is equivalent to the total dose administered in the first week (i.e., 16, 24, or 32 mg Q1W). Opioid effects of an 18 mg hydromorphone injection are completely blocked after a single SC dose of 24 or 32 mg CAM2038 Q1W across the treatment period, and mean BPN plasma concentrations of 1.2 ng/mL are generally sufficient to effectively block the subjective effects of 18 mg hydromorphone [11]. Recent pharmacokinetic/pharmacodynamic (PK/PD) analysis of the relationship between BPN plasma concentration and drug liking indicate that complete blockade of subjective opioid effects can be reached at significantly lower plasma BPN concentrations [32], consistent with the blockade of subjective opioid effects observed after 8 mg SL BPN [33].

Based on our simulations of treatment initiation, BPN concentrations of 1.2 ng/mL were reached approximately 5 h after the recommended starting dose of 16 mg CAM2038 Q1W, supporting rapid onset of opioid blockade after the initial dose. During the first treatment week, one or two additional doses of 8 mg CAM2038 Q1W sustain BPN plasma concentration and are used to individually titrate the dose and determine dosage for subsequent administration of CAM2038 Q1W or CAM2038 Q4W.

For individuals treated with SL BPN, the current dosing recommendation is a starting dose of 16 mg/day, increasing to 32 mg/day if necessary [34,35,36,37,38]. Using the population PK model for simulations, a dose of 24 mg CAM2038 Q1W or 96 mg CAM2038 Q4W gave comparable steady-state \({C}_{\mathrm{max}}\) and \({C}_{\mathrm{trough}}\) values and plasma concentrations well within the range of daily 16 mg SL BPN. Comparable steady-state exposures were also observed for 8 mg/16 mg/64 mg and 24 mg/32 mg/128 mg SL BPN/CAM2038 Q1W/CAM2038 Q4W. It may take two doses of CAM2038 Q1W or Q4W to approach steady state, resulting in initial \({C}_{\mathrm{trough}}\) values that are significantly lower than steady state. Overall, these data support the doses chosen for switching between SL BPN to CAM2038 Q1W or CAM2038 Q4W and between CAM2038 Q1W and CAM2038 Q4W, suggesting that changing between SL BPN and CAM2038 treatments will be well tolerated and efficacious. Importantly, four doses of CAM2038 Q1W (e.g., 4 × 24 mg) results in similar BPN steady-state exposure as one dose of CAM2038 Q4W (e.g., 1 × 96 mg), indicating that the two formulations are interchangeable in terms of efficacy and safety. The cumulative BPN dose is significantly lower for 24 mg CAM2038 Q1W (4 × 24 = 96 mg) and 96 mg CAM2038 Q4W (1 × 96 mg) than for daily 16 mg SL BPN (28 × 16 = 448 mg). This is advantageous, resulting in a ~5-fold reduction in BPN circulating in society. Alongside the fact that CAM2038 is administered by health care professionals, this minimizes the risks of diversion, abuse, misuse, and accidental pediatric exposure to BPN, while avoiding the burden and stigma associated with supervised daily administration of SL BPN, and providing improved treatment outcomes versus SL BPN [4, 12, 19, 39].

It has been suggested that BPN plasma concentrations ≥ 1.25 ng/mL are required for suppression of opioid withdrawal, or 1.2 ng/mL or ≥ 2−3 ng/mL for blocking opioid reinforcement [11, 33]. Our simulations indicate that administration of CAM2038 Q1W (24/32 mg) and CAM2038 Q4W (96/128/160 mg) consistently and rapidly result in BPN plasma concentrations above the suggested therapeutic threshold of 1.25 ng/mL, reaching \({C}_{\mathrm{max}}\) ≥ 2−3 ng/mL. While this remains under debate, the consistent exposure to BPN achieved with the CAM2038 formulations versus SL BPN may lead to improved patient experience by avoiding daily BPN peaks and troughs, during which suppression of withdrawal symptoms is weakest [6, 11, 14, 15].

With SC CAM2038, BPN plasma concentrations decrease slower than for SL BPN due to extended release of BPN from the depot matrix [14]. This could be associated with reduced craving following discontinuation of CAM2038; however, patients may need monitoring for withdrawal signs and symptoms for an extended period. Simulations indicated that BPN plasma concentrations < 0.1 ng/mL were reached about 1 month (21.8–27.3 days) after the last CAM2038 Q1W dose and about 3 months (82.3–104.0 days) after the last CAM2038 Q4W dose.

Based on flexibility and temporary needs of individuals, there is an opportunity for the CAM2038 Q1W dose to be administered up to 2 days before or after the scheduled treatment day, and the CAM2038 Q4W dose may be administered up to 1 week before or after the scheduled treatment day. One intermittent 8 mg CAM2038 Q1W dose may be administered, as supplemental dosing, during the dosing interval for CAM2038 Q1W or CAM2038 Q4W. Overall, CAM2038 should be individually titrated to the desired clinical effect, and small exposure differences, translated to differences in clinical effects, would be addressed by dose adjustments during standard patient care. These conclusions align with European guidelines for treatment of OUD, which recommend individual titration of dose to achieve desired clinical effects [18].

In addition to treatment of opioid dependence, BPN is widely used to treat pain, having proven effective in patients with chronic cancer and non-cancer pain [40, 41], and acute pain [42]. The advantages of BPN over other opioids used to treat pain include dose-dependent analgesia with a ceiling effect on respiratory depression [43,44,45,46,47], slow dissociation of BPN from receptors resulting in a long effect duration and reduced withdrawal symptoms upon discontinuation [48], and a lower abuse potential [49]. CAM2038 is being assessed for treatment of pain with focus on patients with opioid dependence. Based on observed and predicted BPN plasma concentrations relative to published data for approved transmucosal and injectable BPN formulations, CAM2038 may be a candidate for pain management in suitable patient populations.

5 Conclusions

A robust population PK model of BPN plasma concentration was successfully developed for four marketed formulations of BPN and three routes of administration: IV BPN, SL BPN, and CAM2038 Q1W and Q4W, using aggregated data from four clinical trials. The simulations based on the model provide evidence supporting the clinical use of CAM2038 in different treatment phases and scenarios, including treatment initiation, switching between daily SL BPN and CAM2038 Q1W and Q4W, switching between CAM2038 Q1W and Q4W, and tapering after stopping treatment with CAM2038. Importantly, CAM2038 demonstrated extended-release BPN plasma profiles that are matched to SL BPN doses and are within the therapeutic range. The simulations also demonstrate the scope of individualizing treatment with CAM2038 by choosing either 8, 16, 24, or 32 mg doses of CAM2038 Q1W or 64, 96, 128, or 160 mg doses of CAM2038 Q4W, with similar exposure when injected into the buttock, abdomen, thigh, or upper arm. Finally, CAM2038 Q1W and Q4W demonstrated complete BPN bioavailability with lower cumulative BPN dose versus daily SL BPN, thereby minimizing potential misuse and exposure of BPN to children.