Background
Rising cancer incidence, advancements in detection and treatment, and improved survival rates have contributed to an unprecedented number of people living with a diagnosis of cancer [
1,
2], a group known as cancer survivors [
3]. An increasing number of survivors are managing cancer (and its sequelae) as a chronic condition [
4] as adverse long-term and late effects related to the disease and treatment can have debilitating and lifelong consequences [
5]. More than 50% of adult cancer survivors experience persistent physical and functional issues that may limit ability to work and increase healthcare utilisation and costs [
3,
5]. Growing interest in quantifying the impacts of illness and treatment on health, functioning, and quality of life from the patient perspective using patient-reported outcomes (PROs) has led to a proliferation in the number of patient-reported outcome measures (PROMs)—self-report questionnaires about a patient’s health—being used in clinical practice and research [
6].
The potential uses of PRO can be understood using Lipscomb’s framework for cancer outcomes research, which proposes three arenas for PRO application: micro, meso, and macro [
7]. At the micro or individual level, PRO data can be used to support and enhance patient-centred care, patient–clinician interactions, and clinical decision-making [
6]. At the meso or service level, PROs are used to examine the variables that influence health outcomes [
7]. PRO data collected at a macro level can facilitate surveillance of the impacts and burden of cancer on population health [
3,
7]. There is increasing recognition that optimising population-based cancer registries to collect longitudinal, macro-level PRO data would improve understanding of trends in survivorship outcomes and their long-term trajectory [
8,
9].
A challenge in PRO research is the lack of standardisation and comparability of scores from different measures [
10]. Core outcome sets provide one way of addressing these problems. A core outcome set is a recommended, standardised, and minimum set of outcomes to be measured and reported in research on a specific health condition, and which should be consensually agreed to by relevant stakeholders [
11,
12]. By standardising the outcomes that are examined across studies, use of core outcome sets can help to minimise potential bias in outcome selection and reporting, and enable data from different studies to be compared and synthesised [
10]. Various cancer-related core outcome sets have been developed, but there is currently no core outcome set to inform assessment of long-term cancer survivorship concerns at a population level. Most existing cancer-related core outcome sets are tumour-specific and were developed to facilitate standardised assessment of acute treatment-related symptoms and toxicities in clinical trials and clinical research.
This study aimed to obtain consensus on a core outcome set, recommending what PROs should be collected as a minimum for population-based surveillance of cancer survivorship in Australia, and how these outcomes should be measured. The need for a set of outcomes applicable to all cancers is underscored by research identifying issues common across cancers including psychosocial outcomes, fatigue, functional impairment, fear of cancer recurrence, and limitations in healthcare and insurance access [
13,
14].
There have been two previous efforts to standardise assessment of health outcomes across general cancer populations, which differ from this core outcome set in scope and end use. The National Cancer Institute (NCI) identified a core set of symptoms experienced by patients with cancer undergoing treatment [
15], to inform standardised assessment of PROs in adult cancer treatment trials. In the Netherlands, Geerse and colleagues [
16] defined a core set of categories from the International Classification for Disability, Functioning and Health (ICF) representing health-related problems experienced by adult cancer survivors more than a year from diagnosis, for an unspecified end use. Our study is the first to identify a core set of PROs relevant in long-term cancer survivorship for the purpose of routine assessment at a population level, and was developed in Australia.
Major methodological differences also separate the three studies. The initial outcome list for the NCI core set was derived from a systematic review and analysis of datasets [
15], while outcomes in the study by Geerse and colleagues were derived from the ICF [
16]. To obtain consensus on the NCI core set, a meeting with stakeholders was held [
15]. The characteristics and expertise of the stakeholders were not reported. Geerse and colleagues conducted a Delphi study with stakeholder panels for breast, lung, and colorectal cancers to reach consensus on the most important ICF categories for survivors of these cancers in the Netherlands, then undertook a procedure to link the identified categories to existing PROMs [
16]. Our approach uniquely involved consumers from the outset and identified an initial list of outcomes from focus group discussions with cancer survivors. We additionally examined cancer survivorship PROMs to ensure coverage of all outcomes assessed by validated measures. To ensure fit with the literature in terms of how PROMs are described and conceptualised, all domains in our study were classified using a PROMs taxonomy [
17]. Finally, to ensure rigour, we determined consensus on the core outcomes by using comprehensive and recommended Delphi methods that were planned and published a priori [
18].
Discussion
This study used robust methods to generate a core outcome set for national surveillance of the long-term quality of life of cancer survivors in Australia. After two survey rounds and a group consensus meeting with national experts in cancer survivorship practice, research, policy, advocacy, and lived experience, 12 core domains were agreed upon. These were depression, anxiety, pain, fatigue, cognitive problems, fear of cancer recurrence or progression, functioning in everyday activities and roles, financial toxicity, coping with cancer, overall bother from side effects, overall quality of life, and overall health status.
The prioritisation of outcomes in the consensus meeting was driven by several key considerations that emerged as themes in the discussion. Firstly, it was considered important to capture outcomes not well represented in other cancer datasets and with a relatively unknown prevalence at the population level. Secondly, feasibility for the intended application of population surveillance was central to the discussion and prompted considerations regarding the availability, potential overlap, and combined length of prospective measures. Ensuring clear operational definitions for each domain and including broad summary items (e.g., overall quality of life), which could add useful information with little added respondent burden, were agreed avenues for optimising feasibility. Finally, the inclusion of positive and negative items in the core outcome set was deemed important to inform understanding of not only the impacts of cancer, but individual factors that may influence outcomes.
Our core set and the core set developed by the NCI [
15] share 5 outcome domains in common: pain, fatigue, anxiety, depression, and cognitive problems. Similar outcomes are also represented in the core set developed by Geerse and colleagues [
16], including sensation of pain, energy and drive functions, attention functions, and emotional functions. This commonality reflects the broader literature which indicates that pain, fatigue, cognitive limitations, anxiety, and depressive symptoms are consistently present in cancer survivors following primary treatment [
13]. In addition to these, the present core outcome set includes fear of cancer recurrence or progression, functioning in usual activities and roles, financial toxicity, coping with cancer, and overall indicators of health status, quality of life, and bother from side effects. The prioritisation of these domains is consistent with literature highlighting the prevalence of issues related to daily functioning [
29], fear of cancer recurrence [
30], and financial toxicity (due to mounting medical expenses, lost wages, and reduced productivity) among cancer survivors [
31]. The NCI set captures more treatment-related symptoms (e.g., diarrhoea, constipation, dyspnoea, sensory neuropathy) [
15], while the core set developed by Geerse and colleagues includes several environmental factors (e.g., friends; immediate family; health professionals; social security series, systems, and policies) and more specific functional domains (e.g., reading, driving, sexual functions) [
16].
Defining and operationalising multidimensional and subjective constructs were challenging aspects of this exercise, particularly when supporting evidence from the PROMs literature was limited or discordant. For this reason, the panel could not determine consensus on a domain representing the impact of cancer on relationships and families, although this was evidently an important outcome. The issue with this domain was how broadly it could be interpreted and how multifaceted and multidirectional the impacts of cancer on families and relationships could be. This is apparent from the limited survivorship PROMs that assess wide-ranging aspects of this domain, such as marital communication, affection, neglect, and overprotection; parental concerns about both the practical and emotional impacts of cancer on the child; parental efficacy beliefs; support from and communication with close relatives; perceived impacts of cancer on caring for family members, attending family events, time spent with family, and plans to have a family; and interference of illness with family relationships and couple intimacy [
32]. The impact of cancer on family roles, responsibilities, and relationships, and the measurement of these constructs, are areas requiring further research.
The challenge of quantifying the impacts of cancer on relationships was also highlighted in the core outcome set developed by Geerse and colleagues [
16], which included the two related domains ‘basic interpersonal interactions’ and ‘complex interpersonal interactions’. During the process of matching their core outcome domains to questionnaire items from existing PROMs, the authors linked the ‘basic interpersonal interactions’ domain to three items on the Quality of Life in Adult Cancer Survivors (QLACS) [
33] questionnaire and one item on the Impact of Cancer Version 2 (IOCv2) [
34] questionnaire; and the ‘complex interpersonal interactions’ domain to two items on QLACs, 8 items on IOCv2, and three items on the Distress Thermometer and Problems List (DT/PL) [
35]. Although the questionnaire items are not specified, it is evident from examining the measures that the aspects of interactions they assess differ. QLACS contains a ‘social avoidance’ subscale that assesses reluctance to start new relationships and meet new people, and avoidance of social gatherings and friends, as well as a ‘family distress’ domain that assesses worry about family members being at risk for cancer, having cancer-causing genes, or having to undergo genetic testing for cancer [
33]. IOCv2 contains a ‘life interferences’ subscale that captures guilt for not being available to family, feeling alone, feeling misunderstood, and the interference of cancer with usual activities, as well as a ‘relationship concerns’ subscale that assesses self and partner openness and willingness to discuss cancer, relationship problems due to health uncertainty, and worry about being left by partner if cancer was to return [
34]. DT/PL contains a ‘family problems’ scale that assesses problems related to dealing with children, dealing with partner, ability to have children, and family health issues [
35]. Application of the core outcome set may therefore result in conceptually different constructs being assessed depending on the measure(s) used. This example illustrates why further exploration and conceptual clarification of the interpersonal impacts of cancer and their measurement are required.
We based the study methods on recommendations for evidence-based core outcome set development [
25], guidelines for using the Delphi technique to obtain consensus on core outcomes [
36], standards for core outcome set study design [
22], and previous core outcome set studies with cancer populations that employed Delphi methods [
27,
37]. However, there is no agreed-upon methodology for developing a core outcome set. It is therefore unclear to what extent the results from this study would be concordant with those obtained in different settings, using alternate consensus methods, or applying different criteria. Despite these limitations, the study design was considered suitable for the scope and setting of this core outcome set and allowed a large and geographically diverse sample of stakeholders to participate.
Although there is no consensus on the optimal number of Delphi rounds, two or three rounds have been frequently recommended [
38] and commonly used in core outcome set development studies [
25]. Given that we undertook a rigorous process to identify and refine the initial list of outcomes for the Delphi study and included a consensus meeting, two questionnaire rounds were considered sufficient. An advantage of restricting the number of rounds is limiting potential bias due to attrition, which is likely to increase with each round [
36]. A limitation of restricting the study to two questionnaire rounds is that it was not possible to confirm stability of voting, although this is generally thought to be a measure of internal reliability and not consensus [
39]. Instead, we measured the extent to which participants agreed with the statements under consideration (agreement) and the extent to which participants agreed with each other (consensus) [
40]. It is not possible to determine the validity of any specific definition of consensus in Delphi studies, but the proportion of ratings within a range is one of the most commonly employed consensus definitions and the median is considered the most robust measure of central tendency [
40].
Two minor deviations from our study protocol were made for pragmatic reasons. Firstly, because the focus group discussions generated an extremely comprehensive list of outcomes, we elected to conduct a rapid and targeted search for existing review evidence rather than an exhaustive literature review of cancer survivorship PROMs. While it is therefore possible that not all relevant PROMs were identified, the extensive overlap between the outcomes derived from the focus groups, PROMs, and guiding PRO framework suggests that this would not have resulted in any relevant outcome domains being missed. This risk was further mitigated by providing Delphi participants with the opportunity to suggest additional outcomes. Secondly, although we planned to have three stakeholder panels participate in the Delphi survey (cancer survivors, clinicians, and other healthcare and research professionals) and invited a proportionate number of experts accordingly, response was significantly higher among cancer survivors than the other invited groups. Since cancer survivors comprised half of our recruited sample in round 1, we made the decision to group participants into two equally sized panels rather than three. This made it easier to compare results by group and did not compromise the study findings beyond giving additional weight to the views of consumers.
Further work is required to determine how the core outcome set will be assessed and applied in practice. In phase III, valid and reliable measures for assessment of the core outcome domains will be identified, appraised, and selected with expert advice and input from end users [
18]. The final core measurement set will be piloted to ensure its acceptability to cancer survivors and feasibility for routine collection at a population level. As noted during the consensus meeting, further refinement of the core outcome set may occur during this phase of the research and over time as we learn more about the overlap among measures and usefulness of individual concepts.
Publisher’s note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.