Cost-effectiveness analysis of neoadjuvant pertuzumab and trastuzumab therapy for locally advanced, inflammatory, or early HER2-positive breast cancer in Canada

Abstract

Objective:

The NeoSphere trial demonstrated that the addition of pertuzumab to trastuzumab and docetaxel for the neoadjuvant treatment of HER2-positive locally advanced, inflammatory, or early breast cancer (eBC) resulted in a significant improvement in pathological complete response (pCR). Furthermore, the TRYPHAENA trial supported the benefit of neoadjuvant dual anti-HER2 therapy. Survival data from these trials is not yet available; however, other studies have demonstrated a correlation between pCR and improved event-free survival (EFS) and overall survival (OS) in this patient population. This study represents the first Canadian cost-effectiveness analysis of pertuzumab in the neoadjuvant treatment of HER2-positive eBC.

Methods:

A cost-utility analysis (CUA) was conducted using a three health state Markov model (‘event-free’, ‘relapsed’, and ‘dead’). Two separate analyses were conducted; the first considering total pCR (ypT0/is ypN0) data from NeoSphere, and the second from TRYPHAENA. Published EFS and OS data partitioned for patients achieving/not achieving pCR were used in combination with the percentage achieving pCR in the pertuzumab trials to estimate survival. This CUA included published utility values and direct medical costs including drugs, treatment administration, management of adverse events, supportive care, and subsequent therapy. To address uncertainty, a probabilistic sensitivity analysis (PSA) and alternative scenarios were explored.

Results:

Both analyses suggested that the addition of pertuzumab resulted in increased life-years and quality-adjusted life-years (QALYs). The incremental cost per QALY ranged from $25,388 (CAD; NeoSphere analysis) to $46,196 (TRYPHAENA analysis). Sensitivity analyses further support the use of pertuzumab, with cost-effectiveness ratios ranging from $9230–$64,421. At a threshold of $100,000, the addition of pertuzumab was cost-effective in nearly all scenarios (93% NeoSphere; 79% TRYPHAENA).

Conclusion:

Given the improvement in clinical efficacy and a favorable cost per QALY, the addition of pertuzumab in the neoadjuvant setting represents an attractive treatment option for HER2-positive eBC patients.

Keywords::

• Cost-effectiveness analysis
• Cost-utility analysis
• Pertuzumab
• Trastuzumab
• Neoadjuvant
• HER2-positive
• Breast cancer

Introduction

Breast cancer is a heterogeneous disease that can be divided into clinical sub-types based on molecular characterization. Human epidermal growth factor receptor 2 (HER2)-positive breast cancers account for ∼15–30% of all breast cancers1–5. This sub-type has a more aggressive phenotype than the HER2-negative sub-type and is associated with an adverse prognosis, including poor rates of disease-free survival (DFS) and overall survival (OS)3^,5–8 (Supplemental Table 1).

Treatment regimens are largely dependent on the breast cancer sub-type and, thus, tailored to the individual patient. In the case of HER2-positive breast cancers, therapies that specifically target the HER2 protein (anti-HER2 therapies) are particularly important1. Trastuzumab (Herceptin® ), an agent that specifically binds to the extracellular domain IV of HER2, was the first anti-HER2 therapy approved for the treatment of HER2-positive breast cancer, initially in the metastatic setting for advanced disease, and subsequently in the adjuvant then neoadjuvant settings for early-stage disease. When used with chemotherapy, trastuzumab significantly improves both the DFS and OS of patients with HER2-positive breast cancer9–13. However, in recent clinical trials of neoadjuvant or adjuvant trastuzumab and chemotherapy, 5-year progression/relapse rates for early HER2-positive breast cancer ranged from 16–40%, depending on the baseline disease stage and tumor characteristics of enrolled patients13–16. For example, in the NOAH study, patients with locally advanced or inoperable early-stage HER2-positive breast cancer who received neoadjuvant and adjuvant trastuzumab and chemotherapy had a 5-year event-free survival (EFS) of 58%; thus, 42% of patients experienced a progression, relapse, or death13.

Although trastuzumab treatment significantly improves the prognosis of HER2-positive patients, there is still a need to reduce relapse rates and improve OS in these patients. This has led to the development of additional HER2-targeted drugs. Pertuzumab (Perjeta® ) was recently developed for use in combination with trastuzumab and chemotherapy. Treatment with pertuzumab, a monoclonal antibody against the HER2 dimerization domain (sub-domain II), is associated with improved clinical outcomes when used in combination with trastuzumab and chemotherapy (dual HER2 blockade), compared with trastuzumab monotherapy, in both the neoadjuvant and metastatic settings17.

Traditionally, novel agents for the treatment of breast cancer have first been approved in the metastatic setting for patients with limited treatment options and shortened OS. Approval of drugs in the adjuvant setting, for the treatment of early or locally advanced disease, is often not granted for many years, as it requires large randomized trials with long-term follow-up to demonstrate a survival benefit in patients with early-stage disease. New therapies are being tested in the neoadjuvant setting (before definitive surgery), where efficacy can be assessed after several months (6–8 cycles of therapy), thus potentially expediting patient access to promising therapies18.

Assessment of efficacy in the neoadjuvant setting relies on the use of surrogate end-points, as long-term survival data would not yet be available. In order for a surrogate end-point to be an effective predictor of a long-term clinical outcome, the surrogate must correlate with the long-term outcome and fully capture the net effect of treatment on the outcome of interest19. Pathological complete response (pCR; ypT0/is ypN0) has been proposed as a surrogate end-point for the prediction of long-term clinical benefit, EFS, and OS, in patients with early breast cancer. The prognostic significance of pCR in neoadjuvant therapy has been confirmed with several trials, and the correlation between pCR and long-term survival is particularly strong in patients with HER2-positive disease13^,20–27. Recently, the Food and Drug Administration (FDA) spearheaded the formation of the Collaborative Trials in Neoadjuvant Breast Cancer (CTNeoBC) working group to conduct a meta-analysis and address the unresolved issues surrounding the use of pCR as a surrogate end-point for EFS and OS28. In their analysis, the CTNeoBC working group established a patient-level correlation between pCR and long-term survival. This correlation was important for the approval of pertuzumab in the neoadjuvant setting in the US, which was based on pCR data from NeoSphere, the pivotal neoadjuvant pertuzumab trial28^,29.

In the NeoSphere trial, patients receiving pertuzumab in combination with trastuzumab and docetaxel had a significantly increased frequency of pCR compared with patients receiving trastuzumab and docetaxel (ypT0/is ypN0: 39.3% [95% confidence interval (CI) = 30.0–49.2] vs 21.5% [95% CI = 14.1–30.5], p = 0.0063)30. In addition, high pCR rates (54.7–63.6%) were observed in the TRYPHAENA trial, when neoadjuvant pertuzumab and trastuzumab were given either sequentially or concomitantly with an anthracycline (i.e., FEC [fluorouracil, epirubicin, cyclophosphamide]) followed by taxane therapy (i.e., docetaxel), or with carboplatin and docetaxel31.

In Canada, pertuzumab is approved in combination with trastuzumab and docetaxel for the first-line treatment of HER2-positive metastatic breast cancer, with the approval based on results from the CLEOPATRA trial32–35. We report here a Canadian cost-utility analysis (CUA) based on separate analyses conducted with data from the NeoSphere and TRYPHAENA trials for the use of pertuzumab as an add-on therapy to a neoadjuvant treatment regimen containing trastuzumab and chemotherapy.

Methods

Objective

The objective of this economic evaluation was to assess the long-term benefits, costs, and cost-effectiveness of pertuzumab added to a neoadjuvant treatment regimen containing trastuzumab and chemotherapy for patients with locally advanced, inflammatory, or early HER2-positive breast cancer.

Model overview and design

A Markov state transition model was constructed in Microsoft Excel. The analysis was performed from the perspective of the Canadian healthcare payer, with Ontario as a reference province.

The transition model simulated the movement of a hypothetical cohort of patients through three health states—‘event-free’, ‘relapsed’, and ‘dead’ (Figure 1). One arm of the model considered treatment with pertuzumab added to a neoadjuvant regimen containing trastuzumab and chemotherapy, while the other arm considered a neoadjuvant regimen containing trastuzumab and chemotherapy alone. Patients began in the ‘event-free’ state. During each 1 month model cycle, patients could reside in the ‘event-free’ state or move to the ‘relapsed’ or ‘dead’ states. The process was repeated for a 28-year time horizon36. Patients were attributed survival, quality-adjusted survival, and costs for each cycle spent in a given state. Adverse event (AE) costs were also incorporated into the model.

Figure 1. Simplified schematic of the model. The model included three health states—event-free, relapsed, and dead—that were informed from EFS and OS data.

Movement of patients from a living state to ‘dead’ was based on OS data (described below), while EFS data (described below) were used to inform the percentage of event-free patients at each model cycle. The remaining patients at each cycle were in the relapsed state (i.e., relapsed = total – [dead + event-free]).

This CUA included direct medical costs borne by the Canadian payer, including all drugs, treatment administration, AE management, supportive care, and subsequent therapy. Health state utility estimates36 and Canadian costs were taken from the published literature and Ontario cost databases. Where appropriate, costs were inflated to 2014 Canadian dollars using the healthcare component of the consumer price index. Costs and benefits were discounted at a rate of 5% per annum37. All key model inputs are summarized in Table 1 and key model assumptions are provided in Table 2.

Table 1. Summary of key model inputs.

Input parameter	With Pertuzumab	Without Pertuzumab
Patient demographics
Median age, years30	50.0
Weight, kg38	73.33
BSA, m^2^a	1.78
Effectiveness30
NeoSphere analysis
% pCR (ypT0/is ypN0)	39.3%	21.5%
TRYPHAENA analysis
% pCR (ypT0/is ypN0)	63.6%	45. 9%*
Unadjusted Health State Utilities^b
Event-free, year 1	0.97
Event-free, year 2 on	0.99
Local recurrence	0.75
Metastatic disease	0.65
Weighted utility for relapsed^c	0.68
Costs (average cost per patient, CAD 2014)
NeoSphere analysis
Total drug cost for complete regimen^d	$72,468	$56,583
Total administration cost for complete regimen^e	$4114	$3824
TRYPHAENA analysis
Total drug cost for complete regimen^d	$84,728	$62,489
Total administration cost for complete regimen^e	$4232	$3805
Both analyses
Average total cost for management of AEs per patient^f	$624	$543
Supportive care for event-free, years 1–258–60	$238/yr
Supportive care for event-free, years 3–558–60	$160/yr
Supportive care for event-free, year 6 on58–60	$128/yr
1st line subsequent therapy for relapsed^g	$204,671 (one-time cost)
2nd line and beyond subsequent therapy for relapsed^g	$92,471 (one-time cost)

AE, adverse event; BSA, body surface area; pCR, pathological complete response (ypT0/is ypN0).

*Value was estimated based on the absolute difference in pCR between the treatment and comparator arms from the NeoSphere trial.

^aBSA calculated using the Dubois method38^,39.

^bBreast cancer-specific health state utilities were from a published Canadian CEA for adjuvant trastuzumab36; these unadjusted utilities were applied to Canadian age- and sex-matched health utilities43.

^cBased on 25% of relapsed patients having a local recurrence and 75% having metastatic disease11^,69.

^dDrug costs were calculated based on unit costs from Hoffmann-La Roche or previous publications44–47.

^eAdministration costs for the various drug regimens included overhead, nursing, pharmacy, and clinical consultation48^,49^,51–55.

^fIncidence of AEs was from the NeoSphere trial30. The costs of treating AEs were calculated based on the percentage of patients experiencing each AE that would require hospitalization (informed by expert opinion). Costs were obtained from the Ontario Case Costing Initiative, or determined by micro-costing for drugs, physician visits, and additional resources (informed by expert opinion)47^,56^,70.

^gSee Table 2 for assumptions regarding first-line and second-line subsequent therapy.

Table 2. Summary of key model assumptions for the base-case analyses.

Parameter	Assumption
Efficacy	• pCR (ypT0/is ypN0) is an acceptable surrogate for long-term survival
	• EFS and OS of patients treated with pertuzumab are dependent only on whether or not they achieve pCR
	• Survival of patients in the Kim et al.25 study is reflective of pertuzumab-treated HER2-positive patients
	• After 10 years, surviving patients were assumed to have a mortality rate equal to that of the age- and sex-matched general population
	• TRYPHAENA Arm C best reflects predicted Canadian clinical practice for neoadjuvant therapy
	• The absolute difference in pCR observed in NeoSphere (i.e., 17.8%) would have been observed in TRYPHAENA if the hypothetical comparator had been evaluated
Safety	• The adverse events captured in the NeoSphere trial adequately reflect any differences between treatment with or without pertuzumab
	• The TRYPHAENA analysis uses the same incidence of AEs, since data were not available to inform relative AEs for the derived comparator arm
Relapse	• 25% of patients that relapsed had a local recurrence and 75% had metastatic disease11^,69
	• To calculate the total cost of relapse, it was assumed that all patients that died in the first 10 years had first moved through the relapsed state
Subsequent therapy	• 100% of patients that entered the relapsed state, regardless of whether it was local or metastatic, received first-line metastatic therapy and 88% also received second-line metastatic therapy
	• For patients receiving first-line subsequent therapy, 80% received Per + Tra + Doc and 20% received Tra + Doc
	• For patients receiving second-line subsequent therapy, 80% received T-DM1 and 20% received Cap + Tra

AE, adverse events; Cap, capecitabine; Doc, docetaxel; EFS, event-free survival; HER2, human epidermal growth factor receptor 2; OS, overall survival; pCR, pathological complete response; Per, pertuzumab; T-DM1, trastuzumab emtansine (Kadcyla® [Kadcyla is a registered tradename of Hoffmann-La Roche AG, in Basel, Switzerland]); Tra, trastuzumab.

Patient population

The model was populated with data from the NeoSphere or TRYPHAENA trials, which included treatment-naïve women with locally advanced, inflammatory, or early HER2-positive breast cancer (tumor diameter > 2 cm)30^,31. The mean weight and body surface area (BSA) used to calculate dosing were reflective of Canadian breast cancer patients (Table 1)38^,39.

Treatment regimens from NeoSphere and TRYPHAENA

Two separate base-case analyses were conducted using treatment regimens and efficacy data from NeoSphere and TRYPHAENA. In the NeoSphere trial, early breast cancer patients received four cycles of neoadjuvant therapy, which were administered intravenously every 3 weeks, with an additional three cycles of anthracycline chemotherapy (i.e., FEC) provided after surgery (Figure 2A)30. Trial Arm A from NeoSphere was used as the comparator arm in the model, while trial Arm B was used as the pertuzumab treatment arm. The NeoSphere trial provided the most complete data set for neoadjuvant pertuzumab; however, clinical advisors suggested that the treatment regimen used in NeoSphere may not fully reflect current Canadian practice. Specifically, in NeoSphere, FEC was provided after surgery, but patients in Canada are usually treated with 6 to 8 cycles of chemotherapy plus HER2-targeted therapies, all provided prior to surgery. In order to better reflect Canadian treatment patterns, an additional analysis was conducted using data from Arm C of the TRYPHAENA trial (Figure 2B). The rationale for choosing Arm C was that TCH (docetaxel, carboplatin, trastuzumab) is an accepted neoadjuvant treatment regimen in Canada, while Arm A would not be used by most physicians, as there is concern regarding potential cardiotoxicity associated with concomitant administration of an anthracycline (i.e., epirubicin) with trastuzumab and pertuzumab. Although Arm B, with sequential FEC then docetaxel with trastuzumab and pertuzumab, would likely be used in Canada, physicians expected that additional cycles would be used, since this regimen only included three neoadjuvant cycles with the dual HER2 blockade. The TRYPHAENA trial lacks an appropriate comparator arm for modeling cost-effectiveness, as all regimens contained pertuzumab; thus, one was estimated for the analysis (see below).

Figure 2. Trial design of NeoSphere30 (A) and TRYPHAENA31 (B). In NeoSphere (A), patients were given four cycles of neoadjuvant therapy and in TRYPHAENA (B), they were given six cycles of neoadjuvant therapy. From the NeoSphere trial, Arm A was used as the comparator arm for the model and Arm B was used as the model treatment arm. From TRYPHAENA, Arm C was chosen for the base-case analysis to better reflect Canadian clinical practice. The following drug doses were used: pertuzumab loading dose of 840 mg and maintenance doses of 420 mg; trastuzumab loading dose of 8 mg/kg, maintenance doses of 6 mg/kg; docetaxel was given at an initial dose of 75 mg/m², with dose escalation to 100 mg/m² permitted, if tolerated; FEC in the NeoSphere trial was 600 mg/m² 5-fluorouracil, 90 mg/m² epirubicin, 600 mg/m² cyclophosphamide; FEC in the TRYPHAENA trial was 500 mg/m² 5-fluorouracil, 100 mg/m² epirubicin, 600 mg/m² cyclophosphamide; carboplatin was used at a dose of AUC 6 (median 707.5 mg).

Effectiveness assessment

Pathological complete response

The key effectiveness input in the model was pCR (ypT0/is ypN0). This is defined as the complete absence of any residual invasive disease in the breast and all sampled ipsilateral lymph nodes following completion of neoadjuvant systemic therapy, irrespective of residual in situ disease, and is the definition preferred by the FDA28^,40. From the NeoSphere trial, 39.3% and 21.5% of patients achieved pCR in Treatment Arm B (with pertuzumab) and Arm A (without pertuzumab), respectively30. These efficacy values were used in the model to inform the percentage of patients partitioned to the pCR curves for EFS and OS for the treatment and comparator arms, respectively (see below).

In the TRYPHAENA trial, 63.6% of patients in Arm C (TCH + pertuzumab) achieved pCR31. This was used to inform the treatment arm of the TRYPHAENA base-case analysis. As there was no comparator arm from the trial (i.e., TCH), the percentage of pCR for a hypothetical comparator arm was estimated using the absolute difference in pCR observed between treatment regimens with and without pertuzumab in the NeoSphere study (17.8%)30. Thus, the comparator arm was estimated to have 45.9% of patients achieving pCR, corresponding to a relative risk (RR) of 1.39. The absolute difference in pCR was used to calculate the comparator for the TRYPHAENA base-case analysis as it was a more conservative approach than using the RR calculated from the NeoSphere trial (1.83). To address the uncertainty surrounding this estimate, we conducted two alternative analyses and several sensitivity analyses (described below).

Event-free survival and overall survival

As the model required a long-term time horizon and EFS and OS data for patients treated with neoadjuvant pertuzumab were not yet available, separate survival data for patients who did or did not achieve pCR were obtained from a retrospective analysis of HER2-positive patients treated with neoadjuvant trastuzumab and chemotherapy25. Thus, the model assumed that the EFS and OS of patients treated with pertuzumab in combination with trastuzumab and chemotherapy was dependent on whether the patient achieved pCR or not, an assumption that was supported by clinical data13^,20–28 and Canadian clinical advisors.

The EFS and OS curves published by Kim et al.25, comparing survival of patients with or without pCR, were digitized using TechDig software. Data from the digitized curves were used to extrapolate survival beyond the 10-year follow-up period using SigmaPlot v12.5 software. Various parametric fits were considered, with the Gompertz curve best fitting the original Kaplan–Meier data. The survival function of the no pCR curve was used to generate the survival function of the pCR curve, based on the published hazard ratios (HR) for EFS (4.09; 95% CI = 1.67–10.04) and OS (4.15; 95% CI = 1.39–12.38)25. The Gompertz extrapolated curves for EFS and OS had long tails that were not clinically relevant as patients would live indefinitely. To create more meaningful curves, the mortality rates from the Gompertz curves were used for the first 10 years of the model; at which time the general population mortality rates were used for surviving patients. Ten years was chosen as the time to switch to the general population mortality rate, since the published data included up to 10 years and expert opinion suggested that after 10 years patients were less likely to experience disease-related progression or relapse than they were in the first 10 years. A survival curve for the general population of Canadian females was generated using data from Statistics Canada life tables for 2009–2011, accounting for the age of patients at the start of the model (i.e., 50 years)41. These adjusted EFS and OS curves were then weighted for the percentage of patients who did or did not achieve pCR in order to generate the treatment and comparator curves used in the model (Figure 3).

Figure 3. Modeled event-free survival (A) and overall survival (B). The EFS (A) and OS (B) curves for patients achieving pCR or not from the Kim et al.25 analysis were used to calculate the weighted curves for the treatment and comparator arms based on the percentage of patients attaining pCR in the NeoSphere or TRYPHAENA trials30^,31. The general population OS curve for age-adjusted Canadian females is also shown (B).

Quality-of-life associated with disease states

Patients in the event-free state are assumed to have better quality-of-life (QoL) than those in the relapsed state. Health state utilities were obtained from a published Canadian cost-effectiveness analysis of adjuvant trastuzumab in HER2-positive breast cancer36. Hedden et al.36 used adjusted utilities that were derived from a systematic review of CUAs in oncology that included a review of breast cancer-specific health utilities42. These breast cancer-specific health state utilities (summarized in Table 1) were used as multipliers and applied to Canadian age- and sex-matched general population utilities43.

This model did not include utility decrements for AEs, since the AEs reported in the NeoSphere trial were primarily associated with chemotherapy, with very little differences between the pertuzumab and comparator arms30. As such, applying utility decrements for adverse events, which occurred during the initial treatment period, was not expected to influence overall model results.

Costs and resource utilization

Cost of drug treatment

Costs for each treatment arm were based on the treatment regimens delivered in NeoSphere or TRYPHAENA trials30^,31, as shown in Figure 2. The model assumed no wastage of drugs, since pertuzumab is provided at a pre-specified dose (420 mg) and trastuzumab is stable for 28 days, which would allow for sharing with consecutive patients. Unit costs for pertuzumab and trastuzumab were obtained from Hoffmann-La Roche. Unit costs for other chemotherapy agents were obtained from published sources or Ontario databases44–47. The average total drug costs per regimen for the full treatment course, based on a weight of 73.33 kg and BSA of 1.78 m², are summarized in Table 1.

Drug administration costs included overhead, nursing, pharmacy, and clinical consultation costs, which were based on workload times reported for various treatment regimens48–52 and hourly fees53–56. The total drug administration costs per treatment regimen for the full treatment course are summarized in Table 1.

Cost of adverse events

The most common AEs, grade 3 or higher, reported in the NeoSphere trial were incorporated into the model30. The types of AEs and the percentage of patients experiencing each AE from the NeoSphere trial were used for both analyses as there were no data available to inform the AEs associated with the hypothetical comparator arm for the TRYPHAENA analysis. The percentage of patients experiencing each AE was multiplied by the average cost of treatment. The cost of treating each AE was calculated based on the percentage of patients requiring hospitalization for the AE, vs those treated as outpatients (informed by expert opinion). The corresponding costs of treatment based on each setting were determined and used to calculate a weighted average cost for each AE47^,56–58. The average total cost per patient for the treatment of all AEs is shown in Table 1.

Cost of supportive care

In the model, event-free patients were assumed to receive routine follow-up after completing treatment until death. The routine follow-up included in the model was based on Canadian clinical practice guidelines and expert opinion, with the associated costs obtained from the Ontario Schedule of Benefit for Physician Services56^,59^,60. In brief, patients were assumed to require a physician visit every 6 months for the first 2 years post-treatment, followed by an annual visit in subsequent years. All patients were also assumed to require an annual mammogram and 50% were assumed to require bone mineral density scans every 2 years due to an increased risk of developing osteoporosis associated with aromatase inhibitor therapy.

Cost of subsequent therapies

Patients who relapsed in the model were assumed to require subsequent therapies to treat either their local recurrence or metastatic disease. The model applied a cost of subsequent therapy to all patients who relapsed within the first 10 years. Of note, the model assumed that all patients that died within 10 years had first moved through the relapsed state. The rationale for not applying a cost beyond 10 years was that, as discussed earlier, beyond this point, patients were less likely to experience disease-related progression or relapse. Assumptions surrounding the treatment given to patients that progress are summarized in Table 2. The costs of first-line and second-line subsequent therapies and all treatment-related costs were obtained from Hoffmann-La Roche and based on previous economic evaluations32^,61.

Sensitivity analyses

Deterministic sensitivity analyses

In order to account for uncertainty in parameter estimates, we conducted several sensitivity analyses by varying individual parameters or several parameters simultaneously. Analyses were conducted to account for uncertainty surrounding costs, efficacy (ΔpCR), quality-of-life, time horizon, and discount rate. Where alternative values were not available in the literature, the base-case values were increased or decreased by 20% to understand how sensitive the model was to changes in the input parameters. Given that more than 60% of patients were still alive at the end of the 28 year time horizon, a 50 year time horizon was tested in a sensitivity analysis.

Alternative analyses for TRYPHAENA

Since there was no comparator data from the TRYPHAENA trial, we conducted alternative analyses to address the uncertainty surrounding the difference in pCR between the treatment arm and the hypothetical comparator. The first alternative analysis was conducted to address the uncertainty surrounding the timing of anthracycline chemotherapy, before or after surgery, and its impact on the relative efficacy of pertuzumab as an add-on therapy. In particular, the combination of an anthracycline regimen with a taxane and dual HER2 blockade given prior to surgery, as in TRYPHAENA, results in much higher pCR rates (56.2% and 54.7% for trial arms A and B) than when anthracycline is provided post-operatively, as in NeoSphere (39.3%)30^,31. It is possible that the absolute difference in pCR (ΔpCR) observed in NeoSphere for the addition of pertuzumab (17.8%)30 may over-estimate what may have been observed had there been a comparator arm in TRYPHAENA. To evaluate this, we calculated an adjusted ΔpCR based on the relative efficacy observed in other dual HER2 therapy trials with lapatinib plus trastuzumab, namely NeoALTTO and NSABP B-41, that differed only in the timing of the anthracycline therapy. NeoALTTO was similar in design to NeoSphere, with anthracycline provided post-operatively, while NSABP B-41 was similar to TRYPHAENA, with the anthracycline provided prior to surgery27^,62^,63. The adjusted ΔpCR between the hypothetical comparator arm (48.5% pCR) and Treatment Arm C from TRYPHAENA (63.6% pCR) was 15.1% (RR = 1.31).

A second alternative analysis was conducted to explore the possibility that the initial estimate for the TRYPHAENA hypothetical comparator arm may have under-estimated the added benefit of neoadjuvant pertuzumab. Using the ΔpCR of 17.8% from the NeoSphere trial30 to estimate the hypothetical comparator was a more conservative approach than using the calculated RR of 1.83 and may under-estimate the benefit. When the RR of 1.83 was used instead, the hypothetical comparator arm had 34.8% of patients achieving pCR, corresponding to a ΔpCR of 28.8%.

Probabilistic sensitivity analyses

A probabilistic sensitivity analysis (PSA) was performed to determine the impact of uncertainty surrounding model parameters using the variability around point estimates (i.e., costs, utilities, AEs, RR for pCR, and HRs for EFS/OS). Costs were varied using a gamma distribution, RRs and HRs were varied with a log-normal distribution, and utility values and percentages were varied with a beta distribution. Results are presented using a scatter plot of 10,000 iterations and a cost-effectiveness acceptability curve (CEAC).

Results

Effectiveness

The NeoSphere trial results suggested that dual HER2 therapy with pertuzumab in combination with trastuzumab and chemotherapy is more efficacious than trastuzumab with chemotherapy in achieving pCR, defined as the complete absence of invasive disease in the breast and lymph nodes (ypT0/is ypN0)30. In addition, high frequency of pCR was observed in the TRYPHAENA trial, with all treatment arms receiving dual HER2 therapy31. Both the NeoSphere and TRYPHAENA base-case analyses, which were based on an absolute increase in the frequency of pCR of 17.8%30, resulted in discounted gains of 0.333 life-years (LYs) or 0.310 quality-adjusted life-years (QALYs) (Table 3) with the addition of pertuzumab to the neoadjuvant treatment of HER2-positive early breast cancer.

Table 3. Summary of deterministic results of cost-effectiveness for the NeoSphere and TRYPHAENA base-case analyses.

	NeoSphere analysis		TRYPHAENA analysis
	With Pertuzumab (Per + Tra + Doc)	Without Pertuzumab (Tra + Doc)	With Pertuzumab (Per + TCH)	Without Pertuzumab (TCH)
LYs
Undiscounted	23.088	22.426	23.997	23.335
Discounted	13.333	13.000	13.790	13.457
QALYs
Undiscounted	18.926	18.323	19.753	19.150
Discounted	11.042	10.732	11.468	11.158
Costs (Discounted)
Drug costs*	$71,091	$55,062	$83,577	$61,226
Drug administration costs*	$4031	$3728	$4168	$3730
Cost of AEs	$615	$534	$617	$535
Cost of supportive care/subsequent therapy	$49,781	$58,315	$38,062	$46,596
Total	$125,518	$117,638	$126,423	$112,086
Incremental (Discounted)
LYs	0.333		0.333
QALYs	0.310		0.310
Costs	$7879		$14,337
Cost/LY (Discounted)	$23,658		$43,047
Cost/QALY (Discounted)	$25,388		$46,196

AE, adverse event; Doc, docetaxel; LY, life-year; Per, pertuzumab; TCH, docetaxel, carboplatin, trastuzumab; Tra, trastuzumab; QALY, quality-adjusted life-year.

*The reported drug costs and drug administration costs are for all neoadjuvant and adjuvant treatment cycles.

Cost outcomes

Based on the NeoSphere analysis, the pertuzumab treatment regimen was associated with an incremental cost of $7879 (discounted; Table 3). The primary contributor to the difference in costs was the additional drug acquisition cost of pertuzumab. Approximately half of this increased drug cost was offset through the prevention of relapse for pertuzumab-treated patients and the avoidance of the associated cost of subsequent therapy for relapsed patients (Table 3). On the basis of the TRYPHAENA analysis, the addition of pertuzumab to the neoadjuvant treatment regimen resulted in an increased cost of $14,337 (discounted; Table 3). Similar to the NeoSphere analysis, the primary contributor to the difference in costs was the additional drug acquisition cost of pertuzumab, which was partially offset by the prevention of relapse-associated costs (Table 3). In general, there were very little differences between the treatment and comparator arms in the average cost of treating AEs or the drug administration costs for either analysis (Table 3).

Cost-effectiveness

The NeoSphere trial base-case analysis resulted in an incremental cost per LY of $23,658 for the addition of neoadjuvant pertuzumab and an incremental cost per QALY of $25,388 (discounted), while the TRYPHAENA analysis predicted an incremental cost per LY of $43,047 and an incremental cost per QALY of $46,196 (discounted).

Sensitivity analyses and alternative analyses

Results of deterministic sensitivity analyses are summarized in Table 4. Varying inputs to account for uncertainty of parameter estimates suggested that the incremental cost-effectiveness ratio (ICER) for the NeoSphere analysis ranged between $9230–$38,419, while the ICER from the TRYPHAENA analysis was predicted to be between $18,262–$64,421. The first alternative analysis conducted for TRYPHAENA, based on an absolute difference in pCR of 15.1% to adjust for uncertainty surrounding the effect of the timing of chemotherapy, resulted in an incremental cost per QALY of $58,760. In contrast, the second alternative analysis for TRYPHAENA, using the relative risk of 1.83 calculated from the NeoSphere trial, which corresponded to an absolute difference in pCR of 28.8%, resulted in an incremental cost per QALY of $18,262.

Table 4. Summary of results from deterministic sensitivity analyses.

		Alternative values
Analysis/scenario	Base-case value	Low	High	Low value ICER (CAD/QALY)	High value ICER (CAD/QALY)
NeoSphere Base-case ICER	$25,388
NeoSphere sensitivity analyses
Discount rate	5%	0%	3%	$9230	$17,685
Time horizon	28 years	20 years	50 years	$31,109	$22,535
Vary all costs by 20%	–	Decrease by 20%	Increase by 20%	$20,311	$30,466
Vary cost of subsequent therapy by 20%	$297,141	$237,713	$356,570	$30,928	$19,849
Vary absolute difference in pCR by 20%	17.8%	14.2%	21.3%	$38,419	$16,701
Unadjusted utility for event-free, year 2 on	0.99	0.79	1.00	$33,485	$25,082
Unadjusted utility for relapsed, weighted	0.68	0.54	0.81	$24,319	$26,556
TRYPHAENA Base-case ICER	$46,196
TRYPHAENA sensitivity analyses
Discount rate	5%	0%	3%	$19,952	$33,935
Time horizon	28 years	20 years	50 years	$56,638	$40,984
Vary all costs by 20%	–	Decrease by 20%	Increase by 20%	$36,957	$55,435
Vary cost of subsequent therapy by 20%	$297,141	$237,713	$356,570	$51,736	$40,656
Vary absolute difference in pCR by 20%	17.8%	14.2%	21.3%	$64,421	$34,046
Alternative difference in pCR; adjusted based on  NeoALTTO:NSABP B-41 (low) or calculated  using RR from NeoSphere (high)	17.8%	15.1%	28.8%	$58,760	$18,262
Unadjusted utility for event-free, year 2 on	0.99	0.79	1.00	$60,929	$45,638
Unadjusted utility for relapsed, weighted	0.68	0.54	0.81	$44,250	$48,321

CAD, Canadian dollars; ICER, incremental cost-effectiveness ratio; pCR, pathological complete response (ypT0/is ypN0); QALY, quality-adjusted life-year; RR, relative risk.

The results of the PSA are shown in Figures 4 and 5. From the scatter plot for the NeoSphere analysis, 99% of results fell in the North-East quadrant (more costly and more effective), while 98% of those from the TRYPHAENA analysis were more costly and more effective (Figure 4). The CEACs depict the PSA results by showing the probability of neoadjuvant pertuzumab being cost-effective (y-axis) compared with a range of thresholds for maximum acceptable cost-effectiveness ratios (x-axis). For the NeoSphere analysis, at willingness-to-pay thresholds of $100,000 or $150,000 per QALY, the addition of neoadjuvant pertuzumab was 93% and 97% cost-effective, respectively (Figure 5A). At the same thresholds, the TRYPHAENA analysis indicated that the addition of pertuzumab would be cost-effective in 79% and 88% of scenarios, respectively (Figure 5B).

Figure 4. Scatter plots showing the results of the incremental costs and incremental QALYs for 10,000 runs of the PSAs for NeoSphere (A) and TRYPHAENA (B). In both the NeoSphere (A) and TRYPHAENA (B) PSAs, most results fell within the North-East quadrant, suggesting that the addition of pertuzumab results in increased efficacy and increased costs.

Figure 5. Cost-effectiveness acceptability curves for the NeoSphere (A) and TRYPHAENA (B) PSAs. At willingness-to-pay thresholds of $100,000 and $150,000 per QALY, on the basis of the NeoSphere analysis (A), the addition of pertuzumab to neoadjuvant therapy was cost-effective in 93% and 97% of the scenarios; while 79% and 88% of scenarios were cost-effective at the same thresholds for the TRYPHAENA analysis (B).

Discussion

The gold standard for the approval of novel oncology products has been the demonstration of improved disease-free and overall survival from randomized controlled trials. Such a requirement is reasonable for cancers with poor long-term outcomes, but the lengthy time requirements to demonstrate improved OS in cancers where patients can be treated with curative therapies at early stages of disease, such as with breast cancer, means that patients may be denied access to the most promising therapies for extensive periods of time. This concern has pioneered the recent shift in how drugs are assessed for both regulatory and reimbursement purposes, with their evaluation in the neoadjuvant setting playing a central role in expediting the approval of the most promising therapies and their uptake into standard practice. In the neoadjuvant setting, where patients are treated pre-operatively in the earlier stages of disease, efficacy is evaluated over the course of several months through end-points, such as pCR, that have been associated with a clinical benefit13^,20–27. These end-points must then serve as surrogates for long-term survival outcomes that are required to establish both comparative effectiveness and cost-effectiveness.

In the US, the FDA granted accelerated approval of pertuzumab for the neoadjuvant treatment of early HER2-positive breast cancer, with this decision based on the totality of evidence supporting the efficacy of pertuzumab; that is, the positive results from NeoSphere, the demonstrated efficacy in the metastatic setting, the favorable safety profile demonstrated in a large number of patients, and a proven mechanism of action that is complementary to that of trastuzumab18^,29. This accelerated approval was dependent on the pCR data from the NeoSphere trial and the assumption that pCR was a suitable surrogate end-point that correlated with long-term survival and that the treatment effect on pCR was comparable to the effect on survival. A key finding of the CTNeoBC meta-analysis conducted by the FDA was that patients who attained pCR, which was defined as ypT0 ypN0 or ypT0/is ypN0, had improved survival, suggesting a strong correlation at the patient level28. In sub-group analyses, they found that this association was strongest in patients with high-risk tumor types (i.e., triple negative, or HER2-positive). However, in their trial-level analysis, which combined all sub-types of breast cancer, they were unable to show an association between a therapy showing an increased frequency of pCR and an increased EFS or OS for this therapy relative to standard care. This was likely due to heterogeneity in the patient population and the inclusion of trials with ineffective therapies that only demonstrated slight increases in pCR. In fact, the NOAH trial, which examined the efficacy of trastuzumab for HER2-positive early breast cancer, was the only trial included in this analysis that examined a therapy directed to a specific tumor sub-type. Accordingly, the NOAH trial had the largest absolute difference in pCR (20%) and demonstrated the best correlation between the hazard ratio for overall survival and the odds ratio for pCR between the treatment groups13^,28^,64. Despite the inability to show a trial-level correlation, the FDA stated that they believed that ‘if a novel agent produces a marked absolute increase in frequency of pCR compared with standard therapy alone in the intention-to-treat population, that agent could also be reasonably likely to result in long-term improvements in EFS and OS (p. 171)28^,40. Future trials with sub-type-specific populations that are able to demonstrate a large absolute difference in pCR between treatment groups and a correlation with improved survival are needed to fully validate the use of pCR as a surrogate end-point.

The NeoALTTO trial, which examined the efficacy and safety of a dual HER2 blockade with lapatinib and trastuzumab compared with trastuzumab monotherapy, provides additional support for the use of pCR as a surrogate end-point for long-term survival63. Importantly, landmark analyses of all patients included in the trial reported a statistically significant correlation between patients achieving pCR after neoadjuvant therapy and increased EFS (HR = 0.38, 95% CI = 0.22–0.63; p = 0.0003) and OS (HR = 0.35, 95% CI = 0.15–0.70, p = 0.005)27. Also, the addition of neoadjuvant lapatinib to a regimen of trastuzumab and chemotherapy resulted in a significantly higher frequency of pCR compared with trastuzumab alone (46.8% vs 27.6%; p = 0.0007)63. The promising results observed for lapatinib in the neoadjuvant setting were expected to translate into a survival advantage when the same drug combination was given only in the adjuvant setting (ALTTO trial). However, the results of the ALTTO trial suggested that the addition of adjuvant lapatinib to trastuzumab was not associated with a statistically significant improvement in disease-free survival compared with trastuzumab alone (HR = 0.84, 95% CI = 0.70–1.02, p = 0.048)65. Possible reasons for the ALTTO trial not achieving statistical significance are that the trial was under-powered to detect a difference between treatment arms due to a protocol amendment, as well as toxicity issues with lapatinib that led to a significant percentage of under-dosed patients66^,67. Despite the disappointing results for lapatinib in the adjuvant setting, this should not overshadow the fact that many trials, including NeoALTTO, have shown a strong association between patients achieving pCR after neoadjuvant therapy and improvements in EFS and OS.

The primary limitation of this cost-effectiveness analysis (CEA) is that data were not available to inform the long-term survival of patients treated with neoadjuvant pertuzumab. Consequently, it was necessary to use pCR as a surrogate end-point for EFS and OS. The retrospective analysis conducted by Kim et al.25 provided an ideal data source as it included EFS and OS data partitioned for patients achieving or not achieving pCR in the population of interest. The current model was, thus, built on the key assumptions that pCR, defined as ypT0/is ypN0, is an acceptable surrogate end-point for EFS and OS and that the long-term survival data demonstrated by Kim et al.25 would be reflective of pertuzumab-treated patients based on whether or not they achieved pCR following neoadjuvant treatment. Furthermore, in the metastatic setting, the addition of pertuzumab to trastuzumab and chemotherapy resulted in significantly improved progression-free and overall survival34^,68; thus, pertuzumab treatment has already been correlated with improved survival in the metastatic setting.

Another limitation of this CEA is that the NeoSphere trial was conducted to better isolate the treatment effect of pertuzumab as a pre-operative experimental agent, but it may not translate well into Canadian clinical practice. In NeoSphere, the anthracycline regimen (i.e., FEC) was provided after surgery, but in clinical practice, many patients receive it pre-operatively prior to the taxane and anti-HER2 regimen (i.e., docetaxel, trastuzumab, and pertuzumab) (informed by expert opinion). To address this, we conducted an analysis based on the TRYPHAENA trial, which is believed to better represent how pertuzumab will be used in Canadian clinical practice.

The TRYPHAENA analysis also had its own limitations as it lacked a comparator arm without pertuzumab. As such, a hypothetical comparator arm was estimated based on the comparative efficacy observed in the NeoSphere trial. To address the uncertainty surrounding this estimate, we conducted two alternative analyses, as described above, to test a plausible range of costs per QALY. Furthermore, this analysis is limited by the fact that the patients in the trials that were used to inform the model may not behave the same as patients outside the clinical trial setting or may not fully represent Canadian breast cancer patients.

This analysis represents the first Canadian CEA for the addition of pertuzumab to trastuzumab-containing neoadjuvant treatment regimens. In this economic evaluation, the incremental cost associated with the addition of neoadjuvant pertuzumab was $7879–$14,337. A key model driver, apart from the cost and efficacy of pertuzumab, was the choice of subsequent therapy and the resulting cost for patients who experienced a disease recurrence (local or metastatic). The model assumed that the vast majority of patients who relapsed would receive first-line pertuzumab, trastuzumab, and docetaxel, a currently approved regimen which is funded in most of Canada in the first-line metastatic setting32. The model assumed that patients who received neoadjuvant pertuzumab were eligible for re-treatment in the metastatic setting, as is the current practice in Canada for trastuzumab. In addition, the model assumed that most patients who relapsed received trastuzumab emtansine (T-DM1, Kadcyla®) as second-line therapy as it was recently approved in Canada and funded in most provinces for the treatment of HER2-positive unresectable locally advanced or metastatic breast cancer61. As the addition of pertuzumab to the neoadjuvant treatment regimen resulted in fewer patients progressing in the model, there was an avoidance of costs associated with first-line and second-line subsequent therapies. Thus, part of the upfront drug cost for the addition of pertuzumab could be recuperated through avoidance of disease recurrence.

The CEA for the NeoSphere trial resulted in an incremental cost per QALY gained of $25,388. Various deterministic sensitivity analyses, conducted to account for potential uncertainty in the input variables, suggested that the ICER falls within the range of $9230–$38,419 per QALY. The CEA based on the TRYPHAENA trial resulted in an incremental cost per QALY gained of $46,196, with results from sensitivity analyses suggesting that the ICER falls within the range of $18,262–$64,421 per QALY. Both of these analyses suggest that the addition of pertuzumab to neoadjuvant treatment regimens is predicted to be cost-effective for an oncology product in Canada.

Conclusion

This Canadian CEA for the addition of pertuzumab to neoadjuvant trastuzumab and chemotherapy included both a NeoSphere trial analysis and a TRYPHAENA-based analysis and resulted in base-case costs of $25,388 or $46,196 per QALY gained, respectively. To address potential uncertainty surrounding model inputs, sensitivity analyses were conducted and resulted in a combined range between $9230–$64,421 per QALY. Given the improvement in clinical efficacy and a favorable cost per QALY, the addition of pertuzumab in the neoadjuvant setting represents an attractive treatment option for HER2-positive early breast cancer patients.

Transparency

Declaration of funding

The work reported in this manuscript was funded via a consultancy agreement between Cornerstone Research Group, Inc. and Hoffmann–La Roche Ltd. Canada.

Declaration of financial/other relationships

CLA, ANP, STB, and MFT disclose that they are consultants commissioned by Hoffmann–La Roche Ltd. Canada to perform this study. SY and POT disclose that they are employees of Hoffmann–La Roche Ltd. SD, AHP, and GAW have received honoraria for participating in consultant meetings for Hoffmann–La Roche Ltd. Canada. The final manuscript has been read and approved by all authors. JME peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Acknowledgments

The authors would like to acknowledge Dr Jean-Francois Boileau for his clinical advice and assistance with model design and parameterization. The authors would also like to acknowledge Vanja Petrovic and Douglas Millar for their careful review of the manuscript. Work by all contributing non-authors was funded by Hoffmann–La Roche Ltd. Canada.

Notes

*Herceptin is a registered tradename of Genentech Inc, San Francisco, California, US.

†Perjeta is a registered tradename of Hoffmann-La Roche AG, in Basel, Switzerland.