Cancer remains the leading cause of disease-related death in children. For the many children who experience relapses of their malignant solid tumors, usually after very intensive first-line therapy, curative treatment options are scarce. Preclinical drug testing to identify promising treatment elements that match the molecular make-up of the tumor is hampered by the fact that (i) molecular genetic data on pediatric solid tumors from relapsed patients and thus our understanding of tumor evolution and therapy resistance are very limited to date and (ii) for many of the high-risk entities, no appropriate and molecularly well-characterized patient-derived models and/or genetic mouse models are currently available. However, recent regulatory changes enacted by the European Medicines Agency (class waiver changes) and the maturation of the RACE for Children act with the FDA, will require a significant increase in preclinical pediatric cancer research and clinical development must occur. We detail the outcome of a pediatric cancer international multistakeholder meeting whose output aims at defining an international consensus on minimum preclinical testing requirements for the development of innovative therapies for children and adolescents with cancer. Recommendations based on the experience of the NCI funded PPTP/C (www.ncipptc.org) and the EU funded ITCC-P4 public private partnership (https://www.itccp4.eu/) are provided for the use of cell-based and mouse models for pediatric solid malignancies, as well as guidance on the scope and content of preclinical proof-of-concept data packages to inform clinical development dependent on clinical urgency. These recommendations can serve as a minimal guidance necessary to jumpstart preclinical pediatric research globally.

The 5-year survival rate of children and adolescents with cancer has plateaued at 80% in high-income countries, and cancer remains the primary cause of death by disease after age 1. There is an urgent need to introduce innovative therapies to cure more children and to do so with reduced long-term toxicity. Over the past 20 years, European and US regulatory initiatives have incentivized the pharmaceutical industry to develop medicines in the pediatric population (1, 2), although the impact of these incentives has been limited at best in pediatric oncology (3). Currently, the decision to evaluate a new anticancer drug for children with cancer is typically driven by a drug's adult indication rather than by its mechanism of action (MOA) and thus biological relevance to childhood cancer. For example, lung cancer does not occur in children; however, ALK is altered not only in lung cancer, but in numerous pediatric malignancies such as neuroblastoma and anaplastic large cell lymphoma (4).

Recently, the regulatory environment changed significantly. In 2015, the European Medicines Agency revised the class waiver list, and as a result, the number of adult cancer indications for which companies can apply for an automatic waiver for pediatric development was significantly reduced (5). However, concerns were raised over the impact of this revision to incentivize biology-driven development of anticancer drugs in the pediatric population (6). In August 2017, the RACE for Children Act was signed into law as Title V of the US FDA Reauthorization Act to amend the Pediatric Research Equity Act (PREA; 21 U.S. Code 355c) and entered into force in August 2020 (7). This act requires pediatric evaluation of new molecularly targeted drugs “intended for the treatment of adult cancers and directed at a molecular target substantially relevant to the growth or progression of a pediatric cancer.” With this more favorable regulatory environment there must and will be an increase in science-driven pediatric development of anticancer drugs. Recently, knowledge of the biology and heterogeneity of pediatric malignancies has rapidly increased (8, 9), and there is a correspondingly heightened need for thorough preclinical data using better predictive models to develop drugs targeting mechanisms specific to pediatric malignancies.

Research in this space is hampered by the lack of widespread access to comprehensive, well-validated preclinical tools, notably relevant animal models that effectively capture the molecular heterogeneity of pediatric cancer from treatment naïve and relapsed patients. Furthermore, international standards on the development of preclinical data packages necessary to inform clinical decisions do not exist and would greatly facilitate the identification and prioritization of promising agents. Following an international multistakeholder consensus meeting, we propose minimum preclinical testing requirements for the development of innovative therapies for children and adolescents with cancer.

As part of the Innovative Therapies for Children with Cancer—Pediatric Preclinical Proof-of-Concept Project ITCC-P4 (https://www.itccp4.eu/; ref. 10), an Innovative Medicines Initiative 2 (IMI-2) project, a multi-stakeholder workshop was held in late 2018 to define an international consensus on minimum preclinical testing requirements for the development of innovative therapies for pediatric cancer. This meeting was co-organized with the NCI-funded Pediatric Preclinical Testing Consortium (PPTC; www.ncipptc.org; ref. 11), and was unique in that it brought together academia, industry, regulatory agencies and advocacy groups, the outcome of which is the basis for the recommendations.

Current situation of preclinical testing in pediatric oncology drug development

Most preclinical models of pediatric cancer have been developed by academia, primarily from patient tumor material. As a result, models of pediatric cancer are typically decentralized and neither molecular characterization nor benchmarking to standard of care is consistent, limiting the utility of the models for thorough and rationale drug development. This is especially problematic for companies whose access to these models has been limited due to a myriad of factors. Further complicating matters is preclinical evaluation of anticancer drugs for pediatric indications has generally been anecdotic (“proof-of-concept,” POC) rather than systematic.

For 10 years, the forerunner of the PPTC, the NCI Pediatric Preclinical Testing Program (PPTP), attempted a systematic approach to evaluating anticancer compounds using a panel of in vitro and in vivo models (12). The PPTP tested >60 anticancer compounds as single agents or in combination studies. The seven lessons learnt from PPTP are described in Table 1. False-positive results (i.e., preclinical activity without subsequent clinical activity) need to be carefully studied to increase the predictive power of future testing.

Table 1.

Seven lessons from the Pediatric Preclinical Testing Program.

PDX models can be developed that adequately represent the genomic heterogeneity of pediatric cancers. 
Essential to address drug scheduling and systemic exposure for accurate clinical translation. 
Genomic information can be useful for predicting drug sensitivity (BRAF, PALB2, etc.). 
Models can identify novel efficacious agents (selumetinib, eribulin, OBI-3424) and combinations (eribulin/irinotecan, temsirolimus/cyclophosphamide-vinca, etc.). 
False-positive results (i.e., preclinical activity without subsequent clinical activity) need to be carefully studied to increase the predictive power of future testing. 
Single mouse designs will allow more rapid/efficient screening of single agents and possibly combinations if the individual drugs have limited single agent efficacy 
For screening and orthotopic models, “blinded” experimental design is essential. 
PDX models can be developed that adequately represent the genomic heterogeneity of pediatric cancers. 
Essential to address drug scheduling and systemic exposure for accurate clinical translation. 
Genomic information can be useful for predicting drug sensitivity (BRAF, PALB2, etc.). 
Models can identify novel efficacious agents (selumetinib, eribulin, OBI-3424) and combinations (eribulin/irinotecan, temsirolimus/cyclophosphamide-vinca, etc.). 
False-positive results (i.e., preclinical activity without subsequent clinical activity) need to be carefully studied to increase the predictive power of future testing. 
Single mouse designs will allow more rapid/efficient screening of single agents and possibly combinations if the individual drugs have limited single agent efficacy 
For screening and orthotopic models, “blinded” experimental design is essential. 

Models to be used

The projected value and limitations of current preclinical models are summarized (Table 2). Experts concluded that subcutaneous or orthotopically implanted patient-derived xenografts (PDXs), derived from tumors at diagnosis, relapse or following rapid autopsy, as well as genetically engineered mouse models (GEMMs) are the preferred model systems today. Moreover, it is a key to capture intertumor heterogeneity through a saturating repertoire of models/entity rather than relying on “representative models” of a disease (13, 14). Metastatic models should be considered specifically for therapeutics with a MOA targeting the metastatic process; however, such models remain limited. Beyond intrinsic tumor heterogeneity, relapsed disease models are considered important for representation of the evolving tumor milieu and a PDX panel consisting of primary and relapsed tumors is recommended. Although PDX models can impart human context for tumor cell autonomous targets, GEMMs provide an immune-competent setting may confer a more durable desmoplastic component, and enable precise interrogation of discrete genetic events that occur in cancer. Moreover, the use of GEMM-derived tumor material can enable higher throughput testing via syngeneic allografts or in the ex vivo setting. Pediatric organoids may also prove valuable for ex vivo studies, with their inherent recapitulation of tumor heterogeneity and 3D structure, but translatability is uncertain to date (14). Organoids lack tumor–stroma interaction and are developed mainly from epithelial tumors at this time. Despite the advances in organoid technology, comprehensive 2D cell line panels retain utility, especially if large panels reflective of intertumor heterogeneity and molecular subgroups are used. Other models such as zebrafish may have value for the interrogation of specific MOA, though not enough data exist to definitively assess their translatability (15).

Table 2.

Models for pediatric preclinical testing.

Pros for improving clinical predictabilityIssuesRecommendation
In vivo 
PDXs 
  • – Recapitulates heterogeneity of the patients' tumor

  • – Relapse models enable study of tumor evolution

  • – Orthotopic models for brain tumors

 
Tumorigenicity varies according to histology limited utility for IO compound testing; limited microenvironment context Mandatory 
Non-brain orthotopic models 
  • – Reflects tumor microenvironment and supports metastasis (e.g., neuroblastoma)

  • – successful engraftment rate higher

 
Model optimization required Optional 
Metastastic models 
  • – Mechanism of action dependent, such as prevention of metastases

 
  • – No evidence of clinical predictability

  • – Reproducibility concerns

 
Recommended with regard to the mechanism of action, if models are available 
GEMM 
  • – Immune-competent

  • – Immune and stromal components

  • – Interrogation of specific genetic events

 
  • – Lacks heterogeneity of patient tumor

  • – May present cross-species challenges, particularly for biologics

 
Recommended when genetic drivers are defined in a tumor segment 
Zebrafish 
  • – Adult fish: good models for metastasis and angiogenesis

  • – GEMM fish: good models for mechanism of action/on target activity

  • – PDX fish: good models for medium throughput drug screens

 
Technical issues: larvae versus adult fish, sufficient tumor growth Optional and translatability to be defined 
Other in vivo models 
  • – Large syngeneic models: endogenous tumors in dogs/pigs

 
  • – Cost

  • – Experimental tractability and translatability

  • – Drug activity concerns (e.g., antibodies might not work across species)

 
Optional 
In vitro Pros for improving clinical predictability Issues Recommendation 
Organoids 
  • – Screening of larger compound panels (prior to mouse PDX model)

  • – Representation of heterogeneity and 3D tumor biology

 
  • – Relevance to primary tumors

  • – Experimental reproducibility/reliability

  • – Not well suited for antibody evaluation

 
Optional
  • – Could support decision making for tumor types where options are limited (e.g., no PDX)

 
Other:
  • – 2D

  • – 3D models (spheroids and soft agar)

  • – Thick slice/organotypic

  • – Chick chorioallantoic membane (CAM) assay

 
  • – Panels may represent disease/genetic segments underrepresented by in vivo models.

 
  • – Translatability

 
Optional 
Pros for improving clinical predictabilityIssuesRecommendation
In vivo 
PDXs 
  • – Recapitulates heterogeneity of the patients' tumor

  • – Relapse models enable study of tumor evolution

  • – Orthotopic models for brain tumors

 
Tumorigenicity varies according to histology limited utility for IO compound testing; limited microenvironment context Mandatory 
Non-brain orthotopic models 
  • – Reflects tumor microenvironment and supports metastasis (e.g., neuroblastoma)

  • – successful engraftment rate higher

 
Model optimization required Optional 
Metastastic models 
  • – Mechanism of action dependent, such as prevention of metastases

 
  • – No evidence of clinical predictability

  • – Reproducibility concerns

 
Recommended with regard to the mechanism of action, if models are available 
GEMM 
  • – Immune-competent

  • – Immune and stromal components

  • – Interrogation of specific genetic events

 
  • – Lacks heterogeneity of patient tumor

  • – May present cross-species challenges, particularly for biologics

 
Recommended when genetic drivers are defined in a tumor segment 
Zebrafish 
  • – Adult fish: good models for metastasis and angiogenesis

  • – GEMM fish: good models for mechanism of action/on target activity

  • – PDX fish: good models for medium throughput drug screens

 
Technical issues: larvae versus adult fish, sufficient tumor growth Optional and translatability to be defined 
Other in vivo models 
  • – Large syngeneic models: endogenous tumors in dogs/pigs

 
  • – Cost

  • – Experimental tractability and translatability

  • – Drug activity concerns (e.g., antibodies might not work across species)

 
Optional 
In vitro Pros for improving clinical predictability Issues Recommendation 
Organoids 
  • – Screening of larger compound panels (prior to mouse PDX model)

  • – Representation of heterogeneity and 3D tumor biology

 
  • – Relevance to primary tumors

  • – Experimental reproducibility/reliability

  • – Not well suited for antibody evaluation

 
Optional
  • – Could support decision making for tumor types where options are limited (e.g., no PDX)

 
Other:
  • – 2D

  • – 3D models (spheroids and soft agar)

  • – Thick slice/organotypic

  • – Chick chorioallantoic membane (CAM) assay

 
  • – Panels may represent disease/genetic segments underrepresented by in vivo models.

 
  • – Translatability

 
Optional 

Abbreviations: GEMM, genetically modified model; PDXs, patient-derived xenografts.

Preclinical evaluation for brain tumors

The preclinical evaluation of treatment innovations for patients with brain tumors poses additional challenges because most drugs (>90%) are developed to not penetrate the brain to prevent neurotoxic side effects. A misconception that brain tumors typically show a disrupted blood–brain/blood–tumor barrier in all parts of the tumor is not backed up by clinical experience, that is, many drugs that show sufficient plasma levels do not reach clinically relevant exposure in brain tumors, especially in the infiltrative zone that is thought to often form the basis for relapses, particularly for high-grade gliomas. This key challenge must be addressed in the minimal preclinical data package to assess a novel therapeutic concept for neurooncology.

There is utility in using orthotopic brain tumor models beyond blood–brain barrier concerns, based on the very specific microenvironment that brain tumors require, and their tropism toward the CNS. As a matter of practicality, the experts concluded that in vivo testing in subcutaneous flank models may precede testing in orthotopic brain tumor models as a POC for tumor target relevance and efficacy, unless the MOA of the compound calls for direct testing in an orthotopic setting. It was further concluded that for orthotopic models both event-free survival as well as tumor response as assessed by serial in vivo imaging (e.g., by MRI, bioluminescence or PET-CT) were suitable efficacy endpoints, the latter being preferable to monitor objective response rates. Furthermore, pharmacokinetic and pharmacodynamic measurements should ideally occur in an orthotopic setting. There was agreement that negative orthotopic efficacy data should preclude moving forward into a clinical trial. In summary, there was consensus on the mandatory use of several different orthotopic models (preferably a combination of genetically engineered mouse models and PDXs, and preferably utilizing serial in vivo imaging) as part of the minimally recommended preclinical dataset for neurooncology indications.

Preclinical evaluation of immunooncology compounds

Immuno-oncology (IO) agents of potential interest for pediatric preclinical testing are divided into two primary groups: IO agents that enhance adaptive immune responses to tumors (e.g., immune checkpoint inhibitors), and IO agents engineered against specific antigens expressed by cancers as exemplified by T-cell bispecific antibodies (TCB), chimeric antigen receptor (CAR) T cells, and antibody–drug conjugates (ADCs). Immune checkpoint inhibitors as a monotherapy appear to have a limited role for pediatric cancers (16–20) that likely reflects the low prevalence of T-cell infiltrate and low tumor mutational burden of childhood cancers and correspondingly low rate of tumor neoantigens (8). Notable exceptions to this limited activity include patients with rare germline mutations in mismatch repair genes, and patients with Hodgkin lymphoma (21, 22).

For engineered IO agents targeting specific antigens, a first requirement for their application for childhood cancers is identifying the tumors that preferentially express the antigen. It is relevant to explore specific pediatric tumors because they are often of embryonal origin and thus express proteins that are no longer relevant/expressed in differentiated tissues. The availability of gene expression databases allows a rapid assessment of mRNA expression (Supplementary Table S1). As engineered IO agents primarily target proteins expressed on the cell surface, and as gene expression at the RNA level does not faithfully mirror protein expression, it is important to develop a detailed understanding of the surface proteome/surfaceome across childhood cancer preclinical models and patient tumor samples and normal tissues. Nonprotein antigens should also be considered as they may also be clinically relevant (e.g., the GD2-targeted agent dinutuximab).

Candidate target antigens for childhood cancers can be credentialed through testing of either ADCs or TCBs in relevant pediatric cancer preclinical models. Methods for testing ADCs are the same as those for testing standard cytotoxic agents, as illustrated by testing of the anti-CD19 ADC coltuximab ravtansine (SAR3419) in pediatric ALL xenograft lines and the testing of the DLL3-targeted ADC rovalpituzumab tesirine in neuroblastoma xenograft lines (22,–24). Methods for testing TCBs have been developed that use stem cell humanized NSG mice (25), and these can be applied for testing TCBs against relevant pediatric preclinical models, such as PDXs. For IO agents such as checkpoint inhibitors that promote the adaptive immune responses to tumors, GEMMs may generate POC data (26)

Methodology, data, and reporting standards for preclinical testing

It is imperative to generate scientifically sound and comprehensive preclinical data that facilitate the comparison of drugs being considered for evaluation in children, including combinations and conventional radiochemotherapy backbones. In addition to validated and robust platforms using molecularly characterized pediatric preclinical models, careful attention to study design and analysis is critical.

The two primary measures of efficacy used for preclinical testing are “time to event” and “tumor load” (see Supplementary Materials). Methods for assessing prolongation in time to event require defining “event,” with the definition dependent upon the type of model being evaluated (subcutaneous tumors, orthotopic tumors, etc.). The ratio of time to event for treated versus control animals is independent of tumor growth rate when defining an event as a multiple of starting tumor volume (e.g., fourfold) and allows for comparison of results across models and experiments. Tumor load results can be categorized by objective response criteria that mirror clinical RECIST or RANO criteria (12). This presentation of agent activity enhances the translational impact of the data package by identifying agents inducing robust regression, with prolonged regression after cessation of treatment being particularly promising. Slowing down of tumor growth over control is, in most cases, insufficient for advancement (as it would be considered inadequate in patients). Statistical significance between treated and control groups for the distribution in time to event or for tumor volume at a defined time point during the experimental phase is determined using standard methods such as the Gehan–Wilcoxon test or RM-ANOVA (preferred “gold standard,” with the caveat that a statistically significant difference in preclinical experiments may represent a small treatment effect that is unlikely to imply clinical activity). Interpretation of testing results of experimental agents is facilitated by comparisons with levels of activity of standard agents in the same models.

The Single Mouse Trial (SMT) format uses a single mouse per model and treatment arm, thereby enabling the investigation of efficacy in substantially larger panels of well-characterized models (27, 28). SMTs facilitate the identification of biological factors that may influence response to the tested agent (29, 30). An SMT can optimally support efficacy testing across a panel of tumors that reflect the heterogeneity of a disease in children. Historically, data were generated with 40 to 100 models and up to 62 treatment arms (31, 32). A minimal panel size of 10 models seems to be a fair assumption to mirror inter- and intrapatient heterogeneity. The proposed approach to SMTs focuses on the ability of the tested agents to induce regressions for two reasons: Large treatment effects can be reliably detected and agents able to robustly induce regressions are of greatest value in the childhood cancer clinical setting. The experts concluded that drugs yielding tumor growth delay that do not minimally reach the level of stable disease are of limited utility. A recent publication (33) comparing colorectal patient response to cetuximab to an evaluation of cetuximab across a molecularly characterized panel of colorectal PDX models that faithfully represent tumor heterogeneity showed agreement in response to drug treatment using an SMT. To establish in vivo efficacy, this SMTl used three vehicle-treated tumor-bearing mice plus one tumor-bearing mouse per drug treatment. Using this approach, the single mouse PDX trial was a reliable predictor of clinical response. It is worth noting that the authors evaluated 79 PDX tumors representative of the five major known drivers of colorectal cancer and the 6% overall response rate to cetuximab (comparable with clinical reports), was driven by activity only in the subset of KRAS and BRAF WT tumors. In contrast, pediatric tumors typically have far fewer drivers per tumor type (often just a single driver, e.g., a fusion protein) and we are confident that ITCC-P4's goal of upwards of 40 PDX models per tumor type should be more than saturating for the majority of the most common pediatric malignancies.

Sharing data with the research community is crucial and includes accessibility of the data (access of raw and processed data) and a detailed description of the statistical analyses and evaluation criteria. To this end, visualization and data sharing tools are currently being implemented (e.g., through https://r2.amc.nl and http://www.ncipptc.org/genomics). Finally, feedback from clinical data is an important element of preclinical testing, and preclinical datasets should be retrospectively re-evaluated using clinical data (reverse translation).

Recommendations for preclinical POC data packages

Consistency in preclinical evaluation across drugs and across pediatric tumor types will facilitate prioritization for pediatric clinical development of promising agents, and our experts developed guidance for “Recommended POC data packages” (cancer immunotherapy was excluded here due to time restrictions). We used the description of preclinical POC data packages (separated into data modules developed from the SIOP Preclinical Testing Taskforce (Table 3; ref. 34). For drugs selectively targeting kinase-fusions, it was agreed that identification of the kinase-fusion in clinical series of a pediatric tumor type was enough to support clinical testing of the drug. For drugs aimed at nonfusion targets, we agreed on the following three determinants:

  • Step 1, clinical urgency of the tumor type;

  • Step 2, availability of profiling data of a series of tumor samples;

  • Step 3, availability of relevant in vivo.

Table 3.

Preclinical proof-of-concept package modules.

Module 1: Target status and patterns in clinical series 
Module 2: Molecular validation of target dependence in vitro models 
Module 3: Molecular validation of target dependence in vivo models 
Module 4: Drug efficacy patterns in vitro models 
Module 5: Drug efficacy patterns in vivo models 
Module 6: Biomarkers (patient selection/PD) 
Module 7: Resistance mechanisms 
Module 8: Combination testing 
Module 1: Target status and patterns in clinical series 
Module 2: Molecular validation of target dependence in vitro models 
Module 3: Molecular validation of target dependence in vivo models 
Module 4: Drug efficacy patterns in vitro models 
Module 5: Drug efficacy patterns in vivo models 
Module 6: Biomarkers (patient selection/PD) 
Module 7: Resistance mechanisms 
Module 8: Combination testing 

We defined cutoff values for the different subsets in each determinant; see Table 4 for details.

Table 4.

Guidelines for recommended extent of preclinical POC testing packages.

Step 1Step 2Step 3Recommended POC package
Clinical urgencyTumor sample series with molecular profilingMatching vivo model availability(for kinase fusion targets consider clinical development without preclinical POC testinga)
High (<40% survival) Insufficient (<20 samples) n/a POC not possibleb 
 Sufficient (>20 samples) No models POC not possibleb 
  Sufficient models (>1 xeno/PDX, >1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

 
Medium (40%–80% survival)d Insufficient (<100 samples) n/a POC not possibleb 
 Sufficient (>100 samples and >10 in each biological subset) No models POC not possibleb 
  Sufficient models (>1 xeno/PDX, >1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 2 Molecular target validation vitro

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

  • Mod. 8 Combinations vitro

 
Low (>80% survival) Insufficient (<100 samples) n/a POC not possibleb 
 Sufficient (>100 samples and >10 in each biological subset) No models POC not possibleb 
  Limited models (1–5 xeno/PDX, 1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 2 Molecular target validation vitro

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

  • Mod. 8 Combinations vitro

 
  Sufficient models (>5 xeno/PDX, >1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 2 Molecular target validation vitrod

  • (Mod. 3 Molecular target validation vivo)d

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

  • Mod. 8 Combinations vitro

 
Step 1Step 2Step 3Recommended POC package
Clinical urgencyTumor sample series with molecular profilingMatching vivo model availability(for kinase fusion targets consider clinical development without preclinical POC testinga)
High (<40% survival) Insufficient (<20 samples) n/a POC not possibleb 
 Sufficient (>20 samples) No models POC not possibleb 
  Sufficient models (>1 xeno/PDX, >1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

 
Medium (40%–80% survival)d Insufficient (<100 samples) n/a POC not possibleb 
 Sufficient (>100 samples and >10 in each biological subset) No models POC not possibleb 
  Sufficient models (>1 xeno/PDX, >1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 2 Molecular target validation vitro

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

  • Mod. 8 Combinations vitro

 
Low (>80% survival) Insufficient (<100 samples) n/a POC not possibleb 
 Sufficient (>100 samples and >10 in each biological subset) No models POC not possibleb 
  Limited models (1–5 xeno/PDX, 1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 2 Molecular target validation vitro

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

  • Mod. 8 Combinations vitro

 
  Sufficient models (>5 xeno/PDX, >1 GEMM) 
  • Mod. 1 Target patterns

  • Mod. 2 Molecular target validation vitrod

  • (Mod. 3 Molecular target validation vivo)d

  • Mod. 4 Efficacy vitroc

  • Mod. 5 Efficacy vivoc

  • Mod. 8 Combinations vitro

 

Abbreviations: GEMM, genetically modified model; Mod, module; POC, proof-of-concept; PDX, patient-derived xenograft; xeno, xenograft.

aFor kinase fusion genes consider clinical development without preclinical POC testing.

bDevelop sample series and vivo models.

cModule 4 “vitro efficacy testing” used to gate performing module 5 drug efficacy testing in vivo models.

dFor low urgency disease need to perform module 3 “Target validation in vivo models” is gated on results from module 2 “Molecular target validation in vitro models.”

For each resulting category, we then decided on the extent of POC modules needed for the “Recommended POC data package” (Table 4). In some situations, preclinical POC testing is currently not possible due to lack of tumor series with molecular profiling or insufficient availability of relevant models. The level of clinical urgency should influence the extent of POC data packages, for example, for diseases with a very good prognosis, a deeper preclinical POC data package is recommended. The actual extent of the POC data package then depends on the availability of relevant preclinical models.

With regards to POC module 4, the participants agreed that in vitro testing should not factor heavily in determining preclinical POC, but it could be useful as a decision gate for the more complicated module 5 in vivo efficacy testing across a saturated repertoire of models representing the disease heterogeneity.

To accelerate the development of innovative therapies for children with cancer in the changing regulatory environment, easy access to biological information on pediatric tumors and to high quality pediatric preclinical testing must increase. The capacity to generate relevant preclinical information using comprehensive panels of well characterized pediatric tumor models is a key to prioritizing anticancer drugs for pediatric development. At the same time, this should support academic research to understand the biology of pediatric malignancies, to identify mechanism of sensitivity and resistance, and to inform treatment combinations, including combinations with current standard of care. In addition, identification of pathways specific to pediatric tumors will support the development of agents specific to pediatric cancer.

Currently, there are two publicly funded consortia to generate preclinical pediatric cancer data: ITCC-P4 and PPTC. The ITCC-P4 consortium (https://www.itccp4.eu/) was created to address many of the topics identified in this symposium. This precompetitive, preclinically focused public private partnership brings together academia, industry, and contract research organizations to develop a comprehensive translational research platform of molecularly characterized pediatric solid tumor PDXs (aiming at 40 different PDXs per tumor type, including orthotopic brain tumor models for ten histologies with highest clinical need), organoids and GEMMs. Underpinning the ITCC-P4 platform is a powerful, custom-built informatics suite that enables data handling, visualization and interpretation (https://r2.amc.nl). ITCC-P4 testing also supports biomarker-driven patient tailoring, that is, the right drug for the right patient, based on the tumor's molecular profile. The incorporation of GEMMs allows for testing of drugs in tumors within their “naturally” occurring primary location. In addition, humanized PDXs of neuroblastoma and rhabdomyosarcoma are under development as novel preclinical tools for evaluation of immuno-oncology agents. Compound testing, including disease-relevant standard-of-care drugs and targeted agents, began in 2020. By early 2022, the publicly funded portion of ITCC-P4 will end, and the platform will be available to researchers worldwide.

The NCI has supported a pediatric preclinical testing program since 2004, initially through the PPTP, and since 2015 through the PPTC (www.ncipptc.org), with approximately 130 agents from more than 60 companies and representing a broad range of MOA have been tested (>90 publications are available). The use of a model material transfer agreement that is accepted by all PPTP/PPTC institutions has allowed testing to proceed in a timely manner and not be delayed by extended negotiations. PPTP testing typically involved agnostically evaluating novel single agents across a panel of approximately 40 to 50 xenografts reflecting a range of genomically characterized solid tumors and ALL (https://pedcbioportal.org/study/summary?id = pptc). Activity signals were often pursued with dose–response and/or combination testing. Testing through the PPTC has been more hypothesis-driven, with models selected on the basis of the responder-hypothesis for the agent under evaluation. The genomic data facilitate the selection of models for testing against specific targeted agents and help to relate the preclinical testing results to the clinical context. An important consideration is the ability of agents to induce tumor regression in prioritizing agents or combinations for evaluation in children. Several pediatric clinical trials have been influenced by PPTP/PPTC, including the rationale for phase 3 studies of selumetinib for low-grade glioma and of temsirolimus for rhabdomyosarcoma (35, 36). Negative findings from the PPTP/PPTC have also contributed to pediatric drug development by allowing clinical investigators to focus on more promising research.

In the evolving regulatory environment, the capacity to generate relevant biological and preclinical information is essential to the development of new anticancer assets for the pediatric population. International collaboration for global drug development is needed to best address innovation for the treatment of rare cancers in children and adolescents. International standards on the development of preclinical data packages necessary to inform clinical decisions would greatly facilitate the identification and prioritization of promising agents. The proposed requirements, if adopted, have the potential to impact significantly the movement of promising agents into clinical development.

G. Vassal reports grants from Innovative Medicine Initiative during the conduct of the study. S.M. Pfister reports grants from IMI-2 funded project together with Roche, Eli-Lilly, Pfizer, Bayer, and PharmaMar during the conduct of the study. H.N. Caron reports other support from Roche during the conduct of the study as well as other support from Roche outside the submitted work. D.J. Shields reports other support from Pfizer Inc. outside the submitted work. O. Witt reports grants and personal fees from Roche Pharma, personal fees from Novartis, Janssen Research, and SK Life Science outside the submitted work. S. Colombetti reports other support from Roche outside the submitted work. L.F. Stancato reports other from Eli Lilly outside the submitted work. No disclosures were reported by the other authors.

G. Vassal: Conceptualization, resources, writing–original draft, writing–review and editing. P.J. Houghton: Conceptualization, resources, writing–original draft. S.M. Pfister: Conceptualization, resources, supervision, writing–original draft. M.A. Smith: Conceptualization, resources, methodology, writing–review and editing. H.N. Caron: Conceptualization, resources, supervision, writing–original draft. X.–N. Li: Conceptualization, resources, writing–original draft. D.J. Shields: Conceptualization, resources, writing–original draft. O. Witt: Conceptualization, resources, writing–original draft. J.J. Molenaar: Conceptualization, resources, writing–original draft. S. Colombetti: Conceptualization, writing–original draft. J. Schüler: Conceptualization, resources, methodology, writing–original draft. L.F. Stancato: Conceptualization, resources, supervision, writing–original draft, writing–review and editing.

ITCC-P4 is funded by IMI2 grant agreement no. 116064 ITCC-P4. The September 2018 workshop was funded by Solving Kids’ Cancer, Neuroblastoma UK, Sohn Conference Foundation and ITCC-P4. We thanks all participants for their valuable contribution in designing the international consensus: Sandy Azzi-Hatem, Kladiola Berdufi, Patricia Blanc, Trandafir Bogdan-Adrian, Anna-Lisa Böttcher, Stefano Cairo, Hubert Caron, Michela Casanova, Louis Chesler, Silvia Chioato, Marta Colletti, Sara Colombetti, Emilie Dassé, Emmy Dolman, Gregory Donoho, Jarno Drost, Angelika Eggert, Stephen Erickson, David Frost, Patrizia Gasparini, Johannes Gojo, Ralph Graeser, Peter Houghton, Yann Jamin, Kaylee Keller, Leona Knox, Marcel Kool, Jan Koster, Heinrich Kovar, Andrew Kung, Raushan Kurmasheva, Xiao-Nan Li, Caitlin Lowery, Donna Ludwinski, Florence Meyer-Losic, Jan Molenaar, Massimo Moro, Andy Pearson, Stefan Pfister, Gregory Reaman, Ana Rodriguez, Jana Rolff, Gudrun Schleiermacher, Nil Schubert, Julia Schüler, David Shields, Malcolm Smith, Louis Stanaato, Birgit Steckel-Hamann, Didier Surdez, Beverly Teicher, Gilles Vassal, Clémence Vieules, Marco Wachtel, Scott Weinrich, Frank Westermann, Olaf Witt, Dieter Zopf, and Danny Zwijnenburg.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

1.
Bourgeois
FT
,
Hwang
TJ
.
The pediatric research equity act moves into adolescence
.
JAMA
2017
;
317
:
259
60
.
2.
Tomasi
PA
,
Egger
GF
,
Pallidis
C
,
Saint-Raymond
A
.
Enabling development of paediatric medicines in europe: 10 years of the EU paediatric regulation
.
Paediatr Drugs
2017
;
19
:
505
13
.
3.
Vassal
G
,
Geoerger
B
,
Morland
B
.
Is the European pediatric medicine regulation working for children and adolescents with cancer?
Clin Cancer Res
2013
;
19
:
1315
25
.
4.
Pearson
ADJ
,
Pfister
SM
,
Baruchel
A
,
Bourquin
JP
,
Casanova
M
,
Chesler
L
, et al
.
From class waivers to precision medicine in paediatric oncology
.
Lancet Oncol
2017
;
18
:
e394
.
6.
Vassal
G
,
Blanc
P
,
Copland
C
,
Pearson
A
.
Will the revised class waiver list make it?
Lancet Oncol
2015
;
16
:
e425
6
.
7.
US Fodd & DRrug administration
.
PUBLIC LAW 115–52—AUG. 18, 2017 131 STAT. 1005
. Available from: https://www.congress.gov/115/plaws/publ52/PLAW-115publ52.pdf.
Accessed on August 16, 2019
.
8.
Gröbner
SN
,
Worst
BC
,
Weischenfeldt
J
,
Buchhalter
I
,
Kleinheinz
K
,
Rudneva
VA
, et al
.
The landscape of genomic alterations across childhood cancers
.
Nature
2018
;
555
:
321
7
.
9.
Jones
DTW
,
Banito
A
,
Grünewald
TGP
,
Haber
M
,
Jäger
N
,
Kool
M
, et al
.
Molecular characteristics and therapeutic vulnerabilities across paediatric solid tumours
.
Nat Rev Cancer
2019
;
19
:
420
38
.
10.
Innovative therapies for children with cancer—pediatric preclinical proof-of-concept project
. Available from: https://www.itccp4.eu/.
11.
NCI-funded Pediatric Preclinical Testing Consortium (PPTC)
. Available from: http://www.ncipptc.org/.
Accessed on August 16, 2018
.
12.
Houghton
PJ
,
Morton
CL
,
Tucker
C
,
Payne
D
,
Favours
E
,
Cole
C
, et al
.
The Pediatric Preclinical Testing Program: description of models and early testing results
.
Pediatr Blood Cancer
2007
;
49
:
928
40
.
13.
Brabetz
S
,
Leary
SES
,
Gröbner
SN
,
Nakamoto
MW
,
Şeker-Cin
H
,
Girard
EJ
, et al
.
A biobank of patient-derived pediatric brain tumor models
.
Nat Med
2018
;
24
:
1752
61
.
14.
Bleijs
M
,
van de Wetering
M
,
Clevers
H
,
Drost
J
.
Xenograft and organoid model systems in cancer research
.
EMBO J
2019
;
38
:
e101654
.
15.
Deveau
AP
,
Bentley
VL
,
Berman
JN
.
Using zebrafish models of leukemia to streamline drug screening and discovery
.
Exp Hematol
2017
;
45
:
1
9
.
16.
Lee
CH
,
Yelensky
R
,
Jooss
K
,
Chan
TA
.
Update on tumor neoantigens and their utility: why it is good to be different
.
Trends Immunol
2018
;
39
:
536
48
.
17.
Davis
KL
,
Fox
E
,
Reid
JM
,
Liu
X
,
Minard
CG
,
Weigel
B
, et al
.
ADVL1412: initial results of a phase I/II study of nivolumab and ipilimumab in pediatric patients with relapsed/refractory solid tumors—A COG study
.
J Clin Oncol
2017
;
35
:
10526
.
18.
Geoerger
B
,
Kang
HJ
,
Yalon-Oren
M
,
Marshall
LV
,
Vezina
C
,
Pappo
AS
, et al
.
Phase 1/2 KEYNOTE-051 study of pembrolizumab (pembro) in pediatric patients (pts) with advanced melanoma or a PD-L1+ advanced, relapsed, or refractory solid tumor or lymphoma
.
J Clin Oncol
2017
;
35
:
10525
.
19.
Geoerger
B
,
Karski
EE
,
Zwaan
M
,
Casanova
M
,
Marshall
LV
,
DuBois
SG
, et al
.
A phase I/II study of atezolizumab in pediatric and young adult patients with refractory/relapsed solid tumors (iMATRIX-Atezolizumab)
.
J Clin Oncol
2017
;
35
:
10524
.
20.
Pearson
ADJ
,
Scobie
N
,
Norga
K
,
Ligas
F
,
Chiodin
D
,
Burke
A
, et al
.
ACCELERATE and European Medicine Agency Paediatric Strategy Forum for medicinal product development for mature B-cell malignancies in children
.
Eur J Cancer
2019
;
110
:
74
85
.
21.
Bouffet
E
,
Larouche
V
,
Campbell
BB
,
Merico
D
,
de Borja
R
,
Aronson
M
, et al
.
Immune checkpoint inhibition for hypermutant glioblastoma multiforme resulting from germline biallelic mismatch repair deficiency
.
J Clin Oncol
2016
;
34
:
2206
11
.
22.
Mottok
A
,
Steidl
C
.
Biology of classical Hodgkin lymphoma: implications for prognosis and novel therapies
.
Blood
2018
;
131
:
1654
65
.
23.
Carol
H
,
Szymanska
B
,
Evans
K
,
Boehm
I
,
Houghton
PJ
,
Smith
MA
, et al
.
The anti-CD19 antibody–drug conjugate SAR3419 prevents hematolymphoid relapse postinduction therapy in preclinical models of pediatric acute lymphoblastic leukemia
.
Clin Cancer Res
2013
;
19
:
1795
805
.
24.
Sano
R
,
Krytska
K
,
Tsang
M
,
Erickson
SW
,
Teicher
BA
,
Saunders
L
, et al
.
Abstract LB-136: pediatric preclinical testing consortium evaluation of a DLL3-targeted antibody drug conjugate rovalpituzumab tesirine, in neuroblastoma
.
Cancer Res
2018
;
78
:
LB–136–LB
.
25.
Bacac
M
,
Colombetti
S
,
Herter
S
,
Sam
J
,
Perro
M
,
Chen
S
, et al
.
CD20-TCB with obinutuzumab pretreatment as next-generation treatment of hematologic malignancies
.
Clin Cancer Res
2018
;
24
:
4785
97
.
26.
Mao
Y
,
Eissler
N
,
Blanc
KL
,
Johnsen
JI
,
Kogner
P
,
Kiessling
R
.
Targeting suppressive myeloid cells potentiates checkpoint inhibitors to control spontaneous neuroblastoma
.
Clin Cancer Res
2016
;
22
:
3849
59
.
27.
Lock
RB
,
Evans
K
,
Pritchard
T
,
Toscan
CE
,
Mayoh
C
,
Teicher
B
, et al
.
A single mouse trial platform for
evaluation of novel agents in acute lymphoblastic leukemia by the pediatric preclinical Testing Consortium [abstract]
.
Blood In:
Proceedings of the 60th Annual Meeting of the American-Society-of-Hematology (ASH); 2018 December 1–4
;
San Diego, CA
:
ASH
;
2018
.
Abstract nr 4010
.
28.
Murphy
B
,
Yin
H
,
Maris
JM
,
Kolb
EA
,
Gorlick
R
,
Reynolds
CP
, et al
.
Evaluation of alternative in vivo drug screening methodology: a single mouse analysis
.
Cancer Res
2016
;
76
:
5798
809
.
29.
Gao
H
,
Korn
JM
,
Ferretti
S
,
Monahan
JE
,
Wang
Y
,
Singh
M
, et al
.
High-throughput screening using patient-derived tumor xenografts to predict clinical trial drug response
.
Nat Med
2015
;
21
:
1318
25
.
30.
Yao
YM
,
Donoho
GP
,
Iversen
PW
,
Zhang
Y
,
Van Horn
RD
,
Forest
A
, et al
.
Mouse PDX trial suggests synergy of concurrent inhibition of RAF and EGFR in colorectal cancer with BRAF or KRAS mutations
.
Clin Cancer Res
2017
;
23
:
5547
60
.
31.
Bruna
A
,
Rueda
OM
,
Greenwood
W
,
Batra
AS
,
Callari
M
,
Batra
RN
, et al
.
A Biobank of breast cancer explants with preserved intra-tumor heterogeneity to screen anticancer compounds
.
Cell
2016
;
167
:
260
74
.
32.
Williams
JA
.
Using PDX for preclinical cancer drug discovery. The evolving field
.
J Clin Med
2018
;
7
:
41
.
33.
Yao
YM
,
Donoho
GP
,
Iversen
PW
,
Zhang
Y
,
Van Horn
RD
,
Forest
A
, et al
.
Mouse PDX trial suggests synergy of concurrent inhibition of RAF and EGFR in colorectal cancer with BRAF or KRAS mutations
.
Clin Cancer Res
2017
;
23
:
5547
60
.
34.
Schubert
NA
,
Lowery
CD
,
Bergthold
G
,
Koster
J
,
Eleveld
TF
,
Rodríguez
A
, et al
.
Systematic target actionability reviews of preclinical proof-of-concept papers to match targeted drugs to paediatric cancers
.
Eur J Cancer
2020
;
130
:
168
81
.
35.
Kolb
EA
,
Gorlick
R
,
Houghton
PJ
,
Morton
CL
,
Neale
G
,
Keir
ST
, et al
.
Initial testing (stage 1) of AZD6244 (ARRY-142886) by the pediatric preclinical testing program
.
Pediatr Blood Cancer
2010
;
55
:
668
77
.
36.
Houghton
PJ
,
Morton
CL
,
Gorlick
R
,
Lock
RB
,
Carol
H
,
Reynolds
CP
, et al
.
Stage 2 combination testing of rapamycin with cytotoxic agents by the Pediatric Preclinical Testing Program
.
Mol Cancer Ther
2010
;
9
:
101
12
.

Supplementary data