Introduction
Pancreatic cancer (PC) is the seventh most frequent cause of cancer deaths worldwide [
1,
2]. Data published from the International Agency for Research on Cancer suggest that the mortality rate is 96 % of the incidence rate [
3]. The short time from diagnosis to death in most patients makes research challenging due to limited disease-specific follow-up before transition into a general palliative stage. Frequently used outcomes in clinical and epidemiological research include survival, assessment of pancreatic tumors and metastases, soluble biomarkers, symptoms, function, and health-related quality of life. PC presents with symptoms that are most inflicting to patients. Thus, there is a need to measure, document, and monitor the level of health status in dimensions of high relevance and importance to patients in a feasible manner.
Patient-reported outcomes (PROs) have become increasingly emphasized in both clinical research and practice over the last decades and include generic and disease-specific measures [
4,
5], as well as tools to measure utilities [
6]. The Food and Drug Administration (FDA) and the European Medicines Agency recommend that pharmaceutical industry researchers provide sufficient evidence supporting benefit on PROs before granting approval of a new therapeutic agent [
7,
8].
The use of PROs is based on the idea that the patient is the best source of information about the experience of their own condition. This is particularly important when the primary goal is to improve health-related quality of life [
9]. However, there are some general concerns related to the feasibility of PROs. The questionnaires are often extensive which may limit the number of patients who provide their response, especially for longitudinal follow-up [
10]. The lack of relevance of items to patients may also negatively influence the respondent rates.
The implication is that patients also are the best source of information to identify the questions of importance and how they should be presented. In rheumatology, the Rheumatoid Arthritis Impact of Disease (RAID) score, based on patients’ selection of dimensions where the disease has the most important impact, has been developed and validated [
11,
12]. The RAID score has also been used to identify cut points for acceptable symptom state, as perceived by patients [
13], as well as evaluating onset of response to medication [
14].
A number of self-reported measures have been used in research in patients with PC, but until now, only the European Organisation for Research and Treatment in Cancer (EORTC) has developed a disease-specific instrument for PC [
5]. The Quality of Life module for pancreatic cancer (EORTC QLQ-PAN26) has 26 questions and is meant for use in conjunction with the EORTC Quality of Life Questionnaire-C30 (EORTC C-30) which is generic for patients with malignant disease. Eventually, there are 56 questions to complete, which limits the feasibility of the instrument both in research and clinical practice, especially in patients with a severe disease [
10].
The NIH PROMIS initiative [
15] includes a number of items with multiple follow-up questions, which provide detailed and relevant information. Each question has Likert scale response options. PROMIS is a dynamic, promising tool to measure health outcomes, but it has to our knowledge not been validated in PC.
The objectives of this study were to develop a brief, patient-derived, disease-specific tool—pancreatic cancer disease impact (PACADI) score—and to perform a preliminary validation.
Discussion
The aim of this study was to develop a tool to facilitate a seamless assessment and monitoring of PROs in patients with PC in a research setting and in clinical practice. Clinicians have their perspective on collecting information for decision making. The patient perspective is one out of four key issues for decision making in evidence-based medicine [
21]. The PACADI score may help clinicians to identify and monitor dimensions and symptoms that are highly relevant to patients. The score is a global measure that has the potential to be used in both clinical and research settings.
The decision to limit the inclusion of dimensions of health that were reported by more than 20 % of the patients as candidate items for the PACADI was based on the important goal to obtain a feasible score. This aspect is especially important in patients feeling unwell. The number of dimensions was similar to the RAID score (eight versus seven).
The idea behind the longitudinal assessment was to examine whether the selection of domains was stable over time. Overall, the selection was stable but with a reduced number of patients selecting anxiety and itchiness during follow-up (Fig.
1). The reduced focus on anxiety may be related to the adaptation to the disease after proper information and that reduced reporting of itchiness may be explained by procedures like stenting [
22]. However, both these dimensions remained in the PACADI score since they obviously are important in the early phases of the disease, which may be the phases of largest interest when considering current and future therapeutic options.
It was expected that some patients might have problems understanding what was meant by dimensions of health. Therefore, the probe list of 56 dimensions from frequently used questionnaires (Table S
2) was prepared and offered at baseline to exemplify dimensions of health. About half of the selected dimensions for PACADI were in the probe list; this list was mostly used for the initial illustration of examples at a glance. The use of this probe list could potentially have been a limitation to the study by influencing the selection. During follow-up, the probe list was available, but it was rarely used by the patients. The selected dimensions were overall similar at all three time points, even if the patients did not have access to the previous selections (Fig.
1).
An alternative approach to the model of ranks of the weights of each dimension could had been to use the mean weights to construct the final weight (Table
3). However, the weighting accentuates the patient perspective of importance. The aim was to reach a number of patients, which was ten times the number of dimensions [
23]. Even so, only 80 patients from one single center is considered a limitation, since the weighting procedure of the RAID score included more than 500 patients across Europe. The intention for the development of the PACADI instrument was to make it disease-specific, and hence, only patients with PC were included in the primary analyses. However, results were similar both for selection of dimensions and weights when both groups of patients were analyzed (Tables S
1 and S
4).
A particular strength of the PACADI score is that both selection of dimensions (i.e., what to measure) and the weighting procedure (i.e., how important is the dimension to the patients) are based on the patients’ opinion. Inherent in this design is the relevance of the tool to patients.
Several of NIH PROMIS items may be of relevance to patients with pancreatic cancer. However, it seems that only pain, fatigue, and anxiety are covered in both tools, since PACADI reflects only the most important impact on health. While PROMIS is measured on Likert scales, PACADI uses NRS. NRS and Likert scales are often used interchangeably. Likert scales seem to be more frequently used to measure strength of agreement in attitude, whereas NRS seem to be more frequently used for measuring severity/intensity, which suits PACADI. Electronic versions of both tools would increase both feasibility of completing and errors in datasets [
24].
It seems that the three approaches related to the eight selected dimensions of health (selection and giving priority between dimensions of importance, weights, and symptom severity) gave rather consistent results on a group level. For example, the five dimensions reported as important by more than 39 % of the patients (pain/discomfort, fatigue, bowel and/or digestive problems, loss of appetite, and anxiety; Table
1) also had the highest weights (Table
3) and the worst scores on NRS (Table
4). However, the data also indicated that patients may differentiate between severity (the value on the NRS) and the importance (the frequency of reporting dimensions of importance to patients). Even if fatigue is reported as more severe than pain/discomfort on NRS (Table
3), pain/discomfort is still the most important dimension to patients (Table
1). This observation implies that asking patients “what is important to you?” may give complimentary responses to asking patients “how are you?” A similar differentiation between severity and priority in patients with rheumatoid arthritis has previously been reported [
25].
In sample 2, the PACADI profile was tested (
n = 80) in a preliminary validation. The moderate to strong correlations to the global measure of well-being in ESAS and EQ-5D supported that the PACADI score may perform as a global PRO in patients with PC. Further, the weighted PACADI score showed a slightly higher value than the mean of the NRS profile (Table
4), which may support that the inclusion of the weights may accentuate the patient perspective. Interestingly, the PACADI scores differentiated between patients with and without confirmed PC, which may support its role as a disease-specific tool, since a similar differentiation was not seen for ESAS well-being and EQ-5D.
The PACADI score is a brief, patient-derived, disease-specific measure. Preliminary validation regarding construct validity and discrimination encourages further validation in independent patient samples. It has the potential similar to RAID [
13,
14] to be used in both clinical and research settings, but additional testing of reliability and a larger scale validation in a population of at least 200 patients according to requirements from the FDA [
26] should be performed. Sensitivity to change should be examined in intervention studies, for example, by longitudinal comparisons of patients undergoing versus not undergoing surgery [
27]. It is also important that further validation is performed in an international setting.