Incorporation of patient-reported outcomes in pediatric cancer clinical trials: design, implementation, and dissemination

Abstract

Understanding the patient experience of treatment toxicities and their impact on health-related quality of life from cancer treatments requires asking patients using patient-reported outcomes. Over the past 20 years, the National Institutes of Health has sponsored several tools—namely, Patient-Reported Outcomes Measurement Information System measures and the Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events—for precisely this purpose: to ensure valid, reliable tools to collect and detect patient-reported toxicities or adverse events and their impact on health-related quality of life. These patient-reported outcomes measures have been widely incorporated in clinical trials for adults with cancer. Yet, despite considerable work developing and validating developmentally appropriate versions of these measures for pediatric and adolescent self-report, patient-reported outcomes inclusion in pediatric and adolescent and young adult clinical trials has lagged. Here, we discuss optimal strategies to integrate validated patient-reported outcomes tools and sound analytic methodologies in clinical trials for children and adolescent and young adults with cancer, highlighting lessons learned from recent successes and ongoing experiences developing and opening cross-network trials for children and adolescent and young adults through the Children’s Oncology Group for patients with classic Hodgkin lymphoma, osteosarcoma, and acute lymphoblastic leukemia.

Background

Iterative cooperative group trials have led to remarkable improvements in survival for children with cancer, such that more than 80% of children diagnosed with cancer will become long-term survivors.¹ However, childhood cancer treatment is associated with acute and long-term toxicities that can affect short-term and long-term health and health-related quality of life (HRQOL). Comprehensive collection of symptoms, toxicities, and HRQOL plays a crucial role in improving outcomes for children, adolescents, and young adults with cancer. Many current pediatric clinical trials focus on improving outcomes through treatment modifications to reduce short-term and long-term toxicity or inclusion of novel therapeutics with little known about their toxicities in children, adolescents, and young adults. Short-term and long-term symptom reporting is essential to recognizing and managing toxicities, assessing and defining tolerability, and improving HRQOL during and after therapy. True understanding of the patient experience of treatment toxicities and HRQOL requires asking patients themselves.

Patient-reported outcomes are defined by the National Cancer Institute (NCI) as “information about a patient’s health that comes directly from the patient.”² In adult cancer clinical trials, collection of patient-reported outcomes has improved the accuracy of symptomatic adverse event reporting.^3,4 When used clinically to monitor and respond to symptoms, patient-reported outcomes have been associated with improvements in overall survival, symptom control, treatment adherence, and HRQOL.^5-8 However, patient-reported outcomes have been incorporated in less than 20% of pediatric, adolescent and young adult clinical trials.^9-11 Failure to incorporate patient-reported outcomes in the clinical trial setting hinders our ability to improve meaningful patient outcomes as we de-escalate conventional therapies and incorporate novel agents. When novel treatment approaches tested in clinical trials do not improve disease outcomes, patient-reported outcomes still play an important role in our understanding of patient experiences of standard-of-care treatments. The absence of knowledge regarding the patient experience during and after current standard therapies deprives patients and families of information crucial to treatment decision making.

Historic barriers to collecting patient-reported outcomes in pediatric, adolescent, and young adult cooperative group trials have included lack of validated measures, particularly for pediatric self-report across age and developmental spectra; lack of consensus on the appropriate measures and the appropriate rater (patient vs parent); and challenges with obtaining necessary support. Collection of frequent patient-reported outcomes was felt to pose an incumbrance on participants and, given differential research infrastructure and resources, on participating research sites. Studies that collected patient-reported outcomes often had poor completion rates.¹² These concerns resulted in limited experience collecting pediatric patient-reported outcomes in cooperative group clinical trials.

Substantial collaborative work has ameliorated many of these barriers, including development and validation of pediatric-specific patient-reported outcomes measurement tools,^13,14 identification of a core battery of adolescent and young adult recommended instruments,¹⁵ and development of electronic data capture strategies.¹⁶ The Children’s Oncology Group (COG) trial AHOD1331 represents an early success collecting pediatric, adolescent, and young adult patient-reported outcomes within a cooperative group clinical trial. AHOD1331, a phase 3 study of pediatric high-risk classic Hodgkin Lymphoma, compared standard chemotherapy with the incorporation of a novel agent, brentuximab vedotin.¹⁷ An anti-CD30 antibody-drug conjugate with strong antitumor effects, brentuximab vedotin had been associated with elevated rates of peripheral neuropathy in adults. To understand the peripheral neuropathy experience in children and adolescents, serial patient-reported outcomes of this neuropathy and HRQOL were collected. Participation rates in the patient-reported outcomes among age-eligible children and caregivers exceeded 95% and remained high throughout treatment; off-treatment completion rates fell to approximately 70%. Results of this trial support the rationale for embedded patient-reported outcomes studies^18-20 and provided justification for inclusion of these data in a subsequent “all-ages” trial of advanced-stage classic Hodgkin lymphoma led by the SWOG Cancer Research Network clinical trials group.²¹ These clinical trial successes, which also highlighted the distinct information provided by patient-reported outcomes inclusion in the clinical trial setting,²⁰ have inspired and informed subsequent trials, as we discuss below.

Given the ever-improving survival rates for pediatric cancer, known short-term and long-term impacts of cancer treatment, yet limited knowledge of the pediatric experience of novel therapeutics,²² it is crucial that pediatric cooperative group clinical trials collect patient-reported outcomes. Information gleaned from outcomes collected in clinical trials can help identify patients at greatest risk of toxicities and can facilitate early symptom identification to implement supportive care strategies to ameliorate symptoms and support adherence. Furthermore, when faced with treatments with equal efficacy, patient-reported outcomes can aid in shared decision making based on treatment tolerability. Several checklists and consortia offer helpful guidance for researchers to ensure rigorous scientific methods for patient-reported outcomes inclusion and analysis in clinical trials.^23-26 However, these methods are not specific to the unique challenges of patient-reported outcomes collection in pediatric or adolescent and young adult trials. Here, we propose a best-practices approach and key considerations for including patient-reported outcomes in pediatric cancer clinical trials (Figure 1), building on the knowledge gained from recently completed classic Hodgkin lymphoma trials.^20,21,27 We also share examples of our experiences conceptualizing and conducting 3 patient-reported outcomes studies in current clinical trials led by the COG.

Figure 1.

Incorporation of patient-reported outcomes in pediatric and adolescent and young adult clinical trials, from study conception through reporting.

• AOST2031 is a phase 3 clinical trial comparing surgical management strategies for pulmonary metastatic osteosarcoma (ClinicalTrials.gov identifier NCT05235165). This trial, which includes children and adults up to age 50 years, compares an open vs thoracoscopic approach to resection of pulmonary metastases. From study inception, patient-reported outcomes were recognized as important to answering crucial trial questions of surgical tolerability and postoperative pain. The study employs the AYA Patient-Reported Outcomes Task Force–recommended core patient-reported outcomes battery¹⁵ and is the first COG trial to use electronic data capture of patient-reported outcomes measures.

• AHOD2131 is a clinical trial for patients with stage I and II classic Hodgkin lymphoma comparing standard therapy to standard treatment plus brentuximab vedotin and nivolumab (ClinicalTrials.gov identifier NCT05675410). Given that the addition of immunotherapy could affect toxicities and HRQOL in both the short and long terms, and the lack of available data on potential toxicities in children, adolescents, and young adults, this trial includes patient-reported outcomes aims and measures and uses the NCI National Clinical Trials Network (NCTN) adolescent and young adult patient-reported outcomes battery and electronic data capture. As most patients are expected to survive, the study includes long-term patient-reported outcomes assessments 12 years following diagnosis.

• AALL1732 is a phase 3 clinical trial for patients with newly diagnosed high risk B-cell acute lymphoblastic leukemia (ClinicalTrials.gov identifier NCT03959085). This trial, which includes children, adolescents, and adults up to age 25 years, compares a standard chemoimmunotherapy treatment backbone with a regimen that replaces blocks of conventional chemotherapy with the antibody-drug conjugate inotuzumab ozogamicin. Patient-reported outcomes are incorporated to better understand the toxicity and tolerability of this novel regimen compared with conventional chemotherapy.

We collectively learned many lessons developing and designing these pediatric and adolescent and young adult trials, including selection of patient-reported outcomes of interest, identification and refinement of patient-reported outcomes study aims, and determination of the measurement and analytical approach. We share our experience developing these patient-reported outcomes substudies as examples of challenges and successes to inform future pediatric and adolescent and young adult clinical trials that incorporate patient-reported outcomes.

Aligning patient-reported outcomes aims and overarching trial aims

We recommend consideration of which study outcomes may be best captured by direct patient report or a combination of clinical data and patient-reported outcomes during the concept development phase. For AOST2031, metastatic osteosarcoma has an extremely poor prognosis, and it is unknown whether the surgical approach to manage pulmonary metastatic disease has a meaningful impact on long-term survival. Therefore, the impact of surgical approach on postoperative pain, duration of recovery, and functional consequences were primary HRQOL outcomes of interest and clinically important. As these outcomes are optimally assessed through patient report, the strategy for patient-reported outcomes inclusion occurred concurrently with trial development. Based on findings from prior studies comparing surgical approaches in lung cancer,²⁸ we hypothesized that a thoracoscopic approach would be associated with decreased pain interference and less functional impairment than an open surgical approach.

In contrast to the poor long-term outcomes for patients with metastatic osteosarcoma, patients with early-stage classic Hodgkin lymphoma, and pediatric acute lymphoblastic leukemia have favorable long-term survival rates. Yet, their treatments are associated with myriad physical and psychological symptoms, resulting in negative impacts on HRQOL during and after treatment. Little is known about the patient-reported toxicity and HRQOL impacts associated with novel agents and treatment combinations, particularly in pediatrics. In AHOD2131, patient-reported outcomes aims focused on comparing short-term and long-term toxicities with tolerability between treatment arms; if oncologic outcomes are equivalent, treatment tolerability will help determine the standard of care. In AALL1732, given the favorable toxicity profile of inotuzumab ozogamicin in relapsed B-cell acute lymphoblastic leukemia, we hypothesized that the inotuzumab ozogamicin regimen would be associated with improved HRQOL and decreased symptomatic adverse events.^29-31 Further, AALL1732 created an opportunity to characterize the first pediatric, adolescent, and young adult patient-reported toxicity and HRQOL impacts with inotuzumab ozogamicin. Across all 3 of these trials, even if the experimental arm fails to improve survival outcomes, patient-reported outcomes of symptomatic adverse event and HRQOL data will inform understanding of treatment tolerability and help optimize supportive care.

When appropriate, we recommend that patient-reported outcomes aims be included as primary or secondary trial endpoints and that the study enrollment goals be powered accordingly. Exploratory patient-reported outcomes aims risk being omitted from the final trial due to cost, insufficient statistical power, or site burden concerns. Similarly, we favor mandatory patient-reported outcomes completion by all eligible study participants. Optional completion of patient-reported outcomes may introduce bias because individuals who are experiencing more toxicities or who are more vulnerable may be inadvertently excluded. Patient-reported outcomes aims should be structured and written in the same format as other study aims: They must be well defined, specific, measurable, and scientifically justified. We recommend selecting and describing the primary specific patient-reported outcomes domains of interest in the aims, even if more comprehensive patient-reported outcomes will be collected. The strongest patient-reported outcomes aims directly relate to the central study question and design; where appropriate, they may contribute to the determination of the “winning” treatment arm.

Selecting measures and outcomes

Numerous patient-reported outcomes measures can be used to assess domains of interest, but data collection should consider ways to maximize information while minimizing participant, site, and administrative burden. Identification of primary patient-reported outcomes of interest drives patient-reported outcomes measure selection; it is also important to consider the anticipated completion time. Inclusion of some brief patient-reported outcomes batteries can facilitate comparison of patient-reported outcomes across diseases and trials and ensure identification of unexpected toxicities or impaired domains. Perspectives of patient advocates should be incorporated during trial design to ensure selection of priority outcomes for patients and acceptability of measures.

Determination of optimal measures requires considering the age ranges of trial participants and their ability to self-report. Several measurement systems have pediatric versions for children as young as 7 years of age, with corresponding adult measures for older patients (≥18 years of age), and linking metrics for uniform scoring across the age continuum. However, for younger children, caregiver proxy reports are often used or collected concurrently with patient self-reports. Pediatric patient-reported outcomes require consideration of the benefits and challenges of including patients of various ages and developmental phases as well as times where multiple perspectives—parent or caregiver, patient, and clinician—may all provide meaningful though distinct information.^32-34 For adolescent and young adult studies, the NCTN AYA Patient-Reported Outcomes Task Force recommends inclusion of Patient-Reported Outcome Measurement Information System (PROMIS) health status assessment for core HRQOL domains, with added domains depending on the primary HRQOL impact, and the Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE) and Pediatric PRO-CTCAE to capture targeted symptoms and toxicities.⁵ We fully endorse this approach, when age and scientifically appropriate, because these measurement systems support patient self-report and caregiver proxy report for pediatric patients, are available in multiple languages, and have been tested and validated in children with cancer. At times, we have limited eligibility for the patient-reported outcomes aims to ensure patient self-report. Further work is needed to standardize recommended measures when the study population is predominantly below the age of self-report or includes patients with varied cognitive abilities. Studies with wide age ranges may consider different patient-reported outcomes measures for different age participants, which adds statistical and analytic complexity. A plan must also be made for when patients cross patient-reported outcomes measure age cutoffs during the clinical trial; we often use the measures corresponding to the patient’s age at study enrollment.

Across the 3 studies we discuss here, we primarily use PROMIS measures (Table 1)³⁵ and the PRO-CTCAE/Pediatric PRO-CTCAE. PROMIS is a library of validated, reliable tools to measure HRQOL sponsored by the National Institutes of Health and available in the public domain. Importantly, there are pediatric and adult PROMIS measures with well-established psychometric properties and validated self-report measures available for individuals aged 8 years and older in multiple languages.^13,36 Most Pediatric and Adult PROMIS measures are comparable conceptually across the age continuum.³⁷ PROMIS measures are available as short forms of varying lengths that can be administered either as part of larger profile measures or as stand-alone measures. We often use the longer 8-item short form in place of 4-item forms for the primary domain of interest to expand reliability and construct validity for primary HRQOL outcomes.^38,39 For example, as fatigue is prevalent and impactful in classic Hodgkin lymphoma treatment and survivorship, the 8-item Fatigue measure is being used in AHOD2131.^40,41 The PRO-CTCAE/Pediatric PRO-CTCAE were designed for use in clinical trials to capture patient-reported adverse events and have established validity and reliability.^3,14 Study teams are encouraged to select specific items to assess anticipated toxicities from the symptom library. The NCI and US Food and Drug Administration have endorsed the use of the Pediatric PRO-CTCAE in pediatric clinical trials.⁴²

Table 1.

Key considerations and approaches to patient-reported outcomes across example COG studies.

Consideration	AOST2031	AALL1732	AHOD2131
Patient-reported outcomes aims	To determine whether thoracoscopy is superior to open surgical resection for postoperative pain interference To prospectively compare patient-reported functional impairment of the upper extremities, pain intensity, and HRQOL between treatment arms	To compare HRQOL at prespecified time points, by study arm To compare symptomatic adverse events, by study arm, at prespecified time points	To compare patient-reported adverse events, stratified by age group, therapeutic arm, and receipt of radiation therapy over time To evaluate changes in patient-reported fatigue, cognitive functioning, and HRQOL (eg, emotional, physical, and role functioning), by treatment arm
Measures	HRQOL: PROMIS measures (Pediatric and Adult): Pain interference (primary) Upper extremity function Pain intensity Multidimensional health status measures	HRQOL: PROMIS Pediatric Scale Global Health 7 Symptomatic adverse events: Pediatric PRO-CTCAE: Fatigue Nausea Insomnia Sadness Anxiety Appetite	HRQOL: Functional Assessment of Chronic Illness Therapy-Item GP5 PROMIS Pediatric and Adult measures: Depression Anxiety Physical Function/Mobility Emotional Support Sleep Disturbance Fatigue Neuro-QOL Cognition Function Symptomatic adverse events: PRO-CTCAE (16 domains) and Pediatric PRO-CTCAE (15 domains)
Participant ages for patient-reported outcomes	Aged 8-50 y, self-report	Aged 8-17.99 y, self-report	Aged 5-60 y, self-report
Patient-reported outcomes time points of interest	T1: Preoperative baseline T2: 24-72 h postoperative pain T3: 1-2 wk postoperatively T4: 4-6 wk postoperatively	T1: Baseline at random assignment Specific time points in development	Phase 1 assessments: T0: Pretreatment baseline T1: Day 1 of each cycle T2: End of therapy T3, T4, T5: Years 1, 2, and 3 off therapy Phase 2 assessments: T6, T7, T8: Years 4, 8, and 12 off therapy
Mode of data collection	Electronic data collection	Paper and pencil	Electronic data collection
Total No. of patient-reported outcomes time points and anticipated completion time per time point	4 Time points: T1, T3, T4 <15 min; T2 1-2 min	4 Time points for symptomatic adverse events 3 Time points for HRQOL 10 min per time point	Phase 1: 5 Time points for HRQOL and symptomatic adverse events (30 min), plus additional time points for symptomatic adverse events only (15 min)a Phase 2: 3 Time points
Statistical design and analysis considerations	Difference in pain interference between treatment arms at T3 and T4	Primary: Difference in HRQOL between treatment arms Secondary: Differences in symptomatic adverse events between treatment arms Power analysis conducted for the primary aim with co-primary time points	Difference in PRO-CTCAE measures over time between arms and between patients receiving and not receiving radiation therapy Difference in fatigue and cognitive scores between baseline and 1 y after therapy
Challenges and opportunities	Short-interval postoperative assessments create challenges for timing of data collection but opportunities for rapid inclusion of findings in practice	Trial efficacy outcome analysis has hybrid design such that if the experimental arm demonstrates noninferior event-free survival and tolerability, as defined by patient-reported outcomes data along with traditional adverse event reporting, it will be critical in determining the optimal treatment arm	Need for initial paper measures due to short interval between enrollment and treatment initiation Challenging to continue to engage and maintain contact information during long-term follow-up but will provide important novel data

Abbreviations: COG = Children’s Oncology Group; HRQOL = health-related quality of life; PRO-CTCAE = Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events; PROMIS = Patient-Reported Outcomes Measurement Information System.

^aTotal number of cycles/time points depends on treatment arm.

Although the 3 studies discussed focus on symptoms and HRQOL, other domains may be important to assess depending on disease, treatment, and patient characteristics and may require different measures. For example, neurocognitive outcomes can be evaluated with performance measures or with caregiver-reported or patient-reported measures.^43,44 These measures may provide distinct yet complementary information, and some have been used to screen potential participants in at-risk populations.⁴⁵ A multidisciplinary team with diverse areas of expertise, including clinicians, behavioral scientists, psychologists, patients, and statisticians, is critical throughout study development and conduct and may be particularly important in measure selection.

Patient-reported outcomes collection: time points and frequency

Consideration of participant burden requires streamlining of time points and frequency of assessments. The time points for assessment should be directly determined by the research questions and hypotheses and reflect comparable time points when there are differences in trial treatment schedules. For many studies, patient-reported outcomes collection may extend to assess posttreatment long-term outcomes. Unknowns such as the anticipated timeline of emergence or duration of toxicities of novel agents or new treatment combinations as well as long-term implications on HRQOL may make repeated assessments clinically important. Yet, frequent assessments may be overly burdensome and could lead to nonparticipation and missing data. Pairing patient-reported outcomes collection with scheduled clinical assessments or treatment encounters is wise but not always possible, particularly if patient-reported outcomes collection extends beyond routine clinical surveillance, as in AHOD2131. Strategies to promote engagement in patient-reported outcomes assessments in survivorship are needed. Not all patient-reported outcomes need to be collected at every time point. In AOST2031, immediate postoperative pain intensity was a key outcome, but we were mindful of overburdening respondents at a clinically vulnerable moment. So, patient-reported outcomes measurement 24-72 hours postoperatively includes only a single-item pain-intensity measure.

Tight temporal windows for patient-reported outcomes completion require substantial planning and collaboration with research and clinical personnel. In AHOD2131, some participants consent to the study and begin treatment the same day, posing unanticipated challenges for electronic data capture. To provide flexibility, an option to complete the first patient-reported outcomes assessment on paper was added. Scheduled study team meetings, particularly surrounding study opening, help identify and address issues as they occur. In AOST2031, all patient-reported outcomes assessments are anchored on the surgical date, which is often documented in research forms after the surgery has occurred. Our multidisciplinary team developed a procedure to share the anticipated surgical date in advance, with later confirmation of the actual surgical date, which resulted in improved data collection. Regardless of planning, unexpected challenges are likely to arise that require communication and refinement of study procedures. Collaboratively working with clinical research coordinators centrally and at individual study sites is critical, from study design through completion, to identify challenges and solutions. Developing case report forms and mechanisms to update and obtain multiple methods of patient contact in support of patient-reported outcomes collection across diverse sites and patients requires ongoing attention.

Data-collection strategies

Paper-and-pencil data collection has traditionally been used for patient-reported outcomes assessments in clinical trials. This approach is inefficient, may pose a burden on site research staff, and can result in data transcription errors. Although some studies found that adolescent patients may prefer paper-and-pencil survey completion,^46,47 more recent work has shown patient preference for electronic data capture.¹⁶ Electronic data capture supports collection of patient-reported outcomes when patients are not in the hospital or clinic, when outcomes may be more representative of patients’ experiences self-managing toxicities. Electronic data capture allows for central patient-reported outcomes collection, minimizing the burden on participating treatment sites with variable research resources and staff. PROMIS and the PRO-CTCAE have electronic data capture options that can be completed on a patient’s smartphone, tablet, or computer. The NCTN AYA Patient-Reported Outcomes Task Force and the COG have devoted substantial resources to the development of an electronic data capture strategy that uses Research Electronic Data Capture, which is being used in current cross-network trials, including AOST2031 and AHOD2131.

Further research on optimal procedures for electronic data capture, particularly in pediatric clinical trials, is needed because potential issues could arise. Some patients may prefer to complete a survey with in-person research personnel encouragement or support. In addition, it may be more challenging to distinguish between individuals who did not complete a time point due to technical issues such as lack of internet connectivity or email spam filters compared with noncompletion due to high symptom burden. Other unanswered questions include the optimal strategy for ensuring direct patient report from younger children in the absence of caregiver assistance. Similarly, collection of patient-reported outcomes from individuals with intellectual or developmental disabilities or from children with difficulties with reading comprehension, which may be more prevalent among children with cancer who have had prolonged absences from school due to treatment, may require distinct data-collection strategies to facilitate self-report and ensure data integrity. These strategies could include technological innovations such as narration, pictorial representation of the response scale,⁴⁸ or more intensive personnel support. Finally, there are concerns that patients from historically marginalized groups may be less likely to participate in electronic data capture.⁴⁹

Safety plans must be developed for timely communication of concerning findings identified on patient-reported outcomes, such as severe depression or uncontrolled pain, to treating clinicians. Whereas clinicians can see worrisome laboratory results collected in a clinical trial, symptoms or toxicities identified through patient-reported outcomes may not get back to the patient’s care team in a timely fashion. In the studies described, we alert participants that symptoms noted on patient-reported outcomes are for research purposes only and that concerning symptoms should be discussed with clinicians. An a priori safety plan that delineates which measures and scores should trigger a safety flag and the approach to communication with site-specific personnel should be clearly described in the protocol and informed consent process.

Optimizing participation and data completion

It is critical to compute completion rates for patient-reported outcomes assessments and, if possible, collect reasons for noncompletion at each time point. Intermittent evaluation of data missingness is needed during the conduct of the study; if identified, plans to minimize missingness should be implemented. For example, patients who feel too poorly to complete patient-reported outcomes assessments at a given time point may lead to clinicians drawing erroneous conclusions about treatment tolerability. Meanwhile, patients who do not complete assessments due to lack of necessary technology for electronic data capture or because of limited health literacy represent distinct but meaningful concerns. Both are examples of informative missing data resulting in biased results.

We recommend collecting self-reported or family-reported demographic and social information for clinical annotation. The feasibility of obtaining self-reported data from pediatric, adolescent, and young adult patients and their families has been demonstrated in the clinical trial setting.⁵⁰ Recent adult oncology studies have found lower rates of consent to patient-reported outcomes substudies by race and education: Black and Asian participants and individuals with lower levels of educational attainment were least likely to participate in patient-reported outcomes assessments and had lower rates of participation in electronic data capture.⁴⁹ Patient-reported outcomes obtained as part of clinical oncology care have had similar differences in completion rates by race and ethnicity.⁵¹ These concerning trends suggest a risk of missing the voices of patients from historically marginalized groups. Though unexplored in pediatrics, patient-reported outcomes completion in pediatric clinical trials may face similar issues. A study exploring dropout rates in COG clinical trials found that non-Hispanic Black and Hispanic or Latino patients were more likely to be lost to follow-up than were White patients; patients experiencing poverty and individuals with lower levels of education were also more likely to be lost to follow-up.⁵² Optimizing response rates is necessary to incorporate all perspectives; lack of representation may also result in biased estimates and in health data poverty.⁵³ We must continue to develop strategies to ensure the optimal participation of all study participants and pay particular attention to the needs of historically marginalized populations. Ensuring that pediatric patient-reported outcomes measurement tools have validated translations in multiple languages is one strategy to ensure that our knowledge is reflective of many children.

Statistical planning and analysis

Statisticians are integral to determination of the key outcomes and time points of interest, development of an analytical plan that includes an assessment of precision of estimation and power associated with target sample sizes, and the selection of superiority or noninferiority hypothesis testing. Although it may be tempting to obtain patient-reported outcomes from a subsample of the study population, it is important that the selected sample be representative of the population of interest to ensure generalizability of results and that the sample size be sufficient to provide precision in estimates and comparisons of interest. Additional analytic elements that must be specified in the statistical plan include whether to focus on absolute scores from patient-reported outcomes at a specific time point, change scores, or serial scores with which to conduct a longitudinal analysis as well as how to manage missing data. Selection of a co-primary approach for endpoints or time points may require more patients. Preliminary data may aid in hypothesizing when to expect the biggest difference in the patient-reported outcomes of interest between the 2 treatment groups, but if outcomes of interest are too narrow, there is a risk of not capturing the trajectories of patient-reported outcomes throughout key time points of the trial.

Measurement system–specific recommended approaches to scoring and analysis should be used. Compelling and novel strategies have been used in the interpretation of adult patient-reported outcomes to characterize toxicity or tolerability of treatment, but their performance in pediatric cohorts and their reflection of the experiences of children are unexplored. For example, the Toxicity Index, a summary score that incorporates severity and frequency of clinician-graded toxicities, can be enhanced by patient-reported outcomes and is associated with treatment discontinuation in adults with cancer.^54,55 Meanwhile, a single item assessing symptom bother in adults with cancer is associated with an increased number and grade of physician-rated toxicities, poorer physician-reported functional status, and greater likelihood of treatment discontinuation.^56-58 Analyses of patient-reported outcomes data to evaluate the applicability of these summary scoring approaches to pediatric patient experiences should be incorporated in pediatric trials.

Dissemination

Finally, it is crucial that we commit to disseminating the results of patient-reported outcomes so that these meaningful patient data are used to improve care and outcomes. Even when patient-reported outcomes data are collected in pediatric and adolescent and young adult trials, they are rarely published.⁹ Recent efforts focused on timely dissemination of patient-reported outcomes in classic Hodgkin lymphoma should be commended and replicated.^18,20 Prioritization and recognition of the importance of patient-reported outcomes data are required from study investigators; statisticians; governing bodies; stakeholders, including patient advocates and survivors; and funders to ensure timely access to resources and expertise required to analyze and disseminate patient-reported outcomes data.

Conclusions

Patient-reported outcomes provide important and unique data on the pediatric patient experience of treatments and toxicities that have wide-reaching implications. Inclusion of patient-reported outcomes across COG and cross-NCTN cooperative group clinical trials that include children, adolescents, and young adults will enhance our understanding of patient-centered clinical trial outcomes during and after treatment. Patient-reported outcomes are valuable for answering central trial questions of toxicities of contemporary treatments and have increasingly important roles for evaluating the risks and benefits of novel agents or treatment combinations, assessing the HRQOL impact of intensive treatments for diagnoses with poor overall survival, and interpreting primary study outcomes in the context of HRQOL for therapy reductions. Patient-reported outcomes collected in trial settings are valuable in characterizing the toxicities and long-term effects of standard treatments, and these results are crucial to patient education and decision making both within and outside of the clinical trial setting, including for the many patients not enrolled in clinical trials who receive standard-of-care therapies. Partnerships with patients and patient advocates from the trial planning stage is important to ensuring that we evaluate those toxicities and impairments that are most impactful to patients while minimizing the burden of completion. Patient-reported outcomes can be included in pediatric and adolescent and young adult clinical trials with methodologic rigor, and it is essential that findings be analyzed and disseminated to continue to improve care and select treatments that benefit our patients in the ways that matter most to them.

Author contributions

Katie A. Greenzang (Conceptualization, Writing—original draft, Writing—review & editing), Kathleen E. Montgomery (Writing—review & editing), Adam DuVall(Writing—review & editing), Michael E. Roth (Writing—review & editing), Mark Krailo(Writing—review & editing), Michelle M. Nuno (Writing—review & editing), Lindsay Renfro (Writing—review & editing), Natalie DelRocco (Writing—review & editing), John Doski (Writing—review & editing), Kara Kelly (Writing—review & editing), Sharon M. Castellino (Writing—review & editing), Jennifer McNeer (Writing—review & editing), Maureen M. O'Brien (Writing—review & editing), Damon Reed (Writing—review & editing), Katherine Janeway (Writing—review & editing), Pamela S. Hinds (Writing—review & editing), Sue Zupanec (Writing—review & editing), and Susan K. Parsons (Conceptualization, Writing—review & editing).

Funding

This work is supported by NCTN Operations Center Grant (U10CA180886) and NCTN Statistics & Data Center Grant (U10CA180899). Dr Greenzang is supported by the National Institutes of Health, NCI (K08 CA245036)

Conflicts of interest

S. Castellino has received honoraria from Seagen Inc (now Pfizer); and BMS; she serves on the board of directors of the Leukemia and Lymphoma Society (unpaid) and the Scientific Advisory Board of the Lymphoma Research Foundation (unpaid). D. Reed has received honoraria from Springworks and Eisai. M. O’Brien has received honoraria from Pfizer and research funding from Pfizer, AbbVie, and Amgen. K. Kelly serves on the Scientific Advisory Boards of Pfizer (Seagen) and BMS (both unpaid) and the Lymphoma Research Foundation (unpaid); and her institution receives research funds from COG for her participation as a scientific steering committee member for a Merck-sponsored clinical trial run through the COG network. J. McNeer serves on advisory boards for Amgen and Jazz. All other authors declare no conflict of interest.

Data availability

No new data were generated or analyzed for this commentary.