Introduction

Incorporating the patient’s perspective into cancer research is of widely recognized importance [1]. This would seem to be particularly true for evaluations of oncologic treatment. Nevertheless, primary and key secondary endpoints for clinical trials evaluating oncologic treatment tend to be clinical or clinician-reported such as overall survival, tumor response, response duration, or composite clinical benefit endpoints [2]. However, there exist ongoing efforts to develop patient-reported outcome measures (PROs), such as core symptom assessments and functional measures, designed to serve as co-primary or key secondary endpoints [3]. As these PRO development efforts continue and, presumably, PRO data appear in oncologic product labeling, the question arises as to whether oncologists understand, value, and are able to interpret data from these measures in their evaluation of clinical trial results. Therefore, our objectives were: (1) to explore oncologists’ understanding and attitudes toward PRO measures, and (2) to understand the impact of PRO data on clinical decision-making.

Methods

A cross-sectional, qualitative study utilizing semi-structured focus groups was conducted in the United States. All focus groups were conducted face-to-face, with an experienced interviewer (author SA) moderating each group. The focus groups took place in New York, NY and Chicago, IL: two in each city.

A non-probabilistic purposive sampling strategy was used [4]; oncologists from a large database of oncologists willing to participate in research studies were selectively enrolled to represent a diverse sample in terms of gender, years in practice, practice size, practice type, and types of cancers seen. An a priori sample size of 20 oncologists was determined based on available funding. To be eligible for study inclusion, oncologists must have:

  • Practiced 3–30 years;

  • Spent ≥70 % of their time in direct patient care not involving clinical trials;

  • Seen an average of ≥30 cancer patients per week.

A semi-structured discussion guide was developed with specific questions to explore (1) knowledge of and attitudes toward patient-reported concepts (e.g., symptoms) and measures (e.g., patient versus clinician-reported); and (2) influence of PRO data on clinical decision-making (see Table 2 for example, questions). Each focus group lasted approximately 2 hours and was audio-recorded for analysis.

Audio files were transcribed and then coded and analyzed using Atlas.ti software version 5 [5]. Two experienced qualitative coders (author SA and additional coder) conducted thematic analysis of the transcripts based on key questions from the discussion guide.

Results

Twenty oncologists participated in the study, five in each of the four focus groups. Participant characteristics are presented in Table 1.

Table 1 Oncologist characteristics
Full size table

Objective 1: understanding and attitudes toward patient-reported outcome measures

Oncologists’ understanding of PRO measures

Most oncologists were not familiar with the term “PRO” and had difficulty recognizing that PRO measures collect data directly from the patient (see Table 2 for example quotations). They could list patient-reported concepts relevant to their daily clinical practice (e.g., satisfaction with care or hospital procedures), but most had difficulty envisioning patient-reported endpoints appropriate for clinical trials. Nevertheless, a few oncologists were familiar with specific PRO measures (i.e., Functional Assessment of Cancer Therapy [6]; European Organization for Research and Treatment of Cancer [7]).

Table 2 Discussion guide questions and example oncologist quotations
Full size table

Despite some initial difficulty conceptualizing what a PRO measure is, oncologists identified a number of concepts they considered appropriate for patient-reported assessment: symptoms; physical and mental functioning; health-related quality of life (HRQL); burden of treatment; and satisfaction with care or treatment. They also agreed that patients were the best source of symptom data.

Common symptoms across cancer types

Most oncologists agreed that while different cancers or treatment regimens may be associated with different symptoms, several symptoms were universal across cancer types and, as such, could be measured by one instrument. Specific symptoms mentioned included reduced appetite, depression, fatigue, insomnia, nausea, and pain. The oncologists did note, however, that a wide range of responses would be important, as different cancers and stages of disease are associated with different levels of symptom severity.

Comparison of patient-reported versus clinician-reported measures

All oncologists indicated that both PROs and clinician-reported outcomes (ClinROs) were important, but they were divided as to which type of measure was more valuable. Those who placed higher value on ClinROs reasoned that, whereas patients only have insight into their own health in comparison to how it once was, clinicians could more objectively rate a patient’s health state relative to other cancer patients. They also stated that ClinROs, such as performance status (i.e., Karnofsky [8]; Eastern Cooperative Oncology Group [9]), were ingrained into clinical training and practice and therefore meaningful and intuitive to all oncologists. On the other hand, it was acknowledged that while performance status is useful for communicating a patient’s current state to the oncologist, PRO measures more appropriately assess how patients respond to treatment. Oncologists recommended developing PROs that could become as intuitive as the ClinROs they use in clinical practice.

Objective 2: impact of PRO data on clinical decision-making

Where oncologists are exposed to PRO data

When the oncologists were asked where they encounter PRO information, most indicated that they see this type of data in peer-reviewed journals, not in product labeling. In fact, many oncologists stated that they rarely, if ever, look at product labeling, and if they do, it is only for a new drug. The belief was also expressed that they would deem PRO data in a peer-reviewed journal to be more trustworthy than in product labeling.

The influence of PRO data on clinical decision-making

Most oncologists described a hierarchy of outcome data, in which clinical efficacy (i.e., overall survival, progression free survival) was most important, with toxicity, PRO data, and cost/insurance issues following in importance. All oncologists reported that they would not initially look for PRO data when reviewing product labeling for a new treatment, but assumed robust findings would be presented in peer-reviewed journals. Some oncologists felt that patient-reported side effects and symptoms were closely related to toxicity and should supplement toxicity data. Several oncologists also stated that they would weigh patient-reported symptoms and HRQL more highly than toxicity.

Despite the general hierarchy order of clinical efficacy, toxicity, and PROs, most oncologists stated that many factors contribute to treatment decision-making, and this varies by situation. For example, several oncologists stated that PRO data become more important when multiple treatment options with similar levels of efficacy are available. The importance of PRO data also depended on cancer type (e.g., when prominent symptoms are pain or incontinence). Furthermore, almost all oncologists felt that PROs were of paramount importance in advanced or incurable disease and in palliative care. Oncologists mentioned specific cancers where this was often the case: pancreatic, lung, brain, and head and neck.

Interpretation of PRO data

Most oncologists expressed a general lack of familiarity with PRO measures, which prevented them from drawing meaningful conclusions from PRO data. They also expressed concern that the use of PRO measures was not standardized and that there was an abundance of unvalidated or newly created measures. Additionally, they commented that, because of the volume of data with which they were presented, PRO data presentation needed to be simple, short, easy to read and locate, and relevant to their clinical decision-making. Oncologists noted that in addition to statistical significance, they needed to see the magnitude of difference between treatment groups to determine whether differences were clinically meaningful. The notion of a clear definition of a responder, and the presentation of the proportion of responders in each treatment arm, was considered a positive step.

Discussion

The objective of this research was to gauge oncologists’ knowledge of and attitudes toward PRO measures and to understand to what extent PRO data influence clinical decision-making. The oncologists in this sample demonstrated little familiarity with PRO measures and data. This is despite the fact that previous research [10, 11] has shown that oncologists are familiar with the broader concept of HRQL. To the extent that oncologists were exposed to PRO data, this was primarily through peer-reviewed publications, not product labeling.

While clinical efficacy remains of primary importance to oncologists, the results of this study suggest that PRO data become increasingly important in treatment scenarios when (1) many therapeutic options of similar efficacy exist, (2) in advanced or incurable disease, and (3) in palliative care. However, the findings from this study suggest that many practicing oncologists are still not familiar with and have trouble interpreting PRO data, despite initiatives to address the clinical interpretation of patient-reported data [12–15].

The oncologists in this study suggested that methods such as reporting the proportions of responders and non-responders and the magnitude of effect (e.g., effect size) would increase their ability to interpret PRO data. Additional techniques that may increase interpretability include establishing a minimal clinically important difference or presenting a cumulative distribution function [14, 16].

This was a small, qualitative study, and the results should be interpreted in that context. The implications of our study for clinical trial sponsors are that the value of clinical trial PRO data to clinicians will be based on the extent to which PRO measures are familiar to oncologists and have score interpretations that are intuitive and clinically meaningful.