A within-trial cost-effectiveness analysis of panitumumab compared with bevacizumab in the first-line treatment of patients with wild-type RAS metastatic colorectal cancer in the US

Abstract

Aims: This analysis investigated the cost-effectiveness of panitumumab plus mFOLFOX6 (oxaliplatin, 5-fluorouracil, and leucovorin) compared with bevacizumab plus mFOLFOX6 in the first-line treatment of patients with wild-type RAS metastatic colorectal cancer (mCRC).

Materials and methods: The cost-effectiveness analysis was developed from a third-party payer perspective in the US and was implemented using a partitioned survival model with health states for first-line treatment (progression-free), disease progression with and without subsequent active treatment, and death. Survival analyses of patients with wild-type RAS mCRC from the PEAK head-to-head clinical trial of panitumumab vs bevacizumab were performed to estimate time in the model health states. Additional data from PEAK informed the amount of each drug consumed, duration of therapy, subsequent therapy use, and toxicities related to mCRC treatment. Literature and US public data sources were used to estimate unit costs associated with treatment and duration of subsequent active therapies. Utility weights were calculated from patient-level data from panitumumab trials in the first-, second-, and third-line settings. A life-time perspective was taken with future costs and outcomes discounted at 3% per annum. Scenario, one-way, and probabilistic sensitivity analyses were performed.

Results: Compared with bevacizumab, the use of panitumumab resulted in an incremental cost of US $60,286, and an incremental quality-adjusted life-year (QALY) of 0.445, translating into a cost per QALY gained of US $135,391 in favor of panitumumab. Results were sensitive to wastage and dose rounding assumptions modeled.

Limitations: Progression-free and overall survival were extrapolated beyond the follow-up of the primary analysis using fitted parametric curves. Costs and quality of life were estimated from multiple and different data sources.

Conclusions: The efficacy of panitumumab in extending progression-free and overall survival and improving quality of life makes it a cost-effective option for first-line treatment of patients with wild-type RAS mCRC compared with bevacizumab.

Keywords:

• Panitumumab
• wild-type RAS
• outcomes
• cost-effectiveness
• first-line therapy
• metastatic colorectal cancer

JEL CLASSIFICATION:

• D61
• I12
• I18

Introduction

Colorectal cancer (CRC) is the third most common cancer in both men and women in the US and the second leading cause of cancer death¹. Approximately 20% of patients present with metastatic CRC (mCRC) at diagnosis, and metastases eventually develop in 50–60% of patients^1–3. The 5-year relative survival rate is less than 14% in patients with mCRC, indicating a continued need to improve treatment outcomes¹. Surgical resection of metastases is curative in a minority of patients; therefore, the goal of treatment for most patients with mCRC is to prolong survival while maintaining quality of life.

In addition to various chemotherapeutic regimens, current treatments for mCRC include targeted monoclonal antibodies directed against the epidermal growth factor receptor (EGFR), such as panitumumab and cetuximab, and a humanized monoclonal antibody that binds the vascular endothelial growth factor (VEGF-A), such as bevacizumab. Patients with mCRC RAS (rat sarcoma 2 viral oncogene homolog) gene mutations should not be treated with an anti-EGFR agent, as these mutations are predictive of a lack of response to panitumumab and cetuximab^2,3. It is estimated that 50% of mCRC patients have tumors exhibiting RAS mutations⁴, including 40% with KRAS (Kirsten rat sarcoma 2 viral oncogene homolog) (exon 2) mutations⁵. Identifying KRAS status and ultimately RAS status through genotyping allows physicians to target patients who are likely to benefit from treatment with EGFR inhibitors and to minimize the number of patients receiving anti-EGFR agents who are unlikely to respond favorably.

For patients who are likely to benefit from anti-EGFR agents because their mCRC does not present with RAS gene mutations (i.e. RAS wild-type), a question arises as to which treatment is cost-effective between an anti-EGFR agent and bevacizumab in the first-line setting.

The PEAK trial was a phase II, open-label, randomized, multi-center, head-to-head clinical study designed to compare panitumumab plus mFOLFOX6 (oxaliplatin, 5-fluorouracil, and leucovorin; n = 142) vs bevacizumab plus mFOLFOX6 (n = 143) in first-line mCRC. Results from patients with wild-type RAS mCRC, as defined by the extended RAS analysis (n = 88 for panitumumab plus mFOLFOX6, n = 82 for bevacizumab plus mFOLFOX6), showed a statistically significant incremental median progression-free survival (PFS) benefit of 2.7 months (p = 0.03) and a trend in the incremental median overall survival (OS) benefit (8.0 months; p = 0.15) favoring the panitumumab arm^6,7.

To evaluate the cost-effectiveness of panitumumab vs bevacizumab in patients with wild-type RAS mCRC in the first-line setting, we developed a partitioned survival model from a third-party payer perspective in the US based on patient-level data from the PEAK trial^6,7.

Methods

Population

The model population was based on a sub-set of the patient population from PEAK (ClinicalTrials,gov, NCT00819780), the first-line clinical trial of panitumumab plus mFOLFOX6 vs bevacizumab plus mFOLFOX6 in patients with mCRC. This sub-set population was defined as adults (age ≥18 years) diagnosed with mCRC whose genetic profile was wild-type RAS (i.e. no mutation in exons 2, 3, or 4 of KRAS and NRAS) and who had not previously been treated with chemotherapy or investigational agents for mCRC (n = 170; panitumumab plus mFOLFOX6, 88; bevacizumab plus mFOLFOX6, 82)⁷. This population corresponds to that of the current licensed indication for panitumumab in the US⁸.

Model structure

A partitioned survival model structure was used to assess the cost-effectiveness of panitumumab plus mFOLFOX6 relative to bevacizumab plus mFOLFOX6 in the first-line treatment of patients with mCRC.

The model uses a 2-week cycle length; the time horizon is that of the lifetime of a patient with mCRC (assumed to be no more than 20 years after treatment initiation). The model tracks a cohort of patients from first-line mCRC treatment initiation to death. The (partitioned survival) model structure is presented in Figure 1.

Figure 1. Model structure. BSC, best supportive care; mCRC, metastatic colorectal cancer; OS, overall survival; PFS, progression-free survival. Health-state transitions: ¹ Progression-Free to Progressive Disease (Parametric survival modeling of patient-level data from PEAK trial [ie, PFS]). ^a Treat With Subsequent Active Therapy (Percentage of patients utilizing active treatment postprogression from PEAK trial). ^b Treat with BSC (Percentage of patients utilizing BSC postprogression from PEAK trial). ² Progression-Free to Death (Parametric survival modeling of patient-level data from PEAK trial [ie, OS]). ³ Progressive Disease: Treat With Subsequent Active Therapy to Progressive Disease: Treat With BSC (Weighted average of published PFS values for second-line treatment options). ⁴ Progressive Disease: Treat With Subsequent Active Therapy to Death (Parametric survival modeling of patient-level data from PEAK trial [ie, OS). ⁵ Progressive Disease: Treat With BSC to Death (Source: Parametric survival modeling of patient-level data from PEAK trial [ie, OS])

Parametric survival analyses

The partitioned survival analysis was parameterized by parametric survival analyses conducted on patient-level PFS and OS data for patients with wild-type RAS mCRC from the PEAK trial. For both PFS and OS, a series of parametric distributions were first fit (exponential, generalized gamma, log-logistic, log-normal, and Weibull) using PROC LIFEREG in SAS for Windows v9.4 (SAS, Cary, NC). Regressions including and excluding treatment covariates were calculated. Best-fit curves were assessed using overlays of the fitted curves and the observed Kaplan-Meier data and goodness-of-fit tests (Akaike information criterion [AIC] and Bayesian information criterion [BIC]). Additionally, flexible cubic splines were fit using R version 3.4.2 for PFS and OS.

In the OS analysis, the Weibull distribution, with treatment covariate included in the regression, provided the best fit in all criteria (i.e. lowest AIC and BIC values, similar shape to observed Kaplan-Meier data, reasonable long-term projections; Figure 2). In the PFS analysis, it was determined that no distribution adequately modeled PFS for either comparator.

Figure 2. Overall survival Kaplan-Meier plot and best-fit curve. FOLFOX, leucovorin, 5-fluorouracil, and oxaliplatin; KM, Kaplan-Meier.

In order to fit PFS reasonably, a flexible cubic spline model was used with multiple knots tested. Overall, based on the fit criteria outlined above, the Weibull with 2 knots was determined to be the best fit (Figure 3) in the PFS analysis.

Figure 3. Progression-free survival Kaplan-Meier plot and best-fit curve. FOLFOX, leucovorin, 5-fluorouracil, and oxaliplatin; KM, Kaplan-Meier.

In the partitioned survival calculations, the PFS curve was used to model the time patients spent without disease progression, and the OS curve estimated how many patients were alive at each time period. The difference between the OS and PFS curves resulted in the percentage of patients who were alive and receiving subsequent therapy or best supportive care (BSC). The model assumed that no patients would be progression-free after 5 years of treatment.

In the PEAK clinical trial, resection of metastases in patients with wild-type RAS was attempted in 12 of 88 patients (13.6%) treated with panitumumab plus mFOLFOX6 and in nine of 82 patients (11.0%) treated with bevacizumab plus mFOLFOX6⁷. The model assumed that resection occurred at 16 weeks after the start of treatment, similar to resection times from the PLANET-TTD study⁹.

Disease-free survival and OS for successfully resected patients were modeled by digitally extracting data from the disease-free survival and OS Kaplan-Meier plots from Adam et al.¹⁰. Next, the methods of Hoyle and Henley¹¹ were used to re-create a patient-level data set from the digitally extracted data points and the number of patients at risk at each time presented. Interval censored survival analyses were conducted using Stata/MP 15.1 for Windows (StataCorp, College Station, TX). The best-fit parametric curves were deemed to be the log-logistic for disease-free survival and generalized gamma for OS in successfully resected patients.

Subsequent therapy

Following disease progression, patients who have survived may receive either subsequent anti-cancer therapy or BSC. Data from the PEAK trial and a retrospective database analysis examining subsequent therapy within a US population were combined with additional assumptions to estimate the use and duration of subsequent therapy¹².

It was assumed that the same percentage of patients who received subsequent anti-cancer therapy in the PEAK trial (63%) would receive subsequent therapy in the model, as the OS in the model is built on the observed OS in the trial where subsequent therapy contributed to survival. We assumed that patients treated with first-line panitumumab plus chemotherapy would be treated second-line with bevacizumab plus chemotherapy and vice versa. A weighted average duration of therapy for biologic therapy use in second-line therapy was created using data from Parikh et al.¹².

Similarly, we assumed that, once the anti-EGFR and anti-VEGF agents had been used in first- and second-line therapy, third-line therapy would consist of chemotherapy or targeted therapy alone. Data from Parikh et al.¹² are used to inform the percentage of patients who receive third-line therapy given second-line therapy, and a similar weighted average duration of therapy was calculated. After second- and third-line treatment is exhausted, all patients receive BSC until death.

Costs

Drug acquisition costs were modeled using wholesale acquisition costs from Red Book¹³. Drug exposure (defined as the average dose per cycle per patient) and the average number of cycles administered were calculated from data in the PEAK clinical trial for direct treatment comparators. Drug acquisition costs, drug exposure inputs, chemotherapy costs, drug administration costs, subsequent treatment costs, and resection and toxicity incidence are presented in Table 1^7,12–15, along with sources and assumptions. Other medical costs considered by the model include RAS mutation testing, physician visits, diagnostic tests, grade 3/4 toxicity treatment, BSC, and end-of-life treatment (Table 2)^14,16–21.

Table 1. Regimen-specific input parameters.

Input parameter	Panitumumab + mFOLFOX6	Bevacizumab + mFOLFOX6	Source
Drug acquisition cost			Wholesale acquisition cost from the Red Book for listed NDCs^13
Biologic drug	$1,153.97	$758.54	Panitumumab: 100-mg vial, NDC 55513-0954-01 Bevacizumab: 100-mg vial, NDC 50242-0060-01
Chemotherapy drug
Oxaliplatin	$72.35	$72.35	50-mg vial, NDC 61703-0363-18
Irinotecan	$12.34	$12.34	40-mg vial, NDC 61703-0349-16
Leucovorin	$10.00	$10.00	50-mg vial, NDC 25021-0813-10
5-fluorouracil	$2.75	$2.75	50-mg vial, NDC 10139-0063-11
Product exposure/dosing, mg/kg			Product labels
Biologic drug	6.00	5.00
Chemotherapy drug, mg/m^2
Oxaliplatin	85.00	85.00
Irinotecan	180.00	180.00
Leucovorin	400.00	400.00
5-fluorouracil
Bolus	400.00	400.00
Continuous infusion	2,400.00	2,400.00
Drug administration cost per cycle			Administration for mCRC therapies is assumed to occur in the outpatient setting. Panitumumab, mFOLFOX6, and bevacizumab are administered every 2 weeks^14
Biologic + mFOLFOX6	$459.38	$459.38
FOLFOX or FOLFIRI alone	$393.34	$393.34
Mean number of treatment cycles			Observed mean number of administrations from the PEAK trial^15
Biologic	19.30	14.40
mFOLFOX6	11.20	10.50
Patients resected, %	13.6	11.0	Patients with any resection from PEAK final analysis publication^7
Serious toxicities (incidence), %			Incidence of serious adverse events occurring in >2% of patients with wild-type RAS mCRC in either treatment arm (May 30, 2012, data cut-off) from PEAK flash memorandum (Table 8-7) based on the PEAK [NCT00819780] clinical study report^15
Pulmonary embolism	4.7	2.5
Diarrhea	3.5	1.3
Sepsis	3.5	1.3
Dehydration	2.3	0.0
Febrile neutropenia	2.3	2.5
Gastroesophageal reflux disease	2.3	0.0
Pneumonia	2.3	3.8
Deep vein thrombosis	1.2	3.8
Pyrexia	1.2	3.8
Urinary tract infection	1.2	2.5
Vomiting	1.2	2.5
Infection	0.0	2.5
Intestinal perforation	0.0	2.5
Syncope	0.0	2.5
Subsequent therapy			Parikh et al.^12
Drug costs per cycle
Second line	$3,382	$5,937
Third line	$353	$353
Weighted	$2,584.25	$4,465.79
Administration costs per cycle			Model calculation
Second line	$459	$459
Third line	$393	$393
Weighted	$441.98	$441.98

EGFR, epidermal growth factor receptor; FOLFIRI, leucovorin, 5-fluorouracil, and irinotecan; FOLFOX, leucovorin, 5-fluorouracil, and oxaliplatin; mCRC, metastatic colorectal cancer; mFOLFOX6, oxaliplatin +5-fluorouracil + leucovorin; NDC, National Drug Code; PFS, progression-free survival.

All costs inflated to July 2017 US dollars¹⁷.

Table 2. Other input parameters.

Input parameter	Value	Source
KRAS and RAS test	$377.98	CMS final 2014 National Limit Amount for CPT code 81275 (KRAS genetic test)^14. Palmetto GBA Medicare 2014 MoPath Tier 2 Fee Schedule, CPT code 81404 (NRAS) reimbursement rate (inflated to July 2017 US dollars)^16.
General practitioner office visit cost	$108.74	General practitioner visits were assumed to occur every 4 weeks. Costs were from CMS for CPT code 99214^14.
Oncology specialist office visit cost	$146.43	Visits to an oncology specialist are assumed to occur every treatment cycle (2 weeks). Costs were from CMS for CPT code 99215^14.
Computed tomography scan cost	$315.46	It was assumed that disease progression would be monitored every 8 weeks, similar to the protocol from the panitumumab clinical trial. Costs were from CMS for a computed tomography scan of the abdomen and pelvis with contrast (CPT 74177)^14.
MRI for resection surgery	$513.21	Cost of MRI (CPT 74183) of the abdomen and pelvis from CMS assumed to occur once prior to resection^14.
Surgical consult	$146.43	Costs are from CMS^14. Assumed CPT code 99215.
Resection surgery	$22,075.13	Weighted average of DRG 332 (with MCC) = $41,001, DRG 333 (with CC) = $22,185, and DRG 334 (without CC or MCC) = $16,134.
Grade 3/4 toxicity cost (hospital costs)		All costs inflated to July 2017 US dollars^17.
Pulmonary embolism	$13,987	Burudpakdee et al.^18
Diarrhea	$6,814	Assumed similar to vomiting from Burudpakdee et al.^18
Sepsis	$7,623	Assumed similar to anemia from Burudpakdee et al.^18
Dehydration	$6,842	Assumed similar to dyspnea from Burudpakdee et al.^18
Febrile neutropenia	$15,452	Burudpakdee et al.^18
Gastroesophageal reflux disease	$6,814	Assumed similar to vomiting from Burudpakdee et al.^18
Pneumonia	$7,623	Assumed similar to anemia from Burudpakdee et al.^18
Deep vein thrombosis	$10,723	Burudpakdee et al.^18
Pyrexia	$6,842	Assumed similar to dyspnea from Burudpakdee et al.^18
Urinary tract infection	$8,445	Burudpakdee et al.^18
Vomiting	$6,814	Burudpakdee et al.^18
Infection	$7,623	Assumed similar to anemia from Burudpakdee et al.^18
Intestinal perforation	$39,767	Burudpakdee et al.^18
Syncope	$6,842	Assumed similar to dyspnea from Burudpakdee et al.^18
End-of-life cost	$26,788	Estimated from end-of-life costs from Chastek et al.^19 and inflated to July 2017 US dollars^17
Best supportive care costs per day	$164	Cost per day from Shankaran et al.^21 and inflated to July 2017 US dollars^17
Utility weights
Progression-free	0.821	Wild-type RAS from first-line NCT00364013 trial
Progressive disease
Subsequent active treatment	0.782	Wild-type KRAS from second-line 20050181 trial
Best supportive care	0.681	Patients with wild-type KRAS mCRC treated with best supportive care in the third-line 20020408 trial
Disease-free survival following resection
Age, years
35–44	0.890	US general population utility weights from Fryback et al.^20
45–54	0.870
55–64	0.850
65–74	0.860
75–89	0.840

CC, complication or comorbidity; CMS, Centers for Medicare & Medicaid Services; CPT, Current Procedural Terminology; DRG, Diagnosis Related Group code; FOLFOX, leucovorin, 5-fluorouracil, and oxaliplatin; MCC, major complication or comorbidity; MRI, magnetic resonance imaging.

Scheduled physician visits did not include visits for grade 3/4 toxicity treatment or management that occurred on days other than the day of treatment administration. Also, costs of scheduled physician visits during best supportive care after disease progression were captured as part of the best supportive care costs and were not included in these cost estimates because the frequency of office visits was likely to be different.

Utility weights

Utility weights used in the model were based on EuroQol-5 Dimension (EQ-5D) questionnaire responses from patients with wild-type RAS mCRC in the first-line PRIME (NCT00364013) clinical trial^4,22, from patients with wild-type KRAS mCRC in the second-line panitumumab clinical trial²³, and from BSC patients in the third-line panitumumab clinical trial (Table 2)²⁴. For second-line therapy and beyond, we assumed that utility weights for patients with wild-type RAS mCRC were similar to utility weights for those with wild-type KRAS mCRC due to lack of biomarker analysis for wild-type RAS utilities in these trials at the time of the present analysis. Utility weights were calculated by averaging all EQ-5D responses before disease progression in each respective trial and treatment line, using the Dolan²⁵ algorithm. Patients who are disease-free following resection are assumed to have similar utility weights to that of the general population.

Analyses

The model outcomes calculated for each first-line treatment regimen included patient survival (life-years), quality-adjusted life-years (QALYs), costs for healthcare resources, and incremental cost per life-year and per QALY gained. All costs were reported in 2017 US dollars¹⁷, and all costs and outcomes (benefits) in the model were discounted using the suggested discount rate in the US of 3.0% per annum²⁶.

To test the robustness of the model methods, assumptions, and specific parameters, we examined the effect of using alternative methods and data sources for the model inputs in a series of focused scenario analyses conducted around the assumptions and methods used to calculate drug acquisition costs, subsequent treatment, and utility weights.

In addition, a probabilistic sensitivity analysis was performed to examine the effects of joint uncertainty across all parameters of the model. The results of the probabilistic sensitivity analysis were summarized using cost-effectiveness scatter plots (not shown) and cost-effectiveness acceptability curves. Supplementary tables containing the probabilistic sensitivity analysis input parameters modeled are provided.

Results

Table 3 summarizes the cost-effectiveness results of the deterministic analysis (base-case assumptions: Weibull OS, flexible spline PFS, all patients have disease progression at 5 years). The model projected 3.527 life-years for panitumumab plus mFOLFOX6 and 2.950 life-years for bevacizumab plus mFOLFOX6 (both arms discounted 3% per annum). Adjusting for quality of life, panitumumab plus mFOLFOX6 was estimated to produce 2.679 QALYs, and bevacizumab plus mFOLFOX6 was estimated to produce 2.233 QALYs (both arms discounted 3% per annum). For panitumumab plus mFOLFOX6 and bevacizumab plus mFOLFOX6, drug acquisition costs made up 47% and 40% of total costs modeled, respectively, with BSC costs contributing the second greatest proportion of costs (27% and 30% of total). Because of greater PFS (longer duration of therapy) and, hence, higher drug acquisition costs, total drug costs were higher for panitumumab plus mFOLFOX6 than for bevacizumab plus mFOLFOX6 ($132,287 vs $89,428). Similarly, costs for administration, physician visits, monitoring for disease progression, and BSC were higher for panitumumab plus mFOLFOX6 than for bevacizumab plus mFOLFOX6 because of longer survival.

Table 3. Deterministic results.

Outcome/cost category	Panitumumab + mFOLFOX6	Bevacizumab + mFOLFOX6	Difference
Outcome
Patient survival (undiscounted)	3.891	3.214	0.677
Life-years	3.527	2.950	0.577
QALYs	2.679	2.233	0.445
Cost
RAS test	$378	$378	$0
Biologic and chemotherapy drug	$132,287	$89,428	$42,859
Administration	$12,377	$10,482	$1,895
Toxicity treatment and management	$2,390	$3,598	–$1,208
Physician visits	$8,183	$6,763	$1,421
Monitoring for disease progression	$3,214	$2,656	$558
Best supportive care	$76,518	$66,239	$10,278
Resection	$25,530	$20,549	$4,982
End-of-life costs	$23,780	$24,279	–$499
Total costs	$284,658	$224,372	$60,286
Incremental cost-effectiveness ratios of panitumumab + mFOLFOX6 vs bevacizumab + mFOLFOX6
Incremental cost per
Life-year gained	$104,550
QALY gained	$135,391

mFOLFOX6, oxaliplatin +5-fluorouracil + leucovorin; QALY, quality-adjusted life-year.

The incremental cost per life-year gained of panitumumab plus mFOLFOX6 vs bevacizumab plus mFOLFOX6 was estimated to be $104,550, and the incremental cost per QALY gained was estimated to be $135,391.

Scenario analysis conducted around major model assumptions indicated that the model was robust to alternative assumptions of PFS and OS distributions, vial wastage, utilization of RAS testing, and bevacizumab dose (Table 4).

Table 4. Scenario analysis results.

Alternative input parameter scenario	Incremental cost per QALY gained	Change from base case, %
Base case analysis	$135,391
Parametric survival analysis
Weibull PFS	$131,393	–2.95
Flexible spline OS	$136,658	0.94
PFS projections not limited	$104,775	–22.61
Wastage
No vial wastage (vial sharing)	$137,088	1.25
Wastage with no dose rounding	$149,254	10.24
RAS test costs for bevacizumab
No RAS test costs for bevacizumab	$136,240	0.63
Bevacizumab dose, mg/kg
5.5	$131,131	–3.15
6.0	$126,870	–6.29
6.5	$122,609	–9.44
7.0	$118,348	–12.59
7.5	$114,087	–15.73

OS, overall survival; PFS, progression-free survival; QALY, quality-adjusted life-year.

The cost-effectiveness acceptability curve resulting from the probabilistic sensitivity analysis conducted (Figure 4) indicated that 60.7% of simulations were below a willingness-to-pay threshold of $150,000. Additionally, 99% of simulations performed showed panitumumab plus mFOLFOX6 to be more effective and more costly or more effective and less costly than bevacizumab plus mFOLFOX6.

Figure 4. Cost-effectiveness acceptability curve. CE, cost-effectiveness; mFOLFOX6, oxaliplatin + 5-fluorouracil + leucovorin.

Discussion

This analysis used patient-level data from the PEAK study and integrated it with the available information on quality of life and costs to assess the cost-effectiveness of panitumumab plus mFOLFOX6 compared with bevacizumab plus mFOLFOX6 in patients with wild-type RAS mCRC from the perspective of a US payer. By quantifying the comparative value of treatments, this analysis aims to support patients, clinicians, and payers in making informed decisions on which agent to use for the first-line treatment of wild-type RAS mCRC, which may ultimately help maximize patient outcomes.

This economic evaluation has a number of strengths. In particular, the PEAK study is the only head-to-head first-line trial of panitumumab and bevacizumab in patients with wild-type RAS mCRC. Final analyses showed a significant difference in PFS (12.8 vs 10.1 months; hazard ratio [HR] = 0.68; 95% CI =0.48–0.96, p = 0.029) and a trend toward an OS benefit (36.9 vs 28.9 months; HR = 0.76; 95% CI =0.53–1.11, p = 0.15).

Similar to a previously conducted European-focused analysis²⁷, panitumumab plus mFOLFOX6 was found to be a cost-effective option for first-line treatment. Although increased costs were seen with panitumumab plus mFOLFOX6, owing largely to more treatment administrations in the longer PFS period, they were balanced by longer survival and QALYs, resulting in an incremental cost per QALY gained of $135,391. As the US does not have established willingness-to-pay thresholds like many European countries with established health technology agencies, $150,000 per QALY gained has often been cited as a target threshold. This threshold may be considered conservative from some oncology physician perspectives^28–31; therefore, willingness-to-pay thresholds up to $300,000 per QALY gained in oncology have been used and would be generally accepted³². Compared with previous studies²⁷, this analysis is based on more accurate methods to extrapolate PFS beyond the trial duration. PFS survival curves are fitted using flexible cubic splines that resulted in a better fit to the PFS data. Additionally, this analysis used the most mature data to model both PFS and OS; thus, post-trial extrapolations are more robust. Additionally, all patients were transitioned to the progressive disease health state based on the number of patients remaining at risk in the trial⁷, rather than following the PFS curve. Scenario analyses with alternative approaches to PFS modeling (traditional Weibull, not limiting PFS projections) and OS modeling (flexible spline) were also presented. All resulted in incremental cost-effectiveness ratios lower than $150,000.

The model utilized the uncertainty around the efficacy, cost, and utility weight estimates to provide a level of confidence of panitumumab plus mFOLFOX6 being cost-effective via the probabilistic sensitivity analysis. In this analysis, a majority of the iterations (60.7%) indicated panitumumab plus mFOLFOX6 to be under the $150,000 willingness-to-pay threshold.

It should be noted that this analysis was focused for the US market only, and results may be different in other countries where drug acquisition cost differences between the two drugs are different. However, previous cost-effectiveness analyses conducted in France, the Czech Republic, and Spain using data from the PEAK trial have shown that panitumumab is cost-effective^27,33,34 and, given that this analysis evaluated a similarly selected population of patients with RAS wild-type tumors with greater efficacy, similar results to those studies would be expected.

As with any model that uses clinical trial data to extrapolate costs and benefits, there are several limitations, and the results presented should be interpreted within the context of the data inputs and modeling assumptions adopted. First, utility weights used in the model could not be estimated from the PEAK trial, as it did not collect utility weight data. Instead, wild-type RAS utility weights estimated from patient-level data in the first-line panitumumab PRIME trial were used³⁵. These pre- and post-progression weights are similar to those found in the literature for mCRC. Subsequent line utility weights were based on wild-type KRAS utility weights from the second-line 20050181 trial²³. The data on duration of therapy from Parikh et al.¹² represent all patients who received some first-line agent and are not specific to those who received first-line treatment with panitumumab or bevacizumab. Thus, real-world treatment patterns may not perfectly match the assumptions made for the model population. Next, as with the majority of oncology clinical trials, the data on post-progression therapies and resource use were limited in PEAK. A retrospective study of Surveillance, Epidemiology, and End Results (SEER) data was used to model the time and costs of subsequent therapy¹². Assumptions had to be made (e.g. patients would not be treated with the same therapy in subsequent lines of treatment if they had experienced progression on that therapy), and it should be noted that the SEER analysis was based on data from elderly patients only. A change in assumptions and/or practice patterns could potentially affect the cost-effectiveness conclusions. Further scenario analyses are presented assuming a different dosage for bevacizumab, which is often administered at higher doses than its licensed amount in first-line treatment (Quintiles IMS Oncology Services Comprehensive Electronic Records [OSCER] data)³⁶.

As noted previously, the PEAK trial is the only first-line, head-to-head study of panitumumab and bevacizumab in patients with wild-type RAS mCRC. Although the phase II PEAK study is limited by small sample size, its results are generally consistent with larger first-line phase III studies that compared the anti-EGFR therapy cetuximab vs bevacizumab. In the European phase III FIRE-3 study, which compared first-line FOLFIRI (leucovorin, fluorouracil, and irinotecan) plus cetuximab vs FOLFIRI plus bevacizumab, survival was superior for anti-EGFR therapy (HR =0.77; 95% CI =0.62–0.96; p = 0.017)³⁷. In the North American phase III CALGB/SWOG 80405 study, there was no survival difference between cetuximab and bevacizumab; however, follow-up analyses suggest that anti-EGFR therapies are superior for tumors originating from the left side of the colon^38,39. In aggregate, these studies confirm results from the PEAK trial and support the first-line benefit of anti-EGFR therapies for patients with left-sided colorectal tumors.

Conclusions

This model estimated the cost-effectiveness of panitumumab plus mFOLFOX6 compared with bevacizumab plus mFOLFOX6 in the first-line treatment of patients with wild-type RAS mCRC. Results from the model of incremental cost-effectiveness ratios of $104,550 per life-year gained and $135,391 per QALY gained indicate that panitumumab plus mFOLFOX6 is a cost-effective treatment option according to commonly used willingness-to-pay thresholds. Sensitivity and scenario analyses show the robustness of the model to alternative parameters, assumptions, and model uncertainty. Thus, physicians and payers should consider panitumumab plus mFOLFOX6 as a cost-effective option in first-line treatment of patients with wild-type RAS mCRC.

Transparency

Declaration of funding

This study was conducted by RTI Health Solutions (RTI-HS) under the direction of Amgen Inc. and was funded by Amgen Inc. RTI-HS receives funding from pharmaceutical, biotechnology, and medical device companies to conduct pharmacoeconomic and outcomes research studies, including funding for this analysis.

Declaration of financial/other interests

CNG and HNK are employees of RTI Health Solutions. HNK received sponsorship and research funding from Amgen Inc. for this analysis. AC, GH, and TG are employees of Amgen Inc.; AC and GH also own stock in the company. LS is an employee of Amgen GmbH. JHS is a consultant/advisory board member for Amgen Inc., AstraZeneca, Bayer, Celgene, Chugai, and Genentech/Roche, and has received research funding from AbbVie, Exelixis, Genentech/Roche, Gilead Sciences, Leap Therapeutics, Macrogenics, MedImmune, OncoMed, Sanofi Genzyme, and Seattle Genetics. Peer reviewers on this manuscript have received an honorarium from JME for their review work, but have no other relevant financial relationships to disclose.