Health economic evidence for adjuvant chemotherapy in stage II and III colon cancer: a systematic review

A wide range of adjuvant (AC) strategies are currently employed in stage II and III colon cancers (CC) following surgical resection. Single-agent regimens include intravenous 5-fluorouracil (5FU) or oral equivalents such as capecitabine or tegafur-uracil (UFT). Oxaliplatin-based doublet chemotherapy, either FOLFOX (5-FU + oxaliplatin) or CAPOX (capecitabine + oxaliplatin), are standard in stage III but can also be considered in select stage II patients. For patients undergoing oxaliplatin-based chemotherapy, the results of the International Duration Evaluation of Adjuvant Chemotherapy (IDEA) collaboration suggests that the duration of oxaliplatin-based AC may be shortened from 6 to 3 months in those patients assessed to be at lower risk of recurrence based on tumour and nodal staging [1].

As CC incidence is projected to increase over time, questions surrounding the economic costs of AC strategies will become increasingly important [2, 3]. As an illustration, accounting for drug acquisition, delivery, and management of toxicities, a 6 month course of adjuvant FOLFOX chemotherapy costs in excess of 10,000 US dollars (USD) per patient [4, 5]. Strategies to reduce AC utilisation, such as shortening duration of oxaliplatin-based AC from 6 to 3 months, has been projected to result in healthcare system savings of 3.6 to 61.4 million USD over a 5 year period depending on the country [6]. Another strategy to reduce costs is the use of molecular and genomic biomarkers that allows more precise identification of patients at high risk of recurrence that would most likely benefit from AC, reducing overtreatment by avoiding AC in patients who would least benefit [7‐10]. In many countries, such new technologies require reimbursement through taxpayer funds to ensure affordable access. Navigating reimbursement requires demonstration of economic value, which is usually provided by health economic evaluations. As these evaluations are usually derived from clinical trials, more efforts should be made to design trials with future reimbursement considerations in mind.

To appraise the health economic evidence of AC strategies in CC, a systematic review of published evaluations assessing AC treatment strategies in stage II and III CC was undertaken. This review also aimed to identify common assumption and limitations in published evaluations that may inform future trial designs, expediting patient access to new therapies and health technologies.

This review was designed, performed, and reports in line with the Preferred Reporting Items of Systematic Reviews and Meta-Analyses Guidelines, and prospectively registered in the International Prospective Register of Systematic Reviews (CRD42021265063) [11, 12].

The initial literature search identified 568 publications, of which 44 duplicates were removed. Of the 524 unique publications, 459 were excluded following title and abstract screening (Fig. 1). The remaining 65 studies underwent full text review, with 38 meeting the inclusion criteria [4‐6, 18‐52]. Reasons for exclusion during the full text review included studies being reviews or commentaries, pertaining to rectal cancer or neoadjuvant therapies, not being economic evaluations or full text being unavailable. Another study was excluded as it evaluated a drug that is no longer in used in current clinical practice and therefore, the evaluation was of limited relevance. Selected studies were stratified by stage (II or III) and the AC strategies being investigated, with some studies found to examine multiple strategies or report aggregated stage II and III results.

This systematic review included 38 health economic evaluations that compared a number of AC strategies currently employed in stage II and III CC, aiming to report the cost-effectiveness of these strategies and identify areas of potential improvements to inform further trial design and economic evaluations.

Firstly, the evaluations consistently implied that single agent AC is cost-effective compared to no AC. Secondly, most studies reported that oral capecitabine was cost-effective to or dominates (i.e., improved health outcomes at lower cost) 5-FU. In contrast, Soni et al. reported that 5FU dominates both capecitabine and CAPOX. However, this study was modelled on a real-world patient cohort, observing that as older and less fit patients were more likely to receive oral AC, capecitabine-based therapy was also associated with reduced dose intensity (RDI) and a lower probability of 5-year OS [56]. As comparison, an age based analysis of the phase III X-ACT trial reported similar efficacy of capecitabine in all age groups (including ≥ 70 years) despite higher rates of toxicity and dose reduction observed amongst older patients, contradicting the findings noted by Soni et al. [57] Economic evaluations are often based on clinical trials which recruits participants that tend to be younger and fitter than real-world counterparts, resulting in an observable difference between the efficacy observed in-trial and in routine clinical-practice [58, 59]. Modelling purely based on trial data may over-estimate efficacy but also underestimate toxicities, leading to favourable incremental cost-effective ratios (ICERs). One possible solution is the utilisation of real-world data (RWD) to inform model parameters in economic evaluations. RWD could be used to more accurately model control or standard strategies, with relative outcomes from clinical trials being applied. In the case of targeted therapies, RWD could also provide more accurate estimates of uptake based on prevalence of targeted mutations if routinely tested. Well maintained cancer registries could provide robust RWD and improve the generalisability of economic evaluation results. Notably, three of the included studies utilised such registries in modelling transitions between health states such as disease free to recurrence or death. Two studies utilised the Netherlands Cancer Registry (NCR) to develop a health model that simulates patients with stage II colon cancer from diagnosis to death [4, 51]. The third study utilised an Australian-based multi-site colorectal cancer registry (ACCORD) to model the progression of patients following recurrence to death [5]. These studies illustrates the feasibility of using RWD to inform model parameters, perhaps better reflecting real-world outcomes.

The majority of economic evaluations also reported 6-months of FOLFOX as being cost-effective compared to 5-FU in stage III CC. However, following publication of the IDEA collaboration, consensus guidelines now recommend for patients undergoing oxaliplatin-based AC, 3 months duration can also be considered based on further risk stratification [15, 16]. Given all identified economic evaluations demonstrated that 3 months of oxaliplatin-based AC dominates 6 months in stage II and III, shortened AC should be incorporated as a comparator in all economic evaluation as modelling 6 months alone may overestimate costs. Of the 2 biomarker studies published following the IDEA collaboration, only one modelled shortened CAPOX chemotherapy. Additionally, only a small number of oxaliplatin-related evaluations costed or considered long-term disutility from neurotoxicity. Given that the key rationale for shortening oxaliplatin exposure is to reduce neurotoxicity, more consistent consideration of this AE is required [1]. Economic evaluations must be careful in utilising relevant comparators in their assessments as regulators may reject reimbursement submissions that do not provide evidence for cost-effectiveness compared to current standard of care practices. Furthermore, detailed consideration of relevant toxicities and associated costs may more accurately reflect economic benefits.

Current consensus guidelines recommend the use of clinicopathological parameters to guide selection of stage II CC patients for AC. Such an approach results in 20–30% of patients receiving AC but meta-analyses reports only modest reductions in absolute recurrence risk suggesting that a significant proportion of patients are exposed to unnecessary treatment [60‐63]. This has led to the significant interest in risk-stratifying biomarkers that could better define patients that would benefit from AC. We identified 5 studies that investigated a molecular or genomic biomarker approach to patient selection for stage II CC. It is noted that these approaches and their associated costs vary significantly from study to study. Additionally, the modelled efficacy of these biomarkers are based on either retrospective or early phase studies meaning the economic results are to be considered only as early indications of potential cost-effectiveness. Certainly, economic evaluations based on results from larger prospective studies such as the recently published DYNAMIC study, which randomised stage II CC patients to either a ctDNA guided approach to AC patient selection or SOC, is required to confirm cost-effectiveness [64]. Acknowledging these limitations, three of the studies demonstrated that a reduction in AC prescription (range: 14% to 22%) resulted in biomarker informed care dominating SOC, demonstrating the positive economic impact of reducing AC overtreatment [5, 49, 50]. The two remaining studies reported that compared to no AC, biomarker patient selection was cost-effectiveness but notably in these analyses, the uptake of AC in the biomarker strategy was limited (range: 5.8–7.4%). However, as a biomarker’s impact depends on whether it actually can change treatment decisions, economic evaluations should model compliance to biomarker results but only two studies did so. Alberts et al. utilised a survey of clinicians presented with OncotypeDx results and To et al. assumed a low non-compliance rate based on experiences in breast cancer patients [5, 50]. In assessing the economic value of 3 months AC, Hanna et al. utilised a clinician survey to model the clinical uptake of shortened AC, recognising significant variation in clinical practice [6]. Whilst clinician surveys have limitations, they can offer timely insight into the real-world uptake of new biomarkers or treatment recommendations, allowing more robust economic evaluations.

Most studies utilised HUVs from literature, however two of the sources were based on surveys of patients in the 1990s, which may be lower than contemporary values given improvements in the recognition and management of toxicities. Given that cost-utility analysis requires accurate estimation of QALYs, HUVs that are temporally and geographically relevant to the patient cohort under investigation are important. The SCOT study, a randomised phase III trial comparing 3–6 months AC, routinely collected quality-of-life (QoL) questionnaires from recruited patients. The QoL data formed the basis of the HUVs utilised by all economic evaluations assessing shortened duration AC, allowing for more accurate modelling of QALYs, reflecting the outcomes of patients exposed to the treatment under evaluation in a contemporary time period. This demonstrates the importance of collecting patient reported outcomes and QoL data from ongoing clinical trials with the SCOT study serving as a case study. This will allow economic evaluations to provide stronger cost-effectiveness evidence and hopefully expedite reimbursement processes. Additionally, patient surveys can also more accurately capture additional information such as travel costs and patient and carer time-off work, allowing for greater insight into societal costs.

The available health economic evidence suggests that single-agent AC is cost-effective compared to no AC for both stage II and stage III CC, that oral AC is cost-effective compared to intravenous AC for stage III CC, that oxaliplatin-based AC is cost-effective for stage III CC, and that 3-month CAPOX is cost-effective compared to 6-months for high-risk stage II and low-risk stage III CC. The early evidence also suggests that biomarker-driven approaches to refine patient treatment selection for AC for stage II CC may be cost-effective compared to current standard of care, though more robust randomised clinical and economic evidence is warranted. Finally, to further increase the value of future economic evaluations and expedite access to new therapies, clinician and patient surveys should be incorporated into trials to address critical knowledge gaps and improve the robustness of health economic models.