Diabetes and acute care use among metastatic prostate cancer patients treated with androgen receptor signaling inhibitors

Abstract

Purpose:

Androgen receptor signaling inhibitors (ARSIs) are mainstay treatments for metastatic prostate cancer. Hyperglycemia is a common side effect, but limited data exists on outcomes such as acute care use among patients on these medications. This study aimed to assess the impact of diabetes on acute care use in older metastatic prostate cancer patients on ARSIs.

Methods:

We used SEER-Medicare data for patients aged 66 and older with de novo metastatic prostate cancer who were prescribed abiraterone, enzalutamide, or apalutamide from 2010 to 2017. Negative binomial regression calculated incidence rate ratios of acute care use (total hospital or emergency admissions divided by total time at risk) for each model covariate among diabetic and non-diabetic patients after initiation of ARSI.

Results:

2697 patients were included, of which 17.4% had diabetes. The average age was 75.0 years, and most were White (80.3%). Most (85.3%) patients received androgen deprivation therapy prior to ARSI. Acute care use within 6 months occurred in 29.5% of patients, with 39.9% in the diabetes group and 27.4% in the non-diabetes group (p<.0001), Adjusted for covariates, patients with diabetes had an increased rate of acute care use (IRR= 1.38; 95% CI 1.11–1.70, p=0.003) compared to those without diabetes. Additionally, compared to those prescribed only enzalutamide or apalutamide, patients prescribed abiraterone had an increased rate of acute care use (IRR=1.43; 95% CI 1.12–1.82, p=0.005).

Conclusions:

Acute care use was common among patients with metastatic prostate cancer on ARSIs. Patients with diabetes experienced higher rates of acute care use compared to those without diabetes among all ARSI types. Future studies should assess potential interventions in older patients with diabetes on ARSIs.

Introduction

Comprising around 11% of all cancer-related deaths, prostate cancer is the most frequent malignancy in males and the second greatest cause of cancer-related deaths overall.¹ While metastatic prostate cancer is typically diagnosed among older adults that often present with comorbid diabetes, limited concrete data is available to guide treatment decision-making in this specific population. The prevalence of diabetes in older adults with prostate cancer is estimated at around 20%.² Hospitalizations are common among patients with diabetes in general, with up to 27% of hospitalized patients (increased from 17% in 2000) having a diagnosis code of diabetes in a study from the National Inpatient Sample.³ In particular, men, Black patients, and those from poorer zip code regions had greater proportions of hospitalizations.

The advent of novel oral androgen receptor signaling inhibitors (ARSIs) for metastatic prostate cancer has increased overall survival rates significantly.^4–6 However, the clinical trials that lead to drug approval often excluded older patients and patients with pre-existing comorbidities, which are prevalent in the real world setting.^4–7 All of these oral anti-androgen agents are shown to be associated with a high rate of any-grade hyperglycemia (60–80%). In a SEER-Medicare study of older patients with locoregional prostate cancer, 12% had pre-existing diabetes and 11% of those without pre-existing diabetes developed it during treatment with androgen-deprivation therapy (ADT) with GnRH agonists.⁸ Less data is available regarding oral anti-androgen therapies; one subsequent small study showed no association with incident diabetes or cardiac adverse events.⁹ Among the ARSIs, abiraterone has been associated with a 22% increase in hospitalizations after initiation as compared with only 3% with enzalutamide, among US Veterans.⁸ While abiraterone is administered with prednisone to mitigate mineralocorticoid excess syndrome, hyperglycemia from the prednisone alone is about 7%.⁹

Concern exists when patients with diabetes are prescribed ARSI treatments in clinical practice. We hypothesize that pre-existing DM would negatively impact acute care use (i.e. hospital stays and emergency department visits) in older patients with advanced prostate cancer on ARSI treatments.

Methods

Sample Cohort

We used retrospective data from the Surveillance Epidemiology and End Results (SEER)-Medicare database. SEER data are derived from population-based cancer registries representing more than 30% of the U.S. population. Medicare data include demographic and vital status information on all Medicare beneficiaries along with administrative claims on healthcare services provided in the inpatient and outpatient settings including skilled nursing care and cancer therapies. Medicare Part D data files include additional oral prescription drug information.

Inclusion and Exclusion Criteria

Inclusion criteria were patients aged 66 and older with de novo metastatic prostate cancer who were prescribed abiraterone, enzalutamide, or apalutamide from 2010 to 2017, as defined by Medicare Part D claims. Eligible patients must have had continuous Medicare Part A and B coverage 12 months before and at least 6 months after drug initiation, and Part D coverage 6 months before initiation. Patients without HMO/MA coverage during the study period were excluded to ensure complete follow-up data. The date of 2010 was selected due to the approval of abiraterone for metastatic prostate cancer in 2011 and allowing for 1 year of comorbidity assessment. Patients were excluded if the date of death was within 6 months of diagnosis, date of death was unknown, or if chemotherapy was administered prior to ARSI. A flow diagram of the cohort selection process is illustrated in Figure 1.

Figure 1.

Flow chart

Treatments

ARSIs including Abiraterone (Zytiga), Enzalutamide (Xtandi), and Apalutamide (Erledea) were identified using ICD codes for treatments. Due to its late approval and shared mechanism, Apalutamide was combined with Enzalutamide for analysis. Patients were evaluated for treatments received prior to ARSI. This list consists of Androgen Deprivation Therapy (Leuprolide, Bicalutamide, Degarelix, Triptorelin, Flutamide, Nilutamide, Goserelin, Histrelin), Chemotherapy (Docetaxel, Cabazitaxel, Carboplatin, Cisplatin, Oxaliplatin), Sipuleucel, and Radium.

Comorbidities

Pre-existing Diabetes Mellitus (DM) was defined using ICD codes from Medicare claims for diagnosis (250.xx and 362.0x for ICD-9 and E11.x for ICD-10). ICD diagnostic codes from Medicare claims during a 12-month period prior to diagnosis were searched. For other comorbidities, we used the new NCI Combined Index, an adaptation of the Charlson Comorbidity Index for administrative claims databases.¹⁰ We identified newly developed comorbid conditions using ICD codes and Medicare Part D prescription claims. DM was excluded from the NCI score to avoid collinearity.

Covariates

Age, Race, Marital Status, and Region were obtained from SEER Cancer Files. Age was measured as a continuous variable. Race was classified as White, Black, or Other. Marital Status was categorized as Married, Unmarried, or Unknown. Region was grouped into East, Northern Plains, Pacific Coast, and Southwest. Socioeconomic status was approximated based on census tract (used to estimate income with per capita income >30,000 or <30,000) and regional high school education level (non-high school> 10% or non-high school <10%).

Outcomes

Acute care was defined as any inpatient hospital stay or emergency department visit within 6 months of initiation of oral anti-androgen therapies. Emergency visits were identified using 10 HCPCS codes from the physician claims (NCH) database, and hospitalization was classified as any unique short or long inpatient stay (excluding skilled nursing facility stays) from the MEDPAR database. This time frame was selected due to the known risk interval of increased events after starting ADT.¹¹

Statistical Analysis

Covariates of interest were selected a priori based on clinical knowledge and as previously reported.^11,12 Descriptive statistics for the covariates of interest were generated as means and standard deviations for continuous variables and proportions for categorical variables. Univariate associations between categorical covariates and diabetes were assessed using chi-square tests. The primary aim was to investigate the rates of any acute care use among those with and without pre-existing DM who were alive for at least 6 months. To account for the number of acute care events, we used negative binomial regression to calculate incidence rate ratios with 95% confidence intervals, comparing the total number of admissions to the total time at risk. To compare acute care use before and after ARSI initiation, we created a binary variable indicating whether patients had an increase in acute care use in the 6 months before and after ARSI initiation. Logistic regression was used to model the association of diabetes and covariates with this binary outcome. All analyses were performed using SAS 9.4. P-values were two-tailed with a type I error rate α=0.05.

Results

Demographics

A total of 2697 patients with metastatic prostate cancer were included in the analysis, of which 17.4% had pre-existing DM (Table 1). The average age of the cohort was 75.0 years ±7.0 SD, and the majority of patients were White (80.3%). The majority of patients (85.3%) received ADT prior to ARSI and 0.9% received ADT afterwards. Regarding treatments, 28.3% of patients received abiraterone only, 23.8% received only enzalutamide or apalutamide, and 47.6% received both. Patients with DM lived in areas of lower high school education (54.8% in areas with non-high school levels of >10%, compared to 49.5%, p=0.04) and had a higher average NCI comorbidity index (0.58 vs 0.45, p<0.001). Slightly fewer patients with diabetes were on abiraterone compared to those without diabetes (72.7% vs. 77.0%, p=0.049). There were no differences by age, race, marital status, prior ADT treatment, or receipt of enzalutamide or apalutamide between those with and without DM.

Table 1.

Baseline cohort characteristics % (n)

		Diabetes			Acute Care Use (6 mo)
	All (n=2697)	Yes (n=469)	No (n=2228)	P-value	Yes (n=791)	No (n=1906)	P-value
NCI Index (mean±SD)	0.39±0.53	0.49±0.58	0.26±0.45	<0.0001**	0.43±0.57	0.26±0.44	<0.001**
Age	75.0±7.0	75.3±6.9	75.0±7.0	0.32	75.7±7.1	74.6±6.9	0.0008**
Race
White	80.3 (2165)	80.2 (376)	80.3 (1789)	0.64	80.7 (638)	80.1 (1527)	0.78
Black	12.1 (326)	11.3 (53)	12.3 (273)		12.3 (97)	12.0 (229)
Other	7.6 (205)	8.5 (40)	7.5 (166)		7.1 (56)	7.9 (150)
Marital Status
Married	63.1 (1703)	62.5 (293)	63.3 (1410)	0.89	62.7 (496)	63.3 (1207)	0.01*
Unmarried	29.3 (789)	29.4 (138)	29.2 (651)		5.6 (44)	8.5 (161)
Unknown	7.6 (205)	8.1 (38)	7.5 (167)		31.7 (251)	28.2 (538)
Region
East	39.3 (1061)	45.0 (211)	38.2 (850)	0.002**	42.0 (332)	38.3 (729)	0.032*
Northern Plains	10.9 (294)	13.0 (61)	10.5 (233)		11.9 (94)	10.5 (200)
Pacific Coast	44.9 (1211)	38.6 (181)	46.3 (1030)		40.6 (321)	46.7 (890)
Southwest	4.9 (131)	3.4 (16)	5.2 (115)		5.6 (44)	4.6 (87)
Socioeconomic status
Non-High School >10%	50.5 (1358)	54.8 (257)	49.5 (1101)	0.038*	50.0 (396)	50.6 (962)	0.83
Non High School <10%	49.6 (1334)	45.2 (212)	50.5 (1122)		50.0 (394)	49.4 (940)
ARSI Treatments
Abiraterone	76.2 (2056)	72.7 (1715)	77.0 (341)	0.048*	75.3 (596)	76.6 (1460)	0.49
Enzalutamide or Apalutamide	71.8 (1938)	70.6 (331)	72.1 (1607)	0.46	66.3 (525)	74.1 (1413)	0.0003**
Prior ADT
ADT Prior to ARSI	85.3 (2300)	85.3 (400)	85.3 (1900)	0.99	85.2 (674)	85.3 (1626)	0.95
No ADT Prior to ARSI	14.7 (397)	14.7 (69)	14.7 (328)		14.8 (117)	14.7 (280)

^*P<0.05,

^**P<0.01, comparison between groups

Acute Care Use

There were significant differences in demographics among patients who had episodes of acute care use compared to those who did not. Patients with acute care use were older, less likely to be married, and less likely to receive enzalutamide or apalutamide (Table 1). Acute care use was common with 29.5% of patients experiencing at least 1 episode, and among these, 25.7% of patients had more than 1 episode. The most common causes of acute care use included: sepsis (n=32), disease rule out/observation (n=27), acute renal failure (n=21), neoplasm related pain (n=19), pneumonia (n=14), and COPD exacerbation (n=12). Patients with DM had higher rates of acute care use (39.9% vs. 27.4%, p<0.0001) (Figure 2). Before adjustment, patients with DM had a 100% increased rate of acute care use compared to those without DM (p<0.001). In the unadjusted model, patients with DM (vs non-DM) had higher odds of acute care use (OR=1.60; 95% CI 1.30–2.03) from the 6 months pre-initiation to the 6 months post-initiation.

Figure 2.

Diabetes and Acute Care Use Within 6 Months Among Patients on ARSIs

Adjusted for covariates, patients with DM had an increased rate of acute care use (IRR= 1.38; 95% CI 1.11–1.70, p=0.003) compared to those without DM (Table 2). Additionally, compared to those prescribed only enzalutamide or apalutamide, patients prescribed abiraterone had an increased rate of acute care use (IRR=1.43; 95% CI 1.12–1.82, p=0.005). This association of abiraterone and acute care use was independent of having DM (interaction p=0.77). In the adjusted model, patients were more likely (OR=1.22; 95% CI 0.96–1.55) to see an increase in acute care use after compared to before ARSI initiation, but this did not meet statistical significance (p=0.10) (Table 3). Also, patients prescribed abiraterone were more likely (OR 1.42, 95% CI 1.07–1.88) to see an increase in their acute care use in the same window of comparison (p=0.016). Again, there was no interaction of abiraterone use and DM (interaction p=0.17).

Table 2.

Factors associated with acute care use among patients after initiation of androgen receptor signaling inhibitors

Covariates	Incidence Rate Ratio (95% CI)	p-value
Diabetes (Yes vs No)	1.38 (1.11–1.70)	0.003**
Drug type (Abi vs Enza/Apa)	1.43 (1.12–1.82)	0.005*
Age (per year)	1.01 (1.00 −1.02)	0.18
NCI comorbidity index (per unit)	1.49 (1.24–1.80)	<0.001**
Race (Black vs White)	1.01 (0.75–1.36)	0.95
Race (Other vs White)	1.32 (0.91–1.92)	0.15
Marital Status (Unmarried vs Married)	1.25 (1.02–1.53)	0.03*
Region (Northern Plains vs East)	0.93 (0.69–1.24)	0.61
Region (Pacific Coast vs East)	0.88 (0.71–1.09)	0.23
Region (Southwest vs East)	1.11 (0.73–1.60)	0.62
SES (Non HS >10% vs <10%)	0.92 (0.76–1.11)	0.39

^*P<0.05,

^**P<0.01

Table 3.

Factors associated with increase in acute care use 6 months before and after initiation of androgen receptor signaling inhibitors

Covariates	Odds ratio (95% CI)	p-value
Diabetes (Yes vs No)	1.22 (0.96–1.55)	0.10
Drug type (Abi vs Enza/Apa)	1.42 (1.07–1.88)	0.016*
Age (per year)	1.01 (1.00–1.03)	0.10
NCI comorbidity index (per unit)	1.54 (1.26–1.90)	<0.001**
Race (Black vs White)	0.83 (0.57–1.19)	0.31
Race (Other vs White)	1.01 (0.65–1.56)	0.98
Marital Status (Unmarried vs Married)	0.98 (0.77–1.24)	0.87
Region (Northern Plains vs East)	0.81 (0.57–1.15)	0.23
Region (Pacific Coast vs East)	1.05 (0.82–1.34)	0.70
Region (Southwest vs East)	1.05 (0.65–1.70)	0.85
SES (Non HS >10% vs <10%)	0.91 (0.73–1.13)	0.38

^*P<0.05,

^**P<0.01

Sensitivity Analysis

In this cohort, 210 patients of the total 581 patients with acute care use events had emergency visits only. We performed a sensitivity analysis by calculating the adjusted IRRs for emergency visits specifically. In comparison to the main analysis with acute care as an aggregate of emergency visits and hospitalizations, the magnitudes of the IRRs did not change directions, although most of the p-values lost significance. This is likely due to the lack of power in the smaller subgroup.

Discussion

Acute care use was common among patients with metastatic prostate cancer on ARSIs, with just under 30% experiencing a hospital stay or emergency department visit within 6 months of initiation. The rate of acute care use was 41% higher among those with DM after controlling for covariates. Acute care use was also higher after initiation of ARSI compared to before initiation of ARSI among patients with diabetes. Among ARSI types, abiraterone had the highest risk of acute care use compared to patients on apalutamide or enzalutamide, but this association was independent of having diabetes.

Our study was consistent with prior studies demonstrating that abiraterone leads to a higher risk of acute care use compared to the other ARSIs,¹³ potentially secondary to enhanced cardiovascular risk.¹⁴ However, this association was independent of diabetes status, as would be expected given that abiraterone is administered alongside prednisone, which results in enhanced rates of hyperglycemia.¹⁵ This may be due in part to selection bias, as patients who are started on abiraterone likely have milder or well-controlled forms of diabetes. Using claims-based data, it is difficult to account for the granularity of diabetes cases and glycemic control. Notably however, higher A1C levels have not been shown to be a strong predictor of acute care use, particularly among older adults.^16,17 In other studies, ARSIs such as enzalutamide and darolutamide have minimally increased hospitalization rates when added to treatment regimens, among clinical trial and veteran populations^13,18. The current study shows general consistency with these findings, but among patients with diabetes, starting an ARSI regardless of the treatment type (abiraterone vs enzalutamide/apalutamide) is associated with an increase in acute care use. All of the ARSIs commonly results in hyperglycemia, which likely explains this increase in risk for adverse outcomes in diabetes patients.

Notably, the presence of other comorbid conditions as demonstrated by the NCI index also showed a strong association with acute care use among patients receiving ARSIs. Diabetes is seldom an isolated issue in older adults and is commonly associated with other comorbidities including heart disease, kidney disease, neuropathy, retinopathy, and cognition. A higher comorbidity burden may also lead to downstream issues including polypharmacy and falls.¹⁹ Our results support the importance of evaluating the heterogeneity of health status and comorbidities in older adults rather than age alone, since age was not independently associated with acute care use. Because patients with comorbidities such as heart disease are often excluded from the major clinical trials,^4–6 it can be challenging to extrapolate the best plan of care for these patients without more data.

In our cohort 17.4% of patients had a pre-existing diagnosis of diabetes. A two to three-fold increase in the prevalence of diabetes is anticipated by 2050, with the highest rise among adults aged 75 and older.²⁰ Proper management of older adults with diabetes is complex and must take into account whether interventions would be helpful to the individual patient, including factors such as life expectancy, cognitive function, presence of comorbidities, polypharmacy, and vulnerability to hypoglycemia. There is limited clinical trial level data on the utility of glycemic control or lipid-lowering medications in older adults, particularly among those with prostate cancer and other comorbidities, as they are often excluded from clinical trials. Data extrapolated from small subgroup analyses of larger trials suggest limited benefits from intensive glycemic control in frail patients²¹, and individualized A1c goals are recommended based on patient health status. For lipid lowering therapy, statins have been shown to be more effective in the secondary prevention of cardiovascular events in older adults compared to their younger counterparts^22,23. Future interventional studies are needed to address the best management approaches for this high risk and complex population.

There are several limitations to our study. Due to the sample size of patients on ARSIs and the relatively small number of patients with acute care use and diabetes in this subset, the power may be limited. Additionally, during the time period of this study, most patients received ARSI as second line therapy after prior ADT or chemotherapy, as it was approved initially under this indication, while contemporarily, it is now approved in the first line setting. However, our patient population was fairly homogenous as most patients had prior ADT and almost no patients received prior chemotherapy or other treatments, prior to applying the exclusion criteria. Also, misclassification of patient comorbidity data is possible, although we suspect this would be non-differential and bias the results towards the null. Finally, competing risks such as early deaths before hospitalization outcomes may bias the sample, even after exclusion from our dataset, but this number was minimal.

Overall, our study identifies older patients with DM as a high-risk population for acute care use when starting ARSIs, regardless of ARSI type. Diabetes status and comorbidities should be carefully managed before initiation and during treatment with these medications. Future studies addressing the efficacy of potential interventions such as glycemic monitoring and statin adherence will be essential to determining the optimal way to manage these patients. Greater education in the oncology and primary care communities on the negative side effects of ARSIs may lead to better interdisciplinary care and risk management of these patients.

Context Summary.

Key Objective

Do older metastatic prostate cancer patients on androgen receptor signaling inhibitors (ARSIs) experience increased acute care use in the presence of pre-existing diabetes?

Knowledge Generated

In this SEER-Medicare retrospective cohort study, acute care use was increased among older patients with diabetes across all ARSI types. Patients on abiraterone had the highest magnitude of increased risk, but this was independent of having diabetes.

Relevance

Older patients with diabetes and metastatic prostate cancer on ARSIs are at increased risk of acute care use, and interventions should be explored to reduce this risk. Education to oncology and primary care communities on ARSI toxicity may facilitate interdisciplinary management.