Report of the Jumpstarting Brain Tumor Drug Development Coalition and FDA clinical trials clinical outcome assessment endpoints workshop (October 15, 2014, Bethesda MD)

Abstract

On October 15, 2014, a workshop was held on the use of clinical outcome assessments in clinical trials for high-grade glioma of the brain. This workshop was sponsored by the Jumpstarting Brain Tumor Drug Development Coalition, consisting of the National Brain Tumor Society, the Society for Neuro-Oncology, the Musella Foundation for Brain Tumor Research and Information, and Accelerate Brain Cancer Cure. It was planned and carried out with participation from the US Food and Drug Administration. The workshop also included stakeholders from all aspects of the brain tumor community, including clinicians, researchers, industry, clinical research organizations, patients and patient advocates, and the National Cancer Institute. This report summarizes the presentations and discussions of that workshop and the proposals that emerged to move the field forward and toward greater inclusion of these endpoints in future clinical trials for high-grade gliomas.

A number of key issues in brain tumor drug development are related to determining whether a new treatment benefits patients. Although the gold standard for measuring drug efficacy in a clinical trial is overall survival, given the latency required to assess this endpoint and the modest increases in survival afforded with current therapies, it is important to explore the use of alternative endpoints to demonstrate clinical benefit to patients.

The Jumpstarting Brain Tumor Drug Development Coalition (comprising the National Brain Tumor Society, Accelerate Brain Cancer Cure, the Musella Foundation for Brain Tumor Research and Information, and the Society for Neuro-Oncology) sponsored 2 workshops in 2014 to evaluate the state of the field and identify core initiatives for optimizing brain tumor–related clinical trial endpoints, with a goal of advancing the development of treatments for adult high-grade gliomas. The workshops brought together stakeholders from the brain tumor community, including clinicians, researchers, industry, patients and patient advocates, the National Cancer Institute, and the FDA.¹ The aim of these workshops and post-workshop actions has been to present the state of the field in the measurement of clinical trial endpoints used in the evaluation of brain tumors, discuss scientific questions related to endpoint measurement, including those measured by imaging and those measuring clinical benefit, and reach consensus on a set of actions designed to ultimately improve endpoint inclusion and measurement in brain tumor clinical trials that are subject to regulatory review. Ultimately, the goal is to stimulate and increase interest and capacity to pursue clinical trials seeking FDA approval of new therapies for high-grade gliomas.

The first workshop was held on January 30, 2014 and primarily focused on neuroimaging clinical trial endpoints.² Discussions included how to overcome the variables in medical imaging, such as image acquisition parameters, which have hindered the ability to accurately assess brain tumor response to therapies, and how to best incorporate endpoints that rely on imaging into clinical trials. A number of needed projects emerged and have been embraced by the neuro-oncology and neuro-imaging community. This has been demonstrated by the formation of the Brain Tumor Imaging Standardization Steering Committee and its working groups, and has spurred the creation of a standardized imaging acquisition protocol for use in high-grade glioma clinical trials.³

The second of these workshops, and the focus of this report, was held on October 15, 2014 and was designed to enhance understanding of the definition and utility of clinical outcome assessment (COA) endpoints (also referred to as patient-focused outcome measures) as applied to clinical trials for individuals with high-grade gliomas and advance inclusion of such measures into therapeutic trials. Approval of a therapy requires substantial evidence of clinical benefit, defined as improvement in how a patient “feels, functions, or survives.”⁴ Because there are concerns about the clinical relevance of radiographic surrogate endpoints, such as progression-free survival and objective response rate, and biomarker-related endpoints, it is of value to include information on patient symptoms and function in clinical trials more frequently and uniformly to assist in the assessment of whether a treatment demonstrates effectiveness.⁵

The discussions at the COA workshop sought to identify specific, practical action plans to develop new measures, as well as increase the use of existing COA endpoints in future brain tumor clinical trials. Specifically, stakeholders (i) identified multiple concepts of interest addressing both how a patient feels and how the patient functions and began determining which sign(s), symptom(s), and function(s) are of priority to patients, and which can be measured in a clinical trial setting; (ii) examined COAs that are available for use now in brain tumor clinical trials and determined the priorities for refinement of the existing measures and necessary elements needed for development of novel COAs for future use in trials; and (iii) defined how the COA will be used to determine clinical benefit and recommended clinical trial designs for COA validation and use in brain tumor patients.

This report summarizes the presentations and discussions from the COA endpoints workshop and the action plans (Table 5) and recommendations that materialized to move the field forward toward standard, validated, COA measures for use in high-grade glioma trials.

Table 5.

Identified action items

Panel 1 Prioritize symptoms to create a concise list, using patient-friendly language. Consider endpoints in the context of concomitant medications. Function needs to be measured from physical and cognitive perspectives.

Panel 2 (4) Identify the subsets within the currently available measures that are useful in assessing prioritized symptoms and functions prioritized. (5) Perform retrospective analysis of data from clinical trials to confirm that they have utilized existing COA instrument accuracy.  • Using these data, analyze current tools and determine where enhancements can be made. (6) Improve data capture of COAs in clinical trials. (7) Evaluate new instruments in development, including NANO and NFBrSI-24.

Panel 3 (8) Incorporate COAs into earlier-phase trials to enhance understanding of their relationship to other toxicity and efficacy endpoints in various contexts of use. (9) Educate patients and investigators on importance of COAs in clinical trials. (10) Standardize implementation of tools and the analysis of the data. (11) Create statistical paradigms. (12) Develop assessments of patient′s symptoms, physical functioning, and neurocognition and use in future brain tumor clinical trials whenever possible.

Workshop Overview and Summary of Presentations and Discussions

The workshop, moderated by Dr Clifford Goodman of the Lewin Group, began with a formal presentation to establish context for the day's discussions, titled “Use of Clinical Outcome Assessments in Oncology: Successes and Pitfalls.” Following this presentation, the workshop continued with 4 panel-led discussions. After a short overview presentation by each panel, a facilitated audience question-and-answer session followed. Prior to the start of Panels 2 and 3, FDA representatives presented on the topics “What Can Be Learned from the FDA PRO Guidance?” and “Regulatory Experience with COAs and Its Relevance to Brain Tumor Studies,” respectively.

Use of Clinical Outcome Assessments in Oncology: Successes and Pitfalls

Paul G. Kluetz, MD, from the Office of Hematology and Oncology Products, FDA, provided an overview of the regulatory framework, gave examples of successful uses of COAs in oncology, and discussed pitfalls and challenges of COAs in oncology.

Overview of the regulatory framework

Substantial evidence is required to support a marketing claim of treatment benefit that is based on the FDA label. In order to make a marketing claim, adequate and well-controlled studies, which include well-defined and reliable assessments, are needed. Measures are well defined and reliable when empiric evidence demonstrates that the score quantifies the concept of interest in the targeted context of use (Table 1). Two approval tracks were discussed: (i) regular approval is based on prolongation of life, a better life, or an established surrogate; (ii) accelerated approval, for severe or life-threatening diseases, is based on providing meaningful therapeutic benefit over existing therapies typically utilizing a surrogate endpoint. When demonstrating clinical benefit, overall survival is the gold standard in oncology on the basis of a measurement of direct clinical benefit and void of many confounders of interpretation.

Table 1.

Terms^4,6

Term	Definition
Assessment	Measure and evaluation of some aspect of a patient that result in a recorded datum.
Concept of interest	The thing measured by an assessment (eg, pain intensity).
Conceptual framework	An explicit description or diagram of the relationships between the questionnaire or items in a PRO instrument and the concepts measured. The conceptual framework of a PRO instrument evolves over the course of instrument development as empiric evidence is gathered to support item grouping and scores. We review the alignment of the final conceptual framework with the clinical trial's objectives, design, and analysis plan.
Construct validity	Evidence that relationships among items, domains, and concepts conform to a priori hypotheses concerning logical relationships that should exist with other measures or characteristics of patients and patient groups.
Content validity	Evidence from qualitative research demonstrating that the instrument measures the concept of interest, including evidence that the items and domains of an instrument are appropriate and comprehensive relative to its intended measurement concept, population, and use. Testing other measurement properties will not replace or rectify problems with content validity.
Context of use	A comprehensive statement that fully and clearly describes the way the COA is to be used and the drug development–related purpose of the use. The context of use defines the boundaries within which the available data adequately justify use of the COA and describes important criteria regarding the circumstances under which the COA is qualified.
Effectiveness	An essential component of the basis for marketing approval of a drug; drugs must be safe and effective to justify approval. Effectiveness is defined as a benefit to patients in how they feel, function, or survive due to treatment with the drug.
Endpoint	The way an assessment will be used as a study result and statistically compared among treatment groups to assess the effect of treatment. Endpoints are often named by the assessment measured, but a complete statement of the endpoint should include a full description of what data are collected and how they are analyzed to support a specific study objective.
Feels	A patient's physical sensation (eg, symptoms) or perceived mental state. A patient may feel pain, feel feverish, or perceive a severely low mood (as with depression).
Functions	The manner in which a patient can successfully perform tasks and roles required for everyday living. A patient's ability to perform specified activities that are a meaningful (to the patient) part of typical (eg, daily) life.
Score	A number derived from a patient's response to items in a questionnaire. A score is computed based on a prespecified, validated scoring algorithm and is subsequently used in statistical analyses of clinical trial results. Scores can be computed for individual items, domains, or concepts or as a summary of items, domains, or concepts.
Sign	Any evidence of a disease, health condition, or treatment-related effect that can be observed. Signs can be assessed using PRO, ClinRO, ObsRO, or PerfO assessments, as appropriate.
Symptom	Any subjective evidence of a disease, health condition, or treatment-related effect that can be noticed and known only by the patient.

COA is an umbrella term for measures that assess a patient's symptoms, overall mental state, or the effects of a disease or condition on how the patient functions. COAs can be used to determine whether or not a drug provides treatment benefit or safety benefit compared with other treatments.^6,7 There are 4 types of COA measures: (i) patient-reported outcome (PRO), (ii) clinician-reported outcome (ClinRO), (iii) observer-reported outcome (ObsRO), and (iv) performance outcome (PerfO) (for more details, see Blakeley et al, Table 1, this supplement).

Examples of successful uses of COAs

COAs can potentially be used in oncology clinical trials by providing primary support for an indication. The 1996 approval of mitoxantrone (Novantrone) for prostate cancer was largely based on a randomized 161-patient trial. The primary endpoint, time to pain progression, was used to compare mitoxantrone plus prednisone with prednisone alone. Approval was granted based upon demonstration of improved pain, evidence of antitumor effect, and no detriment to overall survival.⁸

COAs can also provide key secondary evidence to support regular approval rather than an accelerated approval. Recently this method was used successfully in the approval of ruxolitinib (Jakafi) for myelofibrosis. In this trial the primary endpoint was radiographic response rate and the key secondary endpoint was PRO total symptom score. Approval was granted upon demonstration of both reduction in spleen size by MRI (response rate) and improvement in total symptom score (assessing abdominal discomfort, pain under left ribs, night sweats, itching, bone/muscle pain, and early satiety).^9,10 The total symptom score helped by providing a correlative to imaging results by demonstrating improvements in how patients were feeling (symptoms); ruxolitinib was granted regular approval.

Lastly, COAs can be utilized as supportive data to better describe the patient experience in clinical trials. This can be a measure of disease symptom(s), treatment symptom(s), physical function, and/or functional assessment. In the trial testing abiraterone acetate (Zytiga) for prostate cancer, time to pain progression was used as an exploratory endpoint. The FDA label for this approval noted that PRO pain results supported demonstration of benefit: “The time to opiate use result was supported by a delay in patient reported pain progression favoring the ZYTIGA arm.”¹¹

Challenges of COAs in oncology

COA data are now being reviewed more closely during new drug application and biologics license application submissions. Therefore, it is essential that sponsors optimize their COA selection, collection, and clinical trial design to incorporate such measures. Although the overall goal is to accurately describe the patient experience, there are a number of challenges associated with using COAs in clinical trials that need to be addressed. There are population issues such as measuring symptoms in asymptomatic or minimally symptomatic groups. There are instrument issues such as the use of those that are too broad and insensitive to reliably measure treatment effects (eg, overall health-related quality of life [QoL] is challenging, narrow concepts have had more success) or tools that are not fit for the purpose (“off the shelf” measures used in a different context than originally intended). There are also trial design issues such as blinding, missing data, and lack of statistical rigor. While all studies will have some degree of missing data, it is imperative that the study be designed and executed in a manner that minimizes missing COA data. Further research is also needed on statistical methods to correct for missing data without introducing bias.

Determining Brain Tumor–Specific Signs, Symptoms, and Functions for Use in Clinical Outcome Assessments (Panel 1)

(Terri S. Armstrong, PhD, ANP-BC, FAANP, FAAN, University of Texas Health Science Center and Texas MD Anderson Cancer Center; Allison Bishof, MSN, Patient Advocate; Paul Brown, MD, University of Texas MD Anderson Cancer Center; Martin Klein, PhD, VU University Medical Center; Joohee Sul, MD, US Food and Drug Administration; Martin Taphoorn, MD, PhD, VU University Medical Center; Christina Theodore-Oklota, PhD, Genentech)

Panel 1 presentation

The aim of Panel 1 began the creation of a consensus for a brain tumor clinical outcomes priority list for inclusion in clinical trials. Dr Armstrong led off by providing an overview of the signs, symptoms, and functions commonly present at diagnosis and occurring throughout the disease, and what patients want. Additional details are summarized in the review of this topic earlier in this supplement (Armstrong et al).

At diagnosis, patients often have multiple symptoms, with headache, paresis/weakness, speech/communication difficulties, seizures, neurocognitive (executive function and memory) deficits, and behavioral issues being the most common.^12–14 The signs and symptoms that are present at the time of diagnosis can improve over time, can be fixed and not improve during the disease course, or can worsen as a consequence of disease progression or the combined impact of tumor and therapy.

It has been reported that 80%–90% of individuals with high-grade gliomas had symptoms that prevented their return to work from the time of diagnosis. Furthermore, patients spend a significant portion of their lives feeling ill and unable to perform usual activities or meet occupational, social, financial, and family obligations. There is also a significant impact on the family/caregiver in terms of psychosocial and financial burden, changes in roles, and impact on physical health.^15–18

To provide better insights into the symptom burden of patients with brain tumors, the Jumpstarting Brain Tumor Drug Development Coalition asked brain tumor patients and caregivers to complete an online survey focused on symptoms patients experience, specifically what symptoms, signs, and functions they feel are the highest priority for inclusion in clinical trials. The survey was designed to evaluate what brain tumor patients and caregivers wanted and help inform stakeholders involved in brain tumor drug development. While 1851 individuals started the survey and 1824 completed it, it was not required that every question be answered, therefore it was common for respondents to skip questions. Out of the 1723 respondents, 54% were caregivers to a loved one who had/has a brain tumor, 34% have/had a first occurrence brain tumor, 11% have/had a recurrent brain tumor, and 1% have/had a metastatic brain tumor. Of these 572 respondents with a primary first occurrence, 41 had a low-grade glioma (grade I or II), 88 had a high-grade glioma (grade III or IV) at the time of diagnosis, and 443 had other brain tumor types. Results from questions are presented in Tables 2–4. Based on these survey results, the panel concluded that impact on function is a priority for patients and caregivers in terms of expectations of therapy benefit.

Table 2.

Jumpstarting Brain Tumor Drug Development Coalition survey—responses from primary high-grade glioma patients to “What symptoms did you have that led to your diagnosis of a brain tumor? (check all that apply)” N = 85

Answer Options	Selected by Percentage of Respondents
Headaches	57%
Seizures or convulsions	40%
Changes in speech, vision, or hearing	26%
Changes in mood or personality	26%
Problems with memory or ability to concentrate	25%
Problems balancing or walking	24%
Nausea and vomiting	22%
Numbness or tingling in the arms or legs	15%
Other	14%
Weakness on one side of my body	13%
Difficulty understanding	11%
Change in bowel patterns	1%
None	0%

Table 3.

Jumpstarting Brain Tumor Drug Development Coalition survey—responses from primary high-grade glioma patients to “In addition to allowing a person to live longer, what would you most like to see future brain tumor treatments do? Please rank the following in terms of importance to you.” N = 87

Answer Options	Average Rating
Maintain ability to walk and perform basic physical tasks	4.63
Improve memory or ability to concentrate	4.33
Retain brain functioning, even if not living longer	4.26
Reduce disease-related personality changes	4.20
Improve mood (ie, anxiety/depression)	4.19
Reduce seizures	4.16
Reduce pain, including headaches	4.14
Reduce weakness/improve strength	4.12
Reduce fatigue	3.99
Improve speech	3.90
Reduce nausea	3.45
Reduce need for corticosteroids	3.42

Respondents were asked to rank the importance of select clinical assessments one would like to see future brain tumor treatments address in addition to living longer on a 1–5 scale: Not at all Important, Neutral, Somewhat Important, Important, Extremely Important.

Table 4.

Jumpstarting Brain Tumor Drug Development Coalition survey—responses from primary high-grade glioma patients to “Of the symptoms you ranked as very important, which would you consider to be the most important?” N = 61

Answer Options	Selected by Percentage of Respondents
Retain brain functioning, even if not living longer	30%
Maintain ability to walk and perform basic physical tasks	28%
Improve memory or ability to concentrate	13%
Reduce fatigue	8%
Reduce weakness/improve strength	5%
Reduce seizures	5%
Reduce disease-related personality changes	5%
Reduce pain, including headaches	3%
Improve mood (ie, anxiety/depression)	3%
Reduce need for corticosteroids	0%
Reduce nausea	0%
Improve speech	0%
Other	0%

Because respondents were asked to pick the most important symptom from the previous question, responses were categorized into one of these 12 symptoms. When multiple answers were given, the first answer was used for categorization purposes. When an answer could not be categorized, it was considered “other.”

Panel 1 assertions leading into the group discussion were

• Steroid use, signs/symptoms, and impact on function are 3 general areas designated as priority COAs for panel discussion.

• Patients primarily experience the impact of steroid use as signs/symptoms and consider the impact of signs/symptoms on how they function on a daily basis.

• Signs, symptoms, and function-related COAs should be considered objective measurements, but also patient/proxy report of the impact.

• Three major domains of interest: (i) steroid use, (ii) signs/symptoms, and (iii) functional status.

Panel 1 discussion

• Although patients clearly wish to live longer, they also want to live “better” while they are alive. Therefore, there is a need for more informed patient participation in medical care.

• There is increasing recognition of the importance of evaluating the impact of therapy on patient-focused outcomes as a measure of clinical benefit.

  • Recent brain tumor clinical trials have demonstrated that symptom-based PRO measures are sensitive to tumor progression and differences in treatment arms and may be related to overall survival and progression-free survival.^{19, 20}
  • Recent brain tumor clinical trials suggest that neurocognitive function may predict overall survival and progression-free survival, as a decline in function often occurs at the time of tumor progression.^19–21

• The relationships among symptoms, signs, and functions are complex, and there is a need to continue to analyze these relationships to determine what is being caused by the treatment and what is being caused by the tumor.

• There is the need for a concise, prioritized list of symptoms, signs, and functions for which new COAs need to be developed. A set of COAs that could be more widely used as brain tumor COAs across trials would allow patients and health care providers to make an educated decision about the course of treatment by comparing outcomes from therapies conducted in a standardized manner.

  • Across currently used brain tumor PRO instruments, there is great redundancy in the symptoms being assessed, the most common being headache and pain, the second being short-term memory; followed by expressive aphasia, or difficulty speaking; hemiparesis or hemiplegia (sometimes worded “weakness” on the instruments); and seizures.
  • There is inconsistent measurement of these common symptoms and signs across trials and a wide variety of analyses used in the interpretation of the measurements. Multiple instruments utilize different measurement tools, and different questions are being asked to measure the same symptom. These inconsistencies make comparison of the measurements extremely difficult if not impossible, particularly from a QoL perspective, and provide little insight as to what impact the treatment in question will have on the patient.
• When determining a priority list for brain tumors, it is important to understand the disease process and natural history, in the context of symptoms at presentation and during the course of the disease trajectory.

  • The most common symptoms at the time of diagnosis are headache, weakness, speech and communication deficits, seizures, neurocognitive issues, and behavioral issues, as documented throughout the published literature.
  • The trajectory of individual patients can vary. Some patients have symptoms that worsen over time, while others have fixed deficits that do not change. Still other patients experience symptoms that wax and wane in severity throughout the course of the disease.

    • Other medications, including corticosteroids, anticonvulsants, and chemotherapy, can impact signs and symptoms.

• In developing COAs, it is important to consider both what is important to the patient as well as what is easily measurable and important within a clinical trial. It is also necessary to consider which of the signs or symptoms are sensitive to change, which cannot be improved upon, and those where stabilization would be considered a success.

• Three general areas were designated at the workshop as priorities to consider when developing COAs (Fig. 1).

• When further prioritizing the above lists, there is a need to work toward the evaluation of the top 6 (or fewer) symptoms and functional measures, including neurocognition, for which there are existing assessments and instrumental activities of daily living (ADL), which need to be developed specifically for brain tumor clinical trials.

Fig. 1.

Areas designated as priorities when developing COAs.

What Can Be Learned from the FDA PRO Guidance?

Virginia E. Kwitkowski, MS, RN, ACNP-BC, a clinical team leader in the Division of Hematology Products, Office of Hematology and Oncology Products, FDA, reviewed the PRO Guidance for Industry.⁶ The final version of this guidance was published in December 2009 and describes the current thinking of the FDA on the topic and how the FDA reviews and evaluates existing, modified, or newly created PRO instruments used to support labeling claims. The PRO Guidance is a framework for optimal instrument development, trial design, and analysis of PRO data. Although this guidance is written specifically for PROs, it can be used as a guidance for all COA use in clinical trials.

In order to demonstrate clinical benefit through improvements in how a patient feels or functions, the measure and trial must be carefully designed, data collection should be complete, and the study must have an impact on “feels/functions.” The measurement properties of the instrument should be well established before enrollment occurs in a trial. Additionally, there should be a clearly defined role that the PRO endpoint will play in the clinical trial (ie, primary, key secondary, or exploratory endpoint).

When choosing a PRO instrument for use in a trial, first determine whether an adequate instrument exists to assess and measure the concepts of interest. If it does not exist, a new tool can be developed, or an existing tool can be modified to fit. The utility of the measure in the target population must be determined; therefore sponsors should provide evidence of patient input during instrument development and performance in that population. FDA reviews of instruments may also evaluate what concepts are being measured and the conceptual framework; the scoring, weighting of items/domains, and response options; the number of items, data collection method, and administration mode; the medical condition, population, and respondent burden; recall period; and translation/cultural adaptation availability.

Assessment of Patient Symptoms, Signs, Neurocognition, and Limitations in Functional Activities in Clinical Trials for High-Grade Gliomas (Panel 2)

(Jaishri Blakeley, MD, Johns Hopkins University; Stephen Joel Coons, PhD, Critical Path Institute; John R. Corboy, MD, University of Colorado School of Medicine; Ruthann Giusti, MD, US Food and Drug Administration; Nancy Kline Leidy, PhD, Evidera; Tito R. Mendoza, PhD, MS, Med, University of Texas MD Anderson Cancer Center; Elektra Papadopoulos, MD, MPH, US Food and Drug Administration; Jeffrey S. Wefel, PhD, ABPP, University of Texas MD Anderson Cancer Center)

Panel 2 presentation

The objective of Panel 2 was to inform stakeholders who are involved in developing therapies for high-grade gliomas about the measurement tools that are available for clinical trials focused on high-grade glioma and to evaluate the strengths and limitations of these instruments for addressing priority signs (including neurocognitive function), symptoms, and limitations in functional activities in these patients. Dr Blakeley provided a review of the types of COAs; described COAs that have been applied in large therapeutic trials for high-grade gliomas; cross-referenced the content of these COAs to the identified priority symptoms, signs (including neurocognitive function), and functional activities for people with high-grade gliomas; and presented the result of the group's analysis of the COAs most commonly applied in high-grade glioma clinical trials (to the extent possible) based on their psychometric properties. Finally, she presented the group's considerations for the “ideal” COAs that can be used for assessing priority signs, symptoms, and functions in people with high-grade gliomas in the setting of therapeutic trials.

There are several COAs, of which many are PROs, that have been developed for brain tumor clinical research (the European Organisation for Research and Treatment of Cancer 30-item Quality of Life [EORTC QLQ30]/20-item Brain Neoplasm questionnaires, Functional Assessment of Cancer Therapy [FACT]–Brain, the MD Anderson Symptom Inventory Brain Tumor Module [MDASI-BT], the 24-item National Comprehensive Cancer Network/Functional Assessment of Cancer Therapy–Brain Symptom Index [NFBrSI-24]) and several that have been applied widely in clinical trials for gliomas (Clinical Trial Battery Cognitive Domains and Tests, Mini-Mental State Examination, Karnofsky Performance Scale, Eastern Cooperative Oncology Group Performance Status). Choosing the “best” COA depends on the question being asked, the population being assessed, and the characteristics of the COA itself. The ultimate goal is to have a measure that assesses well-defined disease-related symptoms, feelings, or functions that have good psychometric properties and is feasible in the setting of therapeutic trials. Several of the COA measures in current use for glioma clinical trials have well-defined metrics and address specific signs, symptoms, and functions. However, many also have important limitations that are both disease specific and treatment related, including nonrobust definition of concepts of interest or a high rate of overlap across items (for example, assessment of mood with items “I feel sad” versus “I am able to enjoy life”). Additional details are summarized in the review of this topic earlier in this supplement (Blakeley et al).

Panel 2 discussion

• No single tool is sufficient to address all “core” signs, symptoms, and functions identified by Panel 1. However, there are tools that have been combined in clinical trials and have provided important treatment- and disease-related insights. The panel recommendation is to analyze the data from previous clinical studies for people with high-grade glioma and determine which symptoms, signs, and functions are assessed reliably within existing instruments and are associated with the disease course and to define the magnitude of change that is considered clinically meaningful for each symptom, sign, and function.

• There is also a need for more targeted endpoint measures, along with defining their relation to use of ClinROs, ObsROs, and PROs. Use of existing instruments and further analysis of these measures may identify something suitable and useful for implementation now while we concurrently develop new instruments that meet the needs of patients.

• A COA that fits the following parameters is needed:

  • Is able to assess specific, well-defined disease-related symptoms.
  • Has good psychometric properties.
  • Is feasible in the clinical trial context to minimize missing data.
  • Has the ability to detect a well-defined, meaningful change in that patient population.

• Important characteristics of an instrument include content and construct validity. These determine if what is being measured is in fact what the instrument is purporting to assess and that the change detected is a real and accurate change in the target concept (eg, physical function). This is required to correctly interpret results and incorporate the results into the treatment labeling in a way that is accurate. The instrument should also be reliable, reproducible, and sensitive enough to detect clinically meaningful change.

• In patients with high-grade glioma, determination of directional movement of the core symptoms provides important insights for interpretation. For example, if all symptoms move in the same direction (ie, unidirectional), this would potentially allow for a total symptom score. Alternatively, symptoms that are highly variable across patients would require that individual patients identify their most problematic symptom and assess this symptom and its progression or resolution with treatment as an endpoint.

• The symptom, sign, or function to be measured should determine the structure of the COA. For the symptom, sign, or function to be measured, assessing the quality of existing instruments and their accuracy of measuring that construct is important before using or deciding to develop a new instrument.

  • The MDASI-BT is the PRO symptom scale that has been applied most widely in brain tumor trials. There is extensive evidence supporting this instrument's validity, reliability, and sensitivity.^{19,20,22–24}

• A major hurdle to the use of health-related QoL as a regulatory endpoint is the difficulty in its measurement, as it is a term without a precise definition. Commonly used QoL measures include some domains that are of little relevance to the therapeutic efficacy of a given agent and domains that have little impact on patient outcomes. Therefore, a symptom inventory may be preferable, as symptoms are more proximal to the disease (and effects of treatment).

• Cognitive and neurologic functions are important components, as they are direct measures of brain function, the site of disease, and target organ of the treatment.

  • The Clinical Trial Battery (Hopkins Verbal Learning Test–Revised, Trail Making Test, Multilingual Aphasia Examination: Controlled Oral Word Association) assess cognitive domains that are frequently problematic for individuals with high-grade gliomas and that interfere with functional ADL, particularly instrumental ADL. This battery has been used in many previous studies and has been validated.^25,26
  • Other PROs used in high-grade brain tumor clinical trials include the Medical Outcomes Study–self-reported Cognitive Functioning, FACT–Cognitive Function, and the Macnair Cognitive Complaint Questionnaire.
  • Neurologic Assessment in Neuro-Oncology (NANO), an extension of the RANO effort, is a ClinRO tool in development that is designed to provide objective functional measures of neurologic function complementary to measures of cognition and PROs.²⁷

• Large gaps of COA data collected in the setting of therapeutic trials for high-grade glioma limit the accurate interpretation of these measures. Electronic data capture has been very successful in reducing missing data and should be considered in future development and assessment of COAs. Furthermore, COAs incorporated into studies must be chosen carefully, and the clear relationship between the COA and the goal of the study must be defined so that patients, caregivers, and clinicians are all invested in the collection of complete COA data.

• Moving forward, there is a need to (i) analyze the existing clinical trial data sets for people with high-grade gliomas to better define the properties of existing COAs in specific therapeutic contexts; (ii) incorporate the “best” available COA instruments into clinical trials in a uniform fashion to increase our knowledge base about the performance of these measures; (iii) improve on the “best” existing measures for specific context of use; and (iv) plan/develop next-generation instruments that more accurately assess the priority symptoms, signs, and functions.

  • Step 1: Incorporate the symptom subscale of MDASI-BT and the Clinical Trial Battery into clinical trials for individuals with high-grade gliomas whenever possible. There is the possibility of prespecifying selected domains of interest and using only these aspects of currently available COAs in trials, to avoid overburdening the investigators and thereby supporting improved compliance with collection of complete data.
  • Step 2: Refine current COAs and consolidate symptom PROs into a single, universally used instrument. This can be guided by retrospective data analysis in which current instruments are mapped to priority domains (from Panel 1), and the best of the existing measures are combined into a single instrument.
  • Step 3: Follow the development of, and assist where needed, new COAs such as the NANO, NFBrSI-24, and brain tumor–specific EORTC instrumental ADL and apply the same analysis to these as suggested in Steps 1 and 2 above.

Regulatory Experience with COAs and Its Relevance to Brain Tumor Studies

Rajeshwari Sridhara, PhD, Director of the Division of Biometrics, Office of Biostatistics, Office of Translational Sciences, Center for Drug Evaluation and Research, FDA, presented on drug approval considerations and statistical considerations in designing a clinical trial with COA as endpoint, using examples to illustrate these points. The presentation emphasized that when using a COA, it is necessary that substantial treatment effect be demonstrated for approval or labeling considerations.

COAs are successfully used for evaluation of drug products for marketing approval in many diseases with double-blind randomized control studies, using well-defined COAs that assess domains that are clinically important, anchored with an objectively measurable outcome, measure a limited number of disease symptoms, and use composite endpoints, weighted scores, change of symptom from baseline to a specific time, and inclusion of patients who have symptoms at baseline. Ruxolitinib (Jakafi) for myelofibrosis provides a prototype for the successful use of a COA in an oncology trial because the sponsors had early interactions and communication with the FDA, there was a clear objective to demonstrate symptom improvement, the trial was double blinded, and the instrument focused on important symptoms (modified from 19 to 7 items).⁹ Additionally, adalimumab (Humira), approved in many indications, including rheumatoid arthritis and Crohn's disease, had approval for rheumatoid arthritis based on reducing signs and symptoms as determined by the well-established American College of Rheumatology 20-symptom scoring criteria. Improvement was needed in any 3 of the 5 core scores: patient global assessment, physician global assessment, pain, disability, and an acute phase reactant.²⁸

Creating Clinical Trial Designs that Incorporate Clinical Outcome Assessments (Panel 3)

(Martha Donoghue, MD, US Food and Drug Administration; Mark Gilbert, MD, University of Texas MD Anderson Cancer Center; Remi Kaleta, MD, EMD Serono; Glenn Lesser, MD, Wake Forest University; Lawrence Rubinstein, PhD, National Cancer Institute; Rajeshwari Sridhara, PhD, US Food and Drug Administration)

Panel 3 presentation

Dr Gilbert presented an overview of ideas for clinical trial designs that incorporate COAs to help determine efficacy and impact of brain tumor treatments. Dr Gilbert stressed the need to establish guidelines for incorporation of COAs and their assessment in brain tumor clinical trials and to develop strategies to enhance acceptance, inclusion, and completion compliance of COAs in therapeutic clinical trials. COAs are an important component of treatment evaluation in clinical trials and should be considered as an essential element of important clinical trials. Additional details are summarized in the review of this topic earlier in this supplement (Gilbert et al).

Panel 3 discussion

• COAs provide an important dimension in evaluating outcomes in clinical trials. There is a need to collaborate and build consensus within the community to identify the best clinical trial designs incorporating COAs.

• There is an opportunity to improve on the current process by educating patients and their families, as well as the clinical investigators, on the importance of incorporating COAs into clinical trials and designing the trial to fully incorporate these endpoints.

• The same principles for incorporation of COAs into clinical trials that are used for other efficacy endpoints, including overall survival or progression-free survival, can be applied. COA-related endpoints should be considered for use as—if not co-primary efficacy endpoints—key secondary efficacy endpoints in clinical trials so that the right resources are applied to capture the data consistently and systematically.

• Maintaining or improving patient functional status based on the COA measure during an early evaluation of experimental treatments can be used to support advancement of the drug for patient care and product labeling.

• Incorporation of COAs in early-phase clinical trials prior to phase III should be supported so that previous experience and learning from using the measure(s) to test the intended hypothesis can be utilized to develop the trial design to confirm treatment effects in later-phase registration trials.

• The COAs to be measured in a clinical trial should be clinically meaningful and have specificity for the symptom being measured, be feasible (including minimizing patient burden), and be designed to account for educational and cultural variability of language.

• The same COA instrument(s) should be used uniformly across trials. Standardization of analysis of COA data would improve interpretation across trials and trial centers.

• Past experience in COA use, particularly from clinical trials, should be interrogated to identify which COAs to use.

• Prospective selection of measures that optimally represent the symptoms or the functional attributes of the patient population that is likely to respond to the therapeutic agent will augment evaluation of the primary endpoint. The brain tumor patient population is heterogeneous, particularly as the disease progresses; therefore, some measures may be more important than others at different stages of disease.

• Trial eligibility criteria can aid in ensuring that treatment arms are well balanced with respect to use of instrument measures. Patients can be designated into low- or high-symptom categories based on predefined criteria thresholds of severity, and this can be used for eligibility or stratification.

• There is the potential for false positive claims, particularly with a multiple measures/comparisons model when a large number of COAs are incorporated. Limiting the number of COAs to be evaluated and defining precisely the critical success factors, prospectively, can mitigate this potential issue. A tiered, sequential approach with a single major (potentially composite) outcome, in which a finite list of concepts is measured, may be optimal. If the primary COA were positive, additional analyses would be undertaken subsequently to examine other individual components.

• Implementation of COA endpoints in a trial requires allocation of resources and logistics. Therefore, capital investment will be required to support training (and materials); as well as electronic data capture platforms. The resources required for monitoring site compliance may be needed to reduce missing data (particularly if the COA is an exploratory endpoint). Incorporation of an interim feasibility assessment would determine if the studies are being conducted appropriately before reaching the end of the trial.

• Early discussion between trial sponsors and the FDA about the use of COAs to assess efficacy, safety, and risk/benefit ratio in brain tumor trials may increase the successful implementation of these measures.

Creating an Action Plan (Panel 4)

(Terri S. Armstrong, PhD, ANP-BC, FAANP, FAAN, University of Texas Health Science Center and Texas MD Anderson Cancer Center; Jaishri Blakeley, MD, Johns Hopkins University; Mark Gilbert, MD, University of Texas MD Anderson Cancer Center; Patrick Wen, MD, Dana-Farber Cancer Institute; Patricia Keegan, MD, US Food and Drug Administration)

Panel 4 discussion

• The panel stressed the importance of the need for more effective treatments, the need to keep open communication with the FDA during both trial development and instrument development, especially the Clinical Outcome Assessments Staff (formerly Study Endpoints and Labeling Development), and—of grave importance—to involve patients in the development process.

• Careful COA selection, thoughtful trial design, and complete data, including treatment-related impact on particular symptoms or signs, make for successful implementation.

• Review of each panel session resulted in the creation of action items intended to represent a starting point for future work in the area of COA use in brain tumor clinical trials (Table 5). Based on these action items, the following recommendations emerged:

  1. Form a working group to further refine the list of disease-related symptoms for assessment in clinical trials and prioritize these symptoms as well as the tool(s) for measuring these symptoms. Ideally, this group will utilize existing clinical trial data and existing (MDASI-BT) and emerging (NFBrSI-24) instruments and work toward development of a symptom PRO tool in which the items correspond to the other concepts of interest being measured in the trial.
  2. Address and develop standards for how and when functional tests, both cognitive and physical, are administered as well as how the results are analyzed and interpreted. Neurocognitive testing includes evaluation of the PerfO Clinical Trial Battery. Physical function assessment requires the development of a COA (PerfO, ClinRO, and/or PRO) by building on select components of existing instruments (EORTC QLQ-C30) or creating a new, well-defined, and reliable instrument.
  3. Foster support for the inclusion of COA measures in brain tumor clinical trials by carefully selecting COAs for each trial and emphasizing the importance of capturing COA data points. This should include adaptation and investment in COAs by academia, pharmaceutical companies, governmental agencies, and patient advocates.

Conclusion

The 2014 COA Endpoints Workshop was designed to enhance understanding of the definition and utility of COA endpoints, with a goal of identifying areas the field needs to address to improve these endpoints. The Jumpstarting Brain Tumor Drug Development Coalition will continue to work with the community to create a strategy to advance the above recommendations, in coordination with the outcomes of the first endpoints workshop on imaging. These efforts are aimed at increasing the frequency of COA data used to support neuro-oncology drug approvals and increasing COA data inclusion in neuro-oncology labels.