Cost-effectiveness analysis of selective first-line use of biologics for unresectable RAS wild-type left-sided metastatic colorectal cancer

Abstract

Background

Evidence from a retrospective analysis of multiple large phase iii trials suggested that primary tumour location (ptl) in RAS wild-type metastatic colorectal cancer (wtRAS mcrc) might have predictive value with respect to response to drug therapies. Recent studies also show a potential preferential benefit for epidermal growth factor inhibitors (egfris) for left-sided tumours. In the present study, we aimed to determine the incremental cost-effectiveness ratio (icer) for the first-line use of an egfri for patients with left-sided wtRAS mcrc.

Methods

We developed a state-transition model to determine the cost effectiveness of alternative treatment strategies in patients with left-sided mcrc:

• ■ Standard of care

• ■ Use of an egfri in first-line therapy

The cohort for the study consisted of patients diagnosed with unresectable wtRAS mcrc with an indication for chemotherapy and previously documented ptl. Model parameters were obtained from the published literature and calibration. The perspective was that of a provincial ministry of health in Canada. We used a 5-year time horizon and an annual discount rate of 1.5%.

Results

Selecting patients for first-line egfri treatment based on left-sided location of their colorectal primary tumour was more effective than the standard of care, resulting in an increase in quality-adjusted life-years (qalys) of 0.226 (or 0.644 life-years gained). However, the strategy was also more expensive, costing an average of $60,639 more per patient treated. The resulting icer was $268,094 per qaly. A 35% price reduction in the cost of egfri would be needed to make this strategy cost-effective at a willingness-to-pay threshold (wtp) of $100,000 per qaly.

Conclusions

Selective use of an egfri based on ptl was more cost-effective than unselected use of those agents; however, based on traditional wtp thresholds, it was still not cost-effective. While awaiting the elucidation of more precise predictive biomarkers that might improve cost-effectiveness, the price of egfris could be reduced to meet the wtp threshold.

BACKGROUND

The treatment of metastatic colorectal cancer (mcrc) has evolved tremendously since about 2009. Cytotoxic chemotherapy regimens such as folfox (5-fluorouracil–leucovorin–oxaliplatin) and folfiri (5-f luorouracil–leucovorin–irinotecan) remain the cornerstone of therapy¹, but the addition of biologic agents—including inhibitors of vascular endothelial growth factor (vegf)¹ such as bevacizumab and of the epidermal growth factor receptor (egfr) such as cetuximab and panitumumab^1–3—has improved outcomes for patients with mcrc.

Primary tumour location (ptl) was recognized as a prognostic factor for patients with mcrc as early as 2001⁴. Post hoc analyses of subsequent clinical trials have confirmed the prognostic importance of ptl for overall survival (os)⁵. The rationale for the prognostic value of tumour sidedness in mcrc is a topic of ongoing research; however, it is thought to be a surrogate for the underlying biologic characteristics of tumours that arise in different locations⁶. Evidence has also emerged indicating that ptl is not only prognostic, but also predictive of a differential response to targeted drug therapy. Two recent systematic reviews and their corresponding meta-analyses based on a retrospective examination of randomized clinical trials confirm the predictive relevance of ptl when using targeted therapies for the treatment of patients with left-sided compared with right-sided KRAS wild-type mcrc^7,8. The results indicated that patients with left-sided ptl experience a significantly greater os benefit from the first-line use of regimens containing egfris (cetuximab or panitumumab) compared with the first-line use of regimens containing the vegf inhibitor bevacizumab (hazard ratio: 0.71; 95% confidence interval: 0.58 to 0.85; p < 0.0003)⁷. In comparison, patients with right-sided ptl were observed not to experience a statistically significant os benefit from the first-line use of a bevacizumab-containing regimen compared with an egfri-containing regimen⁷. Recent guidelines from the U.S. National Comprehensive Cancer Network⁹ also suggest that only patients with left-sided ptl should be offered an egfri-containing regimen as initial therapy, because “there is a preponderance of data to suggest lack of activity of cetuximab and panitumumab in initial therapy for patients whose primary tumours originated on the right side of the colon”⁹. In Canada, similar consensus recommendations based on ptl have also been made for first-line treatment¹⁰.

Although newer targeted therapies have resulted in improved os for patients with unresectable mcrc, they have also directly contributed to the rising cost of treatment^11–13. For example, the cost of a folfox–cetuximab regimen is more than 4 times that of a chemotherapy-only regimen. A cost-effectiveness analysis of various sequences of treatment showed that using egfris in the first line for all patients with KRAS wild-type mcrc is highly cost ineffective—the icer being about $3.177 million in a comparison in which bevacizumab is used in the first line and egfris are deferred to the third line of therapy¹⁴.

Given the emerging data about the effect of tumour sidedness with respect to predicting the relative benefits of first-line treatment with biologic therapies¹⁵, we aimed to determine whether selective use of those therapies based on ptl is cost-effective.

METHODS

Study Design

A state-transition model was developed to determine the cost-effectiveness of alternative treatment strategies in patients treated for mcrc. A cost–utility analysis compared these strategies:

• ■ First-line use of a vegf inhibitor

• ■ First-line use of an egfri

The choice of first-line treatment was based on ptl. Analyses were performed per the guidelines for economic evaluation published by the Canadian Agency for Drugs and Technologies in Health¹⁶. The present report is structured according to guidelines set out by the Consolidated Health Economic Evaluation Reporting Standards¹⁷.

Cohort

The cohort for this study consisted of patients diagnosed with unresectable RAS wild-type (wtRAS) mcrc with an indication for chemotherapy. Patients in the cohort were assumed to have a previously documented left-sided ptl (that is, cancer of the splenic flexure and of regions distal to the splenic flexure, including the rectum)¹⁵.

Strategies

The state-transition model evaluated two different treatment strategies for cost-effectiveness. Strategy 1 (status quo, reflecting the current standard of practice) consisted of first-line use of bevacizumab combined with either folfox or folfiri. Patients who progressed and were eligible for further chemotherapy received second-line folfiri or folfox (alternating the chemotherapy agent, depending on the first-line regimen). Third-line treatment for eligible progressed patients consisted of cetuximab, which was then followed by best supportive care (bsc) for patients who progressed on third-line treatment. Strategy 2 consisted of first-line cetuximab and folfox or folfiri for patients with left-sided ptl. Patients who progressed were given bevacizumab plus folfiri or folfox in the second line. Further progression resulted in third-line treatment with bsc only. In each strategy, the probability of receiving the subsequent line of therapy upon progression was calibrated to generate an os that matched the results of the systematic review and meta-analysis for each patient group in each strategy.

Decision Model

Our cohort-based state-transition model was implemented using the TreeAge Pro software application (2017 release: TreeAge Software, Williamstown, MA, U.S.A.) in the form of a cost–utility analysis over a 5-year time horizon. Our model consisted of health states related to the natural history of mcrc, which included 3 lines of therapy, the corresponding progressed states, and a death state (Figure 1). For each strategy, the patient cohort could move through first-line, second-line, and third-line therapy states based on eligibility criteria. At each transition to a new line of therapy, some patients might not be eligible for the new therapy and would then be transitioned to a progressed health state. In the model, cohort members moved through the predefined health states in monthly cycles.

FIGURE 1.

State-transition model used for the cost-effectiveness analysis. Health states are represented by circles, and permitted transitions are represented by arrows.

Model Parameters

Model parameters (Tables I–IV), including treatment efficacy [using progression-free survival (pfs)]^7,14,18–22, treatment-related adverse events^{20,21,23–46}, mortality (using os)^7,14,18–22, direct medical costs^14,47,48, and utilities¹⁴, were obtained from the published literature. Probabilities for transition to a subsequent line of therapy were obtained by calibration.

TABLE I.

Clinical efficacy of chemotherapeutic agents

Chemotherapy regimen	Median PFS (months)	Source
Bevacizumab
First line in left PTL	11.47
First line in unknown PTL	11.30
First line in right PTL	8.96
EGFRi		Holch et al., 20177; Venook et al., 201718; Venook et al., 201719
First line in left PTL	12.11
First line in unknown PTL	11.40
First line in right PTL	8.36
FOLFOX
Second line in any PTL	4.60	Giantonio et al., 200720; Tournigand et al., 200421
Second line with bevacizumab in any PTL	7.30	Giantonio et al., 200720
EGFRi
Third line in any PTL	3.70	Karapetis et al., 200822

PFS = progression-free survival; PTL = primary tumour location; EGFRi = epidermal growth factor receptor inhibitor; FOLFOX = infusional fluorouracil–leucovorin–oxaliplatin.

TABLE II.

Probability of experiencing grade 3 or 4 adverse events

Regimen and event	Probability (%)	Low (−25%)	High (+25%)	Source
FOLFIRI plus bevacizumab
Diarrhea	13.0	9.8	16.3	Pectasides et al., 201223; Heinemann et al., 201424; Nalluri et al., 200825; Al-Shamsi et al., 201526; Xu et al., 201627; Hapani et al., 200928
Nausea and vomiting	8.0	6.0	10.0
Stomatitis	2.5	1.9	3.1
Febrile neutropenia	5.3	4.0	6.6
Fatigue	2.0	1.5	2.5
Skin toxicity	2.0	1.5	2.5
Sensory neuropathy	0.0	0.0	0.0
GI perforation	2.0	1.5	2.5
Thromboembolic event	7.3	5.5	9.1
FOLFOX plus bevacizumab
Diarrhea	13.0	9.8	16.3	Giantonio et al., 200720; Nalluri et al., 200825; Hapani et al., 200928; Botrel et al., 201629; Saltz et al., 201230; Aprile et al., 201531
Nausea and vomiting	10.0	7.5	12.5
Stomatitis	4.0	3.0	5.0
Febrile neutropenia	4.0	3.0	5.0
Fatigue	11.9	8.9	14.9
Skin toxicity	1.0	0.8	1.3
Sensory neuropathy	18.0	13.5	22.5
GI perforation	1.6	1.2	2.0
Thromboembolic event	7.3	5.5	9.1
FOLFIRI plus cetuximab
Diarrhea	11.0	8.3	13.8	Heinemann et al., 201424; Sobrero et al., 200832; Petrelli et al., 201233
Nausea and vomiting	7.0	5.3	8.8
Stomatitis	4.0	3.0	5.0
Febrile neutropenia	1.5	1.1	1.9
Fatigue	0.5	0.4	0.6
Skin toxicity	26.0	19.5	32.5
Sensory neuropathy	0.0	0.0	0.0
GI perforation	0.0	0.0	0.0
Thromboembolic event	5.0	3.8	6.3
FOLFOX plus cetuximab
Diarrhea	6.0	4.5	7.5	Petrelli et al., 201233; Fernandez-Plana et al., 201434; Bokemeyer et al., 201535; Tveit et al., 201236
Nausea and vomiting	7.0	5.3	8.8
Stomatitis	8.5	6.4	10.6
Febrile neutropenia	5.0	3.8	6.3
Fatigue	7.3	5.5	9.1
Skin toxicity	28.6	21.5	35.8
Sensory neuropathy	13.5	10.1	16.9
GI perforation	0.0	0.0	0.0
Thromboembolic event	0.0	0.0	0.0
FOLFIRI
Diarrhea	14.0	10.5	17.5	Tournigand et al., 200421; Goldberg et al., 200437; Van Cutsem et al., 201238; Colucci et al., 200539
Nausea and vomiting	7.1	5.3	8.9
Stomatitis	1.0	0.8	1.3
Febrile neutropenia	5.0	3.8	6.3
Fatigue	5.0	3.8	6.3
Skin toxicity	1.0	0.8	1.3
Sensory neuropathy	0.0	0.0	0.0
GI perforation	0.0	0.0	0.0
Thromboembolic event	4.0	3.0	5.0
FOLFOX
Diarrhea	11.0	8.3	13.8	Tournigand et al., 200421; Colucci et al., 200539; Bokemeyer et al., 200940; Douillard et al., 201041; Hochster et al., 200842; Cascinu et al., 201743; Cassidy et al., 201144
Nausea and vomiting	7.0	5.3	8.8
Stomatitis	3.3	2.5	4.1
Febrile neutropenia	5.0	3.8	6.3
Fatigue	3.0	2.3	3.8
Skin toxicity	1.1	0.8	1.4
Sensory neuropathy	18.6	14.0	23.3
GI perforation	0.0	0.0	0.0
Thromboembolic event	5.0	3.8	6.3
Cetuximab
Diarrhea	8.5	6.4	10.6	Petrelli et al., 201233; Cunningham et al., 200445; Zhang et al., 201346
Nausea and vomiting	1.5	1.1	1.9
Stomatitis	5.6	4.2	7.0
Febrile neutropenia	0.0	0.0	0.0
Fatigue	9.1	6.8	11.4
Skin toxicity	25.0	18.8	31.3
Sensory neuropathy	0.0	0.0	0.0
GI perforation	0.0	0.0	0.0
Thromboembolic event	0.0	0.0	0.0

FOLFIRI = leucovorin–fluorouracil–irinotecan; GI = gastrointestinal; FOLFOX = infusional leucovorin–fluorouracil–oxaliplatin.

TABLE III.

Management costs associated with metastatic colorectal cancer

Variable	Cost (Canadian dollars)			Source
	Base case	Low	High
Health-state management
Managing 1st line treatment	3,652.25	3,085.62	4,218.87	Riesco-Martinez et al., 201614
Managing 2nd line treatment	1,538.44	564.11	1,965.45
Managing 3rd line treatment	1,717.49	946.94	2,470.63
Palliative care	26,277.20	19,707.90	32,846.50
Drug acquisition and preparation
Chemotherapy with cetuximab	6,713.74	5,035.31	8,392.18	Odette Cancer Centre pharmacy
Chemotherapy with FOLFIRI and bevacizumab	4,023.89	3,017.92	5,029.86
Chemotherapy with FOLFIRI and cetuximab	6,944.32	5,208.24	8,680.40
Chemotherapy with FOLFOX	1,830.97	1,373.23	2,288.71
Chemotherapy with FOLFOX and bevacizumab	5,355.02	4,016.27	6,693.78
Adverse event management
Diarrhea	3,103.21	2,327.41	3,879.02	Riesco-Martinez et al., 201614; Graham et al., 201447; Lathia et al., 201048
Fatigue	3,121.68	2,341.26	3,902.10
Febrile neutropenia	7,058.00	3,401.20	10,203.70
Mucositis	1,170.41	877.81	1,463.01
GI perforation	35,686.55	17,843.30	53,529.80
Rash	295.00	147.50	442.50
Thrombosis	22,387.03	11,193.30	33,580.00
Nausea and vomiting	1,102.75	551.40	1,654.10

FOLFIRI = leucovorin–fluorouracil–irinotecan; FOLFOX = infusional leucovorin–fluorouracil–oxaliplatin; GI = gastrointestinal.

TABLE IV.

Utilities associated with treatments and adverse events during the management of metastatic colorectal cancer

Event	Value	Low	High	Standard deviation	Source
Health-state utilities
Utilityoffirst- treatment	0.819	0.779	0.860	0.0205	Riesco-Martinez et al., 201614
Utility of second-line treatment	0.780	0.738	0.821	0.021
Utility of third-line treatment	0.728	0.670	0.785	0.029
Utility of palliation	0.267	0.165	0.368	0.051
Utility or disutility of adverse events
Utility of diarrhea	0.521	0.452	0.591	0.035	Riesco-Martinez et al., 201614
Utility of fatigue	0.518	0.459	0.578	0.03
Utility of febrile neutropenia	0.720	0.664	0.776	0.028
Disutility of hypertension	0.000	0	0.001	0.0005
Utility of mucositis	0.509	0.421	0.598	0.0445
Utility of neurotoxicity	0.553	0.514	0.593	0.02
Utility of perforation	0.174	0.031	0.316	0.0715
Utility of rash	0.618	0.548	0.688	0.035
Utility of thrombosis	0.230	0.09	0.364	0.07
Utility of vomiting	0.516	0.407	0.626	0.055

Efficacy

Efficacy data for the first-line use of bevacizumab and an egfri in left-sided ptl were based on a systematic review and meta-analysis of the calgb 80405, fire-3, and peak clinical trials^7,18,19. The median pfs for each trial was calculated from the overall hazard ratios reported in the meta-analysis (Table I). Efficacy data for second- and third-line treatments in strategies 1 and 2 were obtained from published phase iii randomized clinical trials for each of the regimens^14,20–22. Duration of therapy was estimated according to the median pfs for each line, with the assumption that treatment was continued until disease progression. The probability of receiving the subsequent line of therapy upon progression was calibrated against the results of the systematic review and meta-analysis for each patient group in each strategy, using a guided search strategy and error sum of squares goodness-of-fit statistic⁴⁹. The calibrated probabilities of receiving second- and third-line treatment after progression were 67% and 50% respectively.

Adverse Events

Common and clinically important grades 3 and 4 adverse events for each therapeutic option at each line of therapy were obtained from published phase iii clinical trials (Table II)^{20,21,23–46}. Adverse events considered in the analysis included diarrhea, nausea, vomiting, stomatitis, febrile neutropenia, fatigue, skin toxicity, sensory neuropathy, gastrointestinal perforation, a thromboembolic event, and treatment-induced hypertension. The probability of each adverse event and its corresponding duration were used to calculate the overall effect on costs and health utilities associated with each treatment. Adverse event durations were assumed to be 1 week for febrile neutropenia, diarrhea, fatigue, hypertension, mucositis, thrombosis, and vomiting; a duration of the entire cycle of treatment was assumed for neurotoxicity, perforation, and rash¹⁴.

Costs and Resource Use

Table III sets out drug costs, health state management costs, and treatment-related adverse event costs. Health state management costs were calculated based on medical resource use and were collected from patient chart reviews at the Odette Cancer Centre in Toronto, Ontario¹⁴. Those costs included outpatient physician costs, diagnostic imaging, medical procedure costs, and laboratory-related costs, which were calculated using the unit cost from the schedule published by the Ontario Ministry of Health and Long-Term Care⁵⁰. Drug regimen costs were obtained from the Odette Cancer Centre pharmacy and included costs for drug acquisition, prophylactic medications, pharmacy time, and nursing time during infusion. Costs for treating a grade 3 or 4 adverse event were based on the published literature^14,47,48. All costs are presented in 2017 Canadian dollars.

Utilities

Utility and disutility values for each health state and adverse event were based on a previously published cost-effectiveness study of egfri use in patients with mcrc¹⁴ (Table IV). Treatment-specific utility values were calculated using the base utilities associated with each line of therapy, weighted according to the probability of the adverse events associated with each treatment and their corresponding durations.

Economic Assumptions

The economic analysis was conducted from the perspective of a provincial ministry of health in Canada and was structured as a cost–utility analysis. Future costs and benefits were discounted at 1.5% annually according to guidelines set by the Canadian Agency for Drugs and Technologies in Health¹⁶.

Analytic Strategy

The base-case analysis began with an estimate of the deterministic results. Our base-case analysis used folfiri as the initial chemotherapy backbone, plus cetuximab. A deterministic one-way sensitivity analysis was then conducted on all model parameters over their plausible ranges (the reported 95% confidence intervals, if available, or ±25% of the reference value). An alternative scenario analysis that used folfox as the initial chemotherapy backbone, plus cetuximab, was then conducted to determine the effect on the icer of selective treatment with an egfri. Similarly, an additional scenario analysis that used panitumumab as the egfri of choice was conducted. For the base-case and each scenario analysis, a probabilistic analysis was also conducted using a Monte Carlo simulation for 10,000 iterations to determine the probability that each strategy would be cost-effective at the willingness-to-pay (wtp) threshold of $100,000. Finally, a threshold analysis was conducted to determine the acquisition cost of cetuximab at which its selective use in left-sided ptl would be cost-effective at a wtp threshold of $100,000.

Validation

We validated our model (Table V) by comparing the median os and the median pfs of the patient cohort allocated to each of the model’s treatment strategies with the results reported in the meta-analysis of the systematic review⁷.

TABLE V.

Validation of the modelled treatment strategies^a

Validation	Reference	Left-side disease		Right-side disease		Otherb
		Bevacizumab	Cetuximab	Bevacizumab	Cetuximab	Bevacizumab	Cetuximab
Progression-free survival
	Venook et al., 201615 (median)	11.2	12.4	9.6	7.8
	Venook et al., 201419 (median)					11.3	11.4
	Venook et al., 201615 (HR)	1.19	0.84	0.61	1.64	NA	NA
	Systematic review (HR)	1.16	0.86	0.65	1.53	NA	NA
	Systematic review (median)c	11.47	12.11	8.96	8.36	11.3	11.4
	Model (median)d	11	12	9	8	11	11
	Model (mean)d	15.638	16.615	12.274	11.55	15.433	15.683
Overall survival
	Venook et al., 201615 (median)	31.4	36	24.2	16.7
	Venook et al., 201419 (median)					31.2	32
	Venook et al., 201615 (HR)	1.3	0.77	0.74	1.36	NA	NA
	Systematic review (HR)	1.41	0.71	0.77	1.3	NA	NA
	Systematic review (median)c	28.95	39.04	23.13	17.47	31.2	32
	Model (median)d	29	39	23	17	31	32
	Model (mean)d	31.122	39.176	25.265	21.083	32.839	34.811

^aPerformed by comparing median overall survival and median progression-free survival for the patient population receiving each treatment strategy as reported in the meta-analysis of the systematic review⁷. The results generated by the model closely matched the results in the published studies.

^bBased on Venook et al., 2014¹⁹.

^cAdjusted value based on the Venook et al., 2016¹⁵ data and the HRs.

^dBased on 10,000 microsimulations.

HR = hazard ratio; NA = not applicable.

RESULTS

The results of the base-case analysis (Table VI) show that selective use of egfris based on ptl (strategy 2) is more effective than using first-line bevacizumab for patients with a left-sided ptl (strategy 1), resulting in an increase of 0.226 quality-adjusted life-years (qalys) (95% credible interval: −0.21 to 0.67)—that is, 0.644 life-years gained. However, that strategy is also more expensive, costing the health care system approximately $60,639 more per patient (95% credible interval: −$27,421 to $150,715) for treatment with cetuximab. The icer resulting from the selective use of cetuximab based on ptl was $268,094. Deterministic 1-way sensitivity analyses (summarized as a tornado diagram, Figure 2) show that varying the parameters used in the model does not significantly change the results of the base-case analysis. The results of the probabilistic analysis (summarized as a cost-effectiveness acceptability curve, Figure 3) also reveal that, under baseline conditions, selective use of egfris based on ptl has a 7.5% probability of being cost-effective at a wtp threshold of $100,000 per qaly.

TABLE VI.

Cost-effectiveness results

Analysis	Strategy			Cost ($)		QALY (n)		ICER ($/QALY)	Acceptability at WTP $100,000 [probabilistic analysis (%)]
	First line	Second line	Third line	Strategy	Δ	Strategy	Δ
Base case
FOLFIRI as the initial chemotherapy backbone, cetuximab at full price
(A)	FOLFIRI–bevacizumab	FOLFOX	Cetuximab monotherapy	255,585		1.610			92.5
(B)	FOLFIRI–cetuximab	FOLFOX–bevacizumab	Best supportive care	316,224	60,639	1.836	0.226	268,094	7.5
Scenario 1
FOLFOX as the initial chemotherapy backbone, cetuximab at full price
(A)	FOLFOX–bevacizumab	FOLFIRI	Cetuximab monotherapy	268,741		1.555			86.9
(B)	FOLFOX–cetuximab	FOLFIRI–bevacizumab	Best supportive care	325,419	56,678	1.728	0.172	328,603	13.1
Scenario 2
FOLFIRI as the initial chemotherapy backbone, panitumumab as the EGFRi
(A)	FOLFOX–bevacizumab	FOLFIRI	Panitumumab monotherapy	252,562		1.610			77.4
(B)	FOLFIRI–cetuximab	FOLFOX–bevacizumab	Best supportive care	292,467	39,906	1.836	0.226	176,428	22.6
Threshold
FOLFIRI as the initial chemotherapy backbone, 35% discount on the price of cetuximab
(A)	FOLFIRI–bevacizumab	FOLFOX	Cetuximab monotherapy	251,011		1.610			50.0
(B)	FOLFIRI–cetuximab	FOLFOX–bevacizumab	Best supportive care	273,525	22,514	1.836	0.226	99,537	50.0

QALY = quality-adjusted life year; ICER = incremental cost-effectiveness ratio; WTP = willingness to pay; FOLFIRI = fluorouracil–leucovorin–irinotecan; FOLFOX = fluorouracil–leucovorin–oxaliplatin; EGFRi = epidermal growth factor receptor inhibitor.

FIGURE 2.

Deterministic one-way sensitivity analyses, summarized in a tornado diagram. Varying the parameters used in the model did not significantly change the results of the base-case analysis. The conclusion about the selective use of first-line epidermal growth factor receptor inhibitors based on primary tumour location (PTL) remained “not cost-effective.” The most sensitive parameters were the cost of the drug therapies, followed by the median progression-free survival (PFS) for each drug therapy. Chemo = chemotherapy; FOLFIRI = fluorouracil–leucovorin–irinotecan.

FIGURE 3.

Cost-effectiveness acceptability curves resulting from probabilistic analyses comparing first-line EGFRi treatment based on left-sided location of the colorectal primary tumour (strategy 1) with standard-of-care treatment (strategy 2). (A) Cetuximab: the acceptability of strategy 1 using cetuximab compared with strategy 2 is 7.5% at a willingness-to-pay (WTP) threshold of $100,000 per quality-adjusted life-year (QALY). (B) Panitumumab: the acceptability of strategy 1 using panitumumab compared with strategy 2 is 22.6% at a WTP threshold of $100,000 per QALY. Cost-effectiveness planes for (C) cetuximab and (D) panitumumab, comparing strategy 2 with strategy 1.

Our scenario analysis (Table VI) indicates that using folfox as the initial chemotherapy backbone with cetuximab does not alter the original conclusions. The icer resulting from the selective use of cetuximab based on ptl is $328,603, with an acceptability of 13% at a wtp threshold of $100,000. Similarly, the scenario analysis (Table VI) indicates that choosing panitumumab as the egfri also does not alter the conclusions. The icer resulting from the selective use of panitumumab based on ptl was $176,428, with an acceptability of 23% at a wtp threshold of $100,000.

Our threshold analysis (Table VI) shows that a 35% price reduction would be required for selective use of cetuximab based on ptl to reach a 50% probability of being cost-effective at a wtp threshold of $100,000 per qaly.

DISCUSSION

We developed a decision model to determine the cost-effectiveness of using chemotherapy plus an egfri as first line therapy for wtRAS mcrc in patients with a left-sided ptl. The results of our model indicated that this chemotherapy-plus-egfri strategy was more effective, but also costlier, than standard care (that is, chemotherapy plus bevacizumab as first-line therapy). On average, the strategy cost an incremental $60,639 per patient with cetuximab, and resulted in a discounted gain of 0.226 qaly, translating into an icer of $268,094 per qaly for cetuximab compared with the standard of care. The icer was driven entirely by the cost of cetuximab. Our threshold analysis showed that a 35% reduction in the price of cetuximab would be required for its first-line use based on ptl to have a 50% probability of being cost-effective at a wtp threshold of $100,000 per qaly. Furthermore, the availability of emerging—and presumably less expensive—bevacizumab biosimilars further reduced the cost-effectiveness of chemotherapy plus an egfri as first-line therapy.

To our knowledge, the present study is the first to look at the cost-effectiveness of treatment selection based on ptl in mcrc. Treatment of wtRAS mcrc with egfris has been found by several studies to be cost-ineffective. A study in the United Kingdom concluded that third-line use of cetuximab, cetuximab–irinotecan, and panitumumab are poor value for money compared with bsc⁵¹. A Canadian study reached a similar conclusion in comparing third-line use of cetuximab with bsc⁵². In both cases, patient survival and drug costs were cited as having the greatest impact on cost-effectiveness. Unselected first-line use of egfris, compared with bevacizumab, in patients with wtRAS mcrc has also been found to be cost-ineffective^14,53. Although compared with other models, the cost-effectiveness model used in the present work is more comprehensive, it is also subject to some of the limitations of its precursors. First, the model inherited the limitations of the published systematic reviews and meta-analyses^7,8. Those analyses could have been subject to potential selection biases, because only the subgroups of patients with available ptl information in the published clinical trials were included in the analyses. Further, some of the trials included in the systematic reviews and meta-analyses had a slight crossover design; egfris might therefore have been used in the second or third line in selected patients originally assigned to the first-line bevacizumab arm, which could affect the os results^54–57. Second, data about the effects of treatment sequencing and the probability that patients would receive the subsequent lines of therapy based on ptl were limited⁹. To overcome those limitations, we conducted careful model calibration and validation to ensure the robustness of the results.

Recent evidence has shown an advantage in terms of the selective first-line use of an egfri in patients with wtRAS and left ptl^7,58,59. In our analysis, the magnitude of the clinical benefit did not offset the additional cost of the egfri. There are two possible reasons for that result: egfris might simply be too expensive compared with vegf inhibitors when considering the benefit that they provide, and we might not have found the optimal biomarker for patient selection. In patients with left-sided crc, ptl is undoubtedly a crude surrogate for biologic factors predicting improved response to egfris¹⁵. Analysis of pooled data from randomized trials is proceeding to try to uncover a biologic fingerprint that more precisely explains the observed phenomenon^18,60,61. A previous study reported an icer of $3.2 million for the first-line use of an egfri compared with bevacizumab for patients with wtRAS mcrc¹⁴. In our analysis, restricting first-line use of egfris to patients with left-sided wtRAS mcrc resulted in a significant decrease in the icer to $268,094. More precise selection of patients for egfri therapy based on elucidation of better predictive biomarkers could further reduce the icer for the first-line use of egfris.

Our results are consistent with the current recommendations from the pan-Canadian Oncology Drug Review’s Expert Review Committee at the Canadian Agency for Drugs and Technologies in Health. In 2014, cetuximab was not recommended for first-line use in unselected patients with unresectable wtRAS mcrc⁶². In 2015, although panitumumab was recommended for first-line use in the same unselected patient group, it was conditional on the drug price being reduced to an acceptable cost-effectiveness level⁶³. More importantly, the Expert Review Committee recently issued a negative recommendation: specifically, that they did not recommend panitumumab for the first-line treatment of unresectable left-sided wtRAS mcrc. The stated reasons were considerable uncertainty concerning the pfs and os results for patients with left-sided ptl and a lack of cost-effectiveness for panitumumab at the current drug price⁶⁴.

CONCLUSIONS

Although some evidence indicates that ptl can be both prognostic and predictive of response to an egfri in mcrc⁹, some limitations are still associated with such evidence. Based on the currently available evidence, our analysis supports the economic aspects of the selective use of first-line egfris based on ptl. Specifically, our analysis demonstrates that selective use of an egfri based on ptl in the first line—although more cost-effective than unselected use of such an agent—is still not cost-effective based on traditional thresholds. While awaiting more evidence or elucidation of more precise predictive biomarkers that might improve cost-effectiveness, the price of the egfris could be reduced to meet a cost-effectiveness threshold.