Abstract
Introduction: More adjuvant treatment options are becoming available for hormone receptor-positive/human epidermal growth factor receptor 2-negative (HR+/HER2−) early breast cancer (EBC) based on results of clinical trials. This study quantified the importance of different attributes of EBC adjuvant therapies to patients and the benefit-risk tradeoffs patients were willing to make. Methods: Women with HR+/HER2− EBC completed an online discrete choice experiment (DCE) survey; the design was informed by clinical data, qualitative interviews (n = 40), and pre-testing interviews (n = 40). Participants (pts) made 10 choices between pairs of hypothetical treatments described by varying levels of 6 attributes. DCE data were analyzed using a correlated mixed logit model. Relative attribute importance scores captured the impact of each attribute across clinically relevant ranges. Benefit-risk tradeoffs were captured as the minimum improvements in 5-year invasive disease-free survival (iDFS) that pts would require to tolerate increases in therapy-associated adverse event (AE) risks. Results: A total of 866 patients from the USA, France, Spain, Canada, the UK, Germany, South Korea, and Australia completed the DCE (mean age: 57.7 years; 76% postmenopausal; 29% stage I disease, 55% stage II, 16% stage III). Improved 5-year iDFS (75.4–82.7% range; associated with combination regimens [CRs] vs. endocrine therapy [ET] alone) contributed the most to treatment preferences (clinically relevant relative attribute importance: 38.4%), followed by reduced risks of venous thromboembolic events (VTEs) (20.4%), neutropenia (20.3%), and diarrhea (15.0%). Treatment type + duration (3.7%) and fatigue (2.3%) were less important. Pts required the largest improvement in 5-year iDFS (3.9%) to tolerate increased risks of VTE (0.7%–2.5%) or neutropenia (5.6%–46%); willingness to accept tradeoffs depended on the AE. Preference heterogeneity was observed across subgroups, but 5-year iDFS improvement was consistently the most impactful on treatment choice in all subgroups. Conclusion: A multicountry sample of patients most valued adjuvant therapies with higher 5-year iDFS and may therefore prefer CRs over ET alone. The value of CRs depends on their specific safety profiles, and shared decision-making should consider this to select treatment options that align with individual preferences.
Introduction
Hormone receptor positive/human epidermal growth factor receptor 2 negative (HR+/HER2–) is the most common subtype of breast cancer (BC), comprising approximately 70% of BCs [1]. Of these, approximately 95% are diagnosed at an early stage [2]. Typically, patients with HR+/HER2− early BC (EBC) undergo lumpectomy or mastectomy and are prescribed adjuvant treatment to reduce the risk of recurrence. Adjuvant endocrine therapy (ET) is standard of care (SOC) in HR+ EBC [3]. However, 27–37% of patients with stage II and 46–57% with stage III EBC experience distant recurrence within 20 years of diagnosis [4].
Cyclin-dependent kinase 4/6 inhibitors (CDK4/6is) + ET had greater efficacy versus placebo + ET in patients with HR+/HER2− advanced BC [5‒9]. Abemaciclib demonstrated an invasive disease-free survival (iDFS) benefit in high-risk, stage II/III, node (N)-positive disease and was approved by the FDA and EMA for high-risk EBC [10‒12]. Ribociclib showed iDFS benefits in a broader patient population with stage II/III disease, including N1-3 and N0 [13]. However, additional adverse events (AEs) are associated with the addition of CDK4/6is to ET [14, 15]. Given the increasing number of adjuvant treatments for HR+/HER2− EBC, each with different AE profiles and duration of therapy, patient preferences should be considered to select the optimal therapy.
Research on patient preferences for treatments in BC has mainly focused on either metastatic disease or ET and chemotherapy in EBC [16‒19]. A study in the USA elicited the preferences of patients with EBC and focused on benefit-risk tradeoffs when choosing between CDK4/6is [20]. Patients generally considered greater 5-year iDFS most important, followed by reduced risks of neutropenia, diarrhea, nausea, and alopecia; administration (dosing schedule) and ECG monitoring were less important [20].
This study aimed to explore patient preferences in a broad EBC population in an international setting. A discrete choice experiment (DCE) was used to quantify patient preferences for attributes of adjuvant treatments and determine the minimum improvements in 5-year iDFS that participants (pts) would require to tolerate increases in therapy-associated AE risks.
Methods
The study protocol was approved by Salus IRB, Austin, TX, USA (Salus IRB study number: 21207).
Study Population
Eligible female pts (aged ≥18 years) with HR+/HER2− EBC who had undergone lumpectomy or mastectomy were recruited from 8 countries via patient databases and online access panels using SurveyEngine. Recruiters screened individuals for eligibility via telephone; diagnosis and surgical history were self-reported. Pts were asked to have medical records or additional information available; recruiters assisted in locating the required information, as needed. Individuals with metastatic BC or other malignancies were excluded. Study pts provided informed consent online before study initiation.
Attribute and Level Development
Literature Review and Clinical Data Extraction
A targeted literature review identified treatment-related attributes (online suppl. Methods; for all online suppl. material, see https://doi.org/10.1159/000543320). Clinical data on ET monotherapy and CDK4/6i + ET combination regimens (CRs) were extracted from the phase 3 monarchE trial (NCT03155997) [21]. Data from the adjuvant NATALEE trial (NCT03701334) were not available during study design, so subgroup data from phase 2 AMALEE (NCT03822468) were used as patients received the same ribociclib dose [22]. Extracted data were used to identify key differentiators in tolerability and informed attribute level ranges for the DCE (Table 1). Once available, data from NATALEE were used to derive clinically relevant ranges [23].
Attributes and levels included in the discrete choice experiment
Attributesa . | Levels, % . | Clinically relevant values, %b . |
---|---|---|
Benefits | ||
5-year iDFSc | 70 | ET: 75.4 [21] |
75 | CDK4/6i No. 1: 82.7d | |
80 | ||
85 | CDK4/6i No. 2: 82.7 [21] | |
Riskse | ||
Diarrhea | 5 | ET: 8.7 [21] |
45 | CDK4/6i No. 1: 14.2 [23] | |
85 | CDK4/6i No. 2: 83.6 [21] | |
Neutropenia | 5 | ET: 5.6 [21] |
25 | CDK4/6i No. 1: 41 [23] | |
50 | CDK4/6i No. 2: 46 [21] | |
Fatigue | 10 | ET: 18 [21] |
25 | CDK4/6i No. 1: 21.9 [23] | |
45 | CDK4/6i No. 2: 40.8 [21] | |
VTEc | 0 | ET: 0.7 [21] |
4 | CDK4/6i No. 1: 1.4 [23] | |
8 | CDK4/6i No. 2: 2.5 [21] | |
Administration | ||
Treatment type + duration | 5 years of ET | ET: 5 years of ET [21, 23] |
5 years of ET, including 2 years of CR | CDK4/6i 1: 5 years of ET, including 3 years of CR [23] | |
5 years of ET, including 3 years of CR | CDK4/6i 2: 5 years of ET, including 2 years of CR [21] |
Attributesa . | Levels, % . | Clinically relevant values, %b . |
---|---|---|
Benefits | ||
5-year iDFSc | 70 | ET: 75.4 [21] |
75 | CDK4/6i No. 1: 82.7d | |
80 | ||
85 | CDK4/6i No. 2: 82.7 [21] | |
Riskse | ||
Diarrhea | 5 | ET: 8.7 [21] |
45 | CDK4/6i No. 1: 14.2 [23] | |
85 | CDK4/6i No. 2: 83.6 [21] | |
Neutropenia | 5 | ET: 5.6 [21] |
25 | CDK4/6i No. 1: 41 [23] | |
50 | CDK4/6i No. 2: 46 [21] | |
Fatigue | 10 | ET: 18 [21] |
25 | CDK4/6i No. 1: 21.9 [23] | |
45 | CDK4/6i No. 2: 40.8 [21] | |
VTEc | 0 | ET: 0.7 [21] |
4 | CDK4/6i No. 1: 1.4 [23] | |
8 | CDK4/6i No. 2: 2.5 [21] | |
Administration | ||
Treatment type + duration | 5 years of ET | ET: 5 years of ET [21, 23] |
5 years of ET, including 2 years of CR | CDK4/6i 1: 5 years of ET, including 3 years of CR [23] | |
5 years of ET, including 3 years of CR | CDK4/6i 2: 5 years of ET, including 2 years of CR [21] |
AE, adverse event; CDK4/6i, cyclin-dependent kinase 4/6 inhibitor; CR, combination regimen; DCE, discrete choice experiment; ET, endocrine therapy; iDFS, invasive disease-free survival; VTE, venous thromboembolic event.
aAttribute definitions used in the study are provided in online supplementary Table S2.
bClinically relevant values were derived from the referenced trial data.
cLevels for these attributes were updated after the pre-testing interviews. Final levels are shown here.
dAssumed equivalent to the 5-year iDFS with CDK4/6i 2; 5 years of ET including 3 years of CR [21, 23] since no head-to-head or indirect comparison data were available.
eClinically relevant values for risks are associated with treatment-emergent AEs reported in clinical trial data, measured over the trial. Risks were presented as the risk while on treatment. Given that AEs are most commonly reported within the first year of use, adapting the risk levels to an annual risk was not deemed relevant.
Qualitative Interviews
Web-based, 1-on-1, qualitative interviews were conducted with 40 patients (n = 5 per country) to identify important attributes in choosing a therapy (online suppl. Methods) [24]. Pts discussed factors that affect their preferences for treatments and completed a best-worst scaling case 1 (BWS1) exercise to inform ranking of treatment attributes (online suppl. Table S1; online suppl. Fig. S1, S2) [25].
Attribute Selection
Attributes were selected by considering insights from the targeted literature review, clinical data, qualitative interviews (including the BWS1), a diverse steering committee, and a pilot study. Attributes were intended to be relevant, cover a wide range of safety aspects, and differentiate between ET monotherapy and the 2 CRs. For the DCE, 6 attributes were selected: 5-year iDFS, 4 treatment AEs (diarrhea, neutropenia, fatigue, venous thromboembolic events [VTEs]), and 1 administration-related attribute (treatment type + duration) (Table 1; online suppl. Methods; online suppl. Table S2). Selected attributes, levels, and overall experimental design were tested and refined using pre-testing interviews (n = 40) and further assessed in an interim analysis (n = 123) (online suppl. Methods).
DCE Design
A D-efficient design with 40 experimental DCE tasks was generated with Ngene (v1.3.0). The 40 tasks were split into 4 blocks of 10 tasks; each pt was randomized to 1 block to limit the cognitive burden of the survey (online suppl. Methods). In each task, pts chose a preferred option in a pair of hypothetical treatments described by included attributes. An opt-out option was not provided as it was not possible to define a fixed opt-out without selecting a specific ET profile for SOC, and it was also expected that the proportion of pts who would choose to opt-out of any adjuvant treatment would be small (online suppl. Fig. S3; Table 1). To further reduce cognitive burden, each choice task included 2 attributes with identical levels (overlap) in both hypothetical treatments [26], marked with gray shading. The order of DCE choice tasks, treatment alternatives, and attributes (including groups) was randomized to mitigate potential ordering effects on treatment preferences [27, 28].
Survey Flow
After screening and consent, pts completed the online survey (online suppl. Fig. S4). In Part 1, pts were given written background information on EBC and treatments and an introduction to the DCE attributes (online suppl. Methods). Pts completed warm-up tasks and 1 full DCE practice task.
In Part 2, pts completed 10 experimental choice tasks. Pts also completed 3 nonexperimental choice tasks: a direct preference elicitation task between 2 CDK4/6i profiles (online suppl. Fig. S5), a stability test (task 13 was a repetition of task 3), and a dominance test [29]. The dominance test presented a choice between 2 profiles, 1 of which had superior efficacy and lower risks (and should thus be chosen), to assess survey engagement. In Part 3, pts completed questionnaires to gather sociodemographic and clinical information and determine health literacy and numeracy.
Data Analyses
DCE data were analyzed using a correlated mixed logit (MXL) model (online suppl. Methods) [30]. The model estimated how pt choices were influenced by changes in each treatment attribute, while accounting for heterogeneity in preferences and choice consistency [31].
Preference weights were generated from MXL model estimates, and clinically relevant level ranges were derived from monarchE and NATALEE (Table 1) [21, 23]. Clinically relevant relative attribute importance (cRAI) scores were generated (summing up to 100%) to estimate the importance of an attribute relative to all others across clinically relevant ranges. Pts’ willingness to make tradeoffs between efficacy and AE risk were calculated as minimum acceptable improvement in 5-year iDFS (online suppl. Methods).
Subgroup analyses examined how preferences varied by sociodemographic and clinical characteristics and assessed overall heterogeneity, including geographic region; age; employment status; education; dependent status; stage at diagnosis; time since diagnosis; menopausal status at survey completion; and experience with ovarian suppression, chemotherapy, and different AEs. The effect of pt characteristics on treatment preferences was investigated by adding interaction effects between characteristics and attributes to the MXL model (online suppl. Fig. S1). Model estimates were used to calculate cRAI scores in different subgroups, and subgroup treatment priorities were compared with those in the overall sample (online suppl. Table S3).
Results
Pt Characteristics
The final sample included 866 pts (online suppl. Fig. S4) who completed the survey between April 29 and May 31, 2023. Most (18%) resided in the USA, followed by France, Spain, Canada, the UK, Germany, South Korea, and Australia (Table 2). The mean age was 57.7 years, and mean time since diagnosis was 4 years. Most (73%) had a university degree or higher. More than half (55%) had stage II BC, 29% had stage I, and 16% had stage III; 76% were postmenopausal at the time of survey completion. Sociodemographic and clinical characteristics were generally similar across countries (online suppl. Table S4).
Participant sociodemographic and clinical characteristics
Characteristic . | Overall (N = 866) . |
---|---|
Age, years | |
Mean (SD) | 57.7 (10.2) |
Median | 58 |
Range | 22–83 |
Time since diagnosis, years | |
Mean (SD) | 4 (2.3) |
Median | 4 |
Range | 0–24 |
Country, n (%) | |
USA | 160 (18.5) |
France | 119 (13.7) |
Spain | 114 (13.2) |
Canada | 113 (13.0) |
UK | 111 (12.8) |
Germany | 106 (12.2) |
South Korea | 72 (8.3) |
Australia | 71 (8.2) |
University degree or higher, n (%) | 634 (73.2) |
Surgical resection for BC, n (%) | |
Lumpectomy | 616 (71.1) |
Mastectomy | 250 (28.9) |
BC stage at diagnosis, n (%) | |
I | 253 (29.2) |
IIA | 251 (29.0) |
IIB | 226 (26.1) |
IIIA | 83 (9.6) |
IIIB | 38 (4.4) |
IIIC | 15 (1.7) |
Menopausal status at time of survey, n (%) | |
Premenopausala | 152 (17.6) |
Postmenopausal | 661 (76.3) |
Otherb | 43 (5.0) |
Not surec | 10 (1.2) |
Previous treatment, n (%)d | |
ET | 853 (98.5) |
Chemotherapy | 315 (36.4) |
CDK4/6i | 69 (8.0) |
Ovarian suppression or ovaries removed | 78 (9.0) |
Characteristic . | Overall (N = 866) . |
---|---|
Age, years | |
Mean (SD) | 57.7 (10.2) |
Median | 58 |
Range | 22–83 |
Time since diagnosis, years | |
Mean (SD) | 4 (2.3) |
Median | 4 |
Range | 0–24 |
Country, n (%) | |
USA | 160 (18.5) |
France | 119 (13.7) |
Spain | 114 (13.2) |
Canada | 113 (13.0) |
UK | 111 (12.8) |
Germany | 106 (12.2) |
South Korea | 72 (8.3) |
Australia | 71 (8.2) |
University degree or higher, n (%) | 634 (73.2) |
Surgical resection for BC, n (%) | |
Lumpectomy | 616 (71.1) |
Mastectomy | 250 (28.9) |
BC stage at diagnosis, n (%) | |
I | 253 (29.2) |
IIA | 251 (29.0) |
IIB | 226 (26.1) |
IIIA | 83 (9.6) |
IIIB | 38 (4.4) |
IIIC | 15 (1.7) |
Menopausal status at time of survey, n (%) | |
Premenopausala | 152 (17.6) |
Postmenopausal | 661 (76.3) |
Otherb | 43 (5.0) |
Not surec | 10 (1.2) |
Previous treatment, n (%)d | |
ET | 853 (98.5) |
Chemotherapy | 315 (36.4) |
CDK4/6i | 69 (8.0) |
Ovarian suppression or ovaries removed | 78 (9.0) |
BC, breast cancer; CDK, cyclin-dependent kinase; ET, endocrine therapy.
aIncluded responses “Premenopausal, still have monthly periods” (n = 132) and “Period stopped by hormonal contraceptive” (n = 20).
bIncluded responses “Periods stopped during chemotherapy” (n = 25) and “Ovaries have been removed” (n = 18).
dNot mutually exclusive.
Overall Sample Preferences
Approximately 90% of pts answered 3 of 5 health numeracy questions correctly (online suppl. Results; online suppl. Table S5). Health literacy responses showed that <20% of pts often or always had trouble reading and understanding medical information and/or had no or little confidence in completing medical forms. Most (86%) passed the stability test, and 98% passed the dominance test. All completed responses were included in the analysis, per standard practice (online suppl. Methods). Data quality (validity, health, and numeracy tests) was in line with those in similar studies [32].
All attributes significantly influenced treatment preferences (p < 0.05) (Fig. 1; online suppl. Table S6). On average, clinically relevant increases in 5-year iDFS had the largest impact on treatment preferences (cRAI: 38.4%), followed by reduced risks of VTE (20.4%), neutropenia (20.3%), and diarrhea (15.0%) (Fig. 2). Treatment type + duration (cRAI: 3.7%) and reduced risk of fatigue (2.3%) were less important.
Attribute effects on preferences at clinically relevant levels and DCE levels. Impact of changes in each attribute on pts’ treatment preferences (N = 866). Utilities link changes in an attribute level (e.g., an increase in 5-year iDFS from 75% to 83%) to preferences. Larger utilities indicate higher desirability. While relative comparisons can be made, the absolute value of utility cannot be interpreted. Attribute ranges are ordered from most favorable on the left to least favorable on the right. Black x-axis labels indicate the clinically relevant range for each attribute, and black solid diagonal bars show the estimated effects of the attribute over this clinically relevant range. Gray x-axis labels indicate the ranges used in the DCE, and gray dotted diagonal bars show the estimated effects of the attribute on the range used in the DCE. Orange bars show the 95% CI for each utility value. The model had a good data fit (adjusted McFadden R2 = 84.7%) and was able to explain the choices that pts made in the DCE. CR, combination regimen; DCE, discrete choice experiment; ET, endocrine therapy; iDFS, invasive disease-free survival; incl., including; pt, participant; VTE, venous thromboembolic event.
Attribute effects on preferences at clinically relevant levels and DCE levels. Impact of changes in each attribute on pts’ treatment preferences (N = 866). Utilities link changes in an attribute level (e.g., an increase in 5-year iDFS from 75% to 83%) to preferences. Larger utilities indicate higher desirability. While relative comparisons can be made, the absolute value of utility cannot be interpreted. Attribute ranges are ordered from most favorable on the left to least favorable on the right. Black x-axis labels indicate the clinically relevant range for each attribute, and black solid diagonal bars show the estimated effects of the attribute over this clinically relevant range. Gray x-axis labels indicate the ranges used in the DCE, and gray dotted diagonal bars show the estimated effects of the attribute on the range used in the DCE. Orange bars show the 95% CI for each utility value. The model had a good data fit (adjusted McFadden R2 = 84.7%) and was able to explain the choices that pts made in the DCE. CR, combination regimen; DCE, discrete choice experiment; ET, endocrine therapy; iDFS, invasive disease-free survival; incl., including; pt, participant; VTE, venous thromboembolic event.
Relative attribute importance at clinically relevant levels. cRAI scores sum to 100% and show the relative impact of improving each attribute from the worst clinically relevant level (Table 1) to the best clinically relevant level (e.g., improving 5-year iDFS from 75.4% to 82.7%). Higher cRAI scores indicate that pts, on average, placed greater importance on improving this attribute. CDK4/6i, cyclin-dependent kinase 4/6 inhibitor; cRAI, clinically relevant relative attribute importance; ET, endocrine therapy; iDFS, invasive disease-free survival; pt, participant; VTE, venous thromboembolic event.
Relative attribute importance at clinically relevant levels. cRAI scores sum to 100% and show the relative impact of improving each attribute from the worst clinically relevant level (Table 1) to the best clinically relevant level (e.g., improving 5-year iDFS from 75.4% to 82.7%). Higher cRAI scores indicate that pts, on average, placed greater importance on improving this attribute. CDK4/6i, cyclin-dependent kinase 4/6 inhibitor; cRAI, clinically relevant relative attribute importance; ET, endocrine therapy; iDFS, invasive disease-free survival; pt, participant; VTE, venous thromboembolic event.
Pts were willing to accept increased AE risk in exchange for improvements in 5-year iDFS, but the magnitude depended on the AE (Fig. 3). To tolerate clinically relevant increases in VTE (0.7%–2.5%) or neutropenia risk (5.6%–46%), pts required improvement in 5-year iDFS (3.9% improvement for both).
Minimum acceptable improvement in 5-year iDFS. The minimum acceptable improvement in 5-year iDFS that pts would be required to tolerate increased risks of AEs or less convenient administration. The values compared for each attribute were based on clinical data for ET monotherapy and CDK4/6i + ET combination therapy from the phase 3 monarchE [21] and NATALEE trials [13]. Values with [a] are based on clinical data for ET monotherapy from monarchE, values with [b] are based on CR data from NATALEE, and values with [c] are based on CR data from monarchE. Comparisons of [a] versus [b] and [a] versus [c] represent the improvements in efficacy needed to tolerate the increased risks with ET versus CR. Comparisons of [b] versus [c] represent the improvements in efficacy needed to tolerate the increased risks with the CR in NATALEE versus the CR in monarchE. Greater minimum acceptable improvement values indicate that pts would require a greater increase in efficacy to tolerate the associated increase in risk or inconvenience. AE, adverse event; CDK4/6i, cyclin-dependent kinase 4/6 inhibitor; CR, combination regimen; ET, endocrine therapy; iDFS, invasive disease-free survival; incl., including; pt, participant; VTE, venous thromboembolic event.
Minimum acceptable improvement in 5-year iDFS. The minimum acceptable improvement in 5-year iDFS that pts would be required to tolerate increased risks of AEs or less convenient administration. The values compared for each attribute were based on clinical data for ET monotherapy and CDK4/6i + ET combination therapy from the phase 3 monarchE [21] and NATALEE trials [13]. Values with [a] are based on clinical data for ET monotherapy from monarchE, values with [b] are based on CR data from NATALEE, and values with [c] are based on CR data from monarchE. Comparisons of [a] versus [b] and [a] versus [c] represent the improvements in efficacy needed to tolerate the increased risks with ET versus CR. Comparisons of [b] versus [c] represent the improvements in efficacy needed to tolerate the increased risks with the CR in NATALEE versus the CR in monarchE. Greater minimum acceptable improvement values indicate that pts would require a greater increase in efficacy to tolerate the associated increase in risk or inconvenience. AE, adverse event; CDK4/6i, cyclin-dependent kinase 4/6 inhibitor; CR, combination regimen; ET, endocrine therapy; iDFS, invasive disease-free survival; incl., including; pt, participant; VTE, venous thromboembolic event.
Preference shares derived from the DCE results are consistent with the direct elicitation task results for the 2 hypothetical CR profiles. Both sets of results showed that most pts (DCE: 100%; direct elicitation: 97%) preferred a CR with 3 years versus 2 years of a CDK4/6i treatment of the same efficacy with higher risk of neutropenia (50% vs. 46%) and lower risk of VTE (1% vs. 4%), diarrhea (9% vs. 83%), and fatigue (10% vs. 41%) (online suppl. Fig. S5).
Preference Heterogeneity
Clinically relevant improvements in 5-year iDFS were the most important driver of treatment preference in all subgroups (online suppl. Table S7). The cRAI of attributes varied in some subgroups versus the overall sample (p < 0.05) (Fig. 4). The 5-year iDFS was relatively more important to pts in Europe (∆cRAI: 10.8%) and less important to pts in Asia-Pacific (∆cRAI: −8.4%). Treatment type + duration was more important to pts in North America than the overall sample (∆cRAI: 6.0%).
Differences in cRAI by attribute and subgroup. Subgroup cRAI scores that deviated from the overall sample (p < 0.05). Bars depict the mean difference in cRAI (%) in the subgroup compared with the overall sample. cRAI, clinically relevant relative attribute importance; iDFS, invasive disease-free survival; VTE, venous thromboembolic event.
Differences in cRAI by attribute and subgroup. Subgroup cRAI scores that deviated from the overall sample (p < 0.05). Bars depict the mean difference in cRAI (%) in the subgroup compared with the overall sample. cRAI, clinically relevant relative attribute importance; iDFS, invasive disease-free survival; VTE, venous thromboembolic event.
The 5-year iDFS was more important to pts who had previously experienced AEs from BC treatment (∆cRAI: 6.4%) versus the overall sample. Compared with the overall sample, the risk of neutropenia was considered less important by those who had previously experienced moderate to severe neutropenia (∆cRAI: −13.0%) but more important by those who had experienced mild neutropenia (∆cRAI: 7.2%). A VTE risk was more important to those aged ≥45 years (∆cRAI: 3.8%) and less important to those aged <45 years (∆cRAI: −3.6%).
Discussion
This is the first multicountry preference study in EBC to evaluate preferences for ET + CDK4/6i CRs in a broad population. Patients with EBC traded off benefits, risks, and administration attributes when choosing a preferred adjuvant treatment. Overall, pts prioritized treatment efficacy, followed by reduced risks of VTE, neutropenia, and diarrhea. Efficacy was significantly more important than these individual risks, with the closest cRAI to 5-year iDFS (38.4%; clinically relevant range: 75.4–82.7%) being to reduce the risk of VTE (20.4%; clinically relevant range: 0.7–2.50%). Pts placed considerably less weight on risk of fatigue and therapy duration (how long CR lasts). While heterogeneity was identified, all analyzed subgroups considered the 5-year iDFS the most important.
Pts’ prioritization of 5-year iDFS is consistent with previous literature on preferences for CDK4/6i CRs in US-based patients with EBC [20]. In both studies, 5-year iDFS was ≈2 times more important than the next attribute (neutropenia [18.2%] for the previous study, VTE [20.4%] and neutropenia [20.3%] in the current study, with preferences based on trading between the lowest and highest level for each attribute). However, the studies included different attributes and levels (e.g., prior studies did not consider VTE); thus, comparing other treatment priorities is difficult.
Combining the preferences elicited in this study with clinical performance data yields consistent preference-share estimates with a direct preference, which supports the validity of this analysis. Using published safety profiles of the CR treatments in EBC, the DCE survey predicted that a 3-year CR with risks of 14.2% for diarrhea, 41% or 62% for neutropenia, 21.9% for fatigue, and 1.4% for VTE would require a smaller iDFS benefit over ET alone to be preferred over a 2-year CR with risks of 83.5% for diarrhea, 46% for neutropenia, 40.8% for fatigue, and 2.5% for VTE [21, 23]. With more mature iDFS data, additional preference comparisons could be explored to better understand patient preferences in the adjuvant setting.
Study limitations must be acknowledged. Although real-life decisions may be more complex, in the DCE, pts made choices between hypothetical treatment profiles based on 6 attributes that were selected as most relevant during the qualitative phase of the study and a pilot. Although the exclusion of certain attributes, such as liver enzyme elevations, may be perceived as a limitation, they were asymptomatic, requiring increased monitoring. The results from the BWS1 during the qualitative interviews suggest that increasing the monitoring burden was only of minor importance to pts, which aligned with previous research [17, 20]. Attributes were also selected to distinguish between ET monotherapy and CRs. As such, these results are unsuitable for understanding preferences for other adjuvant treatments (e.g., chemotherapy). Another limitation is that the study sample did not mirror the clinical populations in monarchE or NATALEE (e.g., median age in the study was 58 years, whereas the median age in the trials was lower (NATALEE, 52 years; monarchE, 51 years); however, preferences were consistent at a high level across stages of BC, age, and menopausal status.
This multicountry analysis demonstrated that patients’ adjuvant treatment choices were most influenced by clinically relevant improvements in 5-year iDFS, followed by reducing the risks of VTE, neutropenia, and diarrhea. These preferences suggest that many patients may prefer CRs for increased efficacy in the adjuvant setting. Shared decision-making should consider varied AE profiles of CDK4/6is and ET to select the best treatment option for individual patients based on their priorities.
Acknowledgments
A multicountry steering committee of 8 members – including clinicians, patient representatives, and experts on preference methods – advised on the study design and provided feedback to develop and refine the DCE. We thank the patients who participated. We also thank Harrison Clarke and Ashley Samuelson for their work on the study. Medical writing was provided by Dr. Jacqueline Janowich Wasserott (Evidera) and funded by Novartis.
Statement of Ethics
The study protocol was reviewed and approved by Salus IRB (TX, USA), Approval No. 21207. Salus IRB is a Central Institutional Review Board based in the USA. Pts provided informed consent online prior to beginning the study.
Conflict of Interest Statement
V.H.: personal fees for consulting, lectures, honoraria, travel expenses from Lilly, Medscape, Novartis, ArticulateScience, AstraZeneca, and Exact Science. C.A.: grants for CBCN advocacy and education funding and honoraria from Novartis. F.B.: advisory boards for Novartis, Lilly, Pfizer, Roche, MSD, and Gilead. G.W.: honoraria from Novartis. M.R. and D.E.M.: nothing to disclose. D.A.M.: travel expenses for meeting attendance from ISPOR and Illumina; honorarium from ISPOR; fees for consulting paid to institution from Analytica and Novartis; research grants to institution from Canadian Institutes of Health Research, Genome Canada, Arthritis Society, and Alberta Innovates; and support by the Svare Chair in Health Economics. C.T. and H.L.: employees of Evidera by Thermo Fisher Scientific, which received funding by Novartis to conduct the work reported in this manuscript. S.H. and N.K.: employees of Evidera by Thermo Fisher Scientific, which received funding by Novartis to conduct the work reported in this manuscript; minority shareholders of Thermo Fisher Scientific as part of employment with Evidera. J.M.P., D.A., A.D., and P.P.: employees and minority shareholders of Novartis, which funded the work described in this manuscript. N.H.: personal fees for consulting and lectures from AstraZeneca, Daiichi Sankyo, MSD, Pierre Fabre, Roche, Sandoz/Hexal, Amgen, Exact Sciences, Gilead, and Seagen.
Funding Sources
This study was funded by Novartis. The funder had a role in the study design but had no role in the data collection, data analysis, or reporting of this study. All authors made the final decision to publish this study.
Author Contributions
Conceptualization, investigation, methodology, and project administration: C.T., S.H., H.L., J.M.P., D.A., A.D., and P.P. Data curation: C.T., S.H., and H.L. Formal analysis: N.K. Interpretation of data: V.H., C.A., M.R., F.B., G.W., D.E.M., D.A.M., C.T., S.H., H.L., N.K., J.M.P., D.A., A.D., P.P., and N.H. Visualization: C.T., S.H., H.L., N.K., J.M.P., D.A., A.D., and P.P. Writing (review and editing): V.H., C.A., M.R., F.B., G.W., D.E.M., D.A.M., C.T., S.H., H.L., N.K., J.M.P., D.A., A.D., P.P., and N.H.
Data Availability Statement
The data that support the findings of this study are not publicly available because the pts did not consent for their data to be analyzed outside this study.