Journal of the National Cancer Institute Advance Access originally published online on February 12, 2008
JNCI Journal of the National Cancer Institute 2008 100(4):234-236; doi:10.1093/jnci/djn031
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
© Oxford University Press 2008.
NEWS |
Quality of Life Researchers Have New Tool and New Focus on Measurement
More attention to quality of life (QOL) has been on the agenda of oncology groups for at least three decades. Most researchers agree that measuring QOL endpoints, such as pain or physical function, is important because they can affect adherence to treatment and outcome. QOL also matters because of increasing survival times among cancer patients.
"If you're looking at 5- to 10-year survival times, it makes sense to be interested in quality of life," said Lori Minasian, M.D., who leads the National Cancer Institute Community Oncology and Preventive Trials Research Group.
|
But using QOL endpoints in clinical trials has been challenging. Problems include the extra burden on patients and staff and the related issue of missing data—many people simply do not fill out QOL surveys.
But the major challenge has been the difficulty of measuring these patient-reported outcomes (PROs)—the umbrella term now used almost interchangeably with QOL endpoints. PRO experts worry that existing measurement tools have not been tested enough in specific patient populations. Trial investigators worry that PRO questionnaires do not reliably capture improvements in QOL. Moreover, trials often use different instruments, so comparing results across studies has been difficult.
Now, a team of researchers has designed a new tool to address these concerns. The PRO Measurement Information System, called PROMIS (http://www.nihpromis.org), lets investigators choose from banks of validated, evidence-based questions to measure standard areas of QOL—pain, fatigue, physical and social functioning, and emotional distress. Four categories relevant to cancer are also in development.
In addition, the U.S. Food and Drug Administration has addressed measurement and methodology issues in detail in a new document that describes how it evaluates studies that include PRO endpoints (see sidebar).
Emphasis on the evidence is a key feature of PROMIS, said David Cella, Ph.D., of Northwestern University in Chicago, a principal investigator on the 5-year project. Each item—a question and its associated response options—has undergone extensive qualitative and quantitative testing with diverse patient groups to ensure that it is a valid measure of a specific symptom or other concept and that it can detect differences and changes in the concept.
"PROMIS items give ultraprecise measurements across diseases," Cella said. "The detailed information behind each item, viewable online, will let researchers select the optimal set of questions to match their needs."
The first version of PROMIS, which went online in November, allows investigators to select a "short form" with about five to 10 items on a specific aspect of QOL. By March, clinical trial researchers will be able to create their own instruments by choosing among items tailored to specific populations.
Also available with PROMIS 1.0 are computerized adaptive testing (CAT) versions of the short forms. CAT selects each question on the basis of the respondent's answers to previous questions. Does the patient report feeling sad most of the time? CAT will follow up with more questions on depression. If the patient reports never feeling sad, CAT will skip to another question.
The effect of both the standard and CAT short forms should be to minimize the need for patients to answer lengthy questionnaires, one of the barriers to QOL measurement, according to Bryce Reeve, Ph.D., a statistician at NCI and the National Institutes of Health science officer for PROMIS. "Our goal was to provide researchers access to well-validated, precise, and standardized measures of PROs with minimal response burden," he said.
Moreover, PRO data collected from any short form or CAT can be compared or combined within the same PRO categories, even when the responders receive different sets of questions, Reeve said.
By March, short forms and CAT protocols for all the standard categories and subcategories will be accessible. These include pain and its effect on functioning and coping behaviors; physical function, including lower-, central-, and upper-body mobility; fatigue and its effect on life; social well-being, including satisfaction with social roles and participation in discretionary activities; and emotional distress, encompassing depression, anxiety, and anger.
PROMIS will eventually include four more categories relevant to cancer—sexual function, perceived cognitive function, sleep, and the negative and positive effect of illness on areas like stress response and coping, shifts in self-concept, social interactions, and spirituality. These NCI-funded additions should be available in fall 2008, Reeve said.
Evidence Base
All the PROMIS items have been developed through a multistep process. The researchers first conducted literature reviews and collected items from existing PRO instruments, winnowing out those that were repetitive and standardizing the response options.
PROMIS researchers then conducted qualitative studies with patients to ensure each item's content validity, i.e., that it actually measures the concept intended. For example, is the statement "I found it hard to focus on anything other than my anxiety [in the past 7 days]" an accurate indicator of anxiety? Both focus groups and cognitive interviews with patients were used to address these questions and to ensure that the items were comprehensible to people of various reading levels and cultural backgrounds. The items also underwent several rounds of field testing with a large representative sample of the U.S. population, including people with cancer and other diseases.
The researchers also used statistical, or psychometric, methods to look at the relationships among items and whether each set of items captured the depth and breadth of the concept, Reeve said. Does asking about being too tired to take a shower measure extreme fatigue more accurately than a question about being too tired to walk around the block? Is the second question more precise for measuring low fatigue levels? Statistical analysis of how items were answered in field tests helped determine such relationships.
To ensure the questions' relevance to oncology, they were tested in 2,000 cancer patients with different tumor types and in different stages of treatment. The questions in the four additional cancer-specific categories are now undergoing the same process. Once PROMIS items are all online, they will be further validated over time in clinical settings.
Years of Discussion
PROMIS addresses issues that have been on the minds of QOL experts for some time. In 2006, the FDA issued draft advice for industry describing how the agency looks at PROs in clinical trials when the results are intended to support drug approval and labeling. The final document is due out soon, said Laurie Burke, director of the FDA's study endpoints and label development office.
Years of discussion lie behind the current focus on PROs, Burke said, beginning with the acceptance of patient-reported data. "It was typical before to assess symptoms through the eyes of the providers," she said. "Now we're realizing we get more well-defined and reliable information from patients about the things they are experiencing."
Also, the broad term QOL has given way to the narrower and more easily measured concept of health-related quality of life (HRQOL). QOL refers to every aspect of a person's life, including those that may have nothing to do with a disease or its treatment. HRQOL instruments aim to measure only the narrowly defined effect of an illness and its treatment.
The first validated HRQOL instrument in oncology, FACT-G (Functional Assessment of Cancer Therapy—General), appeared in the early 1990s. It was followed by FACT subscales for specific tumors and symptoms. FACT instruments are multidimensional—they combine scores on symptoms and functioning, as well as social factors, to come up with an HRQOL score, which indicates the overall effect of the disease and treatment.
Like two other multidimensional HRQOL instruments—the EORTC Quality of Life Questionnaire C30 and the Medical Outcomes Survey SF-36—FACT is widely used in cancer trials. Recently, however, some experts have begun to question using one number to capture the multidimensional concept of HRQOL, arguing that it is too broad a concept to be useful.
But the FDA, which emphasizes measuring changes in individual symptoms or functions, rather than the broader concept of HRQOL, may force change. PROMIS would meet these criteria.
Two recent publications summarize much of the current discussion about PROs: a JNCI Monograph (No. 37, 2007) and a November special issue of the Journal of Clinical Oncology. Articles in both publications discuss the pros and cons of multidimensional HRQOL measurement versus unidimensional measurements of symptoms or functioning, including the FDA perspective.
"We have not ruled out HRQOL claims," Burke said. "But the problem is it's so general that the measures must be sufficiently broad to make sure you haven't missed anything. You have to integrate all risks and benefits, and you don't have a complete picture of that before doing the trial. It's very difficult, very challenging to have the right measure of HRQOL and make sure it actually is capturing all domains."
Does this mean that cancer researchers will abandon HRQOL-based instruments? "No, not at all," said the NCI's Minasian, "HRQOL gives us the opportunity to explore what symptom reductions mean to the patient overall. These instruments let us tease out the connections between symptoms and between symptoms and function," she said. "For instance, what is it about mucositis that detracts from HRQOL? Do the side effects of a pain medication outweigh its benefits?"
On the other hand, in NCI-sponsored phase III trials leading to drug submission to the FDA, the PRO components will be designed to meet the FDA criteria, Minasian said. "When NCI sponsors a trial that incorporates PROs for specific drug labeling, then the trial needs to be in compliance with the FDA guidance," she said.
But otherwise, the institute will use a different and broader set of parameters when deciding whether to approve trials with PRO components. "NCI funds a broad portfolio of clinical studies and clinical trials, many of which do not go to the FDA for specific drug approval," she noted.
Over the next few years, this debate may continue, fueled by both the FDA document and PROMIS-related activities. Three to five clinical evaluation studies will correlate PROMIS questions with established valid questionnaires in the field, like FACT, PROMIS investigator Cella said. In cancer trials, several of NCI's cooperative trial groups are beginning to plan protocols that will use FACT and PROMIS items side by side, Cella said.
NCI is also studying the barriers to PRO use in cooperative group trials to address why measuring HRQOL has not been a priority. It is also funding studies to promote use of HRQOL measurements in practice settings, outside trials.
The PROMIS group has other plans as well. "We will also begin to look at different ways of administering the questions, comparing paper-and-pencil and computerized versions. We want to refine the computerized testing component and eventually make it downloadable," Cella said. "And we want to make PROMIS usable in different cultures and languages. There's a tremendous amount of work left to do."
FDA's New Criteria Stress Single Measures, Documentation
The U.S. Food and Drug Administration is set to issue its final advice on how to incorporate patient-reported outcomes in clinical trials, according to Laurie Burke, director of the FDA's study endpoints and label development office. The FDA's rationale behind the report, PRO Measures: Use in Medical Product Development to Support Claims, is set out in the November–December supplement to the journal Values in Health.
The final document makes five major points, Burke said.
- Start with the claim. When planning the PRO component of a trial, determine what outcome will be on the drug label and how this PRO fits in with other endpoints. Integrate the PRO protocol into the initial overall plan for the trial.
- Establish the conceptual framework for the PRO. Link the desired outcome (e.g., more energy) to the concept you want to measure (fatigue). Then link that concept to the specific symptoms being measured (difficulty getting out of bed, ability to walk around the block).
- Create a PRO instrument with plenty of patient input. Use focus groups and cognitive interviews to establish the questions' validity, ensuring that the instrument covers what patients consider important outcomes and that it is a valid measure of those outcomes. Use psychometric methods to test reliability and responsiveness of the questions. The FDA will review the evidence of the instrument's validity and other properties.
- If using an existing instrument, consider modifications to ensure that it is valid in the population being studied. The FDA will decide whether existing documentation of validity is sufficient on a case-by-case basis, Burke said. The final document will offer examples of different scenarios and the amount of validation required for each.
- Determine how much of a response should be meaningful. Define what amount of change will be considered a response. Use cumulative distribution curves to compare patients in each group of a randomized trial who meet the definition.
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||