Nivolumab and ipilimumab population pharmacokinetics in support of pediatric dose recommendations—Going beyond the body‐size effect

Abstract

Body size has historically been considered the primary source of difference in the pharmacokinetics (PKs) of monoclonal antibodies (mAbs) between children aged greater than or equal to 2 years and adults. The contribution of age‐associated differences (e.g., ontogeny) beyond body‐size differences in the pediatric PKs of mAbs has not been comprehensively evaluated. In this study, the population PK of two mAbs (nivolumab and ipilimumab) in pediatric oncology patients were characterized. The effects of age‐related covariates on nivolumab or ipilimumab PKs were assessed using data from 13 and 10 clinical studies, respectively, across multiple tumor types, including melanoma, lymphoma, central nervous system tumors (CNSTs), and other solid tumors. Clearance was lower in pediatric patients (aged 1–17 years) with solid tumors or CNST than in adults after adjusting for other covariates, including the effect of body size. In contrast, clearance was similar in pediatric patients with lymphoma to that in adults with lymphoma. The pediatric effects characterized have increased the accuracy of the predictions of the model, facilitating its use in subsequent exposure comparisons between pediatric and adult patients, as well as for exposure–response analyses to inform pediatric dosing. This study approach may be applicable to the optimization of pediatric dosing of other mAbs and possibly other biologics.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Body size was often considered to be the primary source of differences in the pharmacokinetics (PKs) of monoclonal antibodies (mAbs) between pediatric patients aged greater than or equal to 2 years and adults. The contribution of age‐associated differences beyond body size to the pediatric PKs of mAbs has not been comprehensively evaluated.

• WHAT QUESTION DID THIS STUDY ADDRESS?

This study identified the tumor‐type–dependent contribution of age‐associated differences beyond body size to the PKs of mAbs in pediatric patients.

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

Pediatric patients with solid tumors or central nervous system tumors had lower mAb clearance and volume of distribution than adults after adjusting for the effects of body size. In contrast, mAb clearance was similar in pediatric and adult patients with lymphoma.

• HOW MIGHT THIS CHANGE DRUG DISCOVERY, DEVELOPMENT, AND/OR THERAPEUTICS?

The study approach described could be extended to support pediatric dose optimization of other mAbs and other biologics.

INTRODUCTION

The use of antibody‐based anticancer therapeutics, including immune checkpoint inhibitors, has shown impressive success in a number of adult patient indications. ¹ However, pediatric drug development faces feasibility and ethical issues, which often lead to delays in regulatory approval of pediatric indications. In 2019, the US Food and Drug Administration (FDA) recommended the inclusion of adolescent patients in adult oncology clinical trials to enable earlier access to effective drugs for adolescent patients. ² An increased understanding of the differences between pediatric and adult pharmacokinetics (PKs) is integral to pediatric drug development. Rigorous and well conducted population PK (PopPK) analyses in combination with clinical data across multiple tumor types and age groups can assist in recommending pediatric dosing regimen.

Historically, the PKs of therapeutic monoclonal antibodies (mAbs) were thought to be primarily affected by body size, and mAb PKs in pediatric patients (aged 2–17 years) were frequently described by PopPK models using an allometric relationship between body size and PK parameters. ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ Earlier studies concluded that mAb PKs in pediatric patients (6–17 years) were readily predictable from adult PK after correction for the body‐weight effect. ³ , ⁸ Similarly, the 2019 FDA guidance indicated that exposure and clearance (CL) of drugs are generally similar between adolescent and adult patients after accounting for the effect of body size. ² Recently, a pembrolizumab PopPK model for pediatric patients (1–17 years) with solid tumors (STs) or classical Hodgkin's lymphoma (cHL) was briefly described. ¹¹ In this model, both age and body weight had effects on pembrolizumab CL and volume of distribution. ¹¹ However, the age effect was assumed to be the same across tumor types despite notable PK difference between STs and cHL. ¹¹ Overall, the contribution of age‐associated effects beyond body‐size differences to mAb PKs in pediatric patients has not been comprehensively assessed in the literature.

Developing nivolumab and ipilimumab PopPK models that can accurately predict pediatric PKs is important in supporting pediatric dosing recommendations. The anti–programmed death‐1 mAb nivolumab has been approved in the United States, the European Union, and many other countries alone or in combination with the anti–cytotoxic T‐lymphocyte antigen‐4 mAb ipilimumab in multiple tumor types in adults, including melanoma (MEL), non‐small cell lung cancer (NSCLC), malignant pleural mesothelioma (MESO), renal cell carcinoma (RCC), and microsatellite instability‐high (MSI‐H) or mismatch repair deficient (dMMR) metastatic colorectal cancer (CRC). ¹² , ¹³ , ¹⁴ Nivolumab with or without ipilimumab is also approved in the United States and other countries for adolescents (aged 12–17 years) with MEL and MSI‐H or dMMR metastatic CRC. ¹³ , ¹⁴ Although there is a good understanding of the quantitative clinical pharmacology characteristics of nivolumab in adults, data from pediatric patient populations are limited. Nivolumab PKs in adults has been well‐characterized by a linear two‐compartment, zero‐order intravenous (i.v.) infusion model with time‐varying CL that decreases with time. ¹⁵ Subsequent PopPK analyses suggested that in patients with MEL (in the adjuvant setting), glioblastoma (GBM), or cHL, nivolumab CL was lower than in patients with metastatic tumors and not time‐dependent in these MEL and GBM settings. ¹⁶ , ¹⁷

Ipilimumab has been approved in the United States, the European Union, and other countries as monotherapy for MEL in the metastatic and adjuvant settings and in combination with nivolumab for MEL, RCC, hepatocellular carcinoma (HCC), NSCLC, MESO, esophageal cancer, and MSI‐H or dMMR CRC. ¹⁸ , ¹⁹ , ²⁰ Ipilimumab is also approved in the United States and the European Union for the treatment of metastatic MEL in adolescents (aged 12–17 years). ¹⁹ , ²⁰ Ipilimumab PKs in adults has also been characterized by a linear two‐compartment, zero‐order i.v. infusion model with time‐varying CL. ²¹ The maximum decrease in ipilimumab CL over time (6%) was much smaller than that of nivolumab (21%). ¹⁵

As discussed above, the possible contribution of age‐associated differences beyond body‐size differences to mAb PKs and the question of which body‐size measure is more relevant to the PopPK of pediatric patients (aged 2–17 years) have not been evaluated in the literature. To enable accurate prediction of pediatric PKs, the effects of age‐related covariates on nivolumab and ipilimumab PK were assessed by PopPK analysis using data from 13 nivolumab clinical studies and 10 ipilimumab clinical studies across multiple tumor types, including MEL, Hodgkin's lymphoma (HL), central nervous system tumors (CNST), and other STs. A separate pediatric PopPK analysis of MEL in the adjuvant setting was also performed. A pooled adult–pediatric PopPK modeling approach was used to estimate the effects of age‐related covariates on nivolumab or ipilimumab PKs. Finally, optimized pediatric PopPK models were successfully applied to support pediatric dosing recommendations.

METHODS

Data

The nivolumab PopPK analysis dataset included 13,104 nivolumab concentration values from 2325 patients in 13 clinical studies (Table S1), of whom 275 were pediatric patients. Three studies (phase I/II) included young pediatric (aged 1–11 years) or adolescent (12–17 years) patients with STs, primary CNST, or HL (including cHL, HL, and non‐HL) who received nivolumab alone or in combination with either ipilimumab or brentuximab vedotin. Ten studies (phase I–III) involved adults with MEL, cHL, CNST (predominately GBM), or other ST. Baseline covariates of patients in this data set are summarized in Table 1. A separate nivolumab PopPK analysis data set for MEL in the adjuvant setting is described in Supplementary Results in Appendix S1.

TABLE 1.

Baseline demographic and clinical covariates in patients who received nivolumab.

Covariate	Adult MEL (N = 993)	Adult HL a (N = 274)	Adult GBM b (N = 556)	Adult other c (N = 227)	Pediatric ST d (N = 79)	Pediatric HL e (N = 46)	Pediatric CNST f (N = 150)	Total (N = 2325)
Sex, n (%)
Male	637 (64.1)	161 (58.8)	374 (67.3)	148 (65.2)	47 (59.5)	27 (58.7)	83 (55.3)	1477 (63.5)
Female	356 (35.9)	113 (41.2)	182 (32.7)	79 (34.8)	32 (40.5)	19 (41.3)	67 (44.7)	848 (36.5)
Race, n (%)
White	969 (97.6)	237 (86.5)	487 (87.6)	190 (83.7)	58 (73.4)	38 (82.6)	120 (80.0)	2099 (90.3)
Black/African American	2 (0.2)	16 (5.8)	10 (1.8)	19 (8.4)	8 (10.1)	4 (8.7)	7 (4.7)	66 (2.8)
Asian	9 (0.9)	8 (2.9)	38 (6.8)	14 (6.2)	7 (8.9)	1 (2.2)	9 (6.0)	86 (3.7)
Other	13 (1.3)	13 (4.7)	21 (3.8))	4 (1.8)	6 (7.6)	3 (6.5)	14 (9.3)	74 (3.2)
Baseline performance status, n (%)
0	726 (73.1)	150 (54.7)	136 (24.5)	61 (26.9)	21 (26.6)	26 (56.5)	55 (36.7)	1175 (50.5)
1	262 (26.4)	124 (45.3)	366 (65.8)	163 (71.8)	47 (59.5)	19 (41.3)	68 (45.3)	1049 (45.1)
2	5 (0.5)	0 (0)	54 (9.7)	3 (1.3)	11 (13.9)	1 (2.2)	26 (17.3)	100 (4.3)
3	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	1 (0.7)	1 (0.0)
Tumor type, n (%)
MEL	993 (100.0)	0 (0)	0 (0)	0 (0)	1 (1.3)	0 (0)	0 (0)	994 (42.8)
HL	0 (0)	274 (100.0)	0 (0)	0 (0)	0 (0)	46 (100.0)	0 (0)	320 (13.8)
CNST	0 (0)	0 (0)	556 (100.0)	0 (0)	0 (0)	0 (0)	150 (100.0)	706 (30.4)
Other ST	0 (0)	0 (0)	0 (0)	227 (100.0)	78 (98.7)	0 (0)	0 (0)	305 (13.1)
Treatment, n (%)
NIVO monotherapy	608 (61.2)	261 (95.3)	184 (33.1)	215 (94.7)	49 (62.0)	15 (32.6)	76 (50.7)	1408 (60.6)
NIVO + IPI 1 mg/kg q3w	0 (0)	0 (0)	6 (1.1)	12 (5.3)	30 (38.0)	0 (0)	74 (49.3)	122 (5.2)
NIVO + IPI 3 mg/kg q3w	385 (38.8)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	385 (16.6)
NIVO + brentuximab vedotin	0 (0)	13 (4.7)	0 (0)	0 (0)	0 (0)	31 (67.4)	0 (0)	44 (1.9)
NIVO + RTx	0 (0)	0 (0)	314 (56.5)	0 (0)	0 (0)	0 (0)	0 (0)	314 (13.5)
NIVO + RTx + temozolomide	0 (0)	0 (0)	52 (9.4)	0 (0)	0 (0)	0 (0)	0 (0)	52 (2.2)
Age (years)
Mean (SD)	59.9 (13.4)	35.9 (12.8)	56.5 (12.5)	58.8 (15.6)	11.2 (4.46)	13.8 (3.46)	9.44 (4.69)	50.3 (20.6)
Median (range)	62 (18–90)	33 (18–72)	58 (18–83)	62 (18–85)	12 (1–17)	15 (4–17)	10 (1–17)	55 (1–90)
Baseline body weight (kg)
Mean (SD)	82.3 (18.1)	76.6 (21.6)	79.3 (17.4)	79.7 (19.6)	44.1 (23.6)	59.7 (22.6)	37.7 (20.6)	76 (22.7)
Median (range)	80.6 (37.4–160)	73.6 (40–168)	78 (41.7–167)	78.7 (32.8–153)	43.2 (9.3–99.6)	58.2 (13.6–121)	33 (9.8–90)	76.2 (9.3–168)
Baseline eGFR (mL/min/1.73 m2)
Mean (SD)	86.7 (18.0)	105 (23)	93.1 (15.1)	83.7 (25)	119 (28.1)	118 (24.2)	118 (27.7)	93.8 (22.8)
Median (range)	88.1 (35.5–139)	107 (32.2–155)	93.7 (41–148)	85.6 (31.2–172)	116 (43.5–202)	125 (70.4–159)	115 (70.8–215)	92.9 (31.2–215)
Missing, n (%)	4 (0.403)	0 (0)	1 (0.18)	2 (0.881)	0 (0)	0 (0)	5 (3.33)	12 (0.516)
Baseline lactate dehydrogenase (U/L)
Mean (SD)	337 (336)	253 (141)	227 (98.5)	225 (147)	N/A	290 (113)	298 (169)	285 (254)
Median (range)	223 (98–2980)	213 (94–1029)	196 (58–827)	185 (91–1106)	N/A	256 (148–558)	223 (144–689)	210 (58–2980)
Missing, n (%)	16 (1.61)	5 (1.82)	25 (4.5)	26 (11.5)	79 (100)	16 (34.8)	126 (84)	293 (12.6)
Baseline serum albumin (g/dL)
Mean (SD)	4.12 (0.55)	4.02 (0.56)	4.06 (0.419)	3.95 (0.447)	3.75 (0.946)	3.65 (0.636)	4.37 (0.381)	4.08 (0.5)
Median (range)	4.2 (2.2–5.1)	4.1 (1.9–5.2)	4.1 (3–5.2)	4 (2.3–4.9)	4 (2.3–5)	3.65 (3.2–4.1)	4.4 (3.3–5.5)	4.1 (1.9–5.5)
Missing, n (%)	872 (87.8)	27 (9.85)	282 (50.7)	17 (7.49)	68 (86.1)	44 (95.7)	7 (4.67)	1317 (56.6)
Baseline tumor burden (cm)
Mean (SD)	7.83 (6.47)	8.38 (5.74)	N/A	11.7 (7.94)	8.38 (6.43)	7.35 (6.43)	N/A	8.48 (6.86)
Median (range)	5.8 (1–38.4)	7.1 (1.9–17.6)	N/A	9.8 (1–61.5)	5.9 (1–26)	4 (0.82–21.7)	N/A	6.7 (0.82–61.5)
Missing, n (%)	4 (0.403)	269 (98.2)	556 (100)	17 (7.49)	14 (17.7)	31 (67.4)	150 (100)	1041 (44.8)
Baseline tumor burden (cm2)
Mean (SD)	N/A	28.6 (29)	9.54 (8.52)	N/A	N/A	30.8 (29.3)	7.74 (7.11)	12.2 (15.2)
Median (range)	N/A	14.6 (4.86–101)	6.89 (0–52.8)	N/A	N/A	24 (1.5–125)	5.64 (1–36.5)	7.7 (0–125)
Missing, n (%)	993 (100)	239 (87.2)	244 (43.9)	227 (100)	79 (100)	15 (32.6)	75 (50)	1872 (80.5)
Baseline lean body mass (kg)
Mean (SD)	57.7 (10.8)	55.9 (11.5)	56.9 (10.3)	57.3 (11.1)	35.4 (16.1)	44.9 (14.3)	30.2 (14.8)	54.5 (13.8)
Median (range)	58.4 (31.7–94.7)	54.1 (36.1–96.5)	57.4 (34.9–94.9)	57.8 (27.8–91.2)	36.3 (7.81–67.9)	45.2 (12.4–80.4)	28 (8.68–63)	55.8 (7.81–96.5)
Missing, n (%)	25 (2.52)	1 (0.365)	5 (0.899)	16 (7.05)	0 (0)	0 (0)	4 (2.67)	51 (2.19)

Notes: Lean body mass (LBM) was estimated using Boer equation and Peter equation. ²⁹ The performance status was based on Eastern Cooperative Oncology Group (ECOG) Performance Status Scale.

Abbreviations: CNST, central nervous system tumors; eGFR, estimated glomerular filtration rate; GBM, glioblastoma; HL, lymphoma; IPI, ipilimumab; MEL, melanoma; N/A, not available; NIVO, nivolumab; q3w, every 3 weeks; RTx, radiotherapy; SD, standard deviation; STs, solid tumors.

^a Includes classical Hodgkin lymphoma (n = 269), Hodgkin lymphoma (n = 4), and non‐Hodgkin lymphoma (n = 1).

^b Includes GBM (n = 542), high‐grade glioma (n = 5), ependymoma (n = 3), diffuse intrinsic pontine glioma (n = 2), medulloblastoma (n = 2), diffuse midline glioma (n = 1), and pineoblastoma (n = 1).

^c Includes non‐small cell lung cancer (n = 139), renal cell carcinoma (n = 35), colorectal cancer (n = 18), Ewing sarcoma (n = 11), prostate cancer (n = 8), osteosarcoma (n = 7), neuroblastoma (n = 3), rhabdomyosarcoma (n = 2), and other STs (n = 4).

^d Includes osteosarcoma (n = 19), neuroblastoma (n = 18), rhabdomyosarcoma (n = 17), Ewing sarcoma (n = 10), melanoma (n = 1), and other STs (n = 14).

^e Includes classical Hodgkin lymphoma (n = 31), non‐Hodgkin lymphoma (n = 9), and Hodgkin lymphoma (n = 6).

^f Includes diffuse intrinsic pontine glioma (n = 34), medulloblastoma (n = 28), high‐grade glioma (n = 20), ependymoma (n = 19), diffuse midline glioma (n = 9), atypical teratoid rhabdoid tumor (n = 7), choroid plexus carcinoma (4), pineoblastoma (n = 3), malignant germ cell tumor (n = 2), anaplastic pleomorphic xanthoastrocytoma (n = 1), embryonal tumor with multilayered rosettes (n = 1), and other CNSTs (n = 22).

The ipilimumab PopPK analysis data set included 6020 ipilimumab concentration values from 1427 patients in 10 clinical studies (Table S1 ), of whom 138 were pediatric patients. Four studies (phase I/II) included young pediatric (aged 1–11 years) or adolescent (12–17 years) patients with MEL, CNST, or other STs who received ipilimumab alone or in combination with nivolumab. Six studies (phase I–III) involved adults with advanced MEL. Baseline covariates of patients in this dataset are summarized in Table 2.

TABLE 2.

Baseline demographic and clinical covariates in patients who received ipilimumab.

Covariate	Adult MEL (N = 1261)	Adult CNST a (N = 6)	Adult other b (N = 22)	Pediatric ST c (N = 43)	Pediatric CNST d (N = 72)	Pediatric MEL (N = 23)	Total (N = 1427)
Sex, n (%)
Male	808 (64.1)	4 (66.7)	14 (63.6)	27 (62.8)	37 (51.4)	9 (39.1)	899 (63.0)
Female	452 (35.8)	2 (33.3)	8 (36.4)	16 (37.2)	35 (48.6)	14 (60.9)	527 (36.9)
Missing	1 (0.1)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	1 (0.1)
Race, n (%)
White	1231 (97.6)	5 (83.3)	17 (77.3)	28 (65.1)	53 (73.6)	19 (82.6)	1353 (94.8)
Black/African American	3 (0.2)	1 (16.7)	1 (4.5)	5 (11.6)	4 (5.6)	1 (4.3)	15 (1.1)
Asian	13 (1.0)	0 (0)	2 (9.1)	4 (9.3)	6 (8.3)	1 (4.3)	26 (1.8)
Other	13 (1.0)	0 (0)	2 (9.1)	6 (14.0)	9 (12.5)	2 (8.7)	32 (2.2)
Missing	1 (0.1)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	1 (0.1)
Baseline performance status, n (%)
0	895 (71.0)	3 (50.0)	4 (18.2)	11 (25.6)	27 (37.5)	10 (43.5)	950 (66.6)
1	363 (28.8)	2 (33.3)	16 (72.7)	28 (65.1)	33 (45.8)	9 (39.1)	451 (31.6)
2	2 (0.2)	1 (16.7)	2 (9.1)	4 (9.3)	12 (16.7)	3 (13.0)	24 (1.7)
3	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	1 (4.3)	1 (0.1)
Missing	1 (0.1)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	1 (0.1)
Tumor type, n (%)
MEL	1261 (100.0)	0 (0)	0 (0)	0 (0)	0 (0)	23 (100.0)	1284 (90.0)
CNST	0 (0)	6 (100.0)	0 (0)	0 (0)	72 (100.0)	0 (0)	78 (5.5)
Other ST	0 (0)	0 (0)	22 (100.0)	43 (100.0)	0 (0)	0 (0)	65 (4.6)
Treatment, n (%)
IPI monotherapy	815 (64.6)	0 (0)	9 (40.9)	12 (27.9)	0 (0)	23 (100.0)	859 (60.2)
IPI + NIVO 1 mg/kg q3w	388 (30.8)	0 (0)	0 (0)	6 (14.0)	0 (0)	0 (0)	394 (27.6)
IPI + NIVO 3 mg/kg q3w	0 (0)	6 (100.0)	13 (59.1)	25 (58.1)	72 (100.0)	0 (0)	116 (8.1)
IPI 10 mg/kg q3w + budesonide 9 mg q.d.	58 (4.6)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	58 (4.1)
Age (years)
Mean (SD)	59.1 (13.3)	19.3 (1.21)	21.2 (2.97)	12.3 (3.96)	9.78 (4.69)	12.4 (3.99)	53.7 (19.5)
Median (range)	61 (18–89)	19.5 (18–21)	20 (18–27)	13 (4–17)	10 (1–17)	13 (2–16)	58 (1–89)
Missing, n (%)	1 (0.0793)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	1 (0.0701)
Baseline body weight (kg)
Mean (SD)	81.5 (17.2)	72.3 (27.3)	69 (17)	48.7 (26.6)	39.1 (20)	50.6 (20.5)	77.7 (21.1)
Median (range)	80.4 (38.6–160)	59.9 (54.3–124)	67 (40.6–96.1)	45 (12.9–151)	37.4 (10.2–87.9)	56.1 (12.2–91.3)	78.1 (10.2–160)
Missing, n (%)	1 (0.0793)	0 (0)	0 (0)	0 (0)	0 (0)	0 (0)	1 (0.0701)
Baseline eGFR (mL/min/1.73m2)
Mean (SD)	86.1 (19.4)	101 (31)	126 (24.2)	128 (30.6)	115 (29.3)	154 (28.6)	90.6 (24.5)
Median (range)	87.9 (21–151)	110 (43.7–126)	129 (91.5–172)	129 (71–192)	111 (70.8–208)	146 (107–230)	90.1 (21–230)
Missing, n (%)	3 (0.238)	0 (0)	0 (0)	0 (0)	3 (4.17)	0 (0)	6 (0.42)
Baseline lactate dehydrogenase (U/L)
Mean (SD)	324 (335)	N/A	214 (95.4)	215 (55.9)	235 (99.1)	323 (225)	321 (330)
Median (range)	213 (83–4539)	N/A	198 (115–414)	206 (123–334)	210 (144–444)	236 (164–1130)	213 (83–4539)
Missing, n (%)	4 (0.317)	6 (100)	13 (59.1)	31 (72.1)	62 (86.1)	0 (0)	116 (8.13)
Baseline serum albumin (g/dL)
Mean (SD)	4.19 (0.449)	4.64 (0.483)	3.86 (0.382)	3.81 (0.599)	4.41 (0.362)	3.97 (0.643)	4.19 (0.466)
Median (range)	4.2 (2.1–5.3)	4.6 (3.9–5.2)	3.9 (3–4.4)	3.8 (2.4–5)	4.4 (3.4–5.2)	3.9 (2.3–4.9)	4.2 (2.1–5.3)
Missing, n (%)	740 (58.7)	1 (16.7)	10 (45.5)	26 (60.5)	4 (5.56)	0 (0)	781 (54.7)
Baseline tumor burden (cm)
Mean (SD)	9.12 (8.01)	N/A	12.2 (8.29)	6.91 (5.41)	N/A	N/A	9.1 (7.97)
Median (range)	6.6 (1–67.2)	N/A	9.7 (3.7–34.1)	5.4 (1–21.8)	N/A	N/A	6.6 (1–67.2)
Missing, n (%)	35 (2.78)	6 (100)	9 (40.9)	14 (32.6)	72 (100)	23 (100)	159 (11.1)
Baseline tumor burden (cm2)
Mean (SD)	N/A	13.1 (9.92)	N/A	N/A	6.88 (7.38)	N/A	7.42 (7.66)
Median (range)	N/A	7.85 (6.86–24.5)	N/A	N/A	4.68 (1–36.5)	N/A	5.25 (1–36.5)
Missing, n (%)	1261 (100)	3 (50)	22 (100)	43 (100)	41 (56.9)	23 (100)	1393 (97.6)
Baseline lean body mass (kg)
Mean (SD)	57.4 (10.4)	51.5 (9.79)	53.2 (9.91)	38.5 (16.2)	31 (14.2)	39.3 (13.9)	55.1 (12.9)
Median (range)	58.3 (32.5–94.7)	50.7 (36.3–62.1)	53.4 (37.6–66.9)	40.9 (11.8–91.6)	30.8 (9.02–63)	39.9 (10.5–57.1)	56.8 (9.02–94.7)
Missing, n (%)	32 (2.54)	0 (0)	0 (0)	0 (0)	2 (2.78)	0 (0)	34 (2.38)

Abbreviations: CNST, central nervous system tumors; eGFR, estimated glomerular filtration rate; IPI, ipilimumab; MEL, melanoma; N/A, not available; NIVO, nivolumab; q3w, every 3 weeks; SD, standard deviation; STs, solid tumors.

^a Includes ependymoma (n = 2), medulloblastoma (n = 2), diffuse intrinsic pontine glioma (n = 1), and high‐grade glioma (n = 1).

^b Includes Ewing sarcoma (n = 5), osteosarcoma (n = 5), rhabdomyosarcoma (n = 3), and other STs (n = 9).

^c Includes osteosarcoma (n = 8), rhabdomyosarcoma (n = 6), Ewing sarcoma (n = 5), renal cell carcinoma (n = 2), neuroblastoma (n = 1), ST (n = 11), and others (n = 10).

^d Includes diffuse intrinsic pontine glioma (n = 16), medulloblastoma (n = 13), high‐grade glioma (n = 9), ependymoma (n = 8), diffuse midline glioma (n = 4), atypical teratoid rhabdoid tumor (n = 3), choroid plexus carcinoma (n = 2), anaplastic pleomorphic xanthoastrocytoma (n = 1), embryonal tumor with multilayered rosettes (n = 1), malignant germ cell tumor (n = 1), and other CNSTs (n = 14).

The distributions for the PK samples per subject are shown in Figure S1.

PopPK model development

Nivolumab and ipilimumab PopPK models were developed in two steps: first as base models and then as full models. PopPK model parameters were estimated using first‐order conditional estimation with an interaction method implemented in NONMEM (version 7.4 or 7.5, ICON Development Solutions). A nonparametric bootstrap (N = 1000) that evaluated the precision of the final estimated parameters was performed using the full model selected to determine parameter uncertainty and estimate 95% confidence intervals.

Nivolumab

Base‐model development consisted of re‐estimating parameters from a previously developed model for adult patients with STs ¹⁵ using the current analysis data set. The previously developed final model was a two‐compartment, zero‐order i.v. infusion model with time‐varying elimination according to a sigmoidal maximum effect (E _max) function (maximum change in CL over time). A proportional residual error model was used that included log‐normally distributed random effects on CL, central volume (VC), and peripheral volume (VP), normally distributed random effects on E _max; and a correlation between CL and VC random effects. The base model included the effects of baseline body weight (WTB), estimated glomerular filtration rate (eGFR), sex, performance status (PS), and race on CL; and WTB and sex on VC. Additionally, the effects of WTB on intercompartmental CL (Q) and VP were constrained to be the same as CL and VC.

A full model was developed by incorporating these additional covariates in the estimation of PK parameters: the effects of age, combination therapy, and patient population on baseline CL; and the effects of PS, patient population, and combination therapy on E _max. The full model with allometric scaling of WTB on CL and VC was initially assessed and then further refined by including alternative measures of body size—lean body mass (LBM) and baseline body surface area (BSA)—and age effects. Potential age effects on CL and VC were evaluated with or without dependence on tumor type; the best model was selected based on Bayesian information criterion (BIC) and plausibility.

For the purposes of this study, the patient populations analyzed (with the specified ages) were termed adult MEL, adult HL, adult GBM, pediatric ST (<12 years or 12–17 years), pediatric HL (<18 years), and pediatric CNST (<18 years). cHL, HL, and non‐HL were combined into a single HL category because of the small sample size. The effects of pediatric ST (<18 years) and pediatric HL (<18 years) on E _max were assumed to be the same as those of adult MEL, which was supported by a sensitivity analysis. The categorical effects of age (pediatric patients <12 years, adolescent patients 12–17 years) on baseline VC were assessed.

To provide dosing recommendations for adjuvant treatment of adolescents (12–17 years) with MEL, a separate nivolumab PopPK model was developed (Supplementary Results in Appendix S1).

Ipilimumab

Base‐model development started with re‐estimating the parameters of a previously developed final model ²¹ using the current PopPK data set. In the prior PopPK analysis, the maximum decrease in ipilimumab CL over time was only ~6%, and ~15 weeks was needed to reach half of the maximum reduction. ²¹ As ipilimumab CL was not expected to change dramatically over a 12‐week dosing period (administered every 3 weeks [q3w] for 4 doses), ipilimumab CL was modeled as stationary. A combined proportional, additive error model was used, which included log‐normally distributed random effects on CL, VC, and a correlation between CL and VC random effects. The base model included the effects of WTB on CL and VC. Additionally, the effects of WTB on Q and VP were constrained to be the same as CL and VC.

The full model was developed by incorporating the following additional covariates in the estimation of PK parameters: the effects on baseline CL of age and body‐size measures (WTB, LBM, and BSA), combination therapy, and patient population. Potential age effects on CL and VC were modeled with the same approach used for nivolumab. The patient populations analyzed were adult MEL, adult CNST, pediatric MEL, and pediatric CNST (<18 years). The categorical effects of age were assessed in the same manner used for nivolumab.

Model evaluation

Model evaluation was performed using standard goodness‐of‐fit plots and prediction‐corrected visual predictive checks (pcVPCs). The pcVPC was performed using 1000 simulated datasets that were obtained using parameter values from the full model, in which the 5th, 50th, and 95th percentiles of observed plasma concentration–time data with their corresponding model‐based 90% prediction intervals were plotted. The pcVPC and bootstrap approaches were conducted using Perl‐speaks‐NONMEM (version 4.9.0); diagnostic plots were prepared using R (version 4.0.2).

Model application

Nivolumab exposures were simulated using stochastic simulations for adolescents (12–17 years) with STs on nivolumab alone or in combination with ipilimumab. Adult MEL exposures were simulated using the approved adult dosing regimens: nivolumab monotherapy (nivolumab 240 mg every 2 weeks [q2w] or 480 mg every 4 weeks [q4w]) or nivolumab plus ipilimumab combination therapy (nivolumab 1 mg/kg plus ipilimumab 3 mg/kg q3w for the first 4 doses followed by nivolumab 240 mg q2w or 480 mg q4w).

Stochastic simulations were performed using an adolescent population created by random sampling from the National Health and Nutrition Examination Survey database (2017–2018). ²² This population included 800 subjects aged greater than or equal to 12 to less than 18 years. The median baseline eGFR of adolescent patients included in the PopPK models was used in the simulation. The distribution of PS was dependent on the tumor type (Table 1). Therefore, the frequency of a PS of 0 (vs. PS >0) in adolescents was randomly assigned to reflect the distribution frequency in adult MEL. An adult MEL population (N = 500) was created by random sampling from the adult patients with MEL included in the PopPK models. All stochastic simulations were performed 100 times, and the repeated simulations were treated as independent patients. One hundred patients were randomly selected for each adolescent or adult body‐weight group. The adolescents from the National Health and Nutrition Examination Survey had similar body weight‐age relationship as the actual adolescent patients (Figure S2).

Ipilimumab exposures in adolescents (12–17 years) with MEL were simulated using an approach similar to that used for nivolumab (except that eGFR and PS were not needed). The approved adult dosing regimen of ipilimumab monotherapy for advanced MEL is 3 mg/kg q3w for four doses.

RESULTS

PopPK model development

Nivolumab models

The nivolumab initial full model (Full1) that was developed did not provide accurate prediction in pediatric populations (Figure S3). Alternative body‐size measures (LBM and BSA) were then evaluated in the initial full model in comparison with WTB (Table 3). Model Full3a had the lowest BIC and was selected for the next step. The effect of age on CL and VC was then evaluated using two approaches. The first approach assumed that the age effect on CL was the same across different tumor types, whereas the second assumed that this effect was tumor‐type–dependent (the age effect on CL for each tumor type was parameterized differently). Models Full2 and Full1c had the lowest BIC values using the first and second approaches, respectively. Parameter estimates in the Full2 model showed that CL increased with age regardless of tumor type. However, this assumption may not hold true because CL in the pediatric HL and the adult HL populations was similar (Figure 1a). Therefore, the Full1c model (NONMEM code in Appendix File S2) was selected for model application.

TABLE 3.

Selection of nivolumab population pharmacokinetic full models.

Model no.	Model description (covariate effects)			Parameter number	OFV	BIC	ΔBIC a
	Effect on baseline CL	Effect on E max	Effect on VC
Starting Full1 model
Full1	WTB, baseline eGFR, sex, PS, race, and patient population (HL, CNST, other STs vs. MEL), combination (I1Q3, I3Q3, and BVCO)	Patient population (adult MEL, adult HL, adult other STs, and pediatric CNST), combination (NIVO + IPI)	WTB, sex	31	74,339	74,653	0
Effect of body‐size parameters (LBM or BSA vs. WTB)
Full3	Same as Full1 except effect of LBM added	Same as Full1	Effect of LBM added	31	74,286	74,601	−53
Full3a	Same as Full1	Same as Full1	Effect of LBM added	31	74,248	74,563	−90
Full3b	Same as Full1 except effect of LBM added	Same as Full1	Same as Full1	31	74,388	74,703	50
Full3c	Same as Full1 except effect of BSA added	Same as Full1	Effect of BSA added	31	74,289	74,604	−50
Effect of age on CL and VC without tumor difference
Full2	Same as Full3a except single numeric effect of age added	Same as Full1	Single numeric effect of age added	33	74,138	74,473	−180
Full2a	Same as Full3a except categorical effect of age (pediatric <12 years, adolescent 12–17 years vs. adult) added	Same as Full1	Categorical effect of age (pediatric <12 years, adolescent 12–17 years vs adult) added	35	74,137	74,493	−160
Full2b	Same as Full3a except separate numeric effects of age (by pediatric <12 years, adolescent 12–17 years, and adult) added	Same as Full1	Separate numeric effects of age (by pediatric <12 years, adolescent 12–17 years, and adult) added	37	74,228	74,604	−49
Categorical effect of age on CL with tumor‐type difference and categorical effect of age on VC
Full1a	Same as Full3a except pediatric (<18 years) ST, HL, and CNST effect added	Same as Full1	Pediatric (<12 years) and adolescent (12–17 years) effect added	36	74,134	74,499	−154
Full1b	Same as Full3a except pediatric (<12 years) and adolescent (12–17 years) ST, HL, and CNST effect added	Same as Full1	Same as Full1a	39	74,120	74,517	−137
Full1c	Same as Full1b except combined pediatric and adolescent HL and CNST effects added	Same as Full1	Same as Full1a	37	74,122	74,498	−156

Abbreviations: BIC, Bayesian information criterion; BSA, baseline body surface area; BVCO, combination with brentuximab vedotin; CL, clearance; CNST, central nervous system tumor; eGFR, estimated glomerular filtration rate; E _max, maximal change in CL over time; HL, lymphoma; I1Q3, combination with ipilimumab 1 mg/kg q3w; I3Q3, combination with ipilimumab 3 mg/kg q3w; IPI, ipilimumab; LBM, lean body mass; MEL, melanoma; NIVO, nivolumab; OFV, objective function value; PS, performance status; ST, solid tumors; VC, central volume; WTB, baseline body weight.

^a Relative to the BIC of the reference model (Full1).

FIGURE 1.

Covariate effects on full nivolumab population pharmacokinetic model parameters: (a) baseline CL and (b) CL_ss/CL₀ and VC. The reference patient was an adult male (white/other race) with MEL and a body weight of 75 kg, lean body mass of 55 kg, performance status of 0, and baseline eGFR of 90 mL/min/1.73 m² who received nivolumab monotherapy. The parameter estimate in the reference patient was considered to be 100% (solid vertical line); dashed vertical lines indicate 80% and 125% of this value. CI values were taken from bootstrap calculations (980 of 1000 successful). The effects of body weight and lean body mass were also added to intercompartmental CL and peripheral volume, respectively, estimates of which were fixed to be similar to those of CL and VC, respectively. CL_ss/CL₀ was calculated using CL_ss/CL₀ = e^EMAX. Adolescent, aged 12–17 years; CI, confidence interval; CL, clearance; CL₀, baseline CL; CL_ss, steady‐state CL; CNST, central nervous system tumors; combo, combination therapy; eGFR, estimated glomerular filtration rate; GBM, glioblastoma; HL, lymphoma; IPI 1 mg/kg q3w, combination of nivolumab with ipilimumab 1 mg/kg q3w; IPI 3 mg/kg q3w, combination of nivolumab with ipilimumab 3 mg/kg q3w; IPI combo, combination of nivolumab with ipilimumab; MEL, melanoma; mono, monotherapy; other, other ST; P05, 5th percentile; P95, 95th percentile; pediatric (<12 years), pediatric patients aged <12 years; ST, solid tumors; VC, central volume.

Covariate effects in the full model are shown in Figure 1 (Figure S4: effects on exposure). Adolescent (12–17 years) and pediatric (<12 years) patients with STs showed 20% and 44% lower baseline CL, respectively, than adult patients with MEL (Figure 1a). Baseline CL in pediatric (<18 years) and adult patients with HL was similar and ~32% to 34% lower than that of adult patients with MEL. Pediatric patients with CNST (<18 years) and adult patients with GBM showed 55% and 44% lower baseline CL, respectively, than adult patients with MEL. In addition, adolescent (12–17 years) and pediatric (<12 years) patients had 24% lower VC than adult patients (Figure 1b). Other covariate effects were consistent with prior PopPK analyses. ¹⁵ , ¹⁶ , ¹⁷

The full model (Full1c) is represented using the following equations with the parameter estimates shown in Table 4:

$${\mathit{CL}}_i\left(t\right)=\mathit{CL}{0}_{\mathit{TV},i}\times e^{\left(\frac{{\mathit{\text{EMAX}}}_i·t^{\mathit{\text{HILL}}}}{T{50}^{\mathit{\text{HILL}}}+t^{\mathit{\text{HILL}}}}\right)}\times e^{{\eta }_{\mathit{CL},i}}$$

where

$$\begin{array}{c}\mathit{CL}{0}_{\mathit{TV},i}=\mathit{CL}{0}_{\mathit{REF}}\times {\left(\frac{{\mathit{WTB}}_i}{{\mathit{WTB}}_{\mathit{REF}}}\right)}^{{\mathit{CL}}_{\mathit{WTB}}}\times {\left(\frac{{\mathit{\text{eGFR}}}_i}{{\mathit{\text{eGFR}}}_{\mathit{REF}}}\right)}^{{\mathit{CL}}_{\mathit{\text{eGFR}}}}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{SEX}}}}^{\left(\text{if female}\right)}\times {e^{{\mathit{CL}}_{\mathit{PS}}}}^{\left(\text{if}\mathrm{PS}>0\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{RAAA}}}}}^{\left(\text{if RACE is African American}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{RAAS}}}}}^{\left(\text{if RACE is Asian}\right)}\times {e^{{\mathit{CL}}_{\mathit{HL}}}}^{\left(\text{if}\mathrm{POP}\text{is adult}\mathrm{HL}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{GBM}}}}^{\left(\text{if}\mathrm{POP}\text{is adult}\mathrm{GBM}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{OTH}}}}^{\left(\text{if}\mathrm{POP}\text{is adult other}\mathrm{ST}\right)}\hfill \\ \times {e^{{\mathrm{C}L}_{\mathit{\text{pedST}}}}}^{\left(\text{if}\mathrm{POP}\mathrm{ped}\left[<12\mathrm{y}\right]\mathrm{ST}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{adoST}}}}}^{\left(\text{if}\mathrm{POP}\mathrm{ado}\left[12-17\mathrm{y}\right]\mathrm{ST}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{pedHL}}}}}^{\left(\text{if}\mathrm{POP}\mathrm{ped}\left[<18\mathrm{y}\right]\mathrm{HL}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{pedCNST}}}}}^{\left(\text{if}\mathrm{POP}\mathrm{ped}\left[<18\mathrm{y}\right]\text{CNST}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{I1Q3}}}^{\left(\text{if NIVO}+\mathrm{IPI}1\mathrm{mg}/\mathrm{kg}\mathrm{Q}3\mathrm{W}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{I3Q3}}}^{\left(\text{if NIVO}+\mathrm{IPI}3\mathrm{mg}/\mathrm{kg}\mathrm{Q}3\mathrm{W}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{BVCO}}}}}^{\left(\text{if NIVO}+\text{brentuximab vedotin}\right)}\hfill \end{array}$$

and

$$\begin{array}{c}{\mathit{\text{EMAX}}}_i={\mathit{\text{EMAX}}}_{\mathit{REF}}+{{\mathit{\text{EMAX}}}_{\mathit{PS}}}^{\left(\text{if}\mathrm{PS}>0\right)}\hfill \\ +{{\mathit{\text{EMAX}}}_{\mathit{\text{COMBO}}}}^{\left(\text{if NIVO}+\mathrm{IPI}\text{combination}\right)}\hfill \\ +{{\mathit{\text{EMAX}}}_{\mathit{HL}}}^{\left(\text{if}\mathrm{POP}\text{adult}\mathrm{HL}\right)}\hfill \\ +{{\mathit{\text{EMAX}}}_{\mathit{OTH}}}^{\left(\text{if}\mathrm{POP}\text{adult other}\mathrm{ST}\right)}\hfill \\ +{{\mathit{\text{EMAX}}}_{\mathit{\text{pedCNST}}}}^{\left(\text{if}\mathrm{POP}\mathrm{ped}\left[<18\mathrm{y}\right]\text{CNST}\right)}+{\eta }_{\text{EMAX},i}\hfill \end{array}$$

TABLE 4.

Parameter estimates in the full nivolumab population pharmacokinetic model.

Parameter	Parameter note	Estimate a	Standard error (RSE% b )	95% CI c
Fixed effects
CL0 REF (mL/h)	Typical baseline clearance	9.66	0.279 (2.89)	8.98 to 10.3
VC REF (L)	Typical central volume	4.01	0.0434 (1.08)	3.93 to 4.10
Q REF (mL/h)	Typical intercompartmental CL	35.9	1.61 (4.47)	32.8 to 39.2
VP REF (L)	Typical peripheral volume	2.77	0.0701 (2.53)	2.62 to 2.93
CL WTB	Body weight on CL	0.630	0.0330 (5.23)	0.566 to 0.693
CL eGFR	eGFR on CL	0.0982	0.0357 (36.4)	0.0206 to 0.174
CL SEX	Sex on CL	−0.0998	0.0184 (18.4)	−0.135 to −0.0630
CL PS	Performance status on CL	0.166	0.0206 (12.4)	0.128 to 0.207
CL RAAA	African American on CL	0.0693	0.0471 (68.0)	−0.0237 to 0.159
CL RAAS	Asian on CL	0.00333	0.0365 (1090)	−0.0693 to 0.0786
VC LBM	Lean body mass on VC	0.932	0.0325 (3.49)	0.872 to 1.00
VC SEX	Sex on VC	0.0195	0.0186 (95.5)	−0.0168 to 0.0578
EMAX REF	Typical maximal change in CL	−0.298	0.0371 (12.4)	−0.385 to −0.184
T50 (h)	Time at which CL achieves half of E max	2670	442 (16.6)	1890 to 3750
HILL	Hill coefficient	2.32	0.375 (16.2)	1.69 to 3.71
CL HL	Adult lymphoma on CL	−0.382	0.0331 (8.66)	−0.447 to −0.321
CL GBM	Adult glioblastoma on CL	−0.578	0.0337 (5.84)	−0.643 to −0.493
CL OTH	Adult other tumors on CL	0.00699	0.0369 (529)	−0.0679 to 0.0789
CL PEDST	Pediatric (<12 years) solid tumors on CL	−0.580	0.0893 (15.4)	−0.761 to −0.406
CL PEDHL	Pediatric (<18 years) lymphoma on CL	−0.411	0.0702 (17.1)	−0.554 to −0.227
CL PEDCNST	Pediatric (<18 years) CNS tumors on CL	−0.801	0.0654 (8.16)	−0.930 to −0.654
CL I1Q3	Combination with ipilimumab 1 mg/kg q3w on CL	0.0973	0.0534 (54.9)	−0.00235 to 0.208
CL I3Q3	Combination with ipilimumab 3 mg/kg q3w on CL	0.349	0.0340 (9.73)	0.278 to 0.416
CL BVCO	Combination with brentuximab vedotin on CL	0.132	0.0634 (48.2)	−0.0572 to 0.282
EMAX PS	Performance status on E max	−0.157	0.0469 (29.8)	−0.263 to −0.0708
EMAX IPICO	Combination with ipilimumab on E max	−0.124	0.0811 (65.4)	−0.361 to 0.0274
EMAX HL	Adult lymphoma on E max	0.132	0.0442 (33.5)	0.0391 to 0.233
EMAX OTH	Adult other tumors on E max	0.118	0.0680 (57.7)	−0.0210 to 0.263
EMAX PEDCNST	Pediatric (<18 years) CNS tumors on E max	0.696	0.133 (19.1)	0.439 to 1.03
CL ADOST	Adolescent (12–17 years) solid tumors on CL	−0.223	0.0832 (37.4)	−0.393 to −0.0655
VC PED	Pediatric (<12 years) on VC	−0.277	0.0458 (16.5)	−0.364 to −0.181
VC ADO	Adolescents (12–17 years) on VC	−0.273	0.0253 (9.30)	−0.322 to −0.222
Random effects
ω 2 CL	Interindividual variability of CL	0.108 (0.329)	0.00604 (5.57)	0.0957 to 0.120
ω 2 VC	Interindividual variability of VC	0.0751 (0.274)	0.00764 (10.2)	0.0607 to 0.0904
ω 2 EMAX	Interindividual variability of E max	0.160 (0.400)	0.0432 (27.0)	0.0816 to 0.262
ω 2 CL:ω 2 VC d	Correlated parameter	0.0220 (0.244)	0.00309 (14.1)	0.0163 to 0.0281
Residual error
Proportional	Proportional error term	0.199	0.00376 (1.89)	0.191 to 0.206

Note: CL0 _REF is the typical value of baseline CL in a reference patient with melanoma receiving nivolumab monotherapy who is a White adult male weighing 75 kg with an LBM of 55 kg and with a normal PS (0). EMAX _REF is a typical value of change in magnitude of CL in a reference adult with melanoma receiving nivolumab monotherapy with a PS of 0. VC _REF , Q _REF , and VP _REF are typical values in a reference patient weighing 75 kg with an LBM of 55 kg. These reference values represent the approximate median values in the analysis dataset. η shrinkage (%): ηCL, 12.2; ηVC, 28.1; ηEMAX, 50.3; ε shrinkage (%): 15.0. The condition number for the full model was 179.

Abbreviations: ADO, adolescents; ADOST, adolescent (12–17 years) solid tumors; BVCO, combination with brentuximab vedotin; CI, confidence interval; CL, clearance; eGFR, estimated glomerular filtration rate; E _max, maximal change in CL over time; GBM, adult glioblastoma; HILL, coefficient for time‐varying CL; HL, adult Hodgkin's lymphoma; I1Q3, combination with ipilimumab 1 mg/kg q3w; I3Q3, combination with ipilimumab 3 mg/kg q3w; IPICO, combination with ipilimumab; LBM, lean body mass; OTH, adult other tumors; PED, pediatric (<12 years); PEDCNST, pediatric (<18 years) central nervous system tumors; PEDHL, pediatric (<18 years) lymphoma; PEDST, pediatric (<12 years) solid tumors; PS, performance status; Q, intercompartmental CL; q3w, every 3 weeks; RAAA, African American race; RAAS, Asian race; REF, reference; RSE, relative standard error; T50, time at which CL achieves half of the maximum value; VC, central volume; VP, peripheral volume; WTB, baseline body weight; ω²CL, interindividual variability of clearance; ω²EMAX, interindividual variability of EMAX; ω²VC, interindividual variability of VC.

^a Random effects and residual error parameter estimates are shown as variance (standard deviation) for diagonal elements (ω _i,i or σ _i,i ) and covariance (correlation) for off‐diagonal elements (ω _i,j or σ _i,j ).

^b Standard error as a percentage of the estimate.

^c Taken from bootstrap calculations (980 of 1000 successful).

^d Correlated parameters.

The value of VC for patient i is given by:

$$\begin{array}{c}{\mathit{VC}}_i={\mathit{VC}}_{\mathit{REF}}\times {\left(\frac{{\mathit{LBM}}_i}{{\mathit{LBM}}_{\mathit{REF}}}\right)}^{{\mathit{VC}}_{\mathit{LBM}}}\times {e^{{\mathit{VC}}_{\mathrm{SEX}}}}^{\left(\text{if female}\right)}\hfill \\ \times {e^{{\mathit{VC}}_{\mathit{ado}}}}^{\left(\text{if}\mathrm{POP}\text{is}12–17\mathrm{y}\right)}\times {e^{{\mathit{VC}}_{\mathit{ped}}}}^{\left(\text{if}\mathrm{POP}<12\mathrm{y}\right)}\times e^{{\eta }_{\mathit{VC},i}}\hfill \end{array}$$

The values of Q and VP for patient i are given by:

$$Q_i=Q_{\mathit{REF}}\times {\left(\frac{{\mathit{WTB}}_i}{{\mathit{WTB}}_{\mathit{REF}}}\right)}^{Q_{\mathit{WTB}}}\times e^{{\eta }_{Q,i}}$$

$${\mathit{VP}}_i={\mathit{VP}}_{\mathit{REF}}\times {\left(\frac{{\mathit{LBM}}_i}{{\mathit{LBM}}_{\mathit{REF}}}\right)}^{{\mathit{VP}}_{\mathit{LBM}}}\times e^{{\eta }_{\mathit{VP},i}}$$

In these equations, CL0 _REF is the typical CL value at time 0 (CL ₀ ) at the reference values of WTB (75 kg), LBM (55 kg), PS (0), baseline eGFR (90 mL/min/1.73 m²), sex (male), race (white/other), and patient population (POP; adult MEL); VC _REF is the typical VC value at the reference values of LBM (55 kg), sex (male), and patient population (all adults); Q _REF and VP _REF are typical Q and VP values at the reference values of WTB and LBM, respectively; and EMAX _REF represents the typical E _max value at the reference value of PS (0) and patient population (adult MEL) on nivolumab monotherapy. T50 represents the time at which the change in CL is 50% of E _max, and HILL represents the sigmoidicity of the relationship with time. The following abbreviations were also used: ado, adolescent; IPI, ipilimumab; NIVO, nivolumab; and ped, pediatric. The interindividual variability in the structural model parameter P (where P is CL, VC, Q, VP, or E _max) is represented by a normally distributed random variable ${\eta }_{P,i}$ that has a mean of zero and a variance of ${\omega }_P^2$.

CL in adolescents (12–17 years) with MEL who received adjuvant nivolumab treatment was 19% lower than that of comparable adult patients (Supplementary Results in Appendix S1). This result was similar to the 20% lower CL observed in adolescent patients with STs compared with adult patients with MEL.

Ipilimumab model

The model selection process for ipilimumab was similar to that of nivolumab (Table S2). The model Full1 (NONMEM code in Appendix File S3) with age as a categorical effect on CL and separated by tumor type was selected for model application. Covariate effects in the full model are shown in Figure 2. Pediatric patients with MEL (<18 years) showed a 29% lower CL than adult patients with MEL. CL in pediatric (<18 years) and adult patients with CNST was similar and ~48% to 49% lower than in adult patients with MEL. Pediatric (<18 years) and adult patients with other STs showed 37% and 50% lower baseline CL, respectively, than adult patients with MEL. In addition, adolescent (12–17 years) and pediatric (<12 years) patients had 19% and 26% lower VC, respectively, than adult patients.

FIGURE 2.

Covariate effects on full ipilimumab population pharmacokinetic model parameters. The reference patient was an adult male with MEL and a lean body mass of 55 kg who received ipilimumab monotherapy. The parameter estimate in the reference patient was considered to be 100% (solid vertical line); dashed vertical lines indicate 80% and 125% of this value. CI values were taken from bootstrap calculations (982 of 1000 successful). The effect of lean body mass was also added to intercompartmental CL and peripheral volume, respectively, estimates of which were fixed to be similar to those of CL and VC, respectively. Adolescent, aged 12–17 years; CI, confidence interval; CL, clearance; CNST, central nervous system tumors; combo, combination therapy; NIVO 1 mg/kg q3w, combination of ipilimumab with nivolumab 1 mg/kg q3w; NIVO 3 mg/kg q3w, combination of ipilimumab with nivolumab 3 mg/kg q3w; MEL, melanoma; mono, monotherapy; other, other solid tumors; P05, 5th percentile; P95, 95th percentile; pediatric (<12 years), pediatric patients aged <12 years; VC, central volume.

The full model (Full1) is represented using the following equations, with parameter estimates shown in Table 5:

$${\mathit{CL}}_i\left(t\right)=\mathit{CL}{0}_{\mathit{TV},i}\times e^{{\eta }_{\mathit{CL},i}}$$

where

$$\begin{array}{c}\mathit{CL}{0}_{\mathit{TV},i}=\mathit{CL}{0}_{\mathit{REF}}\times {\left(\frac{{\mathit{LBM}}_i}{{\mathit{LBM}}_{\mathit{REF}}}\right)}^{{\mathit{CL}}_{\mathit{WTB}}}\hfill \\ \hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{CNST}}}}}^{\left(\text{if}\mathrm{POP}\text{is adult CNST}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{OTH}}}}^{\left(\text{if}\mathrm{POP}\text{is adult other}\mathrm{ST}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{pedOTH}}}}}^{\left(\text{if}\mathrm{POP}\text{is}\mathrm{ped}\left[<18\mathrm{y}\right]\text{other}\mathrm{ST}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{pedCNST}}}}}^{\left(\text{if}\mathrm{POP}\mathrm{ped}\left[<18\mathrm{y}\right]\text{CNST}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{\mathit{\text{pedMEL}}}}}^{\left(\text{if}\mathrm{POP}\mathrm{ped}\left[<18\mathrm{y}\right]\mathrm{MEL}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{N1Q3}}}^{\left(\text{if NIVO}1\mathrm{mg}/\mathrm{kg}+\mathrm{IPI}\mathrm{Q}3\mathrm{W}\right)}\hfill \\ \times {e^{{\mathit{CL}}_{N3Q3}}}^{\left(\text{if NIVO}3\mathrm{mg}/\mathrm{kg}+\mathrm{IPI}\mathrm{Q}3\mathrm{W}\right)}\hfill \end{array}$$

TABLE 5.

Parameter estimates in the full ipilimumab population pharmacokinetic model.

Parameter	Parameter note	Estimate a	Standard error (RSE%) b	95% CI c
Fixed effects
CL REF (mL/h)	Typical clearance	13.5	0.2.07 (1.53)	13.1 to 13.9
VC REF (L)	Typical central volume	3.90	0.0307 (0.786)	3.84 to 3.96
Q REF (mL/h)	Typical intercompartmental CL	35.8	2.33 (6.51)	31.2 to 40.4
VP REF (L)	Typical peripheral volume	3.47	0.0817 (2.35)	3.31 to 3.63
CL LBM	Lean body mass on CL	0.789	0.0536 (6.79)	0.684 to 0.894
VC LBM	Lean body mass on VC	0.874	0.0351 (4.01)	0.805 to 0.943
CL CNST	Adult CNS tumors on CL	−0.661	0.236 (35.7)	−1.12 to −0.199
CL OTH	Adult other tumors on CL	−0.698	0.277 (39.8)	−1.24 to −0.154
CL PEDOTH	Pediatric (<18 years) other solid tumors on CL	−0.462	0.110 (23.7)	−0.677 to −0.248
CL PEDCNST	Pediatric (<18 years) CNS tumors on CL	−0.668	0.191 (28.5)	−1.04 to −0.294
CL PEDMEL	Pediatric (<18 years) melanoma on CL	−0.347	0.107 (30.8)	−0.557 to −0.138
CL N1Q3	Combination with nivolumab 1 mg/kg q3w on CL	0.0417	0.0229 (54.9)	−0.00318 to 0.0866
CL N3Q3	Combination with nivolumab 3 mg/kg q3w on CL	0.316	0.181 (57.3)	−0.0390 to 0.670
VC PED	Pediatric (<12 years) on VC	−0.296	0.0552 (18.7)	−0.404 to −0.188
VC ADO	Adolescents (12–17) years on VC	−0.217	0.0341 (15.8)	−0.283 to −0.150
Random effects
ω 2 CL	Interindividual variability of CL	0.147 (0.383)	0.00853 (5.82)	0.130 to 0.163
ω 2 VC	Interindividual variability of VC	0.0531 (0.230)	0.00720 (13.6)	0.0390 to 0.0672
ω 2 CL:ω 2VCd	Correlated parameters	0.0258 (0.293)	0.00412 (15.9)	0.0178 to 0.0339
Residual error
Proportional	Proportional error term	0.185	0.00708 (3.83)	0.171 to 0.199
Additive (μg/mL)	Additive error term	1.14	0.171 (15.0)	0.805 to 1.48

Note: CL _REF is the typical value of clearance in a reference patient with melanoma receiving ipilimumab monotherapy who is a white adult male with a LBM of 55 kg. VC _REF, Q _REF , and VP _REF are typical values in a reference patient with a LBM of 55 kg. These reference values represent the approximate median values in the analysis dataset.

η shrinkage (%): ηCL, 10.7; ηVC, 22.2; ε shrinkage (%): 16.7.

The condition number for the full model was 162.

Abbreviations: ADO, adolescents (12–17 years); CI, confidence interval; CL, clearance, CNST, adult central nervous system tumors; LBM, lean body mass; N1Q3, combination with nivolumab 1 mg/kg q3w; N3Q3, combination with nivolumab 3 mg/kg q3w; OTH, adult other tumors; PED, pediatric (<12 years); PEDMEL, pediatric (<18 years) melanoma; PEDCNST, pediatric (<18 years) central nervous system tumors; PEDOTH, pediatric (<18 years) other solid tumors; PS, performance status; Q, intercompartmental CL; q3w, every 3 weeks; REF, reference; RSE, relative standard error; VC, central volume; VP, peripheral volume; ω²CL, interindividual variability of CL; ω²VC, interindividual variability of VC.

^a Random effects and residual error parameter estimates are shown as variance (standard deviation) for diagonal elements (ω_i,i or σ_i,i) and covariance (correlation) for off‐diagonal elements (ω_i,j or σ_i,j).

^b Standard error as a percentage of the estimate.

^c Taken from bootstrap calculations (982 of 1000 successful).

The value of VC for patient i is derived by:

$$\begin{array}{c}{\mathit{VC}}_i={\mathit{VC}}_{\mathit{REF}}\times {\left(\frac{{\mathit{LBM}}_i}{{\mathit{LBM}}_{\mathit{REF}}}\right)}^{{\mathit{VC}}_{\mathit{LBM}}}\times {e^{{\mathit{VC}}_{\mathit{ado}}}}^{\left(\text{if}\mathrm{POP}\text{is}12–17\mathrm{y}\right)}\hfill \\ \hfill \\ \times {e^{{\mathit{VC}}_{\mathit{ped}}}}^{\left(\text{if}\mathrm{POP}<12\mathrm{y}\right)}\times e^{{\eta }_{\mathit{VC},i}}\hfill \end{array}$$

The values of Q and VP for patient i are given by:

$$\begin{array}{c}{Q}_i=Q_{\mathit{REF}}\times {\left(\frac{{\mathit{LBM}}_i}{{\mathit{LBM}}_{\mathit{REF}}}\right)}^{Q_{\mathit{LBM}}}\hfill \end{array}$$

$$\begin{array}{c}{\mathit{VP}}_i={\mathit{VP}}_{\mathit{REF}}\times {\left(\frac{{\mathit{LBM}}_i}{{\mathit{LBM}}_{\mathit{REF}}}\right)}^{{\mathit{VP}}_{\mathit{LBM}}}\hfill \end{array}$$

In these equations, CL0 _REF is the typical CL value at the reference value of LBM (55 kg) and patient population (adult MEL); VC _REF is the typical VC value at the reference value of LBM (55 kg) and patient population (all adults); Q _REF and VP _REF are typical Q and VP values at the reference values of LBM, respectively.

Sensitivity analyses, model evaluation, and model application

Sensitivity analyses for the patient populations on E _max and the numeric effect of age on VC in the nivolumab full model (Table S3) supported the selection of the Full1c model for model application based on BIC criteria. The pcVPC and goodness‐of‐fit plots demonstrated that the models appropriately characterized nivolumab or ipilimumab PK from in both the adult and pediatric populations (Figures S5–S8).

Steady‐state nivolumab exposures in adolescent patients with STs in different body‐weight groups who received the approved adult flat dose of nivolumab monotherapy (240 mg q2w) were predicted to exceed those of adults, with the highest exposure differences occurring in the lowest adolescent body‐weight groups (Figure S9). With body‐weight–based dosing at 3 mg/kg q2w (up to a maximum dose of 240 mg), nivolumab steady‐state exposures in adolescents across different body‐weight groups were predicted to be similar to those of adults who received the approved flat dose (Figure S10). Similarly, capping the doses of both nivolumab and ipilimumab when administered to adolescents as a combination regimen using body‐weight–based dosing was predicted to yield nivolumab exposures similar to those of adults, while body‐weight–based dosing without caps did not (Figures S11 and S12; data at nivolumab 480 mg q4w not shown).

With ipilimumab monotherapy at the approved dose of 3 mg/kg q3w, after the fourth dose, ipilimumab exposures in adolescents in different body‐weight groups were predicted to exceed those in adults, with the highest exposure differences occurring in the highest adolescent body‐weight groups (Figure S13). Capping the ipilimumab dose at a maximum of 240 mg q3w in these patients helped mitigate exposures in adolescents weighing greater than or equal to 90 kg (although exposures in adolescents weighing 80–90 kg were not fully mitigated) compared with adult patients receiving the approved regimen (Figure S14). Similar findings were observed with nivolumab plus ipilimumab combination therapy (Figures S15 and S16).

DISCUSSION

It has generally been assumed that age does not have additional effects on the PK of mAbs in pediatric patients aged greater than or equal to 2 years after adjusting for the effect of body size. ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ In prior pediatric PopPK analyses for mAbs, body weight was the most commonly used covariate. ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰ We had expected that allometric scaling of body weight would be sufficient to describe PK in pediatric patients aged greater than or equal to 2 years. However, predictions in pediatric patients based on the nivolumab initial full model with body‐weight effect were unsatisfactory. Alternative measures of body size (LBM and BSA) were then assessed, as other body characteristics may contribute to PK differences between pediatric and adult subjects. ²³ The nivolumab or ipilimumab model that included the effects of LBM on VC showed improved BIC. The finding that LBM is most relevant to VC is consistent with the distribution of mAbs, which are confined largely to the plasma and extracellular fluid. ²⁴ Indeed, LBM has been shown to correlate better with blood volume than other body‐size parameters in adult and pediatric subjects. ²⁵ Additionally, we evaluated the age effect, which significantly improved the pediatric predictions. Prior analyses did not systematically evaluate the effect of age and body size, which may explain the discrepancy between our findings and those of previous analyses. ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ , ⁸ , ⁹ , ¹⁰

The current study has identified the tumor‐type–dependent contribution of age‐associated differences beyond body‐size differences to the PKs of mAbs in pediatric patients. In this analysis, young pediatric (<12 years) and adolescent (12–17 years) patients with STs had age‐dependent decreases in nivolumab CL, with 40% and 20% lower CL, respectively, relative to adults after adjusting for the effect of body size. Pediatric patients (1–17 years) also had lower nivolumab VC (24% lower) than adult patients with MEL. Similar results were also observed with ipilimumab. Nivolumab CL is time‐varying in advanced MEL but stationary in adjuvant MEL. However, similar pediatric effects were observed in both settings. In contrast, nivolumab CL in pediatric patients with HL was similar to that of adult patients with HL. The pediatric effects characterized have enabled the model to make accurate predictions, allowing it to be used for subsequent exposure comparisons between pediatric and adult patients, as well as for exposure–response analyses.

It is noteworthy that only one pediatric patient with MEL was available in the nivolumab analysis data set, and most of the remaining pediatric patients had other STs. Therefore, the pediatric effects observed with nivolumab for MEL may have been confounded by tumor‐type differences. However, ipilimumab CL was 29% lower in pediatric patients (<18 years) with MEL than adults with MEL. The similar magnitude of pediatric effects between nivolumab and ipilimumab suggest that the observed difference in nivolumab CL between pediatric patients with STs and adult patients with MEL was unlikely the result of tumor‐type differences.

Maturity function was used to describe mAbs PKs in children less than 2 years. ²⁶ The mechanisms underlying the PK differences observed between pediatric aged greater than or equal to 2 years and adult patients that were unexplainable by body size are unknown. The differences in disease status and immunobiology between pediatric and adult cancers may explain the PK differences observed. ²⁷ In addition, the expression of neonatal Fc receptor estimated from physiologically‐based PK modeling decreased with age, which may also contribute to the pediatric effects observed. ²⁸

In this analysis, pediatric patients with CNST appeared to have lower nivolumab CL than adults with CNST (mostly GBM), after adjusting for the effects of other covariates, including body size. In contrast, ipilimumab CL appeared to be similar in pediatric and adult patients with CNST. The adult CNST population included in the ipilimumab analysis had a mean age of 19 years, which is much younger than the adult CNST population in the nivolumab analysis dataset (mean 57 years), possibly explaining the difference between nivolumab and ipilimumab.

Due to the observed pediatric effects, adolescents with MEL had higher nivolumab exposures with the approved adult flat dose than adults. Switching to body‐weight–based dosing with a cap yielded similar exposures in adolescents and adults. However, the approved ipilimumab adult dose is body‐weight–based, and applying a dosing cap to that regimen did not fully mitigate the high exposures in some adolescents. Accordingly, exposure–response analysis was conducted for dose recommendations, which will be reported separately.

Nivolumab and ipilimumab show linear PKs in both adult and pediatric subjects, which is likely due to the negligible contribution of target mediated disposition to the total elimination at the clinically relevant dose levels. In the dose ranging study, patients received nivolumab dose from 0.1 to 10 mg/kg every 2 weeks and peripheral receptor occupancy was saturated at doses greater than or equal to 0.3 mg/kg. ³⁰

In conclusion, the current study has identified tumor‐type–dependent contributions of age‐associated differences beyond body size to the PKs of mAbs in pediatric patients (aged 2–17 years). The pediatric PopPK models developed were used to support nivolumab and ipilimumab dosing recommendations for adolescents with MEL in the metastatic and adjuvant settings. The study approach described could be extended to support pediatric dose optimization of other mAbs and possibly other biologics.

AUTHOR CONTRIBUTIONS

Z.H. and L.Z. wrote the manuscript. Z.H., S.L., Y.Z., S.D., L.H., J.S., A.R., and L.Z. designed the research. Z.H., S.L., Y.Z., and L.Z. performed the research. Z.H., S.L., Y.Z., S.D., L.H., J.S., A.R., and L.Z. analyzed the data.

FUNDING INFORMATION

The studies were supported by Bristol Myers Squibb.

CONFLICT OF INTEREST STATEMENT

All authors are employees and stock shareholders of Bristol Myers Squibb.

Supporting information

Appendix S1.

Hu Z, Liu S, Zhao Y, et al. Nivolumab and ipilimumab population pharmacokinetics in support of pediatric dose recommendations—Going beyond the body‐size effect. CPT Pharmacometrics Syst Pharmacol. 2024;13:476‐493. doi: 10.1002/psp4.13098