Attributes and Health Care Resource Utilization of Patients on Enzalutamide or Abiraterone for Metastatic Castration-Resistant Cancer in England

Abstract

PURPOSE

To compare demographics, clinical characteristics, health care resource utilization (HCRU), treatment duration, and overall survival (OS) with enzalutamide (ENZA) or abiraterone acetate (AA) in patients with metastatic castration-resistant prostate cancer (mCRPC) in England.

MATERIALS AND METHODS

This retrospective study analyzed data from the Cancer Analysis System database on patients receiving ENZA or AA (January 2014-March 2020) for chemotherapy-naïve mCRPC (mCRPC was the only funded indication for ENZA/AA during study period). Baseline characteristics were assessed using standardized mean difference (SMD) (<0.1: balanced); differences were adjusted for using propensity score weighting (PSW). Cox proportional hazard models were used for OS and treatment duration. Number needed to treat was calculated from HCRU incidence rate ratios (IRRs).

RESULTS

Overall, 8,485 patients were included (ENZA, 5,330; AA, 3,155). Diabetes mellitus was more prevalent in the ENZA group (SMD, 0.12) at treatment initiation. HCRU was comparable between groups before treatment initiation (SMD < 0.1), but HCRU IRR after treatment initiation favored ENZA. Compared with AA, ENZA was associated with significantly fewer inpatient stays, outpatient or accident and emergency (A&E) visits, and hospitalization days (P < .01), and significantly lower likelihood of treatment discontinuation (adjusted hazard ratio [aHR], 0.90 [95% CI, 0.86 to 0.96]; P < .01) and mortality risk (aHR, 0.92 [95% CI, 0.87 to 0.98]; P = .010). Assuming 8 months' treatment and comparable groups through PSW, 1.9 inpatient admissions, 17.3 outpatient visits, 1.4 A&E visits, and 19.5 hospitalization days could be avoided per 10 patients on ENZA versus AA.

CONCLUSION

Patients with mCRPC on ENZA or AA had generally similar baseline characteristics apart from diabetes prevalence. ENZA was associated with longer OS and treatment duration, and lower HCRU after treatment initiation than AA.

INTRODUCTION

Enzalutamide (ENZA) and abiraterone acetate (AA) are the treatment options for patients with metastatic castration-resistant prostate cancer (mCRPC) in the United Kingdom.^1-4 In the real-world setting, optimal treatment choices may depend on various factors, including comorbidities, concomitant medication, monitoring requirements, and adverse events. As the use of AA necessitates concomitant steroid therapy, ENZA may be preferred in patients with conditions associated with or aggravated by steroid use (eg, diabetes and cardiovascular disease [CVD]).^5-8 Conversely, AA may be preferred in patients with a history or high risk of seizures because seizures have been reported in 0.6% of patients on ENZA.⁹ The European Society of Cardiology guidelines underscore diabetes and hyperglycaemia as potential complications with AA, along with an increased risk of hypertension with AA and ENZA.¹⁰

CONTEXT

• Key Objective

• To compare demographics, clinical characteristics, health care resource utilization (HCRU), treatment duration, and overall survival in patients with metastatic castration-resistant cancer in England who received first-line treatment with enzalutamide (ENZA) or abiraterone acetate (AA).

• Knowledge Generated

• At treatment initiation among 8,485 patients (ENZA, 5,330; AA, 3,155), diabetes was more prevalent in the ENZA group. HCRU after treatment initiation was lower with ENZA, along with fewer inpatient stays, outpatient visits, accident and emergency visits, and days in hospital compared with AA (P < .01). ENZA was also associated with a lower likelihood of treatment discontinuation and mortality risk.

• Relevance

• The results of our large real-world study in the United Kingdom are consistent with other global real-world data that compared AA to ENZA, highlighting the importance of personalized treatment approaches to optimize outcomes and HCRU in patients with metastatic castration-resistant prostate cancer.

Real-world evidence from other countries suggests differences in treatment duration and health care resource utilization (HCRU) between ENZA and AA, with ENZA often associated with longer treatment duration.^8,11,12 Since both are recommended until disease progression or death, prolonged treatment duration may suggest better tolerance and/or delayed disease progression. Previous studies found that ENZA is associated with lower mortality risk, reduced risk of hospitalization, and shorter hospital stays than AA.^12-18 However, there remains an evidence gap regarding differences in patient characteristics, treatment duration, HCRU, and overall survival (OS) associated with ENZA and AA in England's National Health Service (NHS).

This study aimed to describe and compare demographics, clinical characteristics, HCRU during the 24 months preceding treatment initiation, treatment duration, and HCRU and OS after treatment initiation between ENZA and AA in patients with mCRPC in England.

MATERIALS AND METHODS

Study Design and Patient Populations

This retrospective cohort study used data from the Cancer Analysis System (CAS)¹⁹—the national cancer registry—and Hospital Episode Statistics (HES) in England that included adult patients (age ≥18 years at treatment initiation) diagnosed with PC (2012-2020). The treatment initiation period for either drug was January 1, 2014-March 31, 2020. Patients were assumed to have mCRPC, as it was the only NHS-funded indication for ENZA or AA during the study period, before ENZA became available for metastatic hormone-sensitive PC (mHSPC) in April 2020.²⁰ To create comparable groups, the patient population was restricted to chemotherapy-naïve patients (ie, first-line treatment with ENZA/AA), as the data did not indicate when chemotherapy was used. Patient attrition is described in the CONSORT diagram (Fig 1).

FIG 1.

CONSORT diagram. ^aValues have been masked because of small numbers potentially removing anonymization. AA, abiraterone acetate; ADT, androgen-deprivation therapy; CAS, Cancer Analysis System; CDF, Cancer Drugs Fund; ENZA, enzalutamide; HDI, Health Data Insight; ICD-10, International Classification of Diseases 10th Revision; PC, prostate cancer; SACT, systemic anticancer therapy.

The index date was the date of treatment initiation with ENZA or AA. The preindex follow-up period (January 1, 2012-March 31, 2020) was the lookback period from the date of diagnosis to the index date (used to assess HCRU in the 24 months before treatment initiation). The lookback period for identifying comorbidities was from the start of HES records (Fig 2).

FIG 2.

Study design. ^aAny SACT except first-generation antiandrogen, zoledronic acid, or other androgen-deprivation therapy alone. ^bHCRU was measured from the date of treatment initiation to the projected end of treatment +30 days, progression, or death. CCI, Charlson comorbidity index; CDF, Cancer Drugs Fund; COSD, Cancer Outcomes and Services Data Set; ECOG, Eastern Cooperative Oncology Group; HCRU, health care resource utilization; HES, Hospital Episode Statistics; PC, prostate cancer; SACT, systemic anticancer therapy.

Patients were excluded if they received systemic anticancer therapy (SACT) (except first-generation antiandrogen, zoledronic acid, or other androgen-deprivation therapy) before the index date; received concomitant ENZA and AA on the index date; had a documented diagnosis of any other malignancy before the index date; received ENZA or AA in the Cancer Drugs Fund; or participated in a clinical trial before the index date. Follow-up was from treatment initiation until the end of the postindex follow-up period (January 1, 2014-August 31, 2021) or end of data availability. Study data were collected by NHS digital and anonymized, so ethics approval was not required (Fig 2).

Study Objectives

The primary objective was to describe and compare demographic and clinical characteristics of patients with mCRPC initiating first-line treatment with ENZA relative to AA. Secondary objectives included: HCRU in patients with mCRPC 24 months before ENZA or AA were initiated; treatment duration; and post-treatment initiation HCRU. Treatment duration and HCRU were compared after treatment initiation with ENZA or AA until therapy discontinuation (+30 days to account for adverse effects), progression, or death (whichever was earliest). As an exploratory objective, OS was compared in the patients included in the study.

Study Variables

Patient Characteristics

Patient characteristics included demographics on the date of PC diagnosis and/or date of treatment initiation (age, ethnicity, and index of multiple deprivation); and clinical characteristics on the date of PC diagnosis (year of PC diagnosis, disease stage, tumor grade [Gleason score], and tumor morphology) and on the date of treatment initiation (year of treatment initiation, time from diagnosis to treatment initiation, time from treatment initiation to end of study period/death [whichever is first], performance status [Eastern Cooperative Oncology Group (ECOG)], comorbidities, and most common previous SACT). Comorbidities were reported in two groups of interest: CVDs and corticosteroid-sensitive comorbidities (clinical conditions where steroids must be used with caution) (Data Supplement, Table S1, online only).

Treatment Duration

Treatment duration was the time between first treatment cycle initiation (ENZA/AA) and its projected end date. The projected end date was determined by adding the listed prescription length (28 days for ENZA and AA) to the last date of the cycle or administration (date that the prescription was filled) recorded in the CAS. If the time between the regimen end date and the next ENZA or AA start date was less than the allowable gap on the basis of clinical advice (112 days), the two periods were combined and treatment was considered as continuous.²¹

Healthcare Resource Utilization

HCRU from the date of diagnosis to the date of treatment initiation was measured in terms of number and duration of all-cause inpatient stays, all-cause outpatient visits, and all-cause accident and emergency (A&E) department visits.

Statistical Analyses

General Methods

Continuous variables were described using mean, standard deviation (SD), median, and IQR. Categorical variables were described as number and percentage. Where appropriate, 95% CIs were derived.

For head-to-head comparison of treatment duration, HCRU, and OS after treatment initiation, propensity score weighting (PSW) was used to create comparable groups to minimize confounding bias. The propensity score, estimated using a logistic regression model, is the probability of a patient being treated with ENZA versus AA, given their characteristics at treatment initiation (Data Supplement, Table S2).²²

Primary Objective

To assess baseline characteristics between treatment groups, means (for continuous variables) and percentages (for binary/categorical variables) were analyzed using the standardized mean difference (SMD) (<0.1: negligible difference; ≥0.1: meaningful difference).²³ Unadjusted comparisons were made using the unpaired t test for continuous variables (or Mann-Whitney U-test if data were not normally distributed) and chi-square test for categorical variables. Results were stratified by select comorbidities (diabetes, CVD, and hypertension) and period of treatment initiation (2014-2016; 2017-2020) (Data Supplement, Table S1).

Secondary Objectives

The means of inpatient stays, outpatient and A&E visits, and lengths of hospitalizations were analyzed using SMDs to assess differences, with unpaired t test for continuous variables.

Comparable treatment groups were created for treatment duration analysis using PSW. Kaplan-Meier curves were used to evaluate treatment duration between the groups, and Cox proportional hazards regression was used to determine the relationship between each treatment and its respective duration.

HCRU from the date of treatment initiation to the end of treatment with ENZA relative to AA was compared after PSW. The rates of each HCRU outcome were compared using negative binomial regression. For each postindex HCRU outcome, the number needed to treat to prevent one outcome per month of treatment was calculated from the negative binominal model used in the calculation of the incidence rate ratio (IRR).

Exploratory Objective

OS was defined as the time from the index date to the earliest of all-cause death (event date) or censor date (either because of loss to follow-up or end of study period). The association between treatment and OS was analyzed using Cox proportional hazards regression after PSW.

RESULTS

Of 13,699 identified adult patients with PC, 8,485 (62%) met inclusion criteria. Of these, 5,330 (63%) initiated ENZA and 3,155 (37%) initiated AA (Fig 1).

Patient Demographics and Clinical Characteristics

The ENZA and AA groups generally had similar demographics and clinical characteristics. A difference in the year of treatment initiation was observed between the groups (SMD, 0.49): a higher proportion of patients receiving AA initiated treatment in 2014 (13%) versus those receiving ENZA (3%), whereas a higher proportion of patients started treatment with ENZA in 2015 and 2016 (2015: 15%; 2016: 19%) versus AA (2015: 8%; 2016: 10%).

ECOG performance status, disease stage at diagnosis, and Gleason score were similar between treatment groups. Most patients had stage IV disease at diagnosis (ENZA, 70%; AA, 68%). The median (Q1-Q3) time from PC diagnosis to treatment initiation for patients on ENZA was 28.27 months (17.22-43.5) versus 26.58 months (16.13-43.66) on AA. The median (Q1-Q3) time from treatment initiation to the end of follow-up was 14 months (6.97-24.38) for ENZA and 12.65 months (5.62-21.68) for AA.

The prevalence of comorbidities was high, and most patients had corticosteroid-sensitive comorbidities (83%) or CVD (71%). Common individual comorbidities included hypertension (61%), diabetes (23%), and hyperlipidemia (24%) (Table 1 and Data Supplement, Table S3).

TABLE 1.

Baseline Characteristics

Characteristic	Total (N = 8,485)	Enzalutamide (n = 5,330)	AA (n = 3,155)	SMD
Age at index date, years, median (Q1-Q3)	75.6 (69.6-81.1)	75.5 (69.3-81.0)	75.9 (70.1-81.3)	0.05
Clinical characteristics, n (%)
ECOG performance status (at start of regimen)a
0	1,899 (22.4)	1,199 (22.5)	700 (22.2)	0.04
1	4,055 (47.8)	2,569 (48.2)	1,486 (47.1)
2	761 (9.0)	488 (9.2)	273 (8.7)
3-5	*	*	*
Missing	1,723 (20.3)	1,041 (19.5)	682 (21.6)
Year of treatment start
2014	585 (6.9)	164 (3.1)	421 (13.3)	0.49
2015	1,055 (12.4)	795 (14.9)	260 (8.2)
2016	1,337 (15.8)	1,018 (19.1)	319 (10.1)
2017	1,591 (18.8)	988 (18.5)	603 (19.1)
2018	1,724 (20.3)	1,031 (19.3)	693 (22.0)
2019	1,693 (20.0)	1,047 (19.6)	646 (20.5)
2020	500 (5.9)	287 (5.4)	213 (6.8)
Comorbidities,b n (%)
Diabetes	1,914 (22.6)	1,297 (24.3)	617 (19.6)	0.12
Hyperlipidemia	2,033 (24.0)	1,314 (24.7)	719 (22.8)	0.04
Hypertension	5,144 (60.6)	3,305 (62.0)	1,839 (58.3)	0.08
Ischemic heart disease	2,174 (25.6)	1,419 (26.6)	755 (23.9)	0.06
Arrhythmia	2,108 (24.8)	1,280 (24.0)	828 (26.2)	0.05
Congestive heart disease	1,311 (15.5)	836 (15.7)	475 (15.1)	0.02
Chronic renal disease	1,469 (17.3)	942 (17.7)	527 (16.7)	0.03
Urinary tract infection	2,522 (29.7)	1,555 (29.2)	967 (30.7)	0.03
Cataracts	1,284 (15.1)	811 (15.2)	473 (15.0)	0.01
Renal insufficiency	3,439 (40.5)	2,187 (41.0)	1,252 (39.7)	0.03
Comorbidities grouped, n (%)c
CVD	6,052 (71.3)	3,837 (72.0)	2,215 (70.2)	0.04
Corticosteroid-sensitive comorbiditiesc	7,006 (82.6)	4,426 (83.0)	2,580 (81.8)	0.03

Abbreviations: AA, abiraterone acetate; CVD, cardiovascular disease; ECOG, Eastern Cooperative Oncology Group; Q1/Q3, first/third quartile; SMD, standardized mean difference.

^a Values have been masked (indicated with an asterisk) because of small numbers potentially removing anonymization.

^b Comorbidities identified in ≥15% of patients are listed in this table; comorbidities noted in <15% of patients are presented in the Data Supplement (Table S3).

^c A full list of comorbidities for grouped CVD, and corticosteroid-sensitive conditions can be found in the Data Supplement.

A greater proportion of patients on ENZA had diabetes versus AA (24% v 20%; SMD, 0.12), particularly in the later (2017-2020) period versus the earlier period (2014-2016). Differences in CVD and hypertension prevalence were more pronounced in 2017-2020: the SMDs for CVD and hypertension in the later period were 0.06 and 0.09, respectively, while they were 0.01 and 0.06, respectively, in the earlier period (Table 2).

TABLE 2.

Prevalence of Diabetes, CVD, and Hypertension, Stratified by Year of Treatment Initiation

Comorbidity		Overall, No. (%)	Enzalutamide, No. (%)	AA, No. (%)	SMD
2014-2016a
Diabetes	Yes	638 (21.4)	438 (22.2)	200 (20.0)	0.05
	No	2,339 (78.6)	1,539 (77.9)	800 (80.0)
CVD	Yes	2,040 (68.5)	1,359 (68.7)	681 (68.1)	0.01
	No	937 (31.5)	618 (31.3)	319 (31.9)
Hypertension	Yes	1,716 (57.6)	1,158 (58.6)	558 (55.8)	0.06
	No	1,261 (42.4)	819 (41.4)	442 (44.2)
2017-2020a
Diabetes	Yes	1,276 (23.2)	859 (25.6)	417 (19.4)	0.15
	No	4,232 (76.8)	2,494 (74.4)	1,738 (80.7)
CVD	Yes	4,012 (72.8)	2,478 (73.9)	1,534 (71.2)	0.06
	No	1,496 (27.2)	875 (26.1)	621 (28.9)
Hypertension	Yes	3,428 (62.2)	2,147 (64.0)	1,281 (59.4)	0.09
	No	2,080 (37.8)	1,206 (36.0)	874 (40.6)

Abbreviations: AA, abiraterone acetate; CVD, cardiovascular disease; SMD, standardized mean difference.

^a The rationale for stratifying patients by year of treatment was to understand whether the prevalence of comorbidities by treatment group changed over time, as the knowledge about the safety profile of each drug developed. The specific years were chosen so that two time periods similar in length could be examined.

HCRU 24 Months Before Treatment Initiation

HCRU in the 24 months before treatment initiation was similar between the ENZA and AA groups (SMD < 0.1). Patients had a median of one inpatient stay, 17 outpatient visits, and 0 A&E visits per year. The average length of stay among those hospitalized was 11.7 days in the overall population (Data Supplement, Table S4).

Assessment of Propensity Score Weighting

After PSW, patient demographics, clinical characteristics, and preindex HCRU variables were well balanced between treatment groups (ie, SMD < 0.1) (Data Supplement, Table S5). The propensity score (PS) density curves showed that PS-weighted groups were comparable (Data Supplement, Figs S1 and S2).

Treatment Duration

Very few (<1%) patients switched between ENZA and AA. Patients on ENZA remained on treatment longer than those on AA, with an unadjusted median time to discontinuation of 8.97 versus 7.56 months. In the adjusted comparison, the risk of treatment discontinuation was 10% lower with ENZA group than AA (adjusted hazard ratio [aHR], 0.90 [95% CI, 0.86 to 0.96]; P < .01), with an adjusted median time to discontinuation of 8.74 versus 7.95 months (Fig 3).

FIG 3.

Adjusted (propensity scores–weighted) Kaplan-Meier plot of treatment duration in months with ENZA relative to abiraterone. ^aThe count of number of patients at risk reflects the propensity score weighting and not the number of patients observed. AA, abiraterone acetate; ENZA, enzalutamide; HR, hazard ratio.

HCRU After Treatment Initiation With ENZA or AA Until Therapy Discontinuation

In the adjusted comparison, ENZA was associated with a statistically significant lower risk of all-cause inpatient admissions (IRR, 0.94 [95% CI, 0.90 to 0.98]; P < .01), outpatient visits (IRR, 0.88 [95% CI, 0.86 to 0.90]; P < .01), A&E visits (IRR, 0.81 [95% CI, 0.76 to 0.86]; P < .01), and days spent in hospital during follow-up (IRR, 0.81 [95% CI, 0.74 to 0.89]; P < .01) versus AA (Table 3). For the overall population, the most common reason for inpatient admission was corticosteroid-sensitive conditions (3%); the most common International Classification of Diseases 10th Revision code was neoplasms (70%; Data Supplement, Table S6). Assuming a treatment duration of 8 months and that PSW resulted in comparable treatment groups, the number needed to treat with ENZA versus AA was 5.2 patients to avoid one inpatient stay, 0.6 patients to avoid one outpatient visit, seven patients to avoid one A&E visit, and 0.5 patients to avoid 1 day in hospital. Thus, 1.9 inpatient admissions, 17.3 outpatient visits, 1.4 A&E visits, and 19.5 days in hospital could be avoided per 10 patients treated with ENZA instead of AA.

TABLE 3.

Adjusted HCRU After Treatment Initiation Until the Projected End of Treatment (+30 days), by Treatment Group

HCRU After Treatment Initiation (N = 8,215)		IRR (95% CI)a	P	Predicted Rate Per Patient Per Monthb	Number of Patient-Months Needed to Treat With Enzalutamide to Prevent One Outcome (Assuming 8-Month Treatment Duration)
All-cause inpatient stays	AA	0.94 (0.90 to 0.98)	<.01	0.39	5.21
	Enzalutamide			0.37
All-cause outpatient visits	AA	0.88 (0.86 to 0.90)	<.01	1.79	0.58
	Enzalutamide			1.58
All-cause A&E visits	AA	0.81 (0.76 to 0.86)	<.01	0.09	6.94
	Enzalutamide			0.08
Days spent in hospital during follow-up	AA	0.81 (0.74 to 0.89)	<.01	1.31	0.51
	Enzalutamide			1.06

Abbreviations: A&E, accident and emergency; AA, abiraterone acetate; HCRU, health care resource utilization; IRR, incident rate ratio.

^a Groups were compared using a negative binomial regression model with propensity score weighting with abiraterone as the reference.

^b This was predicted from the negative binomial model, whereby it is assumed that all patients receive that treatment.

Overall Survival

Median OS (unadjusted) was 20.11 months in the ENZA group and 18.27 months in the AA group. In the adjusted comparison, ENZA was associated with a reduced mortality risk (aHR, 0.92 [95% CI, 0.87 to 0.98]; P = .01), with an adjusted median OS of 19.84 versus 18.69 months (Fig 4).

FIG 4.

Adjusted (propensity scores–weighted) Kaplan-Meier plot of overall survival in months with ENZA relative to abiraterone. ^aThe count of number of patients at risk reflects the propensity score weighting and not the number of patients observed. ^bValues ≤5 were replaced with an asterisk because of the small number suppression rules used by Health Data Insight. AA, abiraterone acetate; ENZA, enzalutamide; HR, hazard ratio.

DISCUSSION

The ENZA and AA groups had similar demographics and clinical characteristics, except higher prevalence of diabetes at treatment initiation in patients receiving ENZA. As previously reported in literature,²⁴ this study found high prevalence of comorbidities, highlighting the interplay between the disease and preexisting health conditions. Differences in the prevalence of comorbidities between patient groups were more pronounced in later versus earlier time periods, possibly because of changes in treatment access and improved understanding of adverse event profiles with time.^25,26 The study results are consistent with a US retrospective claims analysis in chemotherapy-naïve patients with mCRPC, where patients treated with ENZA were more likely to have hypertension (ENZA, 57.2%; AA, 51.7%) and diabetes (ENZA, 27.5%; AA, 23.1%) versus those on AA⁸; and with a Hong Kong cohort study, in which ENZA and AA cohorts had similar baseline characteristics, except a higher prevalence of diabetes in the ENZA cohort (26% v 13.4%, respectively).²⁷

Median time from diagnosis to treatment initiation was 27.66 months, lower than the 62.7 months in the PREVAIL trial of ENZA in patients with mCRPC before chemotherapy.²⁸ Additionally, patients in this study were older (median, 75.6 v 72 years), had greater comorbidity prevalence (eg, cardiovascular comorbidities, 71% v 20%), and had poorer performance status (ECOG score 1-2, 56.8% v 31.9%).²⁸

Treatment duration was longer for ENZA than AA. This is consistent with previous retrospective mCRPC studies: two in US veterans and one in Sweden found a longer median treatment duration with ENZA than AA (11.7 v 9.1 months, P < .001¹⁶; and 9.93 v 8.47 months, P = .0008¹⁸; median [IQR], 8 months [3.6-16.4] v 4.9 months [2.6-11.7]¹¹).

In this study, the ENZA group had a significantly lower mortality risk than the AA group (adjusted analysis). Similarly, a US study found a higher mortality risk with AA (HR, 1.10 [95% CI, 1.04 to 1.16]; P < .001) versus ENZA¹³; while median OS was 24.2 months with ENZA versus 22.1 months with AA (P = .001) in another.¹⁶ A Taiwanese mCRPC study showed higher mortality rates in the AA group versus ENZA group (53.7% v 40.6%; P = .0004) and a lower mortality risk with ENZA versus AA after PS matching (P = .0013).¹⁴ A recent German real-world study showed that ENZA was associated with a significantly prolonged OS (HR, 0.79 [95% CI, 0.71 to 0.89]; P < .001) versus AA in patients with chemotherapy-naïve mCRPC.²⁹ By contrast, a US study showed no difference in mortality between AA- or ENZA-treated patients, although a significantly increased 6-month mortality was found in patients with one or more preexisting CVDs after starting AA or ENZA.¹⁵

In our study, patients receiving ENZA had 6% fewer inpatient admissions, 12% fewer outpatient appointments, 19% fewer A&E visits, and 19% fewer days in hospital. These are relevant differences, as hospitalization affects patients' quality of life, increases treatment costs, and burdens health care facilities, particularly considering the longer waiting lists in the NHS since the COVID-19 pandemic.³⁰ Our findings align with a review of real-world studies in other countries, showing a lower risk of hospitalization and shorter hospital stay with ENZA.¹⁷ Additionally, Ramaswamy et al³¹ reported a lower mean (SD) number of outpatient visits per person per month with ENZA (2.51 [1.61]) versus AA (2.86 [1.69]; P < .0001). A French study on new users of AA and ENZA reported that potentially serious adverse events leading to hospitalization, such as acute kidney injury, hepatic damage, and atrial fibrillation, were more frequently observed with AA than ENZA.²⁶ Although these findings are significant for mCRPC, further research is warranted to explore the implications for mHSPC, given the longer treatment duration with this disease.

To our knowledge, this is the first and largest study in the United Kingdom to evaluate the clinical characteristics, HCRU, and treatment duration in chemotherapy-naïve patients with mCRPC treated with ENZA or AA. Our results are likely generalizable as they are consistent with other global real-world studies.^{8,11,13-18,26,27,29,31}

Our study leveraged the CAS database, covering all eligible patients in England, with comprehensive cancer data and rigorous quality checks.¹⁹ However, some limitations should be considered. Selection bias may have resulted from inclusion criteria favoring later-stage diagnoses, and comorbidity prevalence may have been underestimated because of reliance on HES, as only those treated or reported in secondary care settings were identified. Despite uncertainties in HES comorbidity recording,^32,33 a 2012 review deemed it suitable for research.³⁴ The mCRPC population was identified using the treatment date based on the National Institute for Health and Care Excellence guidance dates, while some patients may have received AA or ENZA as off-label indications. HCRU in the 24 months before treatment initiation, used as a proxy for unrecorded comorbidities, was similar between the treatment groups and remained comparable after PSW, suggesting a low risk of confounding, although unrecorded patient characteristics may affect treatment selection and outcomes. Economic modeling was not conducted in our study; however, our HCRU findings are consistent with previous real-word studies.^17,26,31 Additionally, a difference in HCRU is expected, given the higher rate of cardiovascular events and diabetes incidence/worsening observed in other studies.^26,29,35,36 Treatment duration was estimated on the basis of projected end dates using assumed gaps between prescriptions, informed by clinical feedback and prescription timing. However, the uncertainty would affect ENZA and AA groups equally, minimizing its effect on the comparisons.

Patients diagnosed before 2012 were excluded, as CAS data were not standardized before this date; therefore, the study focused on patients diagnosed with late-stage PC or rapid disease progression to mCRPC.¹⁹ HCRU in primary care and blood tests were excluded owing to lack of recording in CAS and HES databases.

In conclusion, at treatment initiation with ENZA or AA, the patient groups were similar, except for the prevalence of diabetes, suggesting that comorbidities influence treatment choice. Because ENZA was associated with longer treatment duration, longer OS, and lower all-cause HCRU after treatment initiation versus AA, its use may lead to cost-savings for the NHS. This study highlights the importance of adopting personalized treatment approaches to optimize outcomes and HCRU in mCRPC management.

AUTHOR CONTRIBUTIONS

Conception and design: Amit Bahl, Rita Faria, Robert Snijder, Hanna Sodatonou, Sari Stark, Karla Martins, Nigel Rozario, Andrew Chilelli

Financial support: Andrew Chilelli

Administrative support: Gert Attard, Karla Martins, Andrew Chilelli

Collection and assembly of data: Rita Faria, Hanna Sodatonou, Karla Martins, Nigel Rozario, Jess Ridsdale-Smith, Andrew Chilelli

Data analysis and interpretation: Amit Bahl, Rita Faria, Gert Attard, Robert Snijder, Hanna Sodatonou, Sari Stark, Alessandra Pranzo, Karla Martins, Nigel Rozario, Jess Ridsdale-Smith, Andrew Chilelli

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Attributes and Health Care Resource Utilization of Patients on Enzalutamide or Abiraterone for Metastatic Castration-Resistant Cancer in England

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).