Sie können Operatoren mit Ihrer Suchanfrage kombinieren, um diese noch präziser einzugrenzen. Klicken Sie auf den Suchoperator, um eine Erklärung seiner Funktionsweise anzuzeigen.
Findet Dokumente, in denen beide Begriffe in beliebiger Reihenfolge innerhalb von maximal n Worten zueinander stehen. Empfehlung: Wählen Sie zwischen 15 und 30 als maximale Wortanzahl (z.B. NEAR(hybrid, antrieb, 20)).
Findet Dokumente, in denen der Begriff in Wortvarianten vorkommt, wobei diese VOR, HINTER oder VOR und HINTER dem Suchbegriff anschließen können (z.B., leichtbau*, *leichtbau, *leichtbau*).
Oncolytic viruses (OVs) are a growing immuno-oncology therapeutic class that rely on their capability to activate the dormant endogenous anti-tumor immune response in order to control or eradicate tumor cells. Given their intrinsic mechanisms of action and their biological nature, development of antidrug antibodies (ADA) represents an important aspect to consider during clinical evaluation. ADAs can potentially affect viral kinetics and/or dynamics, ultimately resulting in reductions or even loss of drug efficacy. Here, we present a semi-mechanistic pharmacokinetic/pharmacodynamic model characterizing the interplay between V937 and neutralizing ADA in cancer patients receiving the V937 oncolytic virus.
Methods
The quantitative framework has been developed integrating viral load and ADA titers from 208 cancer patients who received V937 following intratumoral or intravascular administration, in monotherapy or in combination with pembrolizumab.
Results
The model successfully captured both V937 time course and the dynamics of ADAs under the different settings, showing no meaningful impact of ADAs on viral kinetics. Moreover, tumor response was neither affected by the preexistence or development of ADAs, which can be explained by the primary role of the immune system in the response.
Conclusions
This quantitative and (semi-) mechanistic framework can be expanded to other oncolytic viruses and used to explore under which scenarios a relevant impact could be observed, thus supporting the development of novel oncolytic viral therapies.
Iñaki F. Trocóniz and Tomoko Freshwater: Shared senior authorship.
Key Points
The potential impact of developed antidrug antibodies on viral kinetics, viral dynamics and ultimately tumor response is still under debate and has not been addressed in an integrated and quantitative manner in this field.
We present a modelling framework characterizing the impact of neutralizing ADAs on viral kinetics and tumor response of the V937 oncolytic virus, integrating clinical data from different routes of administration, dosing schemas, and even in drug combination.
This quantitative and (semi-) mechanistic framework can be expanded to other oncolytic viruses and used to explore the scenarios under which a relevant impact could be observed, thus supporting the development of novel oncolytic viral therapies.
1 Introduction
Since the approval of the first immune checkpoint inhibitor (ICI) in 2001, there has been an exponential increase in approved and under-development immunotherapies from biological origins. Biologics have an intrinsic capability to induce a humoral response leading to the appearance of antidrug antibodies (ADAs). Hence, the increased use of this type of therapeutics is raising questions regarding their potential immunogenicity, and ultimately, their impact on safety and clinical response [1, 2].
Anzeige
Different therapeutic areas have shown the important consequences of ADA formation for therapeutic proteins, mainly monoclonal antibodies. More specifically, ADAs can hamper clinical efficacy, directly neutralizing the drug activity and/or altering its pharmacokinetics (PK) [3, 4]. However, little is yet known regarding ADA formation and clinical consequences for novel biotherapies such as cell-based therapies or oncolytic virus (OV), among others [2, 5].
OV represents a growing class of anticancer agents that rely on their capability to preferentially replicate within tumor cells, inducing an immunogenic cell death and stimulating the host antitumor immunity [6]. As a counterpart, the virus can also trigger an antiviral immune response, including ADA development, which may alter its own pharmacokinetics, including distribution to the tumor, as well as block the replication and ongoing infection. In addition, many of the viruses used as vectors are naturally occurring, and as such, the target population may already exhibit pre-existing neutralizing antibodies [7].
ADA occurrence should be viewed as a continuous dynamic process (rather than positive/negative) resulting from the interplay between the OV and the immune system of the host. The development of mechanistic-based mathematical models offers an opportunity to tackle complex systems, like the one presented here, integrating information from different sources in order to provide a comprehensive and quantitative assessment of the impact of ADAs on PK, safety, and efficacy [8, 9].
In general, and regardless of the therapeutic area, models describing the dynamics of ADA formation and residence are scarce [10, 11]. This is aggravated in the case of OV where, to the best of our knowledge, no model has been yet proposed.
Anzeige
V937 is a naturally occurring coxsackievirus that has been evaluated in clinical trials for the treatment of patients with advanced solid tumor malignancies. The objective of this work was to develop a semi-mechanistic pharmacokinetic/pharmacodynamic (PK/PD) model characterizing the dynamics of ADA in cancer patients after receiving V937 OV in monotherapy or in combination with pembrolizumab following intratumoral or intravascular administration. In addition, the impact of ADA development on viral kinetics and tumor response was assessed.
2 Material and Methods
2.1 Clinical Studies
For the analysis, V937 mRNA levels and anti-V937 antibody titers were pooled together from four different clinical trials. A detailed description of the protocols, including study design, patients, and dosing schemas can be found here [12‐14]. Briefly, V937 was administered intravenously, MK3475-200 (ClinicalTrials.gov identifier: NCT02043665) [12] in monotherapy, Part A, or in combination with pembrolizumab, Part B. In addition, V937 intratumoral administration was explored in monotherapy, V937-006 (NCT01227551) and its extension (NCT01636882) [13], or in combination with pembrolizumab, V937-007 (NCT02565992) [14]. Table S1 in the electronic supplementary material (ESM) summarizes the main aspects of the different clinical trials and a description of patient characteristics can be found in Table 1.
Table 1
Summary of patient characteristics
Protocol name
V937 monotherapy
V937 combination
All
V937-006
MK3475-200 Part A
MK3475-200 Part B
V937-007
Number of patients
70
18
85
35
208
Age (years)a
64 (15.3) [28–94]
63.2 (12.8) [31–80]
65.9 (9.8) [40.7–83.6]
68.3 (11.8) [43–94]
65.3 (12.5) [28–94]
Sexb
Male
45 (64.3)
13 (72.2)
60 (70.6%)
27 (77.1%)
145 (69.7)
Female
25 (35.7)
5 (27.8)
25 (29.4%)
8 (22.9%)
63 (30.3)
Weight (kg)a
86.9 (22.6) [45.7–146]
78.3 (15.4) [54.7–105]
81.1 (19.1) [68–150]
91.4 (24.4) [59.1–164])
84.6 (21.3) [45.7–164]
Raceb
White
69 (98.6)
17 (94.4)
75 (88.2%)
32 (91.4%)
193 (92.8)
Asian
1 (1.4)
1 (5.6)
6 (7.1%)
1 (2.9%)
9 (4.3)
Hawaiian
1 (1.2%)
1 (0.5)
Black
1 (2.9%)
1 (0.5)
Other
3 (3.5%)
1 (2.9%)
4 (1.9)
Ethnicityb
Hispanic
1 (1.4)
2 (2.4%)
1 (2.9%)
4 (1.9)
Non-Hispanic
69 (98.6)
16 (88.9)
80 (94.1%)
34 (97.1)
199 (95.7)
Not reported
1 (5.6)
1 (1.2%)
2 (1)
NA
1 (5.6)
2 (2.4%)
3 (1.4)
Indicationb
Melanoma
70 (100)
4 (22.2)
35 (100)
109 (52.4)
NSCLC
5 (27.8)
46 (54.1%)
51 (24.5)
BL
5 (27.8)
39 (45.9%)
44 (21.2)
MRPC
4 (22.2)
4 (1.9)
Baseline tumor size (mm)a
54.4 (41.2) [10–205]
85.5 (71.5) [15–272]
65.2 (44.5) [14–234]
63.6 (56.9) [13–274]
63.1 (48.9) [10–274]
Baseline ECOG statusb
0
49 (70)
7 (38.9)
39 (45.9%)
21 (60)
116 (55.8)
1
17 (24.3)
8 (44.4)
46 (54.1%)
13 (37.1)
84 (40.4)
2
3 (16.7)
3 (1.4)
NA
4 (5.7)
1 (2.9)
5 (2.4)
BL bladder cancer, MRPC metastatic castrate-resistant prostate cancer, NA not available, NSCLC non-small cell lung cancer
aMean (standard deviation) [range]
bNumber of patients (percentage)
The study protocols and amendments were approved by local institutional review boards or ethics committees at each participating institution. The studies were conducted in accordance with local laws, Good Clinical Practice guidelines, and the Declaration of Helsinki. All patients provided written informed consent.
2.1.1 Sample Collection and Analytical Method
Blood samples were obtained at different time points (Table S1, see ESM 5) to measure serum for V937 levels and anti-V937 antibody titers. V937 levels were quantified using a polymerase chain reaction as previously described [15] and results were presented as V937 viral RNA copies/mL with a limit of detection (LOD) of 3000 copies/mL. Neutralizing anti-V937 antibody titers, referred to as ADAs (antidrug antibodies), were evaluated using a neutralizing antibody assay [13] reporting titers >1:16, set as LOD.
In addition, best overall response computed using the immune-related Response Evaluation Criteria in Solid Tumors (irRECIST) criteria [16] was available for all studies.
2.2 Data Analysis
Data were analyzed following the non-linear mixed effects modelling approach. For modelling purposes, reciprocal of ADA titers (i.e., dilution factor [DL]) were used and treated as a continuous variable.
V937 mRNA serum levels (n = 3628) and ADA measurements (2163) from a total of 208 patients were available for the analyses (Table S2, see ESM 6). Since measurement values expanded over several orders of magnitude, data were logarithmically transformed and additive residual error models in the logarithmic domain were considered for each measurement type. Inter-individual variability was explored on different model parameters assuming an exponential error model, except for bounded parameters where logit transformation was first applied to ensure individual parameters remained bounded between 0 and 1. Non-diagonal elements of the Ω variance-covariance matrix were tested for significance. Measurements below the limit of detection (BLD) were considered during model building and treated as (left-)censored information, thus maximizing the likelihood that indeed the observation was below the LOD (M3 method) [17]. NONMEM 7.4 and the first-order conditional estimation method with interaction, and the Laplacian method in case of censored data, were used for model building. Results were visualized in R version 4.1.2, through RStudio graphical interface version 2022.02.3.
2.2.1 Pharmacokinetic-Antidrug Antibody (PK-ADA) Model Development
A sequential and integrative modelling approach was followed during model building, increasing model complexity as needed upon inclusion of new information. First, a population PK model was developed using V937 mRNA in the absence of pembrolizumab administration (i.e., monotherapy studies and measurements prior to initial pembrolizumab dose on day 8). At this step, only observations up to day 8 (~ 40%) were used to avoid the potential impact of ADA on viral pharmacokinetics. Different compartmental models were tested, including serum and a theoretical tumor compartment to accommodate for both administration routes (intravenous and intratumoral). First-order transit and elimination processes were assumed, unless otherwise indicated by the data. Additional model complexities such as viral replication were explored as additional drug inputs dependent on the OV levels in the theoretical tumor compartment. In this regard, viral dynamic models [18, 19] explicitly accounting for the capability of the OV to infect tumor cells and produce new virions were not considered due to the semi-mechanistic and data-driven nature of the approach. As a next step, the developed PK model was used to predict the time course of the virus under combination scenarios and at all sampling time points to validate the developed PK model.
The impact of ADAs on viral pharmacokinetics was evaluated taking into consideration the full time course of the continuous variable (hereafter referred to as the ADA model). A turnover-based model where viral levels ultimately induce the appearance of ADA was assumed. A varying number of compartments (i.e., transit compartment models) were explored to account for the delayed appearance of ADA in the different clinical studies. Similarly, the impact of ADA time course on viral PK clearance was explored analyzing both sources of information simultaneously. Finally, the impact of the main design covariates (i.e., route of administration, pre-existence of ADA levels, and combination therapy) on the ADA model was explored using a stepwise approach, first performing a univariate analysis of each of the covariates on different model parameters.
A schematic representation of the final model, together with model equations, can be found in Fig. 1. NONMEM model code is provided in the ESM 2.
Fig. 1
Schematic and mathematical representation of the final model for V937 viral kinetics and antidrug antibody (ADA) development. Upon intravenous or intratumoral administration, the oncolytic virus (OV) can distribute bi-directionally between serum (S) and tumor (TUM), and serum and peripheral tissues (PER), or can be cleared from serum. Note, that following IT administration, only a fraction of the dose will be available for distribution to serum. Subsequently, serum levels of the OV can activate B cells (Bcel1), which, with some delay (Bcel2 and Bcel3), will trigger the appearance of ADAs in the blood stream. Please note that Bcell terminology is chosen to provide some mechanistic interpretation to the system, but Bcel1-3 and ADA compartments have the same unit dimensions. ADA0 individual baseline levels of prior ADA therapy, CL clearance, FIT fraction of the dose administered to the tumor compartment available for distribution to serum, EC50 OV concentration triggering 50% of the effect on Bcell synthesis, KDEG first-order degradation rate constant of ADAs, KSYN first-order synthesis rate constant of Bcells, KTR first-order transit rate constant between Bcell states, PER peripheral tissue, QPER intercompartmental clearance between serum and peripheral tissue, QTUM intercompartmental clearance between serum and tumor, S serum, VPER apparent volume of distribution of peripheral compartment, VS apparent volume of distribution of central compartment, VTUM apparent volume of distribution of tumor compartment
Model selection was driven by the capability of the model to adequately describe the tendency and dispersion of the data, taking into account (i) parameter precision; (ii) classical goodness-of-fit plots such as observations versus model predictions and conditional weighted residuals over time or over model predictions; and (iii) the objective function value (OFV), approximately equal to − 2× log-likelihood. When models were nested, a drop of 3.84 or 6.63 in OFV was considered significant at the levels of 5 and 1%, respectively. Regarding parameter precision, non-parametric bootstrapping was used when relative standard errors (RSE) provided by NONMEM were not obtained or were unrealistically low.
Model performance was evaluated using prediction corrected visual predictive checks (predcorr VPC) [20], where simulated data is graphically compared with observations. Briefly, 200 simulations were performed using the same study characteristics. The 5th, 50th, and 95th predicted percentiles at each time point (bins) were computed for each simulation. Then, the 90% confidence interval (CI) around each predicted percentile was computed, and the area plotted overlaying the observed data. For the data below the detection limit, the predicted percentage of BLD was computed at each simulation and bin, and the 90th prediction interval (PI) was computed and compared with the corresponding observed percentage.
2.3 Exploring Impact of ADA on Tumor Response
Tumor size was not included in the current PK-ADA framework. As such, and to mechanistically evaluate the impact of ADA development on tumor response, the ADA model described in Sect. 2.2.1 (Fig. 1) was integrated into a previously developed physiological model [15]. The latter model described (i) V937 kinetics in serum and tumor through a minimal physiologically based pharmacokinetic (PBPK) model accounting for vascular and interstitial distribution at tumor level; (ii) viral dynamics (i.e., capability of the OV to infect tumor cells in the cellular compartment and transform them to infected tumor cells), and (iii) its subsequent impact on tumor cell growth via the activation of an immune response upon viral-induced tumor cell death. A schematic representation of how models were linked and the model code is provided in ESM 4. More concretely, the serum OV levels predicted after intratumoral administration using the previous PBPK model were used to drive the generation of ADAs. In this regard, we used the developed pharmacodynamic ADA model, where OV in serum activate the production of neutralizing antibodies with some delay, also borrowing the related parameter estimates (KSYN, KTR, KDEG, and EC50). Then, different degrees of neutralization were explored (i.e., capability of the developed ADAs to block viral capacity of the OV to infect tumor cells) and controlled by the β infectivity rate parameter. Given the already high OV clearance, no additional effect of ADA on viral elimination was assumed. To account for ADA distribution to the tumor, where the infectivity takes place, a standard minimal PBPK model for monoclonal antibodies was implemented [21]. The intratumoral administration route was selected for simulation purposes as upon intravenous administration negligible OV levels reach the tumor, making the infectivity process irrelevant and thus precluding the exploration of the ADA impact on viral infectivity.
Model simulations were performed using RsSimulx 2.0.0 R package following a validation of model implementation (data not shown).
3 Results
3.1 PK Model
The time course of V937 mRNA levels stratified by administration route and treatment group can be found in Supplementary Figure 1 (see ESM). A three-compartment model accounting for the central, the (theoretical) tumor and other tissue compartments adequately described the time course of V937 mRNA levels, as well as the percentage of data below the limit of detection (~ 45% of the data used for model building). Similar intercompartmental clearance was observed for both peripheral compartments (tumor and others), and thus a single parameter estimate was used to increase model stability (p > 0.05). Regarding intratumoral administration, a bioavailability (FIT) term was used to acknowledge that only a fraction of the dose is available for further distribution to blood after intratumoral administration, obtaining an estimate of 20%. A summary of key models evaluated during model building can be found in Supplementary Table 3 (see ESM 7).
Anzeige
Final parameter estimates can be found in Table 2. Moderate interindividual variability was only supported on clearance (CL) and FIT; thus, the rest of the variability was captured by the residual error.
Table 2
Final parameter estimates of the selected pharmacokinetic and ADA models
Parameter
Typical (RSE)a [95% CI]
IIV [95% CI]
PK model of V937
VS (L)
87.9 (31.2) [47.2–159]
CL (L/h)
111 (15.5) [78.7–148]
67.5 (14.4) [46.5–90.5]
QTUM = QPER (L/h)
35.3 (33.9) [17.4–67.1]
VTUM (L)
103 (26.9) [58.8–173]
VPER (L)
893 (48.2) [454–1620]
FIT (–)
0.201 (32.0) [0.11–0.377]
4.17 (14.5) [2.22–7.11]b
Error (log copies/mL)
1.33 (3.88) [1.22–1.43]
ADA model
KSYN (DF/h)
113 (10.3) [109–171]
395 (7.9) [211–508]
EC50 (108 copies/L)
0.754 (13.8) [0.554–1.05]
KTR (10-3 h−1)
5.14 (7.98) [4.34–5.92]
95.2 (7.70) [77.1–123]
KDEG (10-3 h−1)
0.250 (12.8) [0.168–0.290]
96.8 (29.3) [81.4–374]
Error (log DF)
1.01 (6.76) [0.900–1.18]
ADA antidrug antibody, CL clearance, EC50 OV concentration triggering 50% of the effect on Bcell synthesis, FIT fraction of the dose administered to the tumor compartment available for distribution to serum, KDEG first-order degradation rate constant of ADAs, KSYN first-order synthesis rate constant of Bcells, KTR first-order transit rate constant between Bcell states, OV oncolytic virus, QPER intercompartmental clearance between serum and peripheral tissue, QTUM intercompartmental clearance between serum and tumor, VPER apparent volume of distribution of peripheral compartment, VS apparent volume of the central serum compartment, VTUM apparent volume of distribution of tumor compartment
aRelative standard errors (RSE) obtained from the bootstrap analysis are reported. For IIV terms reported in %, RSE of the variance over 2 formula is used
bLogit transformed parameter; inter-individual variability (IIV) expressed as CV (%) and computed as \(\sqrt{{e}^{{\omega }^{2}}-1}\), where ω2 represents the estimated variance of the variability parameter
More importantly, the model that was developed was capable of describing not only the data used for model building (i.e., data prior to pembrolizumab administration and before the appearance of ADA), but also V937 mRNA levels observed during combination therapy and after multiple doses (Fig. 1A and Fig. S2, see ESM 1). Although some discrepancies should be acknowledged at late time points after dose, where the model predicted negligible viral levels but instead some measurable levels were attained, these observations (< 1%) were not systematic within the individual nor correlated with dose levels, and thus inclusion of viral replication mechanisms was not supported. Therefore, no further model adjustments were deemed necessary (Fig. 2).
Fig. 2
Prediction corrected (predcorr) visual predictive check of the final pharmacokinetic (A) and pharmacokinetic-antidrug antibody (PK-ADA) model (B). Grey areas in the upper panels represent the 90 % confidence interval around the predicted 5th, 50th and 95th percentiles (dashed lines represent the corresponding observed percentiles). Grey areas in the lower panels represent the 90% prediction interval and the dashed line indicates the observed percentage of data below the detection limit. ADA neutralizing antidrug antibodies, BLD below limit of detection, DL dilution factor, i.v. intravenous, i.t. intratumoral. Note logarithm scale in x axis only in panel A
After drug administration, an increase in neutralizing antibodies was observed in all patients (Supplementary Figure 1, see ESM 1). No major differences between intravascular and intratumoral administration routes were graphically identified when normalized by dose. Median profiles showed a lower tendency after administration in combination compared with monotherapy, although large variability could be noted in ADA development.
As a first approach to characterize the time course of ADAs, previously obtained population PK parameters were fixed, and different structural PD models were explored keeping both PK and ADA observations in the dataset (PPPD method [22]). An adequate description of the data was obtained assuming that (i) serum OV levels trigger the activation of ADA production (i.e., Bcells) in a saturable manner (EC50 = 7.54 × 107 copies/mL) and (ii) using two transit compartments to delay the appearance of circulating ADA improved model performance (p < 0.05). In addition, a degradation rate for circulating ADA slower than the transit rate constant was supported by the data (p < 0.05). Finally, some patients exhibited pre-existing ADAs (n = 20, ~ 10% of the patients). Using a mixture model to account for these two different populations resulted in unstable models. In turn, directly summing the observed pre-existing levels to model predictions for those individuals with ADA values above the LOD (> 1:16) provided a good data description (Fig. 3; ESM 1). A summary of key models evaluated during model building can be found in Supplementary Table 3 (see ESM 7).
Fig. 3
In silico exploration of antidrug antibody (ADA) neutralization effect in oncolytic viral therapy. The impact of different levels of infectivity and replication (columns) on viral (oncolytic virus, OV) serum levels, % of infected tumor cells, CD8 (as a proxy of immune response) and ultimate tumor size response (rows) colored by different levels of neutralization (red: 100%, green: 50%, and blue 0%). A dose of 3 × 108 TCID50 administered intratumorally on days 1, 3, 5, and 8 of the first 21-day cycle and on day 1 of subsequent 21-day cycles was used for the simulation
A simultaneous analysis of viral and ADA levels was attempted as a second step. However, ADAs did not show any significant impact on viral clearance when used as a continuous measure, nor as a categorical time-variant covariate. Final model representation and parameter estimates can be found in Fig. 1 and Table 2.
3.3 Covariate Analysis
In this step, a covariate analysis to evaluate the potential impact of route of administration, treatment (monotherapy versus combination) as well as pre-existing ADA levels on ADA development was undertaken, but none of the evaluated models was deemed superior. Results from this analysis can be found in ESM 3.
3.4 Impact of ADA on Tumor Response
The PK-ADA model was integrated into an existing quantitative systems pharmacology (QSP) model that mechanistically describes tumor volume over time in response to V937. This merged framework allows us to explore in silico a potential impact of ADA on tumor response. In this regard, a simulation exercise was conducted to explore the impact of different degrees of ADA neutralization after intratumoral administration, taking into consideration different infectivity levels (Fig. 3).
Indeed, when OVs exhibit high infectivity (even in the absence of replication), a similar positive response is obtained after repeated intratumoral administration, regardless of the assumed neutralization level (Fig. 3). According to the model, this is due to the fact that a sufficient CD8 response, summarizing the immune system response, is triggered before neutralization starts playing its role, and thus it is barely affected despite infectivity being reduced. A more meaningful impact on tumor dynamics can be foreseen as infectivity decreases and the role of replication to attain a response increases. In those situations, the lower percentage of infected cells is predicted to translate into a lower CD8 response, thus observing how the appearance of neutralizing antibodies could shift the outcome from responder to stable disease or even non-responder. Interestingly, the different simulation scenarios show only a minor impact on OV serum levels, which appears at concentrations below the limit of detection.
Anzeige
Interestingly, and supporting the simulation results, an impact of maximum individual predicted ADAs on best overall tumor response was not identified in our clinical setting, where high infectivity was estimated (Fig. 4).
Fig. 4
Predicted levels of neutralizing antidrug antibodies (ADAs) by best tumor response according to iRECIST criteria. CR complete response, DF dilution factor, i.v. intravenous, i.t. intratumoral, PD progressive disease, PR partial response, SD stable disease
The increasing use of immunomodulatory agents for cancer therapies, such as antibodies targeting immune checkpoints, chimeric antigen receptor [CAR]-T cells or oncolytic virus, is raising questions regarding the potential impact of immunogenicity on treatment response [23]. Indeed, guidelines regarding immunogenicity assessment are already available for therapeutic products [24], and there is active discussion for novel biotherapeutics.
In this work, we present a novel population model to semi-mechanistically describe the interplay between V937 exposure and the development of neutralizing antibodies, integrating data from different clinical trials where two routes of administration (intravascular and intratumoral) were used, in addition to OV administration in monotherapy or in combination with pembrolizumab. In agreement with previous analyses of a subset of the PK data [15], a large volume of distribution (> 1000 L) and fast drug clearance (111 L/h) was identified. A moderate variability in drug clearance was also estimated. This variability was independent of the administration route and the combination scenario, highlighting the robustness of the developed model and the lack of interaction between pembrolizumab and V937 at a pharmacokinetic level. More importantly, no effect of pre-existing neutralizing antibodies on CL was identified.
There is a large debate on whether pre-existing immunity can decrease OV efficacy by enhancing viral clearance, which in many cases has resulted in the inclusion of seroprevalence as an exclusion criteria for clinical trials [13]. Several preclinical studies have shown that pre-existing immunity can indeed increase viral clearance and limit viral spread in mice, especially after systemic administration [25, 26]. There is still limited quantitative clinical information available. However, current knowledge suggests that there is no significant impact of pre-existing neutralizing antibodies on PK at the clinical level (e.g., talimogene laherparepvec [TVEC] [27]). In this regard, a consistent result was observed using V937 studies. Nonetheless, further studies are warranted, as there was a low number of patients with V937 neutralizing antibodies at baseline (~ 10%) and with unbalanced distribution (70% corresponded to the combo intravascular arm).
In addition to pre-existing immunity, newly developed neutralizing antibodies have the potential to impact PK and tumor response. From a modelling perspective, most efforts have been focused on assessing the impact of ADAs as covariates on the PK of monoclonal antibodies [8, 28], with only few works characterizing the ADA time course [11, 29], although there are also several initiatives to develop quantitative systems pharmacology frameworks to address the immunogenicity development after the administration of therapeutic proteins [10, 30]. In this regard, the model presented here represents a valuable scientific contribution, as it addresses the question from a longitudinal perspective.
To account for the dynamics of neutralizing antibody appearance after oncolytic viral therapy, serum OV levels were used as the trigger for neutralizing antibody production via a saturable model (i.e., accounting for a maximum neutralizing antibody production rate). A relatively low EC50 value was attained, therefore rendering a similar capability to trigger a humoral immune response after intratumoral or intravascular administration, despite the lower predicted serum OV levels following intratumoral administration. In addition, two additional transits were introduced to delay drug appearance, reporting a mean transit time of 24.3 days, in line with known antibody development physiology. This model structure is aligned with previous theoretical modelling efforts in the arena of ADA [31], suggesting its generable application in this context. Remarkably, no impact of developed neutralizing antibodies on the OV clearance was identified, regardless of the modelling approach used (e.g., categorical covariate or ADA time course). Nonetheless, large inter-individual variability in line with ADA levels expanding over more than one order of magnitude at the same dose level should be acknowledged. Although some covariate effects on ADA model parameters could be identified at a univariate analysis, such as the impact of pre-existing neutralizing antibodies on ADA synthesis or degradation, these effects inflated variability in other parameters and could not be precisely identified, remaining largely non-conclusive. This could be due to the low representation (~ 10%) and the unbalanced distribution previously discussed, together with the different ADA sampling schemas and follow-ups available for the different clinical studies (longer in the intratumoral arms). In addition, limitations in the ADA bioanalytical assays should also be taken into account. Although reporting antibody levels as titers (i.e., reciprocal of the highest dilution that gives a readout at or just above the cut point of the assay) are considered appropriate and preferred over mass units [32, 33], they also represent a semi-quantitative approach thus associated with a higher variability.
Despite the lack of impact of neutralizing antibodies on OV exposure, one important aspect to elucidate is the potential impact on clinical efficacy. Current preclinical and clinical data suggest that, even if OV replication might be decreased, pre-existing immunity should generally not be considered an obstacle to primary tumor clearance, at least in the setting of intratumoral OV therapy [26]. This is in line with no impact of developed neutralizing antibodies observed in different clinical trials evaluating OV efficacy [34, 35]. However, studies reporting an effect on efficacy and toxicity are also available [36]. Our results add to this collection of knowledge, suggesting little or no impact of neutralizing antibodies on clinical efficacy, assessed as best response RECIST 1.1 criteria. Indeed, when previous knowledge regarding the link between drug exposure, viral dynamics, and tumor response was integrated, it could be observed how neutralizing viral infectivity had limited impact on ultimate response if a sufficient cytotoxic immune response—namely CD8-specific T cells—can be mounted prior to ADA development. However, when replication plays a major role in order to attained sufficient infectivity and triggered immune response, an impact of neutralizing antibodies could be observed. Further studies evaluating the impact of both pre-existing and de novo antiviral immune response against the virus are needed to better understand oncolytic viral response. In this regard, additional efforts should be undertaken to optimize and standardize quantification methods and terminology in order to enable the reliability and comparability of data generated across different studies.
5 Conclusion
We present a pharmacokinetic and pharmacodynamic model capable of integrating information from multiple sources in order to successfully characterize the time course of both V937 and their trigger-neutralizing antibodies, and evaluate their interplay. To the best of our knowledge, this work represents the first efforts to address the dynamics of neutralizing antibodies in the context of oncolytic viral therapy. In this regard, this work provides valuable knowledge regarding the lack of significant impact of neutralizing antibodies not only on viral kinetics, but also on ultimate tumor response. Moreover, it provides a modelling framework to further explore in silico scenarios where immunogenicity could play a relevant role in viral kinetics, thus further supporting the quantitative understanding of the biological mechanisms behind oncolytic viral therapy.
Declarations
Funding
Open Access funding provided thanks to the CRUE-CSIC agreement with Springer Nature. Funding for this study was provided by Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, New Jersey, USA; no grant number is applicable.
Conflict of interest
T.F. is an employee of Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA, at the time of this work. Z.P.P-G. and I.F.T. received research funding from Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA.
Ethics approval
The study protocols and amendments were approved by local institutional review boards or ethics committees at each participating institution. The studies were conducted in accordance with local laws, Good Clinical Practice guidelines, and the Declaration of Helsinki.
Consent to participate
All patients provided written informed consent.
Author contributions and consent for publication
Conceptualization: ZPP-G, TF, and IFT; Formal analysis: ZPP-G and IFT; Resources: TF; Supervision: IFT and TF; Writing – review & editing: ZPP-G, IFT, and TF. All authors read and approved the final version of the manuscript.
Code availability
Original code is provided as supplementary material.
Availability of data and materials
Merck Sharp & Dohme LLC, a subsidiary of Merck & Co., Inc., Rahway, NJ, USA (MSD) is committed to providing qualified scientific researchers access to anonymized data and clinical study reports from the company’s clinical trials for the purpose of conducting legitimate scientific research. MSD is also obligated to protect the rights and privacy of trial participants and, as such, has a procedure in place for evaluating and fulfilling requests for sharing company clinical trial data with qualified external scientific researchers. The MSD data sharing website (available at: http://engagezone.msd.com/ds_documentation.php) outlines the process and requirements for submitting a data request. Applications will be promptly assessed for completeness and policy compliance. Feasible requests will be reviewed by a committee of MSD subject matter experts to assess the scientific validity of the request and the qualifications of the requestors. In line with data privacy legislation, submitters of approved requests must enter into a standard data-sharing agreement with MSD before data access is granted. Data will be made available for request after product approval in the US and EU or after product development is discontinued. There are circumstances that may prevent MSD from sharing requested data, including country or region-specific regulations. If the request is declined, it will be communicated to the investigator. Access to genetic or exploratory biomarker data requires a detailed, hypothesis-driven statistical analysis plan that is collaboratively developed by the requestor and MSD subject matter experts; after approval of the statistical analysis plan and execution of a data-sharing agreement, MSD will either perform the proposed analyses and share the results with the requestor or will construct biomarker covariates and add them to a file with clinical data that is uploaded to an analysis portal so that the requestor can perform the proposed analyses.
Open Access This article is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License, which permits any non-commercial use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by-nc/4.0/.
Boehncke WH, Brembilla NC. Immunogenicity of biologic therapies: causes and consequences. Expert Rev Clin Immunol. 2018;14:513–23.PubMed
2.
Gorovits B, Koren E. Immunogenicity of chimeric antigen receptor T-cell therapeutics. BioDrugs. 2019;33:275–84.PubMed
3.
Chirmule N, Jawa V, Meibohm B. Immunogenicity to therapeutic proteins: impact on PK/PD and efficacy. AAPS J. 2012;14:296–302.PubMedPubMedCentral
4.
Dingman R, Balu-Iyer SV. Immunogenicity of protein pharmaceuticals. J Pharm Sci. 2019;108:1637–54.PubMed
5.
Pan L, et al. 2022 white paper on recent issues in bioanalysis: FDA draft guidance on immunogenicity information in prescription drug labeling, LNP & viral vectors therapeutics/vaccines immunogenicity, prolongation effect, ADA affinity, risk-based approaches, NGS, qPCR. Bioanalysis. 2023;15:773–814.PubMed
6.
Macedo N, Miller DM, Haq R, Kaufman HL. Clinical landscape of oncolytic virus research in 2020. J Immunother Cancer. 2020;8: e001486.PubMedPubMedCentral
7.
Kaufman HL, Kohlhapp FJ, Zloza A. Oncolytic viruses: a new class of immunotherapy drugs. Nat Rev Drug Discov. 2015;14:642–62.PubMedPubMedCentral
8.
Perez Ruixo JJ, Ma P, Chow AT. The utility of modeling and simulation approaches to evaluate immunogenicity effect on the therapeutic protein pharmacokinetics. AAPS J. 2013;15:172–82.PubMed
9.
Gómez-Mantilla JD, Trocóniz IF, Parra-Guillén Z, Garrido MJ. Review on modeling anti-antibody responses to monoclonal antibodies. J Pharmacokinet Pharmacodyn. 2014;41. https://doi.org/10.1007/s10928-014-9367-z
10.
Chen X, Hickling TP, Vicini P. A mechanistic, multiscale mathematical model of immunogenicity for therapeutic proteins: Part 2—model applications. CPT Pharmacometr Syst Pharmacol. 2014;3:1–10.
11.
Liao KH, et al. A mechanistic pharmacokinetic model with drug and antidrug antibody interplay, and its application for assessing the impact of immunogenicity response on bioequivalence testing. Br J Clin Pharmacol. 2020;86:2182–91.PubMedPubMedCentral
12.
Rudin CM, et al. Phase 1, open-label, dose-escalation study on the safety, pharmacokinetics, and preliminary efficacy of intravenous Coxsackievirus A21 (V937), with or without pembrolizumab, in patients with advanced solid tumors. J Immunother Cancer. 2023;11: e005007.PubMedPubMedCentral
13.
Andtbacka RHI, et al. Clinical responses of oncolytic Coxsackievirus A21 (V937) in patients with unresectable melanoma. J Clin Oncol. 2021;39:3829–38.PubMed
14.
Silk AW, et al. A phase 1b single-arm trial of intratumoral oncolytic virus V937 in combination with pembrolizumab in patients with advanced melanoma: results from the CAPRA study. Cancer Immunol Immunother. 2023;72:1405–15.PubMed
15.
Parra-Guillen ZP, et al. Assessment of clinical response to V937 oncolytic virus after intravenous or intratumoral administration using physiologically-based modeling. Clin Pharmacol Ther. 2023;114:623–32.PubMed
16.
Wolchok JD, et al. Guidelines for the evaluation of immune therapy activity in solid tumors: immune-related response criteria. Clin Cancer Res. 2009;15:7412–20.PubMed
17.
Beal SL. Ways to fit a PK model with some data below the quantification limit. J Pharmacokinet Pharmacodyn. 2001;28:481–504.PubMed
18.
Santiago DN, et al. Fighting cancer with mathematics and viruses. Viruses. 2017;9:239.PubMedPubMedCentral
19.
Titze MI, et al. A generic viral dynamic model to systematically characterize the interaction between oncolytic virus kinetics and tumor growth. Eur J Pharm Sci. 2017;97:38–46.PubMed
20.
Bergstrand M, Hooker AC, Wallin JE, Karlsson MO. Prediction-corrected visual predictive checks for diagnosing nonlinear mixed-effects models. AAPS J. 2011;13:143–51.PubMedPubMedCentral
21.
Shah DK, Betts AM. Towards a platform PBPK model to characterize the plasma and tissue disposition of monoclonal antibodies in preclinical species and human. J Pharmacokinet Pharmacodyn. 2012;39:67–86.PubMed
22.
Zhang L, Beal SL, Sheiner LB. Simultaneous vs. sequential analysis for population PK/PD data I: best-case performance. Journal of pharmacokinetics and pharmacodynamics, 30(6), 387–404.
23.
Davda J, et al. Immunogenicity of immunomodulatory, antibody-based, oncology therapeutics. J Immunother Cancer. 2019;7:105.PubMedPubMedCentral
24.
Rosenberg AS, Sauna ZE. Immunogenicity assessment during the development of protein therapeutics. J Pharm Pharmacol. 2018;70:584–94.PubMed
25.
Burnett WJ, et al. Prior exposure to Coxsackievirus a21 does not mitigate oncolytic therapeutic efficacy. Cancers. 2021;13. https://doi.org/10.3390/cancers13174462
26.
Groeneveldt C, van den Ende J, van Montfoort N. Preexisting immunity: barrier or bridge to effective oncolytic virus therapy? Cytokine Growth Factor Rev. 2023;70:1–12.PubMed
27.
Andtbacka RHI, et al. Biodistribution, shedding, and transmissibility of the oncolytic virus talimogene laherparepvec in patients with melanoma. EBioMedicine. 2019;47:89–97.PubMedPubMedCentral
28.
Passey C, Suryawanshi S, Sanghavi K, Gupta M. Reporting, visualization, and modeling of immunogenicity data to assess its impact on pharmacokinetics, efficacy, and safety of monoclonal antibodies. AAPS J. 2018;20:1–13.
29.
Ren Y, et al. A model-based approach to quantify the time-course of anti-drug antibodies for therapeutic proteins. Clin Pharmacol Ther. 2019;105:970–8.PubMed
30.
Kierzek AM, et al. A quantitative systems pharmacology consortium approach to managing immunogenicity of therapeutic proteins. CPT Pharmacometr Syst Pharmacol. 2019;8:773–6.
31.
Chen X, et al. A mathematical model of the effect of immunogenicity on therapeutic protein pharmacokinetics. AAPS J. 2013;15:1141–54.PubMedPubMedCentral
32.
Guidance for industry: immunogenicity assessment for therapeutic protein products. U.S. Department of Health and Human Services Food and Drug Administration Center for Drug Evaluation and Research (CDER) Center for Biologics Evaluation and Research (CBER). 2014. https://doi.org/10.1089/blr.2013.9927.
33.
Guideline on immunogenicity assessment of therapeutic proteins. Committee for Medicinal Products for Human Use (CHMP) guideline. 2017.
34.
Hu JCC, et al. A phase I study of OncoVEXGM-CSF, a second-generation oncolytic herpes simplex virus expressing granulocyte macrophage colony-stimulating factor. Clin Cancer Res. 2006;12:6737–47.PubMed
35.
Heo J, et al. Randomized dose-finding clinical trial of oncolytic immunotherapeutic vaccinia JX-594 in liver cancer. Nat Med. 2013;19:329–36.PubMedPubMedCentral
36.
Van Brummelen EMJ, Ros W, Wolbink G, Beijnen JH, Schellens JHM. Antidrug antibody formation in oncology: clinical relevance and challenges. Oncologist. 2016;21:1260–8.PubMedPubMedCentral
