SUPPLEMENTARY INFORMATIONIn format provided by Morenoet al. (doi:10.1038/nrclinonc.2017.59)

Supplementary Information S1 (table) | Dose-finding designs

Designs / Key points and advantages / Disadvantages
Commonly used designs
3+3 (REF.1) /
  • Simple dose-escalation design for chemotherapy trials that, to date, has been used in the majority of trials of molecularly targeted agents in paediatrics
/
  • Studies remain closed for recruitment a significant amount of time with patients queuing for allocation
  • Only the patients in the last cohort are taken into account in the dose-escalation decisions
  • Design seems to be suboptimal and exceptions cannot be recommended

Rolling six2 /
  • Allows 3 to 6 patients per cohort and therefore reduces the time that the study is closed while information on the first patients is gathered
/
  • More patients are included at the lower dose levels
  • Only the patients in the last cohort are taken into account in the dose-escalation decisions
  • There are still gaps when the study is closed

Continuous reassessment method (CRM)3 /
  • Model-based method where the risk of toxicity is reassessed after each new patient inclusion
  • Flexible as it allows limiting trial suspension, accounting for extra patients, modifying the target risk of dose-limiting toxicity (DLT)
  • It provides a swift way to design expansion cohorts
/
  • If a DLT is not dose-related the CRM risks lowering the recommended phase II dose (RP2D)
  • Needs constant presence of statistician

Time-to-event CRM4 /
  • Takes into account late-onset toxicities and allows for continuous enrolment
/
  • It assumes that patients may stay in study for several cycles or until DLTs appear
  • Requires continuous reporting of toxicity

CRM for multi-groups5 /
  • CRM approach where the risk of DLT accounts for a covariate such as the degree of pre-treatment or the age of the patient
  • Possibility to conduct phase I both in adults and in children
/
  • Some assumption on the ordering of the group and the plausible group difference is required

Bayesian6and Bayesian logistic regression model7 /
  • Model-based approach that formally incorporates prior information drawn from adult data
  • This is the model of choice for bridging studies
  • Expert prior opinion can be combined with aBayesian randomised controlled trial for rare diseases8
/
  • Elicitation and calibration of the prior information is delicate

Less commonly used designs
One out of five9 /
  • Group of 5 or 6 patients are enrolled at each dose. High chance of escalation if the true target response rate is low at the current level. Low chance of escalation otherwise
/
  • Uses dichotomized value of biological endpoint. No formal control of the risk of toxicity
  • More patients are included at the lower dose levels

Supplementary Information S2 (table) | Biomarkers

Biomarkers / Key points and advantages / Challenges and strategies to limit them
Pharmacokinetic /
  • Adequate and sustained druglevels and exposure should be demonstrated in children, and the pharmacokineticmodel of the compound under study, derived from adult trials, can serve to optimize paediatric pharmacokinetics,maximising the information that can be obtained with minimal sampling
  • Young children under the age of 2 years are rare in paediatric oncology, but they have significant differences in drug metabolism;therefore, where pharmacokineticdata for infants is insufficient in the early phase study, further data can be obtained in later phase studies
  • Pharmacokinetic studies should provide information on: i) a correlation between pharmacokinetics and activity; ii) whether the pharmacokinetic profile is age or dose-dependent; iii)if the therapeutic window is wide or narrow; iv) interpatient variability; v) whether the schedule needs modifications
/
  • Often require multiple time points and significant blood volumes. This can be minimised by extrapolating data from adult studies and refining assays to reduce the blood volumes required
  • It is difficult to ensure that patients from all age ranges are adequately represented, particularly young children where pharmacokinetic differences might be more pronounced
  • New formulations might become available at later time points and pharmacokinetic analyses might require repeating as they could differ substantially with the different formulation

Pharmacodynamic /
  • Pharmacodynamic biomarkers demonstrate in patient paired samples that a given target has been inhibited (or modulated) by exposure to a targeted drug and downstream pathway modulation achieved
  • They assist in the go/no-go decisions at the end of phase I and should be required prior to moving the agents into larger phase II trials whenever possible
/
  • Paired tumour samples (or surrogate tissues) are required to show target inhibition and hence, are often invasive or burdensome for participants
  • Assays used to show target inhibition have to be adequately developed, validated and adapted to the paediatric population

Predictive /
  • A biomarker predictive of the response of a specific molecularly targeted therapy allows patient selection for that drug based on their molecular characteristics leading to increased patient benefit
  • Patient selection has to be based on robust preclinical and (where possible) clinical data
  • Biomarkers have to be discovered, developed and qualified for paediatric patients following the appropriate guidelines and regulations. Where no predictive biomarkers are available, early phase I studies should collect tissue to identify potential biomarkers retrospectively
/
  • To prove the robustness of a predictive biomarker, patients without the biomarker have to be treated, to show that they do not respond. However this might not be reasonable in some cases of outstanding clinical benefit in children with the specific genetic alteration and lack of benefit in those without the biomarker
  • While drugs targeting oncogenic drivers are more likely to have a reliable predictive biomarker, for multi-targeted inhibitors or drugs inhibiting some generic cellular functions (for example, antiangiogenics), predictive biomarkers are more challenging. Even if a biomarker is not found, the data obtained with retrospective tissue analyses will facilitate understanding of the biology of the disease/pathway and generation of new hypotheses

Table cont.
Functional Imaging /
  • Functional imaging has been used in adult drug development as a non-invasive pharmacodynamicor predictive biomarker and as a surrogate for response and patient benefit
/
  • Functional imaging might require additional scans or longer scan times, which often requires anaesthesia or sedation in younger children
  • Setting up functional imaging in multicentre studies requires robust quality assurance and quality control procedures

1.Le Tourneau, C., Gan, H.K., Razak, A.R. & Paoletti, X. Efficiency of new dose escalation designs in dose-finding phase I trials of molecularly targeted agents. PLoS One7,: e51039 (2012).

2. Skolnik, J.M., Barrett, J.S., Jayaraman, B., Patel, D. & Adamson, P.C. Shortening the timeline of pediatric phase I trials: the rolling six design. J. Clin. Oncol.26, 190–195(2008).

3.O'Quigley, J., Pepe, M. & Fisher, L. Continual reassessment method: a practical design for phase 1 clinical trials in cancer. Biometrics46, 33–48 (1990).

4. Cheung, Y.K. & Chappell, R. Sequential designs for phase I clinical trials with late-onset toxicities. Biometrics56, 1177–1182 (2000).

5. O'Quigley, J. & Paoletti, X. Continual reassessment method for ordered groups. Biometrics59, 430–440 (2003).

6. Billingham, L., Malottki K., Steven, N. Research methods to change clinical practice for patients with rare cancers. Lancet Oncol.17, e70–80 (2016).

7.Neuenschwander, B., Branson, M. & Gsponer, T. Critical aspects of the Bayesian approach to phase I cancer trials. Stat. Med. J.15, 2420–2439 (2008).

8. Hampson, L.V. et al.Elicitation of expert prior opinion: application to the MYPAN trial in childhood polyarteritis nodosa. PLoS One10, e0120981 (2015).

9. Hunsberger, S., Rubinstein, L.V., Dancey, J. & Korn, E.L. Dose escalation trial designs based on a molecularly targeted endpoint. Stat. Med.24, 2171–2181 (2005).

NATURE REVIEWS |CLINICAL ONCOLOGY