Treatment Sequencing Patterns in Patients with Metastatic Urothelial Cancer Treated in the Community Practice Setting in the United States: SPEAR-Bladder (Study informing treatment Pathway dEcision in bladder cAnceR)

Introduction

In treatment-naïve patients with confirmed locally advanced or metastatic urothelial carcinoma (la/mUC), platinum-based chemotherapy is the standard of care in the first-line (1L) setting.^1,2 For patients who are ineligible for cisplatin, gemcitabine–carboplatin combination chemotherapy, with a lower toxicity profile, has been the mainstay treatment option.³ However, clinical management of la/mUC is challenged by heterogenous patient populations and the need to optimally sequence treatments to yield durable responses.⁴ Treatment with chemotherapy is associated with median overall survival (OS) estimates of 9 to 15 months.^2,5,6 Following relapse on platinum-based chemotherapy, median OS is estimated to be 5 to 7 months.^2,7 Alternative treatment strategies have recently emerged, including immuno-oncology (IO)–based regimens that improve response rates and survival.^2,4,8,9

For patients who progress following 1L chemotherapy treatment, IO-based regimens have emerged as the standard of care.³ IO therapies are checkpoint inhibitors that target programmed cell death 1 ligand 1 (PD-L1) or its receptor (programmed cell death 1 protein [PD-1]). PD-L1 is an immune inhibitory molecule that, when expressed, can bind to T lymphocytes and reduce the ability of the host’s immune system to respond to the tumor.¹⁰ The therapeutic potential of agents that target the PD-1/PD-L1 pathway has been reported in multiple studies for the treatment of advanced cancer.¹¹ For UC, studies have found that approximately 20% of these tumors have a positive PD-L1 expression in the tumor membrane.^12,13

To date, 5 IO therapies have been approved by the US Food and Drug Administration (FDA) for the treatment of la/mUC following progression on platinum-based therapy: atezolizumab (a PD-L1 inhibitor), avelumab (a PD-L1 inhibitor), durvalumab (a PD-L1 inhibitor), nivolumab (a PD-1 inhibitor) and pembrolizumab (a PD-1 inhibitor).^3,4 Based on initial clinical trial results, each of these therapies is associated with favorable efficacy and safety profiles.^2,4,8,9

Based upon the results from the Phase III JAVELIN Bladder 100 trial (NCT02603432), avelumab was approved by the US FDA in June 2020 for use in patients with la/mUC that has not progressed with 1L platinum-containing chemotherapy, making it the first and only FDA-approved maintenance therapy for such patients.¹⁴ With this approval and incorporation into treatment guidelines, platinum-containing chemotherapy regimens followed by 1L avelumab maintenance therapy is a preferred treatment option for the treatment of la/mUC in patients whose disease does not progress with 1L platinum-containing chemotherapy.^3,15 The JAVELIN Bladder 100 trial evaluated avelumab plus best supportive care versus best supportive care alone as 1L maintenance among la/mUC patients whose disease did not progress with 1L platinum-based chemotherapy.^14,16–18 At the planned interim analysis, Powles et al (2020) reported that the JAVELIN Bladder 100 study met its primary endpoint of significantly prolonging OS in the study’s two primary populations: the overall population and the PD-L1+ population.^14,18

With the advent of IO-based therapies, the treatment landscape for la/mUC shifted considerably, and limited studies have examined how these trends manifested in the real-world setting. The aim of SPEAR-Bladder was, therefore, to provide a reflection on the treatment patterns during the pre-IO period as well as insights into the early adoption of IO therapies into the community oncology setting. Particular focus was given to how treatment sequences influence OS.

Patients and Methods

Study Design, Data Source, and Patient Population

This was a retrospective, observational, descriptive study of US patients with la/mUC who had evidence of initiating treatment with either a systemic chemotherapy or IO therapy (as 1L through third line [3L] of therapy) between 01 January 2015 and 30 April 2017 (ie, patient identification period) in the US Oncology Network (USON). Eligible patients were followed through 30 June 2017. The USON includes 1400 affiliated physicians operating in over 450 sites of care across states and provides care for nearly 1 million cancer patients annually.¹⁹

The study protocol was granted an exception and waiver of informed consent by the US Oncology Institutional Review Board. Waiver of informed consent was granted by the US Oncology Institutional Review Board as a deidentified dataset and stringent precautions were taken to reduce risk of reidentification. As such, this study was deemed to have minimal risk to patients. Additionally, it was not feasible to obtain informed consent from eligible patients or provide them with additional information about their participation, as this was a retrospective study of a critically ill population, many of whom were deceased or had not received further care within the USON prior to initiation of the research.

Focus was given to the subset of patients who received the two most commonly observed 1L through 3L treatment sequences: 1L chemotherapy followed by second-line (2L) IO-based regimens without 3L treatment (C-IO subgroup), and 1L chemotherapy followed by 2L chemotherapy without 3L treatment (C-C subgroup). Results of the broader study population have been reported previously.^20,21

Most study data were sourced from the USON’s electronic health record (EHR) system, iKnowMed (iKM; McKesson Corporation, Las Colinas, TX, USA), with death dates sourced from both iKM and the Social Security Administration Death Master File (DMF).

Eligible patients received at least 2 cycles of therapy during the patient identification period and had at least 2 follow-up visits at a USON clinic utilizing the full EHR capacities of iKM and whose data were accessible for research purposes. Patients enrolled in clinical trials and those with other cancer diagnoses were excluded.

Baseline variables were assessed for the 30-day period prior to treatment initiation. All study variables and outcomes were assessed regardless of maximum follow-up, using data available until the end of the study observation period (30 June 2017); no minimum follow-up period was required.

Statistical Methods

Descriptive analyses were conducted to assess demographic, clinical, and treatment characteristics. Computational algorithms were applied to categorize therapy sequences across treatment lines based on start and stop dates as well as the provider-defined line of therapy indicator in iKM. Categorical variables (eg, gender, performance status) were reported as frequencies and percentages. Continuous variables such as age were reported as mean, standard deviation, median, and range. Time to 2L was defined as the time from initiation of 1L treatment to initiation of 2L treatment.

Descriptive analyses were performed to assess associations between categorical and continuous variables. Specifically, Chi-square testing was used to assess associations between categorical variables when patient counts for single cells within the results tables were greater or equal to 5. When distribution could not be assumed to be Chi-square, then the Fisher’s exact test was used. Depending on normality, ANOVA/t-tests or Kruskal-Wallis tests were used for continuous variables.

Kaplan–Meier curves and Cox proportional hazard models were constructed to illustrate time-to-event outcomes with medians and 95% confidence intervals (CIs) and stratified to present results by the most common treatment sequences. Treatment sequence groups were compared using unadjusted hazard ratios (HR) and HR p-values. Time to treatment failure (TTF) was defined as the interval (in months) between treatment initiation and discontinuation of 1L chemotherapy. Treatment-free interval (TFI) was defined as the interval (in months) between the last administration of 1L therapy and the start date of subsequent treatment. OS was defined as the interval (in months) between the start of treatment until the date of death (any cause) as documented in the DMF or the iKM EHR database. For all Kaplan–Meier time-to-event analyses, patients who did not die within the study observation period were censored at the study end date or the last visit date available in the dataset, whichever occurred first.

Results

Study Population

In total, 502 patients initiated a qualifying treatment in the 1L, 2L or 3L setting: 430 patients were followed from initiation of 1L treatment, 66 from initiation of 2L treatment, and 6 from initiation of 3L treatment (Figure 1). Following 1L treatment discontinuation, 152 (35.3%) patients proceeded to 2L treatment. Of the patients who proceeded from 1L to 2L treatment during the study observation period, 131 did so in less than 6 months, 17 did so within 6 to 12 months and 4 after 12 months.²¹ Compared to the 278 patients who did not proceed, those who received 2L treatment were younger (median age 68 vs 74 years; P<0.0001), had more advanced disease at diagnosis (59.8% vs 47.1% with stage III/IV, respectively; P=0.0127) and had a higher number of documented metastatic sites (34.9% vs 17.6%, respectively; P=0.0256).

Figure 1 Study attrition.Notes: *These groups of patients were selected for the treatment sequencing subanalyses: patients who received 1L chemotherapy, followed by either 2L chemotherapy (n=38) or IO-based regimens (n=79) without a 3L of treatment.

The most common treatment sequences were 1L chemotherapy followed by either IO-based regimens (C-IO subgroup) or chemotherapy (C-C subgroup), without a subsequent treatment (Figure 1). In total, 79 (18.4%) patients received the C-IO treatment sequence and 38 (8.8%) the C-C treatment sequence.

The distribution of the most common treatment regimens received by the study population by line of therapy is presented in Figure 2.

Figure 2 Regimen distribution by line of therapy.

Patient Characteristics

Across the C-IO and C-C subgroups, the median age was 69 years (38–90+), with 74.4% male and 88.0% Caucasian (Table 1). In contrast, the median age of the overall study population at initiation of 1L treatment was 72 years (31–90+), with 78.4% male and 87.2% Caucasian (Table 2). Additional baseline characteristics of the study population have been previously reported.^20,21

Table 1 Demographic and Clinical Characteristics Among Treatment Sequence Subgroups

Table 2 Demographic and Clinical Characteristics at the Start of 1L and 2L Therapy

At initiation of 1L treatment, the baseline characteristics of the C-IO and C-C subgroups were similar (Table 1). The median hemoglobin level was lower among C-C patients (10.8 vs 12.2 g/dL for C-IO patients, P=0.0091). The proportion of patients with aggressive disease status was numerically higher among patients who received the C-C treatment sequence (65.8% vs 58.2% in the C-IO subgroup; P=0.7151). Otherwise, no notable differences in demographic or clinical characteristics were observed. C-IO patients had a longer time from initiation of chemotherapy to start of IO therapy than those in the C-C subgroup had from initiation of 1L chemotherapy to start of 2L regimen (ie, median 6.7 vs 4.6 months, respectively; P=0.0047).

Clinical Outcomes

Clinical outcomes for the overall study population are available in Table 3.²²

Table 3 Clinical Outcomes Across Lines of Therapy Among the Overall Study Population (n=502)

The median TTF from initiation of 1L treatment was 2.5 months (95% CI 2.3–3.0) among patients who received the C-IO treatment sequence and 2.3 months (95% CI 1.4–3.7) among those who received the C-C treatment sequence (unadjusted HR=0.832 [95% CI 0.562–1.232]; HR P=0.3582; Table 4; Figure 3). The median TFI was 2.2 months (95% CI 1.6–3.5) among C-IO patients and 1.4 months (95% CI 0.9–1.6) among the C-C subgroup (unadjusted HR 1.434 [95% CI 0.965–2.130]; HR P=0.0742). Lastly, the median OS was 19.2 months (95% CI 11.9-not reported [NR]) among patients who received the C-IO treatment sequence and 11.9 months (95% CI 8.5–20.8) among those who received the C-C treatment sequence (unadjusted HR 0.627 [95% CI 0.342–1.150]; HR P=0.1318).

Table 4 Clinical Outcomes by Most Common Treatment Sequences

Figure 3 Kaplan–Meier curves for clinical outcomes by treatment sequence.Notes: Time to treatment failure (A) and overall survival (B).

Discussion

The baseline characteristics of the C-IO and C-C subgroups considered for analysis were similar to those in previously reported real-world la/mUC patient populations.^23,24 In particular, Fisher et al (2018) performed a retrospective assessment of adult patients diagnosed with stage IV bladder cancer between 01 January 2008 and 01 June 2015 in US community oncology clinics.²³ Among the 508 patients included in the analysis, the authors reported that the mean age was 70 years among a mostly male (75.2%) and Caucasian (79.1%) population. These demographic characteristics are comparable to those observed in this study (median age of 69 years, with 74.4% male and 88.0% Caucasian). Likewise, Flannery et al (2019) reported that the median age of metastatic bladder cancer patients identified in the USON database between January 2010 and June 2014 was 72 years (range 26–90+).²⁴

Approximately 35% of 1L patients in this study proceeded to a subsequent treatment. Previous studies have similarly reported high attrition rates across lines of therapy among la/mUC patients in real-world settings. Based on an analysis of the Surveillance, Epidemiology, and End Results (SEER)-Medicare database, Galsky et al (2018) reported that only 35% of advanced or metastatic bladder cancer patients who initiated 1L treatment proceeded to a subsequent treatment.²⁵ Similarly, Fisher et al found that 265 out of the 508 (52%) 1L patients included in their study received a next line of treatment.²³ High attrition rates across lines of therapy suggest an unmet treatment need in this patient population, particularly for older and more frail patients who may not tolerate additional cytotoxic regimens.

It was observed that the time between initiation of the two treatments (time to 2L) was longer among C-IO patients (P<0.001). This trend may have been influenced by the higher proportion of patients with aggressive disease in the C-C subgroup relative to the C-IO subgroup (65.8% vs 58.2%, respectively, although this difference was not statistically significant). It was not possible, however, to determine if this difference in the proportion of patients with aggressive disease was due to underlying variation in patient characteristics or another factor, such as diagnostic testing requirements associated with use of IO regimens. In addition to the time to 2L difference observed between groups, the median OS among patients who received the C-IO treatment sequence was 19.2 months, while the median OS was 11.9 months among those who received the C-C treatment sequence.

Randomized clinical trials have investigated the role of IO therapies as a 1L maintenance strategy among patients who respond to platinum-based 1L treatment.^16,17,26,27 The Phase III, open-label JAVELIN Bladder 100 trial evaluated the safety and efficacy of avelumab plus best supportive care as 1L maintenance therapy among la/mUC patients, regardless of PD-L1 status, whose disease did not progress with 1L platinum-containing chemotherapy.^14,16,17 At the first planned interim analysis, patients who received 1L maintenance therapy with avelumab had significantly longer median OS than who received best supportive care alone.¹⁴ The results from this trial have led to avelumab becoming a recommended approach as maintenance therapy in patients whose disease does not progress on 1L platinum-containing chemotherapy.³ Results from a Phase II randomized controlled trial of pembrolizumab (NCT02500121) vs placebo suggests that transitioning patients to an IO maintenance therapy following response on 1L platinum-based treatment can extend PFS.²⁸

Several prognostic factors, including performance status, sites of metastases, white blood cell counts, hemoglobin levels, and time from prior treatment, are positively associated with improved clinical outcomes among la/mUC patient populations.^29–31 Similarly, the development of predictive biomarkers will be crucial to identify patients with la/mUC who respond to 1L chemotherapy and are suitable for treatment with IO-based therapies in the 1L maintenance setting.

The results of this study should be considered in the context of the strengths and weaknesses of its design. As a retrospective observational study, treatment selection was based on clinical considerations, and underlying patient differences may have contributed to observed variation across groups, particularly for clinical outcomes. For example, physicians may have transitioned more frail patients and/or those experiencing toxicities to IO therapies for 2L treatment. Additional research should be undertaken to understand the specific factors that influenced physician’s choice of treatment sequence and how these may change with expanded approvals of IO-based therapies.

As PD-L1 expression testing was not routinely performed in clinical practice at the time of study, we were unable to determine how PD-L1 tumor expression levels or PD-L1 testing methods affected the choice of therapy and patient outcomes. Over time, PD-L1 expression testing may become more widespread in clinical practice, and future studies should consider how expression levels influence outcomes.

Study data were sourced from the USON’s EHR, iKM, which was originally populated for clinical practice purposes, not research. Consequently, there may have been data entry errors or missing values that introduced some level of misclassification bias, and some variables of interest were not complete across the study population. Cost data were unavailable for healthcare resource use assessments and, as such, it was not possible to evaluate the direct and indirect economic implications of the treatment sequences. Lastly, the results of this study may not be generalizable to community oncology clinics outside the USON or those within the USON that do not utilize the full EHR capabilities.

Conclusions

This study provides real-world insight into treatment sequences and clinical outcomes of la/mUC patients who received care in US community oncology clinics as IO therapies were being adopted into clinical practice. Consistent with treatment guidelines, community oncology clinics appear to be adopting IO-based regimens as a subsequent treatment as the standard of care. As reported by other real-world studies, high attrition rates across lines of therapy were observed, suggesting an unmet need in this patient population. This study found notable improvements in OS among patients who received the C-IO treatment sequence versus those who received the C-C treatment sequence. As further evidence from ongoing trials of IO therapies emerges, including the potential role of 1L maintenance approach among patients who respond to platinum-based chemotherapy, it will be important to consider how these trends and outcomes manifest in the real-world community oncology setting. The results of this study may be used as a benchmark for these future studies.