First-line cyclin-dependent kinase 4 and 6 inhibitors in combination with an aromatase inhibitor for HR+/HER2- metastatic breast cancer: A real-world cost-effectiveness assessment in a US Medicare-eligible population

Abstract

BACKGROUND:

In hormone receptor–positive, human epidermal growth factor 2–negative (HR+/HER2-) metastatic breast cancer (mBC), cyclin-dependent kinase 4 and 6 inhibitors (CDK4/6is) in combination with an aromatase inhibitor (AI) are the preferred first-line (1L) treatment. Although prior cost-effectiveness models comparing CDK4/6is palbociclib, ribociclib, and abemaciclib have used data from placebo-controlled clinical trials, no analyses in the United States have been conducted using real-world evidence (RWE) for a US Medicare-eligible population.

OBJECTIVE:

To estimate the long-term clinical outcomes and health care costs of 1L CDK4/6i treatment in patients aged 65 years and older using RWE.

METHODS:

We developed a partitioned survival model to project patient time in progression-free and progressed disease health states. Progression-free survival (PFS) and overall survival (OS) curves for palbociclib + AI were obtained from an analysis of patients aged 65 years and older treated 1L for HR+/HER2- mBC using the Flatiron Health Analytic Database (Flatiron). Adjusted comparative effectiveness estimates vs palbociclib + AI for both ribociclib + AI (PFS hazard ratio = 0.98 [95% CI = 0.86-1.13]; OS hazard ratio = 1.01 [95% CI: 0.87-1.18]) and abemaciclib + AI (PFS hazard ratio = 0.99 [95% CI = 0.86-1.15]; OS hazard ratio = 1.00 [95% CI = 0.84-1.19]) were obtained from the same analysis. All-cause medical costs and CDK4/6i drug costs were based on an analysis of patients aged 65 years and older in Optum Clinformatics DataMart. We used a Medicare perspective over a lifetime horizon for a cohort of patients with mean age of 73.7 years. Sensitivity analyses were performed to assess the robustness of results to plausible variation in input values.

RESULTS:

Projected life-years (LYs) with palbociclib + AI, ribociclib + AI, and abemaciclib + AI were similar: 5.16 (95% credible range [CR] = 4.94-5.35), 5.12 (95% CR = 4.53-5.82), and 5.16 (95% CR = 4.49-5.90), respectively. Total lifetime health care costs were also similar ($865,000 [95% CR = $807,400-$925,000], $866,800 [95% CR = $786,000-$965,000], and $901,000 [95% CR = $809,000-$1,004,600], respectively). Sensitivity analyses further supported no differences in LYs or total costs between CDK4/6is.

CONCLUSIONS:

Based on effectiveness and cost estimates from real-world data, our analyses suggest that palbociclib, ribociclib, and abemaciclib produce similar life expectancy and health care costs in US patients aged 65 years and older with HR+/HER2- mBC.

Plain language summary

This is the first study to use real-world evidence (RWE) to compare total health care cost and effectiveness of palbociclib, ribociclib, and abemaciclib in US patients aged 65 years and older treated first-line (1L) for hormone receptor–positive, human epidermal growth factor 2–negative (HR+/HER2-) metastatic breast cancer (mBC). We found no differences in life expectancy or costs when comparing the 3 commercially available cyclin-dependent kinase 4 and 6 inhibitors (CDK4/6is) in combination with aromatase inhibitors (AIs).

Implications for managed care pharmacy

Based on contemporary RWE, our analyses suggest no difference in long-term survival or health care costs between palbociclib, ribociclib, and abemaciclib when used in combination with an AI for treating patients aged 65 years and older with HR+/HER2- mBC. For a Medicare-eligible population, this implies no clinical or economic rationale to differentiate between CDK4/6is in benefit design or formulary placement.

Metastatic breast cancer (mBC) is associated with a significant patient and economic burden. By 2030, it is estimated there will be more than 246,000 prevalent cases of mBC in the United States.¹ Of all subtypes of breast cancer, hormone receptor–positive, human epidermal growth factor receptor 2–negative (HR+/HER2-) is most common, comprising about 70% of cases.² Five-year survival among those diagnosed with HR+/HER2- de novo mBC is only 35.4%.³ As prevalence rises, economic burden is also expected to increase.¹ Similar to other solid tumor cancers, mBC is associated with high cost of care.⁴ In a study of adults with cancer, mean all-cause costs for patients with mBC ranged from $172,700 to $249,200 annually during the first 4 years after diagnosis.⁴ Other solid tumor cancers had first-year costs ranging from $147,200 to $418,600.⁴

Cyclin-dependent kinase 4 and 6 inhibitors (CDK4/6is) in combination with aromatase inhibitors (AIs) are preferred for first-line (1L) treatment of HR+/HER2- mBC. In the United States, CDK4/6is palbociclib, ribociclib, and abemaciclib in combination with an AIs have all been approved by the US Food and Drug Administration (FDA) based on favorable progression-free survival (PFS) outcomes vs AIs alone in phase 3 clinical trials.^5–7 However, randomized controlled trial (RCT) results have been mixed regarding overall survival (OS).^8–10 To date, there have been no head-to-head RCTs comparing efficacy or safety of CDK4/6is, although 1 study comparing ribociclib to palbociclib is ongoing.¹¹ In addition, indirect treatment comparisons using data from RCTs have reported mixed findings, with multiple recent studies indicating no significant differences in PFS and OS across CDK4/6is.^12–14

In the United States, real-world evidence (RWE) has played a growing role in regulatory and payer decision-making, in part because these data are more reflective of what payers would experience in terms of outcomes and costs. Although RCTs are the gold standard for establishing causation and measuring efficacy, RWE is useful for providing outcomes in a postapproval, nonexperimental setting. Once a drug is commercially available, measures of real-world effectiveness may be more important and relevant to benchmark. More recently, the Centers for Medicare and Medicaid Services have explicitly stated in guidance for the Medicare Drug Price Negotiation Program that RWE from a Medicare population would be particularly important for price negotiation.¹⁵ Given that breast cancer is most frequently diagnosed between the ages of 65 and 74 years, a population with substantial comorbidity burden and often underrepresented in clinical trials, management of mBC is likely to impact Medicare most significantly among US payers.³

Recently, several real-world comparative effectiveness studies have shown generally similar effectiveness between the CDK4/6is, including 2 studies in the United States.^16–22 Moreover, only 1 cost-effectiveness analysis (CEA) comparing CDK4/6is has used data from RWE.²³ One US CEA has leveraged RWE in comparing CDK4/6is in combination with endocrine therapy to endocrine therapy alone but has not made comparisons between CDK4/6is.²⁴ The objective of this study was to estimate the long-term clinical outcomes, costs, and cost-effectiveness for a Medicare-eligible population using real-world data on comparative effectiveness and costs.

Methods

GENERAL APPROACH

We used decision-analytic techniques to estimate long-term clinical outcomes and health care costs for a Medicare-eligible patient population aged 65 years and older treated with 1L CDK4/6i + AI combination therapy for HR+/HER2- mBC. All clinical and cost data were sourced from US RWE studies, where available (Table 1).

TABLE 1.

Model Inputs

Parameter	Value	Source
Baseline characteristics
Age, years (mean)	73.7	P-VERIFY subgroup analysis in patients aged ≥65 years
Sex, %
Male	1.2%
Female	98.8%
Clinical inputs
Ribociclib + AI vs palbociclib + AI, PFS HR	0.98	P-VERIFY subgroup analysis in patients aged ≥65 years
Ribociclib + AI vs palbociclib + AI, OS HR	1.01
Abemaciclib + AI vs palbociclib + AI, PFS HR	0.99
Abemaciclib + AI vs palbociclib + AI, OS HR	1.00
Resource use and costs
Palbociclib + AI drug cost per month	$12,818	Pluard et al 202343
Ribociclib + AI drug cost per month	$11,428
Abemaciclib + AI drug cost per month	$11,765
Palbociclib + AI, PF cost per month	$6,717
Ribociclib + AI, PF cost per month	$8,002
Abemaciclib + AI, PF cost per month	$8,682
PD cost per month	$6,612	Hao et al 201644
Terminal care cost, 1-time cost at death	$4,732	Chastek et al 201245
Subsequent treatments a
Proportion who receive any subsequent treatment, %	44.4%	P-REALITY X subgroup analysis
CDK4/6i uptake in 2L	43.2%
Chemotherapy uptake in 2L	20.6%
Endocrine therapy (alone) uptake in 2L	22.6%
Other anticancer treatment uptake in 2L,	19.8%
Maximum duration of 2L CDK4/6i, months	8.3	Ring et al 202347
Maximum duration of 2L chemotherapy, months	4.3	Li et al 201646
Maximum duration of endocrine therapy alone, months	4.5	Swallow et al 201448
Maximum duration of other anticancer treatment, months	6.2	Li et al 201646
Utilities (scenario analysis only)
PF	0.736	Rugo et al 201851
PD	0.630

^a Based on 2L therapies after 1L treatment with palbociclib + AI in P-REALITY X subgroup analysis of patients aged 65 years and older over 24 months of median follow-up.

1L = first-line; 2L = second-line; AI = aromatase inhibitor; CDK4/6i = cyclin-dependent kinase 4 and 6 inhibitor; HR = hazard ratio; OS = overall survival; PD = progressed disease; PF = progression-free; PFS = progression-free survival.

MODEL STRUCTURE

We developed a 3-state partitioned survival model with 1-month cycle lengths (Figure 1). All patients entered the model in the progression-free (PF) state and transitioned to either progressed disease (PD) or death based on PFS and OS demonstrated in RWE studies. Both life-years (LYs) and costs were discounted at 3% per year and outcomes were estimated over a lifetime horizon.²⁵

FIGURE 1.

Model Structure

OS = overall survival; PD = progressed disease; PF = progression-free; PFS = progression-free survival.

POPULATION

The baseline age and sex of the modeled cohort were based on the subpopulation of patients aged 65 years and older in the RWE studies as discussed below.

TREATMENTS

Modeled patients received either palbociclib + AI, ribociclib + AI, or abemaciclib + AI for 1L treatment and remained on therapy until progression. Adherence was not explicitly modeled, but the use of real-world drug costs (and outcomes) inherently reflects the effects of adherence.

CLINICAL INPUTS

To estimate long-term survival, we conducted a targeted literature review of RWE studies that compared PFS or OS of ribociclib or abemaciclib with palbociclib in 1L treatment of HR+/HER2- mBC. Both published literature and conference proceedings were included. We assessed the quality of evidence in accordance with published guidance.^26,27 When available, we preferred adjusted to unadjusted comparisons. We also evaluated studies based on sample size, adjustment for clinically appropriate variables, and generalizability of the study cohort to a US Medicare-eligible population. There were 2 adjusted comparisons^16,20 of US patients (Supplementary Table 1 ^{(341.2KB, pdf)} , available in online article), 5 ex–US-adjusted comparisons^{17–19,21,22} (Supplementary Table 2 ^{(341.2KB, pdf)} ), and 6 unadjusted analyses (Supplementary Table 3 ^{(341.2KB, pdf)} ).^28–33 None of the unadjusted studies reported hazard ratios comparing PFS or OS between CDK4/6is; these studies were not further considered (Supplementary Table 3 ^{(341.2KB, pdf)} ).

Two recently presented US-adjusted comparisons, the P-VERIFY (Palbociclib Verifying Evidence of Real-world Impact Study; NCT06495164) study²⁰ and a study by Brufsky et al,¹⁶ both compared OS in adult patients (aged ≥18 years) treated with 1L CDK4/6i + AI treatment for HR+/HER2- mBC using the Flatiron Health electronic health record–derived deidentified longitudinal database (Flatiron). We identified P-VERIFY as a more appropriate source for deriving 1L OS curves given larger sample size, longer follow-up, use of stabilized inverse probability of treatment weighting (sIPTW) in primary analysis, and exclusion of patients treated with fulvestrant, which is more commonly used in later lines. Notably, both P-VERIFY and Brufsky et al report no significant difference in OS between the 3 CDK4/6is.^16,20

The P-VERIFY study²⁰ used Flatiron, a US nationwide database containing patient-level structured and unstructured data, curated using natural language processing with machine learning and technology-enabled abstraction.^34–36 PFS data were derived using natural language processing and manual abstraction. The data were subject to obligations to prevent reidentification and protect patient confidentiality. Study variables were validated using Flatiron Health’s quality and performance assessment frameworks.^37–39 The primary analysis used sIPTW to balance patient characteristics at baseline, including age, sex, race, practice type, Eastern Cooperative Oncology Group (ECOG) performance status, disease stage at initial diagnosis, visceral metastasis, bone-only disease, number of disease sites, and disease-free interval (from initial breast cancer diagnosis to mBC diagnosis).²⁰

To model a Medicare-eligible population, we conducted a subgroup analysis using patient-level data from the P-VERIFY study to derive survival outcomes for patients aged 65 years and older. Similar to the main study, the subgroup analysis included patients with HR+/HER2- mBC treated with CDK4/6i in combination with an AI as 1L therapy.²⁰ Patients in the subgroup analysis were included if they were aged 65 years and older at mBC diagnosis. OS was defined as time from start of 1L treatment to death, and PFS was defined as time from start of 1L treatment to earliest occurrence of death or disease progression (based on clinical assessment, radiographic scan, or tissue biopsy). Both OS and PFS were estimated with Kaplan-Meier analysis using sIPTW to balance the same baseline demographic and clinical characteristics between treatment groups, similar to the published primary analysis of OS.²⁰

Survival results of the P-VERIFY subgroup analysis after sIPTW adjustment are shown in Table 1, with additional information in Supplementary Table 4 ^{(341.2KB, pdf)} . The palbociclib cohort size (n = 3,723) was substantially larger than that of ribociclib (n = 689) or abemaciclib (n=561) and had longer follow-up (31.6 months vs 15.5 or 20.1 months, respectively). Similar to the overall population in P-VERIFY,²⁰ there were no significant differences in OS when comparing ribociclib + AI to palbociclib + AI OS hazard ratio [1.01 (0.87;1.18)]) or abemaciclib + AI to palbociclib+ AI hazard ratio [1.00 (0.84;1.19)] in patients aged 65 years and older. In addition, there were no differences in PFS between the CDK4/6i regimens.

Given that the P-VERIFY subgroup analysis had a larger palbociclib + AI cohort with longer follow-up, we modeled PFS and OS for palbociclib + AI as the reference case, deriving long-term survival from patient-level data. Various curve fits were considered for long-term extrapolation of PFS and OS for palbociclib + AI, including standard parametric distributions. The Weibull distribution was chosen for extrapolating both PFS and OS for palbociclib + AI based on goodness-of-fit (Supplementary Table 5 ^{(341.2KB, pdf)} ) and reasonable long-term survival projections (Supplementary Table 6, Supplementary Figure 1, and Supplementary Figure 2 ^{(341.2KB, pdf)} ).

The survival outcomes for ribociclib+ AI and abemaciclib + AI were modeled by applying the relevant sIPTW adjusted hazard ratios to the fitted palbociclib + AI survival curves. Based on visual inspection of the log-cumulative hazard plot and Schoenfeld residuals, there was no clear evidence to suggest the proportional hazards assumption was violated between ribociclib + AI and palbociclib + AI. Therefore, applying a hazard ratio for these treatments was deemed appropriate. Although there was some minor evidence that the proportional hazards assumption could be violated between palbociclib + AI and abemaciclib + AI, fitting separate survival curves to the abemaciclib + AI cohort that fit the tail of the Kaplan-Meier resulted in overestimated long-term survival (approximately 10% survival at 30 years). Therefore, applying a hazard ratio in this case gives a more plausible long-term survival estimate.

We also assumed that modeled OS would remain equal to or lower than background age-adjusted survival in the US general population at all time points.⁴⁰ In addition, PFS would not exceed OS.

RESOURCE USE AND COST INPUTS

We modeled lifetime all-cause medical costs and drug costs based on US RWE (Table 1). Monthly PF costs for patients treated 1L with palbociclib, ribociclib, or abemaciclib in combination with AI were based on the total inpatient, outpatient, and emergency department costs for a cohort of patients aged 65 years and older and enrolled in Medicare Advantage (n = 2,220) in Optum Clinformatics DataMart.⁴¹ Costs for adverse events were also captured by total medical costs and thus not modeled separately. PD health state costs were modeled based on the monthly cost of care for patients aged 65 years and older who received second-line (2L) therapy after progressing on AI.⁴² The study reported costs for patients who received either chemotherapy or everolimus in subsequent lines and predated the approval of CDK4/6is.

We calculated a weighted average based on the proportion of patients who received either chemotherapy or other anticancer treatment following palbociclib + AI in the P-REALITY X subgroup analysis of patients aged 65 years and older as discussed below. Terminal care costs were assigned as a 1-time cost at transition to death based on end-of-life hospice service for breast cancer patients.⁴³

Drug costs for CDK4/6i + AI regimens were calculated using wholesale acquisition cost (WAC) and real-world utilization accounting for dose adjustments and wastage over the course of treatment.⁴¹ We assumed that all patients receiving a 1L CDK4/6i + AI regimen would discontinue treatment after progression.

P-VERIFY did not report subsequent treatment patterns following discontinuation of 1L CDK4/6i + AI. To derive this input, a subgroup analysis of patients aged 65 years and older in the P-REALITY X study (NCT05361655) was conducted.⁴⁴ Using Flatiron, P-REALITY X assessed PFS and OS in patients aged 18 years and older at diagnosis of HR+/HER2- mBC who were treated 1L with either palbociclib + AI or AI alone and reported subsequent treatment patterns. The primary analysis adjusted for differences in baseline characteristics using sIPTW. A subgroup analysis in patients aged ≥75 years has previously been published.⁴⁵ The baseline characteristics of patients aged 65 years and older at mBC diagnosis for the unadjusted palbociclib + AI cohort is shown in Supplementary Table 7 ^{(341.2KB, pdf)} . Subsequent treatment patterns (Table 1) were assumed to be identical for ribociclib + AI and abemaciclib + AI.

Duration of subsequent treatment was based on time to treatment discontinuation for later lines of CDK4/6i, chemotherapy, endocrine therapy alone, or other anticancer treatment, represented by everolimus-based therapy.^46–48 Drug acquisition cost for other subsequent treatments were calculated using WAC and maintenance dosing according to prescribing information.⁴⁹ All costs were inflated to 2024 US dollars.⁵⁰

SENSITIVITY ANALYSES

One-way sensitivity analyses were conducted to estimate the impact of individual parameter uncertainty and to identify key drivers of the model. Parameters tested include hazard ratios for PFS and OS, drug cost, monthly cost of PF or PD, cost of terminal care, treatment mix following progression, and duration of 2L treatments. Key drivers of the model are identified by assessing the impact of input values, by testing an upper and lower bound, on output values. All parameters were varied by the reported 95% CI, where available, or a range of ±10% from the base-case value.

Overall uncertainty in model results was also tested in a probabilistic sensitivity analysis (PSA) by jointly varying all model parameters over 1,000 simulations. The 95% credible range (CR) estimates are reported for total costs and LYs. In addition, we calculate the probability that each CDK4/6i + AI regimen is cost-effective across willingness-to-pay (WTP) thresholds up to $300,000 per LY.

SCENARIO ANALYSES

In addition to outcomes based on a 3.0% discount rate, we report nondiscounted LYs for each CDK4/6i + AI regimen.

We also conducted a scenario analysis using health-related utility to estimate quality-adjusted survival (QALYs). Utility values for the PF and PD health states are based on an analysis of EuroQol-5 dimension health-related quality of life outcomes in the phase 3 PALOMA-2 study.⁵¹

MODEL VALIDATION

We assessed the clinical validity of the model by comparing our findings to the available literature. We compare modeled PFS and OS to that of RCTs, RWE studies, and published CEAs. We also compared modeled lifetime all-cause costs to total costs estimated in previously published studies.

Results

Aggregate and incremental model results are shown Table 2 and Table 3, respectively. Palbociclib + AI, ribociclib + AI, and abemaciclib + AI were associated with a total discounted life expectancy of 5.16 (95% CR = 4.94-5.35), 5.12 (95% CR = 4.53-5.82), and 5.16 (95% CR = 4.49-5.90) LYs, respectively. Total all-cause medical costs, including drug costs, were $865,000 (95% CR = $807,400-$925,000), $866,800 (95% CR = $786,000-$965 000), and $901,000 (95% CR = $809,000-$1,004,600), respectively. Although mean total health care costs associated with abemaciclib + AI are numerically higher, 95% CRs derived from the PSA indicate no difference in life expectancy or total health care costs between 1L palbociclib + AI, ribociclib + AI, and abemaciclib + AI (Table 3).

TABLE 2.

Model Results

Intervention	Base case									Scenario
	1L drug costs	PF costs	2L drug costs	PD costs	End-of-life costs	Total costs	LYs, PF	LYs, PD	Total LYs	QALYs, PF	QALYs, PD	Total QALYs
Palbociclib + AI	$440,800 ($394,900-$488,100)	$231,000 ($207,400-$256,400)	$7,100 ($5,600- $8,800)	$182,000 ($156,100-$207,200)	$4,000 ($3,700-$4,400)	$865,000 ($807,400-$925,000)	2.87 (2.75-2.99)	2.29 (2.05-2.51)	5.16 (4.94-5.35)	2.11 (2.02-2.20)	1.45 (1.28-1.61)	3.55 (3.41-3.70)
Ribociclib + AI	$400,400 ($341,400-$470,000)	$280,300 ($240,800-$329,200)	$7,300 ($4,900-$9,600)	$174,800 ($114,500-$236,300)	$4,000 ($3,600-$4,400)	$866,800 ($786,000-$965,000)	2.92 (2.58-3.33)	2.20 (1.47-2.95)	5.12 (4.53-5.82)	2.15 (1.89-2.45)	1.39 (0.91-1.86)	3.54 (3.16-3.96)
Abemaciclib + AI	$408,300 ($344,300-$482,500)	$301,300 ($255,000-$355,700)	$7,400 ($4,800-$9,700)	$179,900 ($113,100-$250,200)	$4,000 ($3,600-$4,400)	$901,000 ($809,000-$1,004,600)	2.89 (2.54-3.29)	2.27 (1.45-3.12)	5.16 (4.49-5.90)	2.13 (1.87-2.42)	1.43 (0.91-1.97)	3.56 (3.14-4.01)

Shown are point estimates based on deterministic means and 95% CRs based on a probabilistic sensitivity analysis with 1,000 iterations.

1L = first-line; 2L = second-line; AI = aromatase inhibitor; CR = credible range; LY = life-year; PD = progressed disease; PF = progression-free; QALY = quality-adjusted life-year.

TABLE 3.

Incremental Model Results

Pairwise comparisons	Base case		Scenario
	Incremental total costs (95% CR)	Incremental LYs (95% CR)	Incremental QALYs
Ribociclib + AI vs palbociclib + AI	$1,818 (−$93,600 to $104,700)	−0.04 (−0.60 to 0.59)	−0.02 (−0.37 to 0.39)
Abemaciclib + AI vs palbociclib + AI	$35,986 (−$59,300 to $143,900)	0.00 (−0.64 to 0.69)	0.00 (−0.39 to 0.43)

AI = aromatase inhibitor; CR = credible range; LY = life-year; QALY = quality-adjusted life-year.

In addition to discounted outcomes, we estimated the total nondiscounted LYs associated with palbociclib + AI, ribociclib + AI, and abemaciclib + AI (5.95 [95% CR = 5.67-6.24], 5.90 [95% CR = 5.13-6.80], and 5.95 [95% CR = 5.09-6.92] years, respectively).

One-way sensitivity analyses findings indicate that when accounting for uncertainty in key individual parameters, total life expectancy and costs were similar across CDK4/6is. Total LYs gained compared with palbociclib + AI were most sensitive to uncertainty in OS hazard ratio and the palbociclib + AI OS reference curve. When varying OS hazard ratio by the reported 95% CIs, the resulting incremental LY ranges for ribociclib + AI (−0.61 to +0.55) and abemaciclib + AI (−0.64 to +0.69) indicate no difference in life expectancy compared with palbociclib + AI. Incremental total costs compared with palbociclib + AI were most sensitive to estimates of relative OS and PFS and CDK4/6i drug costs. Varying individual parameters indicated no difference in total costs for ribociclib + AI or abemaciclib + AI compared with palbociclib + AI.

Based on the PSA, palbociclib + AI, ribociclib + AI, and abemaciclib + AI each have a probability of 48.2%, 38.3%, and 13.5%, respectively, of being cost-effective at a WTP of $150,000 per LY (Figure 2). Across a range of WTP thresholds up to $300,000 per LY, palbociclib and ribociclib have a similar probability of being cost-effective. However, abemaciclib + AI is less likely to be cost-effective given that it is associated with higher total cost of care. The cost-effectiveness distribution for ribociclib + AI and abemaciclib + AI compared with palbociclib + AI is shown in Supplementary Figure 3 ^{(341.2KB, pdf)} .

FIGURE 2.

Cost-Effectiveness Acceptability Curve

The probability that each treatment strategy is cost-effective at varying willingness-to-pay thresholds are represented.

AI = aromatase inhibitor; LY = life-year.

In a scenario analysis accounting for the utility of PF and PD states, 1L palbociclib, ribociclib, and abemaciclib resulted in 3.55 (95% CR = 3.41-3.70), 3.54 (95% CR = 3.16-3.96), and 3.56 (95% CR = 3.14-4.01) QALYs, respectively, showing no significant difference (Table 2).

Discussion

This study provides a timely analysis of the long-term clinical and economic trajectory of patients aged 65 years and older treated 1L with CDK4/6i + AI for HR+/HER2- mBC in a real-world setting. It is the first analysis using both real-world clinical outcomes and costs in a US Medicare-age population for an indication most frequently diagnosed in older adult individuals. Our analyses show that long-term outcomes are similar between 1L palbociclib + AI, ribociclib + AI, and abemaciclib + AI. In the base-case analysis, 1L CDK4/6i + AI was associated with life expectancy of over 5 years, with less than a month’s nonstatistically significant difference between the 3 regimens. For US Medicare-eligible patients with HR+/HER2- mBC, this suggests that 1L treatment with any combination of CDK4/6i + AI may provide similar long-term survival benefits. For a US Medicare payer, the lifetime economic costs associated with 1L CDK4/6i + AI were also comparable across the 3 regimens.

This study uses RWE of PFS and OS of 1L palbociclib + AI in US patients aged 65 years and older and best available comparative effectiveness evidence of ribociclib + AI and abemaciclib + AI in HR+/HER2- mBC. The P-VERIFY subgroup analysis leverages the Flatiron database, which includes US nationwide representation of more than 721,000 patients with breast cancer, originating from approximately 280 cancer clinics (approximately 800 sites of care).²⁰ Analyses were specific to patients aged 65 years and older, which reflects the US Medicare-eligible population.

Our model results indicate that there is no difference in life expectancy between 1L palbociclib + AI, ribociclib + AI, and abemaciclib + AI, which is generally consistent with published RWE across all ages. In US patients, both P-VERIFY and the study by Brufsky et al reported no difference in PFS or OS between the CDK4/6i + AI regimens.^16,20 Among adjusted comparisons in ex-US populations, findings have been mixed, and many have limitations in study design and concerns regarding validity of the results. Among ex-US RWE studies comparing ribociclib to palbociclib, an analysis of the German OPAL registry data adjusted for the greatest number of clinically relevant variables using sIPTW.²¹ This study reported relative survival estimates that indicate no difference between ribociclib and palbociclib (PFS hazard ratio = 1.01 [95% CI = 0.80-1.27]; OS hazard ratio = 0.96 [95% CI = 0.71-1.28]). Only one study of Polish cancer center outcomes reported relative survival estimates comparing abemaciclib to palbociclib, which also indicated no difference (PFS hazard ratio=0.80 [95% CI = 0.55-1.18]; OS hazard ratio = 0.90 [95% CI = 0.56-1.44]).¹⁸ Although none of the RWE studies were specific to patients aged 65 years and older, the median age across all studies was generally older, from 66 to 68 years in the German study and from 59 to 62 years in the Polish study.^18,21

Modeled long-term survival outcomes were also clinically reasonable in the context of phase 3 clinical trial findings. Median survival estimates based on fitted curves for palbociclib + AI (median PFS [mPFS] 25.3 and median OS [mOS] 53.8 months) fell within the 95% CIs reported for the PALOMA-2 trial (mPFS 22.4-30.3 and mOS 49.8-60.8 months).^8,53 Similarly, median survival based on fitted curves for ribociclib + AI (mPFS 25.8 months and mOS 53.3 months) were also consistent with results from MONALEESA-2 (mPFS 25.3 months [95% CI=23.0-30.3] and mOS 63.9 months [95% = CI 52.4-71.0]).^9,54 Lastly, the abemaciclib + AI fitted curves generated mPFS and mOS estimates (25.5 and 53.8 months, respectively) that were similar to but numerically lower than those of MONARCH 3 (29.0 and 66.8 months, respectively), for which 95% CIs have not yet been reported.^10,54

There are no US CEAs comparing CDK4/6is to each other based on RWE of clinical outcomes and costs. One US perspective CEA has compared CDK4/6is in combination with ET as a class to ET alone.²⁴ Despite several years of treatment with CKD4/6is in clinical practice, available CEAs are solely based on RCTs. However, RCT evidence may not reflect outcomes in a real-world clinical setting, given differences in patient population, treatment setting, and follow-up. Three US payer perspective analyses have compared CDK4/6is based on RCT outcomes in adults of all ages.^55–57 Both Mistry et al and Zhang et al modeled OS for palbociclib using outcomes from the phase 2 PALOMA-1 study (mOS 37.5 months),^56–58 which reported worse survival than the P-VERIFY subgroup analysis of patients aged 65 years and older informing this model (mOS 53.8 months). As such, the LYs and QALYs estimated by Mistry et al (4.64 LYs and 3.07 QALYs) and Zhang et al (3.77 LYs and 2.56 QALYs) were lower than our analysis (5.16 LYs and 3.55 QALYs).^56,57 Mistry et al and Zhang et al also reported cancer-specific costs only ($432,100 to $549,200), which are lower than our modeled all-cause costs. Mistry et al conclude that use of ribociclib + letrozole is dominant when compared with palbociclib + letrozole. The authors assumed equivalent OS but better PFS for ribociclib.⁵⁶ Zhang et al conclude that neither palbociclib + letrozole nor ribociclib + letrozole are cost-effective compared with letrozole alone.⁵⁷ However, these findings have limited applicability given that both analyses used immature OS data from MONALEESA-2 and outdated phase 2 PALOMA-1 results. Lastly, Masurkar et al reported substantially greater QALYs (9.23 to 10.12) than our model (3.54 to 3.56), which is unrealistic given that the phase 3 trials used to inform their analyses each report an mOS of less than 6 years.^9,55,59,60 The overestimated long-term survival is likely due to the authors’ choice of exponential functions and also likely explains the high cancer-specific costs reported ($1,121,400 to $1,176,400).⁵⁵

Although there are no US CEAs that leverage RWE to compare CDK4/6is, Al-Ziftawi et al compare palbociclib to ribociclib using RWE from a Qatari perspective (palbociclib: 5.97 LYs, 3.06 QALYs; ribociclib: 6.33 LYs, 3.16 QALYs).²³ Compared with the 95% CRs of our analysis, the Qatari study estimated greater LYs associated with both palbociclib and ribociclib. Importantly, it is unclear if event rates were adjusted for differences in baseline characteristics.

LIMITATIONS

The primary limitation of our research is uncertainty in the available RWE. There were 13 comparative effectiveness studies identified and reviewed. We did not conduct a network meta-analysis given significant heterogeneity between studies. Given that this model is specific to the US Medicare-eligible population, our findings may not be generalizable to patients younger than 65 years or non-Medicare health plans that include patients of all ages. It is noted, however, that P-VERIFY reported similar hazard ratios for OS across age groups from 18 to 49 years, 50 to 64 years, 65 to 74 years, and 75 years and older.

PF disease was associated with a monthly cost derived from Pluard et al.⁴³ Given that AE-related costs were included in total medical costs reported by Pluard et al, our model does not compare AE-related burden between CDK4/6is. In addition, this study included both 1L and 2L use of CDK4/6i + AI and skewed sample sizes of 3,182 patients receiving palbociclib, 149 patients receiving ribociclib, and 286 patients receiving abemaciclib. Monthly medical costs reported by Pluard et al for patients with Medicare Advantage ($6,255 to $8,085) are similar to those reported in an analysis of all adult women treated with CDK4/6is in any line of therapy ($7,136 to $12,378).⁶¹ In a more recent study by Behan et al, medical costs for Medicare patients in the year following initiation of 1L CDK4/6is were notably lower at $3,292 per month.⁶²

Monthly medical costs associated with PD are based on that of patients treated with 2L everolimus or chemotherapy after progressing on AI.⁴⁴ However, these costs may not accurately reflect that of patients who progress on 1L CDK4/6i + AI. In addition, 2L treatment is assumed to be similar across CDK4/6is and is based on the P-REALITY X subgroup analysis of patients who discontinue 1L palbociclib + AI. In the P-REALITY X subgroup analysis, over half of patients did not initiate a 2L treatment during a median follow-up of 24.3 months and 19% of patients initiated a 2L CDK4/6i. To date, no studies report all-cause medical costs for US patients following progression on a 1L CDK4/6i.

In addition, 1L CDK4/6i drug costs were based on an analysis that used WAC and utilization data, including dose adjustment and wastage.⁴³ Given WAC does not account for rebates or discounts, true drug costs may be overestimated. Because we do not adjust for adherence, the cost of CDK4/6is, which are oral, may also be overestimated. Further, given, that some patients may discontinue drug prior to progression, duration of 1L CDK4/6i use and cost may be overestimated. Finally, our analysis did not account for future loss of patent protection for the CDK4/6is. Although palbociclib is anticipated to lose exclusivity in 2027, ribociclib and abemaciclib may not have generic formulations until 2031 or later.^63–65 This would likely reduce the total cost of care significantly given the large share of drug costs.

Conclusions

Recent RWE studies combined with cost-effectiveness modeling provide the opportunity to assess long-term effectiveness and costs of CDK4/6is in clinical practice in the United States. Based on RWE, our analysis suggests that palbociclib + AI, ribociclib + AI, and abemaciclib + AI demonstrate similar life expectancy and health care costs in patients aged 65 years and older with HR+/HER2- mBC.

Disclosures

Drs Sandin, Stergiopoulos, and Liu are employees of Pfizer Inc., and own stock in Pfizer, Inc. Funding for the study was provided by Pfizer Inc. to Curta Inc. Drs Veenstra, Sun, Sullivan, and Mr Williams are employees of Curta Inc. Drs Brufsky and Pluard served as consultants to Pfizer Inc.