Healthcare resource utilization and associated costs in patients with metastatic urothelial carcinoma: a real-world analysis using German claims data

Abstract

Aims

This retrospective claims data study characterized real-world treatment patterns, healthcare resource utilization (HCRU), and costs in patients with metastatic urothelial carcinoma (mUC) in Germany.

Materials and methods

Continuously insured adults with incident mUC diagnosis (=index; ICD-10: C65–C68/C77–C79) in 2015–2019 were identified from two German claims databases. Patients who received first-line (1 L) treatment within 12 months of index were divided into three mutually exclusive sub-cohorts: platinum-based chemotherapy (PB-CT), non–PB-CT, and immunotherapy (IO). Patient characteristics were assessed during a 24-month baseline period; treatments, HCRU, and costs (of the health insurance fund) per patient-year (ppy) were described during 12-month follow-up.

Results

We identified 3,226 patients with mUC (mean age, 73.8 years; male, 70.8%; mean Elixhauser Comorbidity Index, 17.6); 1,286 (39.9%) received 1 L treatment within 12 months of index. Of these, 825 (64.2%) received PB-CT, 322 (25.0%) non–PB-CT, and 139 (10.8%) IO. On average, treated patients had 5.1 hospitalizations ppy. Most UC-related hospitalizations ppy were observed in the PB-CT cohort (5.8), followed by the non–PB-CT (4.2) and IO (2.3) cohorts. Mean UC-related hospitalization costs ppy were €22,218 in the treated cohort, €24,294 in PB-CT, €19,079 in IO, and €18,530 in non–PB-CT cohorts. Cancer-related prescription costs ppy averaged €6,323 in treated patients, and €25,955 in IO, €4,318 in non–PB-CT, and €4,270 in PB-CT cohorts.

Limitations

We recognized limitations in our study’s sample selection due to unavailable mUC disease status data. We addressed this through an upstream feasibility study conducted in consultation with clinical experts to determine a suitable proxy. Proxies were also used to delineate treatment lines, switches, and discontinuations due to data absence. Furthermore, due to data restrictions, collective dataset analysis was not possible, prompting a meta-analysis for pooled results.

Conclusions

The study shows that mUC is associated with significant HCRU and costs across different types of 1 L systemic therapy.

Keywords:

• Metastatic urothelial carcinoma
• real-world treatment
• healthcare resource utilization
• cost
• disease burden
• claims data

JEL CLASSIFICATION CODES:

• I15
• C82
• I13

Introduction

Bladder cancer is the most common malignant disease of the urinary tract, and more than 90% of all bladder cancers in Europe are urothelial carcinomas (UCs)^1–3. In 4–10% of UC cases, the disease originates from the upper urinary tract, which includes the renal pelvis and ureter³. In about a quarter of UC cases, patients present with metastatic disease at diagnosis^4,5; at this stage, the overall survival rate is low (at about 8–15 months), especially if the patient is left untreated^6–8.

In Germany, approximately 30,000 people are newly diagnosed with bladder cancer annually (∼18,000 patients with invasive carcinoma and ∼12,000 patients with non-invasive carcinoma)⁹, of whom only 26% are women^9,10. Median age at onset is 75 years, and the disease is generally known to occur more frequently in elderly males with a greater number of comorbidities and thus greater risks associated with treatment^9,11. Traditionally, platinum-based chemotherapy (PB-CT) has been used to treat locally advanced or metastatic UC (la/mUC) and has been the standard first-line (1 L) treatment for all eligible patients since the 1980s¹². For patients who cannot tolerate PB-CT due to renal impairment or other comorbidities, non–PB-CTs are available^3,13.

In recent years, immune checkpoint inhibitors have been shown to improve outcomes compared with standard therapies in the second line (2 L) and are indicated as monotherapy for the treatment of la/mUC in adults not suitable for PB-CT or after failure of prior PB-CT¹⁴. Both atezolizumab and pembrolizumab were approved in Germany in 2017 as 1 L monotherapy in patients with la/mUC who are not eligible for treatment with PB-CT and whose tumours have high programmed cell death 1 ligand 1 (PD-L1) expression and as 2 L therapy after failure of PB-CT^15,16. In the same year, nivolumab was authorized as a 2 L treatment option after disease progression following PB-CT¹⁷. Avelumab was approved in 2021 for use as 1 L maintenance therapy in patients whose disease has not progressed after 1 L PB-CT¹⁸. The administration of systemic therapies has been linked to improved survival outcomes in patients in Germany¹⁹, and the approval of these immunotherapies in 1 L can provide more options to those previously deemed ineligible for systemic treatment²⁰.

Considering these recent approvals and in anticipation of new therapies (e.g. enfortumab vedotin combined with pembrolizumab and nivolumab combined with cisplatin/gemcitabine²¹) for the treatment of la/mUC in the near future and in the coming years, payers, hospitals, and clinicians are interested in understanding healthcare resource utilization (HCRU) and costs associated with existing therapies as well as key cost drivers impacting patient care. A previous study estimating the economic burden of bladder cancer in Europe found that around €4.9 billion is spent annually on treatment, with direct healthcare costs accounting for €2.9 billion (59%) of the total cost. The five most populous countries (i.e. France, Germany, Italy, Spain, and the United Kingdom) alone accounted for €3.6 billion, 73% of all bladder cancer treatment costs throughout Europe²². However, limited research has been conducted on the economic burden or HCRU in patients with la/mUC specifically. Therefore, this study aims to investigate the real-world HCRU and costs (of the statutory health insurance [SHI]) for a representative sample of patients with mUC in Germany. In addition, since a previous analysis found that almost 59% of patients with mUC do not receive systemic treatment¹⁹ although treatment itself has a decisive impact on healthcare utilization, the study further aims to describe HCRU and costs in a cohort of treated patients with mUC.

Methods

Study design and data sources

This study was a retrospective, non-interventional, cohort analysis in patients with incident mUC based on German administrative data provided by SHI companies covering the years 2013 to 2020.

Two anonymized SHI databases were used in this study. Data were provided by AOK PLUS, one of the largest German SHI companies, covering about 3.5 million insured persons in the regions of Saxony and Thuringia, and by GWQ ServicePlus AG, an institutional association of several smaller SHI companies. The anonymous GWQ ServicePlus research dataset covers various regions throughout Germany and includes around 5 million persons insured by 19 different SHI companies. Together, both datasets represent about 10% of the total German population²³.

Study periods and population

Patients were identified between 1 January 2015 and 31 December 2019, allowing for a 24-month baseline and 12-month follow-up period (Figure 1). Insured individuals were considered patients with UC if at least one inpatient and/or two confirmed outpatient diagnoses of UC (International Classification of Diseases, Tenth Revision [ICD-10] codes: C65–C68) were documented by a specialist in two different quarters within 365 days of the identification period. This method for identifying diagnoses is considered best practice for working with German claims data and has been accepted by the German health authorities²⁴. Furthermore, an individual was assumed to have an incident metastasis diagnosis if at least one inpatient or one confirmed outpatient diagnosis of metastasis (ICD-10 codes: C77, C78, or C79 excluding C79.0 and C79.1) was documented by any specialist within the period from 3 months before the first observed UC code to 6 months after the last observed UC code and no diagnosis of metastasis was made in the previous 24 months. The date of the incident metastasis diagnosis was defined as the index date for the identified patient with mUC. Patients included in the mUC sample were required to be at least 18 years old at index date and continuously insured for at least 24 months before and 12 months after the index date (except for those who died during follow-up). Patients with one inpatient or two confirmed outpatient specialist diagnoses with ICD-10 codes C34 (malignant neoplasm of bronchus and lung), C18 (malignant neoplasm of the colon), C19 (malignant neoplasm of the rectosigmoid junction), and/or C20 (malignant neoplasm of the rectum) within 12 months before and after the index date were excluded from the study sample, as the aforementioned tumours are similarly aggressive and lethal and therefore may require a similar treatment regimen.

Figure 1. Study overview. Abbreviation. UC, urothelial carcinoma.

Patients who received at least one outpatient prescription or inpatient administration of a systemic UC treatment approved or recommended in Germany²⁵ within the first 12 months after the index date were assigned to the treated cohort (Supplementary Table S1). Furthermore, three mutually exclusive sub-cohorts were delineated according to the type of the first observed treatment during this time (i.e. 1 L mUC therapy: PB-CT, immunotherapy [IO], and non–PB-CT). Patients who did not receive recommended treatment for mUC within the first 12 months after the index were excluded from this study (Supplementary Table S2).

Study outcomes

Patient characteristics of the identified mUC cohort and the treatment sub-cohorts observed in follow-up were descriptively assessed during the 24-month baseline period (e.g. for comorbidities) or at index. Age, sex, comorbidities based on the Charlson Comorbidity Index (CCI)²⁶ and Elixhauser Comorbidity Index (ECI)²⁷ were reported.

HCRU included the number of all-cause and UC-related hospitalizations (hospitalization with a main or primary ICD-10 code of C65–C68) and the number of days hospitalized. The proportion of hospital days was reported as the number of hospitalization days divided by the total observation days within the 12-month follow-up period. The number of general practitioners, oncologists/hematologists, and the 15 specialists most frequently visited in the outpatient setting were approximated by counting the dates of invoiced codes according to the uniform valuation scheme (Einheitlicher Bewertungsmaßstab [EBM]). In addition, the number of patients with UC-related surgical procedures and interventions was assessed by respective inpatient operation and procedure key (OPS) (Supplementary Table S3). The number of patients receiving palliative care, defined as institutional care of terminally ill patients, was evaluated by OPS and EBM codes.

For AOK PLUS only (as the data were not available for GWQ), the number of patients with at least one aid (Hilfsmittel) or remedy (Heilmittel) and the 15 aids and remedies most frequently prescribed in an outpatient setting were analysed. Furthermore, the number of non-retired patients was estimated based on the insurance status “retiree,” which was only available in the AOK PLUS dataset. For non-retired patients, the number of all-cause and UC-related (defined as ICD-10 codes C65–C68) work absence days were reported based on the number of sick leave days paid by the SHI company.

Direct costs were estimated for all-cause hospitalizations and outpatient prescriptions, as well as UC-related hospitalizations and cancer-related treatments, where cancer-related treatments were defined as all antineoplastic and immunomodulatory agents included in the Anatomical Therapeutic Chemical (ATC) categories L01/L02. The costs for outpatient care were omitted as no direct monetary values are available; they can only be derived and approximated by weighted points from the EBM codes, which could lead to under- or over-estimation of the actual costs. Furthermore, not all cost items were available for the combined database, and the observable costs were negligible (e.g. rehabilitation cost, sick leave days). However, since a fairly large proportion of AOK PLUS patients received prescriptions for medical aids and remedies, the costs of these were reported.

Data analysis

Frequency analysis was performed for all categorical variables, and results are reported by number and percentage for each category. Summary statistics, including mean and standard deviation (SD), were specified for continuous variables.

HCRU and associated costs were assessed based on the 12-month follow-up, unless a patient was censored due to death. To account for differences in follow-up length, HCRU and cost items were reported per patient-year (ppy). HCRU and costs were calculated from the perspective of the SHI company. Patients’ co-payments or out-of-pocket payments were not taken into account. Costs were inflation-adjusted for price variations throughout the study period.

Due to data protection regulations and internal restrictions by the database owners, we analysed data from each database separately. However, we conducted a fixed-effects meta-analysis to combine the available data from both the AOK PLUS and GWQ databases for outcomes that were present in both datasets. As an outcome, cumulative distributions were reported for categorical variables and combined/weighted means (SD) for continuous variables, along with pooled two-sampled t-tests for the calculation of p values.

Regulatory aspects and general considerations

As the study examined a retrospective, anonymized dataset, neither ethical review nor informed patient consent was needed. However, the study protocol was reviewed by a scientific steering committee and the data owners. Work on the dataset conformed to all social security data protection requirements.

All relevant data queries for extraction of data from the claims databases were performed using Microsoft SQL Server 2019 (AOK PLUS) or 2016 (GWQ). Outcomes analyses were carried out using Stata (AOK PLUS: version 17.0, GWQ: version 15) and R (version 4.2.0), while Microsoft Excel (AOK PLUS: version 2021, GWQ: version 2019) was used for reporting purposes.

Results

Patient identification and baseline characteristics

Of the observed 35,109 patients with UC, 4,668 were diagnosed with incident metastatic disease. After applying all other exclusion criteria, 3,226 continuously insured adults with mUC were identified and included in the analysis. Of these, 1,286 patients (39.9%) received treatment in the first 12 months. Among those who received treatment, 825 (64.2%), 139 (10.8%), and 322 (25.0%) were allocated to the PB-CT, IO, and non–PB-CT cohorts, respectively. A detailed attrition chart is presented in Figure 2.

Figure 2. Study cohort overview. Abbreviations. 1L, first line; ICD-10, International Classification of Diseases, Tenth Revision; IO, immunotherapy; mUC, metastatic urothelial carcinoma; PB-CT, platinum-based chemotherapy; UC, urothelial carcinoma.

The mean (SD) age of the identified patients with incident mUC was 73.8 (10.8) years, and 70.8% were male. The comorbidity level was considerable, with a mean (SD) CCI of 6.3 (3.8) and an ECI of 17.6 (11.4). The cohort of treated patients was, on average, younger (mean [SD], 68.8 years [10.4]), mostly male (74.0%), and had a lower comorbidity level (mean [SD] CCI, 5.5 [3.5]; mean [SD] ECI, 15.2 [10.7]). Among the treated sub-cohorts, patients receiving IO had the highest mean (SD) age and comorbidity level (age, 72.7 [10.0] years; CCI, 6.5 [3.7]; ECI, 19.0 [11.8]), followed by patients who received non–PB-CT (age, 72.0 [8.8] years; CCI, 6.2 [3.7]; ECI, 17.3 [10.7]) and PB-CT (age, 66.9 [10.5] years; CCI, 5.1 [3.4]; ECI, 13.8 [10.3]). The lowest proportion of male patients was observed in the IO cohort (71.9%), followed by the PB-CT (73.7%) and non–PB-CT (75.8%) cohorts (Table 1).

Table 1. Patient characteristics and comorbidities.

	Patients with mUC	Treated patients	PB-CT cohort	IO cohort	Non–PB-CT cohort
	N = 3,226	n = 1,286	n = 825	n = 139	n = 322
Average observation length within 12 months of follow-up, mean (SD), years	0.6 (0.4)	0.8 (0.3)	0.8 (0.3)	0.7 (0.3)	0.7 (0.3)
Total patient-years within 12 months of follow-up	1,885.7	1,014.7	685.4	98.7	230.7
Age (at index), mean (SD), years	73.8 (10.8)	68.8 (10.4)	66.9 (10.5)	72.7 (10.0)	72.0 (8.8)
Males, %	70.8	74.0	73.7	71.9	75.8
CCI (24-month baseline)
Mean (SD)	6.3 (3.8)	5.5 (3.5)	5.1 (3.4)	6.5 (3.7)	6.2 (3.7)
Median (range)	6 (0–19)	5 (0–18)	4 (0–18)	6 (0–16)	5 (0–18)
ECI (24-month baseline)
Mean (SD)	17.6 (11.4)	15.2 (10.7)	13.8 (10.3)	19.0 (11.8)	17.3 (10.7)
Median (range)	16 (−7 to 66)	14 (−3 to 53)	11 (−3 to 48)	17 (−3 to 53)	16 (0–52)

Abbreviations. CCI, Charlson Comorbidity Index; ECI, Elixhauser Comorbidity Index; IO, immunotherapy; mUC, metastatic urothelial carcinoma; PB-CT, platinum-based chemotherapy.

The total observation time was 1,885.7 patient-years in the 12-month follow-up period, and the mean (SD) observation time was 0.6 (0.4) years; 60.8% of patients did not complete the 12-month follow-up period due to death. In the cohort of treated patients, the observation time was 1,014.7 patient-years, and the mean (SD) 12-month observation time amounted to 0.8 (0.3) years; 43.9% of treated patients did not complete the 12-month follow-up period due to death.

Healthcare resource utilization

Hospitalizations

Of the identified patients with mUC, 97.2% had at least one all-cause hospitalization in the 12-month follow-up period, and UC was documented as the main diagnosis in most of these hospitalizations (66.6%); in treated patients, the corresponding values were 99.1% and 80.9%, respectively (Table 2). The proportion of patients with at least one all-cause hospitalization was similar across treatment sub-cohorts (PB-CT, 99.9%; IO, 95.7%; non–PB-CT, 98.8%).

Table 2. Hospitalizations during the 12-month follow-up period.

	Patients with mUC	Treated patients	PB-CT cohort	IO cohort	Non–PB-CT cohort
	N = 3,226	n = 1,286	n = 825	n = 139	n = 322
All-cause hospitalizations
Patients with ≥1 hospitalization, n (%)	3,136 (97.2)	1,275 (99.1)	824 (99.9)	133 (95.7)	318 (98.8)
Number of hospitalizations ppy, mean (95% CI)	7.3 (7.1–7.5)	9.1 (8.8–9.4)	9.6 (9.2–10.0)	6.5 (5.8–7.4)	8.6 (8.3–9.5)
Number of hospitalization days ppy, mean (95% CI)	81.0 (77.8–83.4)	89.9 (85.1–93.3)	89.4 (84.0–93.4)	81.2 (66.7–98.3)	95.1 (90.8–109.6)
Hospitalization length for admitted patients, mean (SD)	48.7 (47.6)	71.5 (59.7)	74.4 (58.7)	60.0 (65.5)	68.9 (59.5)
Average proportion of hospitalization days, mean (95% CI)	0.39 (0.37–0.40)	0.30 (0.29–0.31)	0.28 (0.26–0.29)	0.28 (0.24–0.33)	0.34 (0.33–0.39)
Hospitalizations with UC as main diagnosis
Patients with ≥1 hospitalization, n (%)	2,148 (66.6)	1,040 (80.9)	718 (87.0)	97 (69.8)	225 (69.9)
Number of hospitalizations ppy, mean (95% CI)	3.6 (3.4–3.8)	5.1 (4.9–5.4)	5.8 (5.6–6.3)	2.3 (1.8–2.7)	4.2 (3.7–4.8)
Number of hospitalization days ppy, mean (95% CI)	40.0 (38.0–41.8)	46.7 (43.7–49.3)	50.3 (47.1–53.8)	29.7 (23.1–39.1)	43.2 (39.8–52.4)
Hospitalization length for admitted patients, mean (SD)	35.1 (33.0)	45.5 (40.0)	47.9 (40.3)	30.5 (36.4)	44.2 (39.4)
Average proportion of hospitalization days, mean (95% CI)	0.20 (0.18–0.21)	0.15 (0.14–0.16)	0.16 (0.16–0.18)	0.13 (0.11–0.18)	0.17 (0.14–0.19)

Abbreviations. IO, immunotherapy; mUC, metastatic urothelial carcinoma; PB-CT, platinum-based chemotherapy; ppy, per patient-year; UC, urothelial carcinoma.

The average number of all-cause hospitalizations ppy was 7.3 in all patients with mUC and 9.1 in treated patients. Among the treatment-specific sub-cohorts, the lowest number of all-cause hospitalizations ppy was observed in the IO cohort (6.5), while similar numbers were observed in the PB-CT (9.6) and non–PB-CT cohorts (8.6). This trend was supported when observing hospitalizations with UC as the main diagnosis only (IO, 2.3; PB-CT, 5.8; non–PB-CT, 4.2).

The average number of all-cause hospitalization days ppy in the first year after incident mUC diagnosis was 81.0 days in patients with mUC and 89.9 days in treated patients. Patients in the non–PB-CT cohort recorded the highest number of all-cause hospitalization days ppy (95.1), followed by patients receiving PB-CT (89.4). Patients receiving IO recorded a lower number, at 81.2 all-cause hospitalization days ppy. For UC-related hospitalizations, patients who received IO had the lowest observed number of hospitalization days ppy (29.7), followed by patients who received non–PB-CT (43.2) and PB-CT (50.3).

In all identified patients with at least one hospitalization, the mean (SD) hospitalization length was 48.7 (47.6) days. Treated patients had a mean (SD) length of 71.5 (59.7) days. Patients receiving IO spent, on average, 60.0 (65.5) days in the hospital, while patients receiving PB-CT and non–PB-CT spent 74.4 (58.7) and 68.9 (59.5) days, respectively. In addition, the average proportion of all-cause hospitalization days in the total observation time and across patients was 0.39 in all identified patients and 0.30 in treated patients. Patients receiving non–PB-CT had the highest average proportion of hospitalization days (0.34), followed by patients receiving IO (0.28) and PB-CT (0.28). When observing UC-related hospitalizations only, patients in the IO cohort spent 30.5 (36.4) days hospitalized, patients receiving PB-CT 47.9 (40.3) days, and patients receiving non–PB-CT 44.2 (39.4) days. The average proportion of UC-related hospitalization days was 0.13 in the IO cohort, 0.16 in patients receiving PB-CT, and 0.17 in patients receiving non–PB-CT.

Outpatient visits

During the follow-up period, 88.8% of patients with mUC visited a physician in the outpatient setting, most frequently general practitioners (81.3%), urologists (62.5%), radiologists (38.0%), and oncologists (16%) (Figure 3). In comparison, 93.7% of treated patients visited a general practitioner, 78.6% a urologist, 57.3% a radiologist, and 29.2% an oncologist. Among the treatment sub-cohorts, 36.0% of the IO cohort saw an oncologist, followed by the PB-CT (31.2%) and non–PB-CT cohorts (21.7%).

Figure 3. Percentage of patients with outpatient visits to general practitioner or specialists during follow-up. Abbreviations. IO, immunotherapy; mUC, metastatic urothelial carcinoma; PB-CT, platinum-based chemotherapy.

Palliative care

Among patients with mUC, 41.4% received palliative care; a similar percentage was observed in treated patients (42.8%). The lowest proportion of patients receiving palliative care was observed in the PB-CT cohort (40.6%), followed by the non–PB-CT (45.3%) and IO (49.6%) cohorts.

Aids and remedies

Of all identified AOK PLUS patients, 81.3% had a prescription for at least one aid or remedy, while the proportion among treated patients was 91.1%. Fewer patients in the IO cohort received UC-related aids and remedies than patients in the other treatment sub-cohorts, despite having the highest mean age (IO, 82.1%; PB-CT, 93.4%; non–PB-CT, 89.3%). The most frequently prescribed therapeutic appliances and aids for patients with mUC were incontinence aids (product group 15, 35.3%), followed by application aids (product group 03, 21.9%) and walking aids (product group 10, 17.9%).

Work absence

The proportion of AOK PLUS–insured patients without “retiree” insurance status (i.e. the assumed working population) was 14.0%, and 24.8% of treated patients were non-retired (Table 3). The proportion of non-retired patients was highest in the PB-CT cohort (29.0%), followed by the non–PB-CT cohort (18.6%) and the IO cohort (16.4%), in parallel with the average age observed in each cohort. The number of all-cause work absence days ppy among non-retired patients was 98.9 in patients with mUC and 115.3 in treated patients. The proportion of patients with work absence days varied notably between the three sub-cohorts. Patients receiving IO had a low number of work absence days during follow-up (52.8 days ppy); however, the proportion of non-retired patients in the cohort was generally small (16.4%). The non–PB-CT cohort had 77.3 and the PB-CT cohort 130.3 work absence days ppy. Similar results were observed with regard to UC-related work absence days ppy (IO, 51.5; non–PB-CT, 53.3; PB-CT, 127.5).

Table 3. Number of non-retired patients and work absence.

	Patients with mUC	Treated patients	PB-CT cohort	IO cohort	Non–PB-CT cohort
	N = 1,852^a	n = 637	n = 393	n = 67	n = 177
Non-retired patients, n (%)	260 (14.0)	158 (24.8)	114 (29.0)	11 (16.4)	33 (18.6)
Work absence
Patients with ≥1 work absence, n (%)^b	135 (37.5)	104 (65.8)	82 (71.9)	4 (36.4)	18 (54.5)
Number of work absence days ppy, mean (95% CI)	98.9 (77.3–120.5)	115.3 (88.7–141.8)	130.3 (96.7–163.8)	52.8 (−24.6 to 130.2)	77.3 (36.1–118.4)
Work absence related to UC
Patients with ≥1 work absence, n (%)	129 (49.6)	100 (63.3)	80 (70.2)	4 (36.4)	16 (48.5)
Number of work absence days ppy, mean (95% CI)	93.0 (71.9–114.1)	108.0 (82.2–133.9)	127.5 (94.6–160.3)	51.5 (−25.6 to 128.7)	53.3 (18.1–88.5)

Abbreviations. IO, immunotherapy; mUC, metastatic urothelial carcinoma; PB-CT, platinum-based chemotherapy; ppy, per patient-year; UC, urothelial carcinoma.

^aBecause the data were not available in the GWQ database, analyses were performed using only the AOK PLUS data.

^bPercentages were calculated in the working population only.

UC-related interventions

Of patients with mUC, 65.9% received at least one of the specified UC-related interventions in the first 12 months after index (Table 4). Of the treated patients, 69.3% received at least one UC-related intervention.

Table 4. UC-related surgical interventions.

	Patients with mUC	Treated patients	PB-CT cohort	IO cohort	Non–PB-CT cohort
	N = 3,226	n = 1,286	n = 825	n = 139	n = 322
Any UC-related surgical intervention, n (%)	2,126 (65.9)	891 (69.3)	582 (70.6)	93 (66.9)	216 (67.1)
Transurethral resection of the bladder	1,666 (51.6)	719 (55.9)	488 (59.2)	65 (46.8)	166 (51.6)
Continuous irrigation of the bladder	1,054 (32.7)	444 (34.5)	309 (37.5)	40 (28.8)	95 (29.5)
Cystoscopy	723 (22.4)	296 (23.0)	183 (22.2)	45 (32.4)	68 (21.1)
Ureterorenoscopy	723 (22.4)	113 (8.8)	69 (8.4)	19 (13.7)	25 (7.8)
Ureter splint	721 (22.3)	296 (23.0)	183 (22.2)	45 (32.4)	68 (21.1)
Transfusions	574 (17.8)	192 (14.9)	102 (12.4)	28 (20.1)	62 (19.3)
Cystectomy	430 (13.3)	211 (16.4)	129 (15.6)	29 (20.9)	53 (16.5)
Cutaneous urinary drainage	394 (12.2)	177 (13.8)	96 (11.6)	29 (20.9)	52 (16.2)
Percutaneous nephrostomy	378 (11.7)	134 (10.4)	85 (10.3)	20 (14.4)	29 (9.0)
Nephrectomy	204 (6.3)	105 (8.2)	52 (6.3)	23 (16.6)	30 (9.3)
Surgical evacuation of a urinary bladder tamponade	195 (6.0)	70 (5.4)	49 (5.9)	3 (2.2)	18 (5.6)
Lymph node dissection	163 (5.1)	84 (6.5)	56 (6.8)	10 (7.2)	18 (5.6)
Radiotherapy	111 (3.4)	35 (2.7)	15 (1.8)	8 (5.8)	12 (3.7)
Bladder replacement	89 (2.8)	65 (5.1)	52 (6.3)	4 (2.9)	9 (2.8)
Ureterocutaneostomy	73 (2.3)	22 (1.7)	10 (1.2)	4 (2.9)	8 (2.5)

Abbreviations. IO, immunotherapy; mUC, metastatic urothelial carcinoma; PB-CT, platinum-based chemotherapy; UC, urothelial carcinoma.

The proportion of patients receiving at least one intervention during follow-up was 70.6% in the PB-CT cohort, 66.9% in the IO cohort, and 67.1% in the non–PB-CT cohort. The most often reported UC-related interventions in patients with mUC were transurethral resection of the bladder (51.6%), continuous irrigation of the bladder (32.7%), cystoscopy (22.4%), and ureterorenoscopy (22.4%); in treated patients, the proportions were 55.9%, 34.5%, 23.0%, and 8.8%, respectively.

UC-related inpatient operations/procedures

Based on OPS, the two most common inpatient operations in patients with mUC during follow-up were “operations on the urinary bladder” (OPS 5-57, 38.0%) and “other operations on blood vessels and additional information on operations on blood vessels” (OPS 5-39, 25.9%). In patients who received treatment, the proportions were 43.6% and 54.4%, respectively. The two most common imaging procedures in the identified patients were “computed tomography with contrast medium” (OPS 3-22, 61.2%) and “native computed tomography” (OPS 3-20, 46.1%). In treated patients, 72.3% had undergone computed tomography of the skull, and 47.6% had undergone native computed tomography.

Finally, the two most common non-operational therapeutic measures in patients with mUC were “other multimodal complex treatment” (OPS 8-98, 44.0%) and “urinary tract manipulations” (OPS 8-13, 42.9%). Similar proportions of treated patients underwent these procedures (“other multimodal complex treatment,” 45.1%; “urinary tract manipulations,” 48.1%). Based on the treatment coded by OPS, “cytostatic therapy, immunotherapy, and antiretroviral therapy” was observed most frequently (OPS 8-54, 75.7%).

Healthcare costs

In patients with mUC, all-cause hospitalization costs amounted to €36,099 ppy, and most of the cost was UC related (€20,117 ppy) (Figure 4). In treated patients, €36,483 ppy all-cause hospitalization costs were observed in addition to €22,218 ppy related to UC. The lowest all-cause hospitalization costs were observed in the IO cohort (€34,970 ppy), followed by the PB-CT (€36,492 ppy) and non–PB-CT cohorts (€37,471 ppy). However, the lowest UC-specific hospitalization costs were observed in the non–PB-CT cohort (€18,530 ppy), directly followed by the IO cohort (€19,079 ppy) and then by the PB-CT cohort (€24,294 ppy).

Figure 4. All-cause outpatient prescriptions costs and all-cause hospitalization costs ppy during the 12-month follow-up period. Abbreviations. IO, immunotherapy; mUC, metastatic urothelial carcinoma; PB-CT, platinum-based chemotherapy; ppy, per patient-year.

Total outpatient prescription costs were €8,249 ppy for patients with mUC and €13,420 ppy for treated patients. In the treatment sub-cohorts, costs with IO were €42,896 ppy, while PB-CT and non–PB-CT had costs of €10,235 ppy and €10,432 ppy, respectively. Cancer-related outpatient prescription costs amounted to €3,495 ppy in all identified patients and €6,323 ppy in treated patients. Costs were highest in the IO cohort (€25,955 ppy), followed by the non–PB-CT cohort (€4,318 ppy) and PB-CT cohort (€4,270 ppy). Figure 5 shows the percentage share of UC-related costs in the combined all-cause hospitalization and prescription costs.

Figure 5. Percentage of UC-related costs in all-cause hospitalization and prescription costs. Hospitalizations were considered UC related if the main or primary International Classification of Diseases, Tenth Revision code was UC. Prescriptions were considered UC related if the patient had a cancer-related ATC code or OPS in the outpatient setting (all “L” ATC codes [antineoplastic and immunomodulating agents] and corresponding OPS medication codes [6-00]; additionally, OPS 8-54 [cytostatic chemotherapy/immunotherapy] was reported, without consideration of 8-541 [locoregional therapy] and 8-548 [highly active antiretroviral therapy]). Abbreviations. ATC, Anatomical Therapeutic Chemical; IO, immunotherapy; mUC, metastatic urothelial carcinoma; OPS, operation and procedure key; PB-CT, platinum-based chemotherapy; UC, urothelial carcinoma.

For the patients in the AOK PLUS database for whom costs for remedies and aids were available, costs amounted to €2,578 ppy, while costs for treated patients were €2,466 ppy. Among the treatment sub-cohorts, costs were very similar (IO, €2,642 ppy; PB-CT, €2,577 ppy; non–PB-CT, €2,147 ppy).

Discussion

To our knowledge, this is one of the first studies evaluating real-world mUC treatment, associated HCRU and costs, featuring data from two claims databases covering a geographically diverse SHI population in Germany. These data cover about 8.5 million insured persons, representing more than 10% of the German population insured by SHI funds. In addition, by using datasets from different kinds of German sickness funds, this study is also representative in terms of the SHI type of observed patients. Therefore, there is a high level of population coverage and assumed representativeness regarding patients and care received.

Results from this study indicate that mUC is associated with substantial HCRU and associated costs in comparison to average healthcare expenditure per patient in Germany²⁸, regardless of whether patients receive treatment, that are driven primarily by hospitalizations in a predominantly elderly patient population with substantial comorbidities. Due to costs associated with mUC-targeting therapies, treated patients incurred higher outpatient prescription costs ppy than all patients with mUC. Similar findings were observed in a previous US study by M. Kearney et al., which found that median all-cause medical costs in patients with la/mUC amounted to $76,952 per patient per year, mainly driven by inpatient costs²⁹. Other studies from the US reinforce these findings^7,30.

In addition, a small indirect economic burden was identified in association with days absent from work in non-retired patients. The number of days absent from work ppy was twice as high in non-retired treated patients than in all identified non-retired patients, which may be related to the time treated patients spend away from work while receiving treatment in the in- or outpatient setting, combined with possible treatment side effects. Notably, this concerns only a small proportion of the patient population; mUC mainly affects patients aged >65 years, and therefore, most individuals are retired. Consequently, assessments of burden related to work absence (and forgone earnings, which are not measured in the study) may be less relevant in this patient population. Future studies should seek to contextualize the economic burden related to indirect impacts on caregivers, society, and patients.

A descriptive assessment found notable differences in HCRU and costs among the three treatment sub-cohorts; patients in the IO cohort especially differed from the other cohorts. Many of these differences are likely due to differences in patient characteristics. On average, the IO therapy cohort was slightly older, had a higher comorbidity index, and contained fewer male patients than the other two cohorts. Previous studies have hypothesized that use of IO in patients diagnosed with bladder cancer may correspond with improvements in clinical outcomes and concurrent cost savings³¹. This study partially supports the hypothesis, as it found that total hospitalization costs were lower in patients receiving IO compared with PB-CT or non–PB-CT. However, it also observed higher outpatient prescription costs in this patient cohort.

One study from the US used the National Inpatient Sample database to assess outcomes in patients with gastric cancer and showed a reduction in hospitalization length and mortality, possibly in relation to use of adjuvant IO³². In our study, patients receiving IO had the lowest number of UC-related hospitalizations ppy and a lower number of days per hospitalization on average. One reason for the fewer hospital days observed in our IO patient population could be a decrease in treatment burden due to using only one agent and the occurrence of fewer severe adverse events³³. The combination of less severe side effects of these therapies and the preference for outpatient treatment (over inpatient treatment) described in § 39 SGB V³⁴ of the German Social Code may facilitate reimbursement in the outpatient setting and thus reinforce the result that fewer patients receiving IO are treated in hospitals.

All-cause work absences ppy were lowest in the IO sub-cohort, although only a small proportion of these patients were included in the working population. However, despite their positive HCRU outcomes, patients receiving IO incurred the highest all-cause prescription cost ppy in this study, and previous research has shown poorer overall survival outcomes in patients receiving IO compared with patients receiving PB-CT or non–PB-CT¹⁹. Based on these findings, future health economic and epidemiologic studies are needed to investigate the relationship between novel treatment options and HCRU, cost, and treatment outcomes.

Among the treatment sub-cohorts, patients receiving IO were found to visit oncologists and urologists most often, suggesting a possible correlation between the treating physician group and choice of therapy. One US study by J.D. Tariman et al. investigated factors related to treatment decision-making in older adults with newly diagnosed multiple myeloma³⁵. Findings from this study suggested that the physician’s expertise and the type of practice might be related to treatment choice. However, further evidence is needed to understand the external validity of these results in the German setting and to explore drivers of treatment selection in patients with mUC specifically, given a generally low observed treatment rate within the first 12 months after index mUC diagnosis¹⁹.

We know of no other German studies that have attempted to characterize and compare the direct costs incurred by patients with mUC in the first year after diagnosis. Consequently, this study addresses current gaps in the literature and provides a basis for future research using claims data to capture HCRU and cost outcomes in patients with mUC receiving treatment options approved after the observation period of this study (i.e. avelumab in 1 L maintenance monotherapy). Nevertheless, given the ever-evolving treatment landscape providing additional treatment options for elderly patients with comorbidities and mUC, further research is needed to assess indirect costs (to patients and caregivers) and to understand the societal burden faced by patients with mUC in general and those receiving different types of 1 L therapies.

Strengths and limitations

The main strength of this study is the large population from two databases including approximately 8.5 million statutory insured patients and covering all inpatient and outpatient care recorded for these patients.

Our study design made it highly improbable that any selection bias would impact the findings. Furthermore, this design is less prone to missing data, as the German claims databases contain information on almost all filled prescriptions and coded diseases, irrespective of the prescribing/diagnosing physician. We do, however, acknowledge that our analysis has certain limitations.

We recognize a potential limitation in our study’s sample selection methodology due to the use of codes C65–C68, which may encompass patients with conditions other than UC, given the absence of a specific ICD-10 code for UC. Moreover, no data on tumour, node, metastasis (TNM) staging system³⁶, laboratory values, or therapy lines are available in claims data, which are used primarily for billing purposes. Accordingly, information on clinical characteristics and treatment irrelevant for reimbursement purposes is not captured. Therefore, we acknowledge some limitations due to unavailable data and some risk of potential misclassification of the condition and disease status. To overcome these challenges, we conducted a thorough feasibility study guided by input from clinical experts. This study aimed to define our patient identification strategy and optimize our approach, despite the absence of both a specific ICD-10 code for UC and detailed disease stage information. Nevertheless, we acknowledge that this algorithm was not formally validated, and no previous studies using German claims data were identified for comparison, given that this study was the first of its kind. Moreover, the specific codes used to identify patients with mUC (i.e. OPS) may reflect German-specific clinical practices only and limit the findings’ generalizability to other countries. Linking claims to registry data could help future researchers fill information gaps and overcome some of the mentioned limitations³⁷.

Furthermore, our strategy for delineating treatment sub-cohorts is based on the first observed treatment received during 12 months of follow-up and does not account for treatment switches, which could lead to some misattribution. However, as the average follow-up within the first 12 months was short and the standards of the treatments used to define the sub-cohorts specify multiple cycles/applications when initiating mUC therapy, it can be assumed that the impact of this is minor^3,38–40. We also acknowledge that comparisons between the sub-cohorts are only descriptive, and therefore, we cannot infer statistically significant differences between sub-cohorts from our results.

This study did not assess the indirect financial and social costs for patients with mUC but was limited to the direct costs available in the databases. In addition, since no detailed information on retirement status was available for patients with family insurance, retired and employed patients could not be distinguished in this cohort. Consequently, the assumption that patients with family insurance are part of the working population may have led to incorrect allocation of patients in work absence calculations. However, since only patients with no or very low income are eligible for family insurance, it can be assumed that this does not affect many of the (generally older) patients included in this study⁴¹.

Moreover, there may be some underestimation of the number of patients receiving palliative treatment, given special reimbursement situations/rules in Germany (i.e. outpatient nursing service or hospice care), resulting in a lack of observable data within the claims databases used in this study. In addition, due to systemic differences in the variables included in the AOK PLUS and GWQ databases, for some outcomes, reporting is limited to the AOK PLUS database only (e.g. work absence, aids/remedies). Due to data protection regulations and internal restrictions by the database owners, we analysed data from each database separately. However, recognizing the value of synthesizing data for robust analysis, we conducted a fixed-effects meta-analysis to combine the available data from both the AOK PLUS and GWQ databases for outcomes that were present in both datasets. While this approach allowed us to aggregate the findings from the individual databases, enhancing the statistical power and generalizability of our results, it has the limitation that no median values could be reported; however, all other measures could be aggregated within the framework of the meta-analysis. Finally, we acknowledge the potential impact of the COVID-19 pandemic on the reported HCRU in our study, particularly within the longitudinal design.

Conclusions

mUC is a burdensome disease mainly affecting elderly males with comorbidities. Information on HCRU and related costs is limited, especially in Germany. Our study provided initial evidence using German claims data and showed that mUC is associated with considerable real-world HCRU and costs across different types of therapy and irrespective of whether treatment was received. However, the treatment landscape in mUC is dynamic, with new treatment options becoming available. Therefore, the development of real-world HCRU and costs should be further investigated, particularly within the context of the German healthcare system, considering the impact of novel therapeutic strategies. In addition, an attempt should be made to present the economic burden with a focus on the indirect effects on caregivers, society, and patients.

Transparency

Declaration of financial/other relationships

GN conducted symposia for Roche Pharma, MEDAC, Pfizer, Bristol Myers Squibb, and AstraZeneca; was part of the following advisory boards: Roche Pharma, Sanofi, Bristol Myers Squibb, Merck, Pfizer, MEDAC, and Janssen; and received reimbursement for travel cost and congress registration from Roche Pharma, Pfizer, Merck, and Bristol Myers Squibb.

MOG reports consulting or advisory roles for AstraZeneca, Bristol Myers Squibb, Ipsen, MSD, ONO, Pfizer, Astellas Pharma, and EUSA; has received reimbursement for travel and accommodations expenses from Bristol Myers Squibb and Merck; has received honoraria from Astellas Pharma, AstraZeneca, Bristol Myers Squibb, MEDAC, MSD, ONO, Novartis, Pfizer, Ipsen, Merck, and EUSA; and has received research funding from Bayer (Inst), Bristol Myers Squibb (Inst), and Intuitive Surgical (Inst).

FH, JK, and AS participated in this study as staff members of Cytel; the work of Cytel in this study was funded by Merck, as part of an alliance between Merck and Pfizer.

UO is employed by Merck Healthcare Germany GmbH, Weiterstadt, Germany, an affiliate of Merck KGaA.

BD is an employee of GWQ and has nothing to declare.

UM was employed by AOK PLUS at the time the study was conducted and has no conflict of interest to declare other than the ones related to the affiliation.

TW is the managing director of IPAM e.V. and has received honoraria/consulting fees from Cytel.

MK is employed by Merck and holds stock/shares in Merck, Novartis, and UCB Biopharma SPRL.

Peer reviewers on this manuscript have no relevant financial or other relationships to disclose.

Author contributions

All authors contributed to the study’s conception and design. FH, JK, and AS performed material preparation, data collection, and analysis. JK wrote the first draft of the manuscript, and all authors commented on previous versions. All authors read and approved the final manuscript and agree to be accountable for all aspects of the work.

Acknowledgements

Editorial support was provided by Katherine Quiroz-Figueroa on behalf of Nucleus Global and funded by Merck and Pfizer.

Data availability statement

The datasets generated and/or analyzed during the current study are not publicly available since the findings of this study are extracted from individual patient records. Data were available for research purposes from the sickness fund upon request, in an anonymized form. Due to restrictions around revealing patients' confidential information, data were used under license for the current study, and so are neither publicly available nor can be shared further.

Ethics approval

No ethical review was required due to the non-interventional, retrospective nature of this study and the anonymity of the analysed dataset.

Consent form

Informed consent was not obtained from the patients due to the anonymized and retrospective nature of the data.

Previous presentations

Some of the results in this manuscript were presented in a poster at ISPOR 2023, 7–10 May 2023; Boston, MA, USA (Poster No. RWD173).

Additional information

Funding

This research was funded by Merck (CrossRef Funder ID: 10.13039/100009945) and was previously conducted under an alliance between Merck and Pfizer. Authors who were employees of the funders were involved in the data analysis, manuscript preparation, and the decision to publish.