Unplanned hospitalization among advanced prostate cancer patients by diabetes status: a population-based study

Abstract

Background

Older adults with advanced prostate cancer and type 2 diabetes mellitus are underrepresented in trials of androgen receptor pathway inhibitors. This study examined changes in unplanned hospitalization rates in patients receiving androgen receptor pathway inhibitors by type 2 diabetes mellitus status and assessed if unplanned hospitalization varies according to androgen receptor pathway inhibitors.

Methods

This population-based study of advanced prostate cancer patients aged older than 66 years used Surveillance, Epidemiology, and End Results–Medicare data. Prepost androgen receptor pathway inhibitor initiation changes and androgen receptor pathway inhibitor differences in unplanned hospitalization rates were estimated by adjusted incidence rate ratio with considerations for interactions between period, androgen receptor pathway inhibitor, and type 2 diabetes mellitus status. Linear contrasts were used to estimate and test conditional incidence rate ratios. Tests were 2-sided, and a P value less than .05 was considered statistically significant.

Results

The study included 12 240 patients: 3160 (25.8%) with type 2 diabetes mellitus, 7191 (58.8%) received abiraterone acetate with prednisone, and 5049 (41.2%) received enzalutamide. Unplanned hospitalization rates increased after androgen receptor pathway inhibitor initiation by 65% among patients with type 2 diabetes mellitus complications (adjusted incidence rate ratio = 1.65, 95% confidence interval [CI] = 1.37 to 1.98) and 109% in nondiabetics (adjusted incidence rate ratio = 2.09, 95% CI = 1.94 to 2.26). Among patients with type 2 diabetes mellitus without complications, the increase in unplanned hospitalization rates depended on the androgen receptor pathway inhibitor initiated: 103% after abiraterone acetate with prednisone (adjusted incidence rate ratio = 2.03, 95% CI = 1.70 to 2.43) and 47% after enzalutamide (adjusted incidence rate ratio = 1.47, 95% CI = 1.21 to 1.80) and a 38% greater increase in unplanned hospitalization rates after abiraterone acetate with prednisone than enzalutamide (ratio of abiraterone acetate with prednisone adjusted incidence rate ratio divided by enzalutamide adjusted incidence rate ratio = 1.38, 95% CI = 1.06 to 1.80).

Conclusions

All patients had higher unplanned hospitalization rates after androgen receptor pathway inhibitor. Our findings highlight the importance of using real-world data to better understand the interplay between preexisting health conditions and treatment outcomes, a critical step toward precision medicine.

Introduction

Prostate cancer is the most common nonskin cancer in the United States.¹ Advanced prostate cancer is often a disease of older adults² with multiple comorbidities, including type 2 diabetes.^3-7 Abiraterone acetate with prednisone and enzalutamide are 2 androgen receptor pathway inhibitors with different mechanisms of action used for advanced prostate cancer.^8-16 Glycemic events have been reported in patients receiving androgen receptor pathway inhibitors.^17,18 Unplanned hospitalization associated with therapy-related adverse events can cause interruptions in cancer therapy and affect the patient’s health.¹⁹ Because acute inpatient care is the number 1 cost driver in Medicare patients with advanced cancer,²⁰ unplanned hospitalization rate following treatment is a clinically meaningful endpoint.^21-23

Although advanced prostate cancer and type 2 diabetes mellitus are common, to our knowledge, to date, no studies have examined the interplay between preexisting type 2 diabetes mellitus and androgen receptor pathway inhibitor use for predicting unplanned hospitalizations in advanced prostate cancer patients. Zaorsky et al.²⁴ reported on treatment-related toxicity in patients with advanced prostate cancer receiving definitive radiation therapy. Antidiabetic medication was used to stratify patients as nondiabetics, metformin users, other oral antiglycemic users, insulin users, and no-medication users. They found that patients with type 2 diabetes mellitus not on any medication and those treated with insulin have a greater chance of late toxicities than patients without diabetes. However, in their study, patients did not have an advanced-stage disease and were treated only with radiation, and the categorization of type 2 diabetes mellitus differs from the current study. A study by Lee et al.²⁵ examining type 2 diabetes mellitus patients with advanced prostate cancer utilizing metformin vs not using metformin found that those using metformin had fewer hospitalizations. Nevertheless, the Lee study²⁵ failed to report on any other antidiabetic medications used. Using the Veteran’s Health Administration data, Schoen et al.²⁶ showed that among veterans with advanced prostate cancer and cardiovascular disease or diabetes, those treated first with enzalutamide had longer median overall survival (23.2 vs 20.5 months, P < .001) compared with those treated with abiraterone acetate with prednisone.

Given the lack of outcomes data specific to type 2 diabetes mellitus, current treatment guidelines do not provide specific treatment strategy based on the preexisting type 2 diabetes mellitus status. The objectives of this study were to (1) quantify the changes in unplanned hospitalization rates following abiraterone acetate with prednisone or enzalutamide as a function of type 2 diabetes mellitus status and (2) test whether changes in unplanned hospitalization rates vary with the abiraterone acetate with prednisone used, according to type 2 diabetes mellitus to facilitate the development of tailored treatment strategies.

Methods

This population-based study used Surveillance, Epidemiology, and End Results linked with the Medicare (SEER-Medicare) database. This report followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.²⁷ SEER registries cover approximately 36.7% of the US population as of 2017.²⁸ SEER is a widely used cancer registry, with a case ascertainment rate of 98%.²⁹ Medicare is the primary health insurance provider among US patients aged 65 years and older.³⁰ This study received expedited review approval from the institutional review board at Thomas Jefferson University, as it involved minimal risk and the analysis of existing deidentified data; further, cell sizes are limited to at least 11 to maintain patient confidentiality, according to National Cancer Institute (NCI) guidelines.

Study participants

The study cohort consisted of patients aged 66 years and older diagnosed with primary advanced prostate cancer between January 1, 1999, and December 31, 2019, and received their first abiraterone acetate with prednisone or enzalutamide between January 1, 2013, and December 31, 2020. The first date of abiraterone acetate with prednisone or enzalutamide constituted the index date. To ensure the study cohort has complete claims data, we included patients with continuous Medicare Parts A, B, and D coverage, without health maintenance organization enrollment, from 12 months before until 6 months after the index date. Existing type 2 diabetes mellitus, with and without complications, was identified using International Classification of Diseases, Ninth or Tenth Revision, codes during the year before the index date.³¹ The NCI comorbidity macro was used to identify those with type 2 diabetes mellitus complications.³² Although type 1 diabetes mellitus and type 2 diabetes mellitus are related diseases, their course and average date of onset and, therefore, duration are different; the effects of type 1 diabetes mellitus on advanced prostate cancer are not fully understood and are outside the scope of this study³³; thus those with type 1 diabetes mellitus were excluded using International Classification of Diseases, Ninth and Tenth Revision codes according to a tested algorithm.^34,35 Patients who received chemotherapy during the 12 months before the index date were also excluded. A detailed Consolidated Standards of Reporting Trials for the patient selection process is found in Figure 1.

Figure 1.

Inclusion and exclusion criteria for all participants (created with BioRender.com). Abbreviations: AA = abiraterone acetate; ENZ = enzalutamide; HMO = health maintenance organization; PEDSF = Patient Entitlement and Diagnosis Summary File; seq = sequence; T1DM = type 1 diabetes mellitus; T2DM = type 2 diabetes mellitus; w/o, without.

Drug exposure

Drug exposure was identified from the Medicare drug event file using “abiraterone acetate” (Zytiga) and “enzalutamide” (Xtandi). Following the intention-to-treat protocol, the classification of drug exposure was based on the first treatment received. The abiraterone acetate with prednisone group included patients treated with only abiraterone acetate with prednisone or abiraterone acetate with prednisone followed by other treatments, including enzalutamide. Similarly, the enzalutamide group included patients treated with only enzalutamide or enzalutamide followed by abiraterone acetate with prednisone or other medications. To exclude patients who had used chemotherapy in the 12 months before index, the following chemotherapies were identified from event and part D files: docetaxel (using “J9171” and “J9170”) cabazitaxel (using “J9043” and “J9064”), mitoxantrone (using “J9293”), carboplatin (using “J9045”), and estramustine (using “estramustine phosphate sodium” and “Emcyt” and “00013-0132”).

Primary endpoint

The primary endpoint was a change in the rate of unplanned hospitalization, defined as a hospital admission initiated in the emergency department (ED). To account for the baseline rate, we measured the number of unplanned hospitalizations 6 months before the index date (first date of treatment) and compared it with 6 months after the index date. ED visit definition derives from the Centers for Medicare and Medicaid Services Research Data Assistance Center definition.³⁶ Unplanned hospitalization rates were calculated as the total number of ED admissions divided by the time at risk during each 6-month period. Time at risk was defined as total time alive minus time spent in the hospital.

Descriptive variables

Demographic variables included age at index date (66-74, 75-84, 85 years and older), race, marital status (married, unmarried, unknown), SEER region (Northeast, South, Central, West), state buy-in (a program in which the state pays Medicare premiums to expand care to eligible people otherwise unable to afford them), modified comorbidity index, and socioeconomic status including census-tract level median household income and education (percentage with a high school diploma or less); socioeconomic status also included dual Medicare and Medicaid eligibility (yes, no) during the study period. Race was classified as non-Hispanic White, non-Hispanic Black, and Other (including Native American, Alaskan Native, Asian, and Pacific Islander). Modified comorbidity index was based on NCI’s Comorbidity Index, which excludes solid tumors, leukemias, and lymphomas and is similar to the original Charlson Comorbidity Index assigning weights to a series of comorbid conditions based on the likelihood of 1-year mortality.²⁹ The modified comorbidity index measure in this study further excluded diabetes and was categorized as scores of 0, 1-2, and 3 or higher. Type 2 diabetes mellitus status was defined as those without type 2 diabetes mellitus, those with type 2 diabetes mellitus but no diabetes mellitus–related complications, and those with type 2 diabetes mellitus and diabetes mellitus–related complications. The NCI comorbidity macro was used to identify those with type 2 diabetes mellitus complications.³² The stage at diagnosis was categorized as local, regional distant, or unknown, and the year of diagnosis was split between those diagnosed before 2013 and those diagnosed from 2013 to 2019.

Statistical analysis

Descriptive statistics for various risk groups were generated and compared by type 2 diabetes mellitus status using 2-sided χ² tests. Incidence rates were estimated as the total number of unplanned hospitalizations across patients divided by total time at risk. To evaluate whether androgen receptor pathway inhibitor drug type was associated with the degree of changes in unplanned hospitalization rates, we evaluated 2-way interaction terms between androgen receptor pathway inhibitor drug type and period (post- vs pretreatment); type 2 diabetes mellitus status with androgen receptor pathway inhibitor and period; and a 3-way interaction term between type 2 diabetes mellitus, androgen receptor pathway inhibitor, and period in covariate-adjusted negative binomial regression models based on generalized estimating equations^37,38 that produce standard error estimates accounting for the correlations in paired hospitalization rate data within patients. These models were adjusted for age, race, marital status, geographic region, year of diagnosis, cancer stage at diagnosis, modified Charlson Comorbidity Score, dual eligibility status, state buy-in, neighborhood median household income, and percentage of high school education. Non-statistically significant interaction terms (P ≥ .05) were removed in a hierarchical stepwise fashion. Two-way interactions between type 2 diabetes mellitus and androgen receptor pathway inhibitor and androgen receptor pathway inhibitor and period were statistically significant, so separate regression models for each type 2 diabetes mellitus status were constructed to estimate incidence rate ratios and 95% confidence intervals (CIs) comparing 6 months after vs before androgen receptor pathway inhibitor initiation. The interaction between androgen receptor pathway inhibitor and period was only statistically significant among type 2 diabetes mellitus patients without complications (P < .01) for whom conditional associations were presented. Only main effects were presented for other type 2 diabetes mellitus status patients. Analyses were conducted in SAS v. 9.4 (SAS Institute Inc., Cary, NC, USA).

Results

Population characteristics

Our cohort included 12 240 patients with advanced prostate cancer (Figure 1). Among these, 74.2% (9080) did not have type 2 diabetes mellitus, 18.0% (n = 2204) had type 2 diabetes mellitus without complications, and 7.8% (n = 956) had type 2 diabetes mellitus with complications (Table 1). The most common age group in the cohort (42.3%) was 75-84 years. The cohort was 84.3% non-Hispanic White, 7.1% non-Hispanic Black, and 8.7% other racial and ethnic backgrounds. More patients used abiraterone acetate with prednisone (58.8%) than enzalutamide (41.2%). Use of abiraterone acetate with prednisone ranged from 80.6% of the study population in 2013 to 47.8% in 2020 (Table S1). Among patients without type 2 diabetes mellitus, 60.5% used abiraterone acetate with prednisone vs 39.5% enzalutamide (P < .01), and among those without type 2 diabetes mellitus complications, 55.9% used abiraterone acetate with prednisone vs 44.1% enzalutamide (P < .01). However, among patients with type 2 diabetes mellitus complications, the distribution of androgen receptor pathway inhibitors was almost even (48.8% abiraterone acetate with prednisone vs 51.2% enzalutamide).

Table 1.

Study cohort characteristics overall and stratified by type 2 diabetes mellitus status

Characteristic	Overall, No. (%) (N = 12 240)	No type 2 diabetes mellitus, No. (%) (n = 9080)	Type 2 diabetes mellitus without complications, No. (%) (n = 2204)	Type 2 diabetes mellitus with complications, No. (%) (n = 956)	P
Novel hormone therapy					<.001
Abiraterone acetate	7191 (58.8)	5492 (60.5)	1232 (55.9)	467 (48.8)
Enzalutamide	5049 (41.2)	3588 (39.5)	972 (44.1)	489 (51.2)
Age at index date, y					<.001
66-74	4852 (39.3)	3645 (40.1)	854 (38.7)	353 (36.9)
75-84	5173 (42.3)	3728 (41.1)	991 (45.0)	454 (47.5)
85 and older	2215 (18.1)	1707 (18.8)	359 (16.3)	149 (15.6)
Race					<.001
Non-Hispanic Black	865 (7.1)	528 (5.8)	220 (10.0)	117 (12.2)
Non-Hispanic White	10 316 (84.3)	7844 (86.4)	1740 (78.9)	732 (76.6)
Other	1059 (8.7)	708 (7.8)	244 (11.1)	107 (11.2)
Marital statusa					<.001
Married	5729 (46.8)	4297 (47.3)	1022 (46.4)	410 (42.9)
Not married	1472 (12.0)	1130 (12.4)	253 (11.5)	89 (9.3)
Unknown	5039 (41.2)	3653 (40.2)	929 (42.2)	457 (47.8)
Region					<.001
Northeast	5163 (42.2)	3706 (40.8)	1027 (46.6)	430 (45.0)
South	2205 (18.0)	1614 (17.8)	364 (16.5)	227 (23.7)
Central	1488 (27.6)	1128 (12.4)	277 (12.6)	83 (8.7)
West	3384 (27.6)	2632 (29.0)	536 (24.3)	216 (22.6)
Stage at diagnosis					.002
Localized	7371 (60.2)	5376 (59.2)	1393 (63.2)	602 (63.0)
Regional	2044 (16.7)	1552 (17.1)	347 (15.7)	145 (15.2)
Metastatic	2376 (19.4)	1834 (20.2)	370 (16.8)	172 (18.0)
Unknown	449 (3.7)	318 (3.5)	94 (4.3)	37 (3.9)
Year of diagnosis					<.001
Before 2013	7271 (59.4)	5324 (58.6)	1438 (65.2)	509 (53.2)
2013-2019	4969 (40.6)	3756 (41.4)	766 (34.8)	447 (46.8)
Modified comorbidity indexb					<.001
0	6862 (56.1)	5481 (60.4)	1119 (50.8)	262 (27.4)
1-2	3933 (32.1)	2768 (30.5)	811 (36.8)	354 (37.0)
≥3	1445 (11.8)	831 (9.2)	274 (12.4)	340 (35.6)
State buy-inc					<.001
No	11 956 (97.7)	8892 (97.9)	2129 (96.6)	935 (97.8)
Yes	284 (2.3)	188 (2.1)	75 (3.4)	21 (2.2)
Educationd					<.001
<25%	2521 (20.6)	1943 (21.4)	369 (16.7)	209 (21.9)
25%-49%	4162 (34.0)	3038 (33.5)	769 (34.9)	355 (37.1)
50%-74%	3612 (29.5)	2636 (29.0)	694 (31.5)	282 (29.5)
≥75%	1945 (15.9)	1463 (16.1)	372 (16.9)	110 (11.5)
Median household income					<.001
<$35000	2030 (16.6)	1472 (16.2)	406 (18.4)	152 (15.9)
$35000-$49999	2468 (20.2)	1811 (19.9)	450 (20.4)	207 (21.7)
$50000-$74999	2846 (23.3)	2110 (23.2)	502 (22.8)	234 (24.5)
≥$75000	3153 (25.8)	2376 (26.2)	510 (23.1)	267 (27.9)
Dual eligibilityd					<.001
No	11 744 (95.9)	8766 (96.5)	2069 (93.9)	909 (95.1)
Yes	496 (4.1)	314 (3.5)	135 (6.1)	47 (4.9)

^aSome persons missing this variable. State buy-in: program where the state pays Medicare premiums to expand care to eligible people.

^bDual-eligible persons are those who may receive both Medicare and Medicaid.

^cModified comorbidity index excluded cancer and diabetes mellitus.

^dEducation: percentage in the census tract with a high school diploma or less.

Changes in unplanned hospitalization rates following androgen receptor pathway inhibitor treatment initiation

Crude unplanned hospitalization rates and rate ratios by type 2 diabetes mellitus status and androgen receptor pathway inhibitor

Table 2 and Figure 2 summarize how crude unplanned hospital admission rates increased from the 6 months before vs the 6 months after androgen receptor pathway inhibitor initiation. Although the absolute increase in crude unplanned hospitalization rates was slightly larger among patients with preexisting type 2 diabetes mellitus status (both with and without complications) than those without type 2 diabetes mellitus (Figure 2), the relative increase was larger in patients without type 2 diabetes mellitus because of lower baseline risk (Table 2 and Figure 2). Those with the highest baseline risk (patients with type 2 diabetes mellitus complications) had the lowest relative increase in crude unplanned hospitalization rates. Figure 2 also demonstrates that the crude unplanned hospitalization increased slightly more after abiraterone acetate with prednisone vs enzalutamide: 104% after abiraterone acetate with prednisone vs 84% after enzalutamide among those without type 2 diabetes mellitus, 85% vs 41% among those with type 2 diabetes mellitus without complications, and 78% vs 47% among those with type 2 diabetes mellitus complications. The crude incidence rate ratio by medication and year is found in Table S2.

Table 2.

Crude unplanned hospitalization incidence rate ratios^a

	No type 2 diabetes mellitus	Type 2 diabetes mellitus without complications	Type 2 diabetes mellitus with complications
Population considered	Incidence rate ratio (95% CI)	Incidence rate ratio (95% CI)	Incidence rate ratio (95% CI)
Overall	1.96 (1.82 to 2.12)	1.68 (1.47 to 1.91)	1.61 (1.35 to 1.92)
Abiraterone acetate with prednisone	2.04 (1.85 to 2.24)	1.86 (1.57 to 2.21)	1.78 (1.39 to 2.28)
Enzalutamide	1.84 (1.62 to 2.09)	1.41 (1.16 to 1.72)	1.47 (1.15 to 1.88)

Abbreviation: CI = confidence interval.

^aIncidence rate ratio represents the relative change in rates comparing the 6-month period after treatment initiation to the 6-month period before initiation.

Figure 2.

Unplanned hospitalization incidence rates before and after treatment initiation. Incidence rate per 6 person-months represents an inpatient hospitalization initiated by an emergency department visit divided by days at risk (when the individual was alive and not hospitalized). Abbreviations: AAP = abiraterone acetate with prednisone; ENZA = enzalutamide; T2DM = type 2 diabetes mellitus.

Covariate adjusted changes in unplanned hospitalization rates and differences associated with androgen receptor pathway inhibitor by type 2 diabetes mellitus status

Table 3 summarizes the adjusted post- vs pretreatment period unplanned hospitalization rates: 65% increase among patients with type 2 diabetes mellitus complications (adjusted incidence rate ratio = 1.65, 95% CI = 1.37 to 1.98) and 109% increase among nondiabetics (adjusted incidence rate ratio = 2.09, 95% CI = 1.94 to 2.26). Because the interaction between treatment and period was statistically significant among those type 2 diabetes mellitus patients without complications, their increase after androgen receptor pathway inhibitor depended on the type of androgen receptor pathway inhibitor initiated: a 103% increase following abiraterone acetate with prednisone (adjusted incidence rate ratio = 2.03, 95% CI = 1.70 to 2.43) and a 47% increase following enzalutamide (adjusted incidence rate ratio = 1.47, 95% CI = 1.21 to 1.80). This was a notable difference in the increase in unplanned hospitalization after abiraterone acetate with prednisone vs enzalutamide among type 2 diabetes mellitus without complications: a 38% larger increase in unplanned hospitalization for abiraterone acetate with prednisone compared with enzalutamide (ratio of abiraterone acetate with prednisone adjusted incidence rate ratio divided by enzalutamide adjusted incidence rate ratio = 1.38, 95% CI = 1.06 to 1.80). The full models, including adjusted incidence rate ratios for each level of confounder along with 95% confidence interval, are available in Table S3.

Table 3.

Adjusted unplanned hospitalization incidence rate ratios (IRR)^a

Characteristic	No type 2 diabetes mellitus		Type 2 diabetes mellitus without complicationsb		Type 2 diabetes mellitus with complications
	Adjusted incidence rate ratio (95% CI)	P	Adjusted incidence rate ratio (95% CI)	P	Adjusted incidence rate ratio (95% CI)	P
No interaction model
AAP vs ENZA	1.39 (1.26 to 1.53)	<.01			1.03 (0.83 to 1.27)	.81
Change (post vs pre)	2.09 (1.94 to 2.26)	<.01			1.65 (1.37 to 1.98)	<.01
Interaction modelb
AAP vs ENZA in pre period			1.20 (0.94 to 1.52)	.14
AAP IRR			2.03 (1.70 to 2.43)	<.01
ENZA IRR			1.47 (1.21 to 1.80)	<.01
AAP IRR vs ENZA IRR			1.38 (1.06 to 1.80)	.02

Abbreviations: AAP = abiraterone acetate with prednisone; CI = confidence interval; ENZA = enzalutamide.

^aAdjusted incidence rate ratio represents the relative change in rates between periods or the relative difference in rates between treatment types adjusted for possible confounding using all characteristics listed in Table 1.

^bThere was a statistically significant interaction between androgen receptor pathway inhibitor and period only among those having type 2 diabetes mellitus without complications, as such the main effect for AAP vs ENZA only applies to the preinitiation period, and the main effect of post vs pre only applies to those receiving AAP.

Discussion

Type 2 diabetes mellitus is common among advanced prostate cancer patients; however, data on the interplay between type 2 diabetes mellitus status and treatment outcomes in advanced prostate cancer are limited. In this population-based study, we examined the changes in unplanned hospitalization following 2 widely used androgen receptor pathway inhibitor treatments—abiraterone acetate with prednisone or enzalutamide—for advanced prostate cancer by type 2 diabetes mellitus status.^26,39 This study showed that the rates of unplanned hospitalizations increased at least 47% within 6 months of the index date, regardless of preexisting type 2 diabetes mellitus status. Furthermore, among those with type 2 diabetes mellitus without complications, the degree of increase in unplanned hospitalization depended on the type of androgen receptor pathway inhibitor received with higher rates after abiraterone acetate with prednisone vs enzalutamide. Our findings highlight the importance of monitoring potential adverse events when initiating abiraterone acetate with prednisone or enzalutamide and tailoring treatment approaches based on preexisting type 2 diabetes mellitus status.

Unplanned hospitalization is a critical factor in choosing cancer therapy because it reflects treatment safety,⁴⁰ patient quality of life,^41,42 and health-care costs.^40,43 In general, therapies with lower rates of unplanned hospitalizations indicate better tolerability and allow for more consistent treatment, potentially improving other outcomes. Reducing unplanned hospitalizations also eases financial and logistical burdens on patients and health-care systems. Furthermore, it has been noted that the extent of ED use among cancer patients has been understudied,⁴⁴ and an NCI workshop on this topic recommends future research to characterize individual, system, and societal drivers of unscheduled cancer-related medical care.⁴⁵ Finally, more than 67% of hospitalizations in the first year following cancer diagnosis originate in the ED.⁴⁰

Our study showed a substantial increase in unplanned hospitalization within 6 months of abiraterone acetate with prednisone or enzalutamide (Figure 2). The majority of our patients had metastatic castration–resistant prostate cancer because the indication for metastatic castration–sensitive prostate cancer occurred in 2018 for abiraterone acetate with prednisone⁴⁶ and 2019 for enzalutamide;⁴⁷ however, some of the patients may have had metastatic castration–sensitive prostate cancer. Although the medications were indicated for those with metastatic castration–resistant prostate cancer before 2017, it is possible that increased hospitalization rates were due to disease progression from metastatic castration–sensitive prostate cancer to metastatic castration–resistant prostate cancer. It is also possible that increases in rates were due to worsening metastatic castration–resistant prostate cancer. Although we did find that at the beginning of the study period, more patients used abiraterone than enzalutamide, this trend evened out by the end of the study period, likely as enzalutamide became more familiar and its indications expanded. Likewise, the incidence rate ratio for hospitalization also became stable from 2016 onward to the end of the study period. It is plausible that some of the increases in hospitalization rates were due to several effects, including cancer progression, secular changes in prescribing patterns, and differences in underlying comorbidity.

Differential impacts on metabolic disturbance between abiraterone acetate with prednisone and enzalutamide might have contributed to the observed 38% higher increase in unplanned hospitalization rates among those having type 2 diabetes mellitus without complications receiving abiraterone acetate with prednisone compared with enzalutamide.

Our findings are consistent with published literature showing that risks of ED visits, hospitalization, and costs were higher after abiraterone acetate with prednisone than enzalutamide.^48-51 A scoping review showed that abiraterone acetate with prednisone was associated with higher ED visits and hospitalizations than enzalutamide.⁴⁹ Another study found that patients using abiraterone acetate with prednisone were 15% more likely to have inpatient hospitalization (adjusted incidence rate ratio = 1.15, 95% CI = 1.32 to 1.01; P = .03) than enzalutamide.⁴⁹ Although this study did utilize a comparison between pre- and posttreatment, it didn’t report on inpatient admissions originating in the ED. Increased unplanned hospitalization within 6 months of treatment initiation may be related to therapy-related adverse events and could have major implications for health-care costs.

The varying associations of abiraterone acetate with prednisone and enzalutamide with unplanned hospitalization rates might, in part, result from their distinct pharmacological effects. Abiraterone acetate with prednisone functions by inhibiting CYP17A1, an enzyme crucial for producing androgens, which not only lowers androgen levels to hinder the growth of prostate cancer but also leads to an overproduction of mineralocorticoids. This imbalance may result in high blood pressure, low potassium levels, fluid retention, and various electrolyte disturbances, exacerbating cardiovascular issues prevalent in type 2 diabetes mellitus patients, thereby increasing the risk of unplanned hospital admissions^52-54; prednisone (a corticosteroid) is administered with abiraterone to counteract these side effects. However, the prednisone administered with abiraterone acetate with prednisone can add to and exacerbate diabetic complications by promoting insulin resistance and elevating glucose production in the liver, which can make it challenging for type 2 diabetes mellitus patients to keep steady blood glucose levels and may lead to serious hyperglycemic events that necessitate hospitalization.⁴⁹ Additionally, the hepatotoxic effects of abiraterone acetate with prednisone are well known, leading to more pronounced metabolic disturbances and potentially increasing hospitalizations, particularly in type 2 diabetes mellitus patients who are already more vulnerable to metabolic instability.⁴⁹ In contrast, enzalutamide has minimal impact on steroid hormone production. Its main mechanism is direct blocking androgen receptor signaling, leading to a more favorable metabolic side effect profile than prednisone’s potential effects on blood sugar. Further, as noted by La et al.,⁵⁵ abiraterone acetate with prednisone works upstream of enzalutamide, and a differential response to the medications may be due to an overactive or mutated androgen receptor, which is found in some cases of advanced prostate cancer, which would act independently (downstream) of abiraterone acetate with prednisone’s mechanism of action.

It is also likely that cancer progression accounts for the increase in hospitalization. Although each patient serves as their own pre and post control, the event of interest (new drug initiation) likely correlates with another patient-level event: clinical progression of disease. The increase in hospitalization likely represents events caused by drug toxicity and clinical complications, and this ratio should therefore not be misinterpreted as representing solely the effect of drug toxicity. The lower rates over time could be because of earlier treatment, as the approved indication for abiraterone acetate with prednisone and enzalutamide moved toward earlier stages of advanced prostate cancer and therefore less sick patients with less advanced disease. It could also be that the secular trend is because of clinicians knowing better how to manage drug toxicity. If the latter is true, it suggests that more hospitalizations today are due to complications of the cancer. Indeed, there are several studies indicating an increase in adverse events and metabolic disturbance following administration of either androgen receptor pathway inhibitors, indicating further study to discern the best candidate for each medication is warranted.^56-59 In short, although we cannot fully explain why there are more hospitalizations, it is important to note that an increasing trend clearly exists.

This study has several strengths. First, the study used each patient as their own control by comparing post- to pretreatment periods and measured changes in unplanned hospitalization rates. Second, the SEER-Medicare database is a large population-based database with a broad representation of the US population. As a result, the findings are likely to be generalizable to most patients aged 66 years and older in the United States. Importantly, this study included patients who are underrepresented in clinical trials, such as older patients with serious comorbidities. Third, the SEER-Medicare database provided longitudinal follow-up, enabling us to track health-care encounters over time. The findings on the interplay between type 2 diabetes mellitus and unplanned hospitalization offer new insights and fill knowledge gaps that clinical trials cannot provide.

This study had some limitations. The first was the retrospective observational nature of the data. Second, SEER-Medicare does not contain laboratory values or all relevant clinical information to fine-tune our adjustments. Third, although we attempted to remove type 1 diabetes mellitus patients from our cohort, some misclassification might exist. However, as previously indicated by our group and others, the misclassification was unlikely to change the conclusions,^53,60 as the proportions of type 1 diabetes mellitus in our population are small and the misclassification was unlikely to be related to the treatment choice. Nonsystematic misclassification such as this would likely attenuate the observed associations. To counteract the influences of confounding, we adjusted for a broad range of variables that included demographic variables (age, race, marital status, dual eligibility status, residence), clinical variables (stage at and year of diagnosis, comorbidity burden), and small-area measures (education, household income, state buy-in). However, even with adjustment, the study was retrospective, and there was a lack of clinical details, such as serum lab values tracking both advanced prostate cancer and type 2 diabetes mellitus, which may have influenced androgen receptor pathway inhibitor choice and likelihood of unplanned hospitalization. Because of this, the estimates may have been biased to some degree from unmeasured confounders. Finally, although we adjusted for the stage of disease at diagnosis, we were unable to adjust for whether someone had hormone-sensitive or resistant disease, possibly affecting outcomes associations. However, it should be noted that for the majority of the study period, the medications were approved for metastatic castration–resistant prostate cancer. Post hoc analysis indicated that from 2016 through the end of the study period, the rate of hospitalization leveled off. This is an important finding and one that bolsters the idea that men were receiving medications earlier in their disease state, as it follows the landmark study (NCT01715285) of abiraterone acetate with prednisone for metastatic castration–sensitive prostate cancer in 2018.⁶¹

Unexpected hospitalizations not only affect the patient’s quality of life but also add a substantial financial burden on the health-care system. The varying effects of abiraterone acetate with prednisone and enzalutamide on hospitalization rates in type 2 diabetes mellitus patients without complications underscore the importance of tailored treatment. It is essential to thoroughly assess the metabolic profiles and the comorbid conditions of their patients when selecting these androgen receptor pathway inhibitors to reduce the likelihood of adverse outcomes and related hospitalizations. It is also important to consider polypharmacy and possible drug–drug interactions in avoiding unplanned hospitalizations.^49,62,63 This underscores the role of multidisciplinary teams reviewing prior and current therapies, comorbidities, toxicity profiles, and patient preferences to inform clinical decision making.

This population-based study revealed that the rates of unplanned hospitalization increased substantially after initiating androgen receptor pathway inhibitor regardless of type 2 diabetes mellitus status. Furthermore, changes in unplanned hospitalization rates were greater with abiraterone acetate with prednisone than enzalutamide among men with preexisting type 2 diabetes mellitus. These findings highlight the importance of using real-world data to monitor cancer treatment outcomes based on preexisting health conditions. Further studies to confirm the findings and better understand the underlying reasons for unplanned hospitalization are critically important to develop tailored treatment strategies to optimize treatment outcomes.

Author contributions

Amy L. Shaver (Conceptualization, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Writing—original draft, Writing—review & editing), Krupa Gandhi (Formal analysis, Writing—review & editing), Scott W. Keith (Methodology, Writing—review & editing), Nikita Nikita (Writing—review & editing), Christopher C. Yang (Writing—review & editing), Felix J. Kim (Writing—review & editing), Hushan Yang (Writing—review & editing), William Kevin Kelly (Writing—review & editing), Stephen J. Freedland (Writing—review & editing), and Grace Lu-Yao (Conceptualization, Data curation, Funding acquisition, Project administration, Resources, Supervision, Writing—review & editing)

Supplementary material

Supplementary material is available at JNCI Cancer Spectrum online.

Funding

DoD W91XWH-05-1-0235, PA CURE: 4100088563, NCI Cancer Center Support Grant: 5P30CA056-036, NCI LRP 1L30CA284329-01, NCI 1R01CA244749, NCI R01CA255792.

Conflicts of interest

The authors declare that they have no conflicts of interests.

Data availability

SEER-Medicare Data are available after registration and payment.