Pharmacokinetic Simulation Analysis of Less Frequent Nivolumab and Pembrolizumab Dosing: Pharmacoeconomic Rationale for Dose Deescalation

Abstract

Nivolumab and pembrolizumab, anti–programmed cell death protein 1 monoclonal antibodies, have revolutionized oncology but are expensive. Using an interventional pharmacoeconomic approach, these drugs can be administered less often to reduce costs and increase patient convenience while maintaining efficacy. Both drugs are good candidates for less frequent dosing because of long half-lives and no evidence of a relationship of dose to efficacy. Established population pharmacokinetic models for both nivolumab and pembrolizumab were used to simulate profiles for multiple dosing regimens on 1000 randomly generated virtual patients. Simulations were initially performed on standard dose regimens to validate these in silico predictions. Next, simulations of nivolumab 0.3 mg/kg every 3 weeks revealed that >95% of patients maintained ≥1.5 μg/mL at steady state, which was inferred as the minimum effective concentration (MEC) for both drugs. Various alternative dosing regimens were simulated for both drugs to determine which regimen(s) can maintain this MEC in >95% of patients. Extended dosing regimens of nivolumab 240 mg every 4 weeks and 480 mg every 8 weeks along with pembrolizumab 200 mg every 6 weeks were simulated, showing that >95% of patients maintained MEC or greater. These simulations demonstrate the potential to reduce drug exposure by at least 50%, thus substantially reducing patient visits (as well as costs), while maintaining equivalent efficacy. These models provide the scientific justification for an ongoing prospective randomized clinical trial comparing standard interval fixed dosing with extended interval fixed dosing, and ultimately an efficacy-driven comparative trial.

The anti–programmed cell death protein 1 (PD-1) monoclonal antibodies nivolumab and pembrolizumab represent a major advance in the treatment of many cancers. These drugs block the interaction of the PD-1 receptor on the T cell with its major ligand, PD-L1, on the tumor cell to prevent tumor-mediated suppression of the T cell.¹ Nivolumab and pembrolizumab were both initially approved by the Food and Drug Administration (FDA) in 2014 for metastatic melanoma. As of August 2021, nivolumab has been approved for 16 indications, while pembrolizumab has gained over 30 separate FDA approvals to treat a wide array of malignancies,^2,3 although Prasad et al⁴ have suggested that the 2 drugs are clinically interchangeable. To that end, it is estimated that nearly half of patients with advanced cancer in the United States are currently eligible for a checkpoint inhibitor.⁵ However, there is an abundance of evidence that lower doses of these two PD-1 inhibitors may retain efficacy while reducing drug and financial toxicity, much of which has been reviewed previously.^6,7 There are several additional studies reporting efficacy of low-dose pembrolizumab or nivolumab in Hodgkin lymphoma^8–10 and non–small cell lung cancer (NSCLC).¹¹

Interventional pharmacoeconomics is a new discipline that aims to actively disrupt the rising cost of drugs through the application of clinical pharmacology principles.¹² One such approach is less frequent dosing, which is especially applicable to a monoclonal antibody with a long half-life, flat dose-response relationship, and time-dependent clearance (ie, decreasing over time). A reduced dose burden will not only decrease costs but also require fewer patient visits to the clinic, which improves convenience and compliance. In fact, in August 2021, the FDA released a Draft Guidance for Industry to use pharmacokinetic modeling and simulation to support approval for alternative dose regimens (eg, extended interval) for immune checkpoint inhibitors to decrease the number of patient visits to infusion centers.¹³ Furthermore, this strategy may also reduce the duration and/or severity of immune-related adverse events, which may not occur for months after initiation of therapy.¹⁴

Neither nivolumab nor pembrolizumab exhibits a significant dose-response or exposure-response relationship,^15,16 and there are abundant clinical data demonstrating that lower doses are effective in a variety of tumor types.^17–22 These data have been thoroughly summarized previously^6,7; however, a few findings are worth highlighting. In a phase 1 study on melanoma, patients with disease progression on 0.1 mg/kg or 0.3 mg/kg were escalated to 1.0 mg/kg without evidence of benefit, suggesting that (1) 0.1 mg/kg is part of the dose-response plateau and (2) factors other than exposure may be relevant to poor outcomes in nonresponders.¹⁸ Moreover, in the dose-ranging phase 2 study of nivolumab 0.3 to 10 mg/kg every 3 weeks in renal cell carcinoma (RCC), responses were observed in the 0.3 mg/kg every 3 weeks regimen as early as 6 weeks, before the third dose, to the same degree as the 10 mg/kg dose.²² In this trial, both the overall survival (OS) and progression-free survival were equivalent across the range of doses. Unfortunately, no pharmacokinetic data are available from this latter study.

The flat exposure-response for nivolumab was further demonstrated through statistical modeling, where average nivolumab concentrations were not significantly predictive of OS in melanoma or NSCLC, in contrast to baseline clearance and body weight.^20,23 Baseline drug clearance has repeatedly been demonstrated to be an important prognostic variable for OS for both nivolumab^20,23 and pembrolizumab.²⁴ While exact physiologic mechanisms that cause fast baseline clearance have not been fully elucidated, this seems to be a general attribute of monoclonal antibodies administered to patients with cancer, even if the drug is inactive.²⁵ Increased monoclonal antibody clearance is also associated with cachexia, hypoalbuminemia, and weight loss. While greater tumor burden would be expected to impact the magnitude of target-mediated drug disposition, this mechanism plays a minor role on baseline clearance, given that the pharmacokinetics of both agents are linear due to target saturation at doses studied clinically. Additionally, both nivolumab and pembrolizumab exhibit time-dependent clearance to very similar degrees, with clearance decreasing as disease status (eg, tumor burden, cachexic state) improves.^26,27

The intent of this in silico simulation study was to employ interventional pharmacoeconomics coupled with quantitative population pharmacokinetics (popPK) to identify model-informed dose regimens capable of reducing dose/exposure as much as possible, while still maintaining clinically effective serum concentrations. Such pharmacokinetic simulations commonly use the 90% to 95% threshold for the percentage of virtual patients who can maintain a particular exposure,^28–31 and our approach seeks to use this same general paradigm. In this study, pharmacokinetic simulations of low-dose regimens aim to identify extended nivolumab and pembrolizumab dosing regimens that can be predicted to maintain this inferred MEC in ≥95% of patients.

Methods

For all simulations, dosing regimens were applied to a 1000-virtual-patient data set, uniquely and randomly generated for both the nivolumab and pembrolizumab models, including the drug-specific covariates^26,27 using R version 4.0.3 (R Foundation for Statistical Computing, Vienna, Austria). All in silico experiments presented here were performed using the Phoenix NLME modeling and simulation package (Certara, Princeton, New Jersey). Model codes and R scripts for virtual patient generation for both models and data sets are provided in the Supplemental Information.

Nivolumab Simulations

For nivolumab, we used the model of Liu et al,²⁷ developed and published by authors from the FDA (Liu et al) to verify the popPK findings from the sponsor’s popPK model (Bajaj et al) and, most importantly, included the highly relevant covariate of baseline albumin for its impact on baseline clearance and time-dependent changes in clearance, which were not included in the sponsor’s model.³² This model also incorporated body weight, renal function via the estimated glomerular filtration rate (eGFR; Cockcroft-Gault), baseline performance status (PS; as measured using the Eastern Cooperative Oncology Group), baseline tumor size (in mm), and tumor type (RCC or not) as covariates for baseline clearance. Body weight and sex were relevant to the central compartment volume. To estimate the time-dependent changes in clearance, baseline albumin, tumor size, and best overall response were also incorporated in a maximum effectiveness (E_max) model. Although Bajaj et al included covariance estimates for clearance, central compartment volume, and E_max, Liu et al did not estimate covariance among random effects.²⁷

Based on the covariates included in the Liu et al²⁷ model, R version 4.0.3 was used to randomly generate 1000 unique values for age (years), sex (1:1, M:F), body weight (in kg) that was generated based on both sex and age, body surface area (m²), eGFR (mL/min/1.73 m²), baseline tumor size (in mm), baseline albumin (g/dL), PS at baseline (0 or ≥1), and RCC tumor type (0/1, 35% probability). Each variable was randomly generated using the mean and standard deviation statistics from Bajaj et al,³² which was the model used by the sponsor to simulate the 480 mg every 4 weeks regimen that was ultimately approved by the FDA based purely on in silico simulations (the Liu et al²⁷ publication did not contain this patient information). The resulting 1000-virtual-patient data set was used to simulate varying nivolumab dosing regimens (Table S1).

To verify that (1) the model was replicated accurately and (2) the simulated virtual patient variables were relevant, the simulation model was validated by simulating the approved 480 mg every 4 weeks regimen and comparing the results to published data.³³ These comparisons were made following the ninth dose, with the peak concentration assessed 30 minutes following the ninth dose (end of infusion; week 32), the minimum steady-state trough value (C_min,ss) assessed just before the 10th dose (week 36), and the average steady-state concentration during the ninth dose. Geometric means of all 1000 virtual patients at those designated times were assessed and compared to Long et al,³³ which provided the same metrics at the same time points. Once the simulation model was validated, a low-dose regimen known to be as effective as the labeled nivolumab dose, 0.3 mg/kg every 3 weeks,²² was simulated in order to estimate the minimum effective concentration (MEC), defined as the steady-state trough maintained by ≥95% of virtual patients at this dose and schedule.

When exploring extended dosing regimens, we considered simple modifications to the 480 mg every 4 weeks schedule for ease of clinical implementation: 240 mg every 4 weeks (half the dose) and 480 mg every 8 weeks (half as frequent). For each simulated dose regimen, the predicted concentrations were output at specific times encompassing 11 doses for graphical analyses. Additionally, trough values were analyzed following each simulated extended regimen for statistical analysis. Based on the early observations seen in RCC at the second dose trough on week 6 from the 0.3 mg/kg every 3 weeks regimen,²² statistics on early exposures (before reaching steady state) were analyzed for each alternative dosing regimen simulated to assess the percentage of virtual patients predicted to maintain an inferred effective exposure level.

In all simulated regimens, predicted nivolumab serum concentrations, as well as each virtual patient’s post hoc baseline clearance, were predicted. It has been demonstrated that baseline clearance rates >0.5 L/day are associated with treatment failure, regardless of the administered dose.^27,34 Therefore, of the 1000 virtual patients, those with predicted baseline clearance (based on first dose) >0.5 L/day were excluded from both statistical and graphical analyses.

Pembrolizumab Simulations

For pembrolizumab, we used the model of Li et al,²⁶ which incorporates time-dependent clearance as modulated by baseline tumor size and albumin. This model also incorporated body weight, sex, bilirubin, eGFR, Eastern Cooperative Oncology Group, baseline tumor size, and tumor type (melanoma or not) as covariates for baseline clearance. Albumin and sex were also incorporated as covariates for the central compartment volume. To estimate the time-dependent changes in clearance, baseline albumin, tumor size, and best overall response were also incorporated in an E_max model.

Based on these covariates, R was used to randomly generate 1000 unique values for age (years), sex (1:1, M:F), body weight (kg; based on age and sex), eGFR (mL/min/1.73 m²), baseline tumor size (in mm), baseline albumin (g/L), PS at baseline (0 or ≥1), bilirubin (μmol/L), and melanoma cancer type. The resulting 1000-virtual-patient data set (unique for pembrolizumab simulations) was used to simulate varying pembrolizumab dosing regimens (Table S2). To ensure accuracy of predictions, the replicated simulation model was validated by simulating the approved regimens of 200 mg every 3 weeks and 400 mg every 6 weeks and comparing the results to published data.^35–37 Low-dose pembrolizumab (1 mg/kg every 3 weeks) was also simulated to understand the minimum steady-state exposures achieved and maintained by that regimen.

The only extended pembrolizumab dosing regimen explored was 200 mg every 6 weeks, as other schedules were deemed not feasible for clinical practice. In all simulated regimens, pembrolizumab serum concentrations and post hoc baseline clearances were predicted, and, like nivolumab, those with predicted baseline clearance >0.5 L/day were excluded from both statistical and graphical analyses.

Statistical Analyses

For both nivolumab and pembrolizumab simulations, descriptive statistics, including geometric means, medians, and various percentiles over time, were calculated following the first dose as well as on varying weeks approaching and during steady state, based on the regimen using R. Predicted serum concentrations were also output and graphically analyzed by plotting the median ± 90% prediction interval (shaded) using the ggplot2 package in R.

Results

Validation of the Nivolumab Simulation Model

Before the exploration of extended-interval dosing, the nivolumab simulations were validated through comparison of our simulated results (480 mg every 4 weeks for 1000 virtual patients) with those reported by Long et al.³³ Ultimately, the simulation results matched very well with the literature (Table 1). The 480 mg every 4 weeks simulations performed in this study were also graphically depicted (Figure 1) and show a comparable serum concentration vs time profile as expected, providing confidence in the results following simulations of alternative dosing regimens without literature precedence.³³

Table 1.

Validation of Simulation Models

Nivolumab 480 mg Every 4 Weeks
Metric (μg/mL)	Simulationsa	Published Simulationsb
Cmin,ss	54.7 (73.1)	55.2 (62.9)
Cavg,ss	85.2 (60.9)	90.0 (46.4)
Cmax,ss	189 (44.8)	184 (57.7)
Pembrolizumab 200 mg Every 3 Weeks
Metric (μg/mL)	Simulationsc	Published Simulationsd
Cmin,ss	26.0 (14.3–47.9)	27.6 (14.9–46.2)
Cmax,ss	88.5 (66.4–119)	89.1 (66.4–124)

C_avg,ss, average steady-state concentration; C_max,ss, maximum steady-state concentration; C_min,ss, minimum steady-state trough concentration

N = 1000 virtual patients.

^aValues represent geometric mean (%CV) serum concentrations following the ninth dose (day 252).

^bLong et al.³³

^cValues represent median (10th-90th percentile) serum concentrations following the 10th dose (day 210); N = 1000 virtual patients.

^dFreshwater et al.³⁶

Figure 1.

Simulated concentration vs time profile for various nivolumab dosing regimens.

Predicted nivolumab serum concentration vs time during the first dose and steady state for the regimens (left to right) 480 mg every 4 weeks (Q4W), 0.3 mg/kg every 3 weeks (Q3W), 480 mg every 8 weeks (Q8W), and 240 mg every 4 weeks (Q4W). Solid line represents to median (50th percentile). Shaded region represents the 90% prediction interval (5th-95th percentiles of predicted concentrations at each time point).

Identification of an MEC

An MEC was inferred from simulations of a known effective, low-dose nivolumab regimen. After exclusion of 96 virtual patients (9.6%) with baseline clearance >0.5 L/day (the subset that is unlikely to benefit regardless of the administered dose), the geometric means of the steady-state minimum concentration (ie, trough, C_min,ss) following the first, second, and seventh doses were tabulated, along with various percentiles of trough concentrations (Table 2). By steady state, 95% of virtual patients maintained 1.5 μg/mL (Table 2, Figure 1). Further, responses to nivolumab 0.3 mg/kg every 3 weeks were observed after only 2 doses (before the third dose at week 6),²² at which 90% of virtual patients simulated here maintained 1.1 μg/mL (Table 2). Given the accuracy with which this simulation model predicted steady-state exposures from 480 mg every 4 weeks during model validation (Table 1) and the comparable clinical activity of 0.3 mg/kg every 3 weeks with higher doses, these predicted values seemed very reasonable.

Table 2.

Descriptive Statistics of Nivolumab Serum Trough Levels After Selected Doses

Scenario	Dose Regimen	Dose	Weeka	Geometric Mean (95%CI)	5th Percentileb	10th Percentilec	20th Percentiled	50th Percentilee
Standard regimen
1	480 mg every 4 weeks	First	4	22.8 (22.1–23.5)	10.2	12.3	15.2	23.6
		Third	12	45.0 (43.4–46.7)	17.5	21.1	28.6	46.8
		Fifth	20	57.1 (54.9–59.4)	20.3	25.0	34.8	60.1
Low-dose regimen
2	0.3 mg/kg every 3 weeks	1st	3	1.33 (1.29–1.37)	0.67	0.78	0.93	1.34
		2nd	6	2.06 (1.99–2.12)	0.94	1.13	1.39	2.11
		7th	21	3.81 (3.67–3.96)	1.51	1.91	2.34	3.87
Extended interval regimen
3	480 mg every 8 weeks	1st	8	9.89 (9.31–10.5)	1.87	2.99	5.22	11.7
		2nd	16	15.8 (14.9–16.8)	3.05	4.78	8.26	18.2
		3rd	24	19.2 (18.0–20.4)	3.68	5.71	9.77	22.1
4	240 mg every 4 weeks	1st	4	11.4 (11.0–11.8)	5.10	6.16	7.62	11.8
		3rd	12	22.5 (21.7–23.3)	8.74	10.5	14.3	23.4
		5th	20	28.5 (27.4–29.7)	10.2	12.5	17.4	30.1

CI, confidence interval.

96/1000 (9.6%) patients excluded due to baseline clearance >0.50 L/day.

^aWeek when trough from that specific dose was observed.

^bRepresents the serum concentration maintained by 95% of virtual patients.

^cRepresents the serum concentration maintained by 90% of virtual patients.

^dRepresents the serum concentration maintained by 80% of virtual patients.

^eRepresents the serum concentration maintained by 50% of virtual patients.

Experimental Nivolumab Dosing Regimens

Based on these simulations, subsequent alternative dosing regimens focused on maintaining a proposed MEC of ≥1.5 μg/mL. This is probably a conservative MEC, since a nivolumab dose of 0.1 mg/kg every 2 weeks is also effective in metastatic melanoma.¹⁷ The first extended nivolumab regimen simulated was 480 mg given every 8 weeks (Figure 1). This regimen prolonged the period until the next dose of 480 mg as clearance decreased over time (time to 50% of the change in clearance from baseline was ≈8 weeks), maintaining C_min,ss (troughs) above our inferred MEC in ≥95% of virtual patients after a single dose (Table 2). An alternative to this 480 mg every 8 weeks regimen is 240 mg every 4 weeks (Figure 1), which predicted even higher serum trough concentrations (Table 2). These simulations provided confidence that the regimens of nivolumab 480 mg every 4 weeks or 240 every 4 weeks were reasonable clinical options to maintain inferred efficacious drug levels in a clinical trial.

Validation of the Pembrolizumab Simulation Model

We simulated 1000 unique virtual patients at pembrolizumab 200 mg every 3 weeks, finding our results were very comparable to published results (Table 1).^26,36,37 The pembrolizumab simulations performed were graphically depicted (Figure 2) and also show a comparable serum concentration vs time profile.

Figure 2.

Simulated concentration vs time profile for various pembrolizumab dosing regimens.

Predicted pembrolizumab serum concentration versus time during the first dose and steady-state for the regimens (left to right) 200 mg every 3 weeks (Q3W), 400 mg every 6 weeks (Q6W), 1 mg/kg every 3 weeks (Q3W), and 200 mg every 6 weeks (Q6W). Solid line represents to median (50th percentile). Shaded region represents the 90% prediction interval (5th-95th percentiles of predicted concentrations at each time point).

Experimental Pembrolizumab Dosing Regimens

It was an assumption of this in silico study that nivolumab and pembrolizumab have the same MEC that was inferred for nivolumab from low-dose published data. However, there has been no published experience with very low doses of pembrolizumab. We first simulated pembrolizumab 1 mg/kg every 3 weeks, a known effective dose (graphically represented in Figure 2), which demonstrated that 95% of a unique set of virtual patients after the first dose (at week 3) achieved a C_min of ≥2.4 μg/mL.

For the dosing regimen of pembrolizumab 200 mg every 6 weeks, 97.5% of virtual patients maintained drug exposure ≥1.5 μg/mL over the first 6 weeks (Table 3, Figure 2). Thus, pembrolizumab 200 mg every 6 weeks was considered appropriate for further investigation, and less frequent dosing schedules were considered unlikely to maintain the MEC.

Table 3.

Descriptive Statistics of Pembrolizumab Serum Trough Levels After Selected Doses

Scenario	Dose Regimen	Dose	Weeka	Geometric Mean (95%CI)	5th Percentileb	10th Percentilec	20th Percentiled	50th Percentilee
Standard regimens
1	200 mg every 3 weeks	1st	3	11.1 (10.8–11.3)	6.59	7.22	8.35	11.3
		5th	15	23.7 (23.1–24.3)	12.5	14.0	16.7	23.5
		10th	30	27.7 (26.9–28.4)	14.4	16.1	19.3	27.1
2	400 mg every 6 weeks	1st	6	9.13 (8.82–9.45)	3.69	4.37	5.58	9.46
		3rd	18	15.8 (15.2–16.3)	6.63	7.63	9.74	15.9
		5th	30	17.9 (17.3–18.6)	7.39	8.81	11.1	17.9
Low-dose regimen
3	1 mg/kg every 3 weeks	1st	3	4.12 (4.04–4.21)	2.38	2.69	3.06	4.24
		5th	15	8.83 (8.59–9.07)	4.49	5.08	6.05	8.85
		10th	30	10.3 (10.0–10.6)	5.20	5.81	6.97	10.3
Extended interval regimen
4	200 mg every 6 weeks	1st	6	4.57 (4.41–4.73)	1.85	2.18	2.79	4.73
		3rd	18	7.88 (7.61–8.15)	3.31	3.81	4.87	7.94
		5th	30	8.97 (8.66–9.29)	3.69	4.40	5.55	8.93

CI, confidence interval.

65/1000 (6.5%) patients excluded due to baseline clearance >0.50 L/day.

^aWeek when trough from that specific dose was observed.

^brepresents the serum concentration maintained by 95% of virtual patients.

^cRepresents the serum concentration maintained by 90% of virtual patients.

^dRepresents the serum concentration maintained by 80% of virtual patients.

^eRepresents the serum concentration maintained by 50% of virtual patients.

Discussion

This in silico study demonstrated that when the standard dosing interval for both nivolumab and pembrolizumab is doubled, clinically effective exposures are maintained in a vast majority (>95%) of virtual patients. To arrive at this conclusion, this study meticulously replicated the published popPK models used by the FDA for drug approval and dose regimen changes^26,27 and, to ensure accuracy, independently validated the simulation results of standard dosing regimens that existed in the literature. After verifying that the model simulations were accurate and robust, we identified a conservative MEC of 1.5 μg/mL, based on the evidence that low-dose nivolumab is as effective as the labeled dose.²² Similar data were not available for pembrolizumab, but Merck has indicated to the FDA that target saturation is predicted at pembrolizumab doses of 0.1 to 0.3 mg/kg, and that the dose-response relationship is expected to be independent of tumor type.³⁸ Thus, we believe that our assumption that pembrolizumab and nivolumab have the same MEC is reasonable, although we acknowledge that further studies are required.

It is noteworthy that while both the nivolumab²⁷ and pembrolizumab²⁶ popPK models used for these simulation exercises included tumor type as covariates on clearance, the magnitude effect of tumor type on clearance was very small. This was verified clinically by Bi et al³⁹ that demonstrated no exposure differences with 480 mg every 4 weeks dosing across a half-dozen different tumor types. From this, we conclude that despite small changes in clearance models based on tumor types, no clinically relevant exposure differences would manifest, and therefore the MEC should be globally applied to all tumor types, consistent with Merck’s representations to the FDA.³⁸

The experimental extended-interval nivolumab regimens of 240 mg every 4 weeks and 480 mg every 8 weeks are predicted to yield serum trough concentrations that are comfortably above our inferred MEC of 1.5 μg/mL. A 50% dose reduction of pembrolizumab (200 mg every 6 weeks) also easily maintained this target in >95% of virtual patients after just a single dose. Additionally, lower doses may result in less frequent treatment discontinuation due to immune-related adverse events³⁹ and fewer patient clinic visits for infusions will increase convenience and compliance.¹³

Although in silico simulations are powerful, they are not a complete substitute for clinical trials. Randomized clinical trials with efficacy (eg, overall response rate) as the primary end point are needed to confirm that these alternate dosing regimens have comparable efficacy to the labeled regimens in all patients, regardless of baseline clearance. However, before planning such definitive trials (potentially using a near-equivalence design⁴⁰), we first aim to demonstrate PK noninferiority (defined by the ability to maintain our proposed MEC, 1.5 μg/mL) by comparing standard vs extended interval dosing of both drugs in patients with advanced or metastatic cancer. In this ongoing study (ClinicalTrials.gov NCT04295863), eligible patients are randomized to a standard dose arm (nivolumab 240 mg every 2 weeks or 480 mg every 4 weeks, or pembrolizumab 200 mg every 3 weeks) or a matching investigational arm (nivolumab 240 mg every 4 weeks or 480 mg q8w, or pembrolizumab 200 mg every 6 weeks). To not confound the primary pharmacokinetic end point, evaluable patients will have post hoc estimated baseline clearance rates < 0.5 L/d. These investigational regimens were selected on the basis of the in silico data presented in this study, where each of these regimens maintained the inferred MEC of 1.5 μg/mL in >95% virtual patients and also provided the most logical and convenient regimen for physicians and patients. This institutional review board–approved trial ensures patient safety by minimizing the risk of underdosing on the investigational arm through an external monitoring committee. Enrolled patients sign informed consent acknowledging this risk.

The paradigm described herein of using popPK modeling and simulation to identify alternative dosing regimens to minimize dose, toxicity, and financial burdens while maintaining an MEC can be applied beyond the examples of nivolumab and pembrolizumab. For example, atezolizumab is dosed such that the minimum exposure yielded is >20-fold higher than the published MEC and is therefore an obvious candidate.^41,42 Atezolizumab also demonstrates a flat exposure-efficacy relationship, yet presents a positive relationship (P = 0.026) between exposure (steady-state AUC) and the proportion of AEs in patients with NSCLC.⁴³ In conclusion, the work presented here provides an interventional pharmacoeconomic opportunity to explore alternative dosing regimens for therapies that, with the inclusion of new or additional safety and efficacy data, may benefit from dose reoptimization.

Data Availability Statement

The simulated data can be shared by sending requests to cody.peer@nih.gov.