Patient choice in colorectal cancer treatment – A systematic review and narrative synthesis of attribute‐based stated preference studies

Abstract Aim The global burden of colorectal cancer (CRC) is set to increase by 60% by 2030. An aging population and increasing treatment complexity add difficulties for patients and clinicians in CRC management. Patient preferences can be investigated using attribute‐based stated preference (AbSP) techniques to explore trade‐offs between different treatments. These techniques include discrete‐choice experiments (DCEs), conjoint analysis and time‐trade off (TTO) methods. This systematic review with a narrative synthesis aimed to determine the use and design of AbSP studies in CRC treatment and to identify patient choice themes. Methods The searches were performed using MEDLINE, Embase, PsycInfo and Cochrane Library in March 2021. All manuscripts featuring the use of AbSP techniques in CRC treatment were included. Data synthesis was performed using a narrative approach. Results The search strategy returned 271 articles. Eighteen AbSP studies were included featuring 1890 patients and 296 clinicians. AbSP techniques compromised DCE (38.9%, n = 7), TTO (38.9%, n = 7) and conjoint analysis (22.2%, n = 4). Eleven studies (61.1%) involved piloting of tasks and the average task completion rate was 75%. CRC treatments included chemotherapy (33%, n = 6), combined treatments (33%, n = 6), surgery (17%, n = 3), targeted therapy (11%, n = 2) and radiotherapy (6%, n = 1). The most examined domain was physical health, investigated with 49 (59.8%) attributes. Conclusions Life expectancy was the main attribute in chemotherapy treatment. With surgery, patients were willing to trade life‐expectancy to avoid adverse outcomes or a permanent stoma. Communication skills, treatment cost, and clinicians' views were important attributes for patients in cancer services. Further research in the elderly population, and other quality of life domains, are needed to deliver patient‐centred CRC care.


INTRODUC TI ON
Colorectal cancer (CRC) involves 11% of cancer diagnosis globally and is the third most common cancer in the world [1]. Novel treatment modalities, such as targeted therapies, in addition to the classic approaches of surgery, radiotherapy and chemotherapy, have added to the complexity of CRC management [2]. An aging population has resulted in almost half of CRC cases occurring in patients over 75 years, demanding careful patient selection for each treatment modality [3,4]. With the global burden of CRC expected to increase by 60% by 2030, a clear understanding of patient choice is required to ensure that management is aligned to patients' expectations [5].
Patient-centred care has been adopted as the preferred approach by healthcare systems and in the UK is a key high-quality care indicator [6]. This model places the patient at the centre of the decision-making process. In CRC management, patients are often required to make difficult choices. The choice they make is dependent upon their preferences and how they weigh-up different aspects of treatment and trade-off certain attributes. For example, an adult patient with moderate frailty might have to decide between a right hemicolectomy and potential cure, but with the risk of postoperative morbidity, against conservative management with inevitable cancer progression, but without immediate impact on quality of life.
To investigate preferences that patients have towards CRC treatments, attribute-based stated preference (AbSP) techniques can be utilised to explore trade-offs between different treatments. These techniques include discrete-choice experiments (DCEs), conjoint analysis, best-worst scaling (BWS) studies, and time-trade off (TTO) methods. DCEs and conjoint analyses present respondents with a series of choices between two or more (treatment) options, each of which is described in terms of attributes (which may be outcomes and risks). The attributes are split into levels, often describing the severity of risk within healthcare research. Respondents weigh up the pros and cons of each option and choose which, on balance, they feel offers them the greatest value [7,8]. The results are used to calculate the significance of each attribute and can be used for economical estimates for the willingness to pay for an attribute unit change. BWS studies establish respondents' relative preference for treatment and service attributes by asking them to rank and rate aspects or state which aspects are the worst and best [9]. The responses enable an exploration of how individuals value different characteristics of a service or treatment and if they are willing to trade between those characteristics. The TTO method focuses on F I G U R E 1 PRISMA flow diagram for study selection. Adapted from [11] Reports assessed for eligibility (n = 5)

Study design and participants
The protocol for this systematic review was guided by the PRISMA and AMSTAR 2 guidelines and was prospectively registered with PROSPERO (registration no. CRD42021245077) [11,12]. identified studies were reviewed against the inclusion and exclusion criteria to assess eligibility. Referenced studies within identified literature were accessed and considered for inclusion.
The systematic search screening was performed by two independent investigations (MK and FD) using the databases described.
All studies published up until 29 March 2021 inclusive were considered for eligibility. Studies identified were analysed for relevance to the systematic review prior to full inspection. Any discrepancies between the independent investigators were addressed by a third investigator (DM) until consensus was achieved. The search strategies used are displayed in full in Appendix S1.

Primary and secondary outcomes
The primary outcome was the use of AbSP studies in CRC treatment. This was investigated through a focus on study design, with specific reference to study type, patient inclusion criteria, survey design, statistical design, sample size and analysis methods. The checklist created by the conjoint analysis task force of the International Society for Pharmacoeconomics and Outcomes Research (ISPOR) was used to standardise methodology assessment [13]. The ISPOR checklist is a 40-item guideline reviewing general and specific aspects of AbSP studies such as attribute and level setting and preference elicitation. It is designed to highlight good research practises within the field of AbSP studies. The feasibility of preference elicitation in this study group was assessed as evidenced by completion rates, missing data and assessments of validity.
The secondary outcomes focused on the attributes identified collectively by all AbSP studies. These were extracted from each identified study and categorised according to chemotherapy, combination treatments, surgery, targeted therapies and radiotherapy.

Data extraction
Two independent investigators (MK & FD) extracted data using a standardised data collection proforma, which included the following data fields:

Data synthesis
Due to study heterogeneity and the nature of research examined, a narrative synthesis approach was chosen to analyse a wide range of studies in a meaningful manner. We did not perform a meta-analysis of effect estimates. The synthesis was conducted in line with the Guidance on the Conduct of Narrative Synthesis in Systematic Reviews from the Economic and Social Research Council [14].
Included studies were tabulated and grouped according to chemotherapy, combination treatments, surgery, targeted therapies and radiotherapy. Attributes investigated by individual studies were categorised using the World Health Organisation (WHO) domains of health-related quality of life (HRQoL) [15]. The quality assessment was examined for each study using the ISPOR checklist. The evidence provided by the included literature was then synthesised to provide a structured narrative that was relevant to the research question. Due to study heterogeneity, the application of a standard scoring tool for risk of bias assessment was not possible.

Included and excluded studies
The search strategy returned 271 articles after duplication removal.  Table 1.

Feasibility of preference elicitation
We observed an average AbSP task completion rate of 75% (range 43%-97%) across all studies. Difficulties with data collection were uncommon, with one case by Pieterse et al. [18] of missing recording data due to logistical reasons. Validity was formally assessed and presented in five studies (28%) with the use of an additional control task. Out of these, three studies (17%) declared the use of test retest reliability calculated using paired t-tests.

Patient preferences in colorectal cancer treatment
The most investigated CRC treatments were chemotherapy and combinations of treatments in six studies (33%), with the latter involving a mixture of surgical and chemoradiation therapy. Three studies focused on surgery alone (17%), two investigated targeted therapy such as anti-vascular endothelial growth factor (VEGF) agents (11%), and one examined the use of standalone radiotherapy (6%). The full results of patient preferences are displayed in Table 2.
In chemotherapy treatments, life expectancy was the most important attribute when choosing treatments in most studies (4 out of 6 studies specifically when examining personal factors such as age and gender. We have only included one AbSP study which focused on different age groups; thus we identify this as an area for further research in the field [27]. A further systematic review by Currie et al. [28].
identified eight studies concerning patient preferences in CRC. We support their conclusions in patients' willingness to trade life expectancy for adjuvant therapy and stoma avoidance. In addition to this, we identify the disparity with clinician's views, which is likely to impact on patient decision-making and their satisfaction with outcomes and personal choices.
With regards to future research, we have identified a need for further analysis of patient preferences related to surgical interventions, with specific reference to elderly patients (age over 65).
The average age across all included studies was 62 and no study focused on an inclusion criterion of age or frailty index score to capture this patient group. In addition to this, we recommend future research to involve a comprehensive assessment of HRQoL.

ACK N OWLED G EM ENTS
We thank Judy Wright, senior information specialist from the University of Leeds, for her advice and support with formulating the literature search.

FU N D I N G I N FO R M ATI O N
The systematic review described in this study received no funding.

E TH I C A L A PPROVA L
No ethical apporval was required for this study.

CO N FLI C T O F I NTE R E S T
None declared.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing is not applicable to this article as no new data were created or analyzed in this study.