Real-world ramucirumab and immune checkpoint inhibitor sequences in US patients with advanced gastroesophageal cancer

Gastric and esophageal (and gastroesophageal junction, GEJ) cancers constituted up to 1.4 and 1.0% of the total cancer cases in the USA. More than 60% of the cases were diagnosed at advanced or metastatic stages [1,2]. The five-years survival rates for gastric and esophageal cancers in the USA are 32.4 and 20.0%, respectively [1,2]. Globally, there is an observed trend of decrease in incidence and mortality of gastric and esophageal cancers; however, the mortality due to gastric and esophageal cancers remains high in East Asia compared with other geographies due to higher incidence rates [3,4].

Chemotherapy is the standard of care for advanced gastroesophageal cancers, with platinum plus fluoropyrimidine-based regimens being the most preferred regimens in first-line (1L) therapy [5,6]. Treatment options beyond the 1L therapy have only been established in the last decade [7–9]. Ramucirumab (RAM), a VEGFR2 inhibitor, was approved by the US FDA as monotherapy or in combination with paclitaxel (PAC) (RAM + PAC) in patients with advanced gastric or GEJ adenocarcinoma previously treated with fluoropyrimidine/platinum [7,10,11]. However, RAM in combination with fluoropyrimidine/platinum is not recommended in 1L gastric or GEJ adenocarcinoma treatment [10]. The development and availability of immune checkpoint inhibitors (ICI) have significantly changed the treatment dynamics of solid cancers, including that of gastroesophageal cancers [12,13]. The findings from clinical studies of pembrolizumab initially led to its approval and use in third-line (3L) or in later for gastric/GEJ adenocarcinoma [14–16], and second-line (2L) esophageal squamous-cell cancer (SCC) [17,18], with nivolumab also approved for 2L esophageal cancer [19,20]. Recently, pembrolizumab [21–23] and nivolumab [9,24] were approved as 1L regimens in combination with chemotherapy for gastroesophageal cancers [21,24] or chemotherapy plus trastuzumab [22] for patients with HER2-positive gastroesophageal adenocarcinoma. With multiple studies demonstrating the benefits of ICI in combination with chemotherapy in gastroesophageal cancer, these combinations are expected to become a new standard of care in 1L therapy [25].

In addition to chemotherapy options, the National Comprehensive Care Network (NCCN) guidelines recommend RAM ± chemotherapy as 2L regimen for gastroesophageal adenocarcinomas, with ICIs recommended in 2L therapy for microsatellite instability-high (MSI-H) and tumor mutational burden (TMB) high gastroesophageal adenocarcinoma (pembrolizumab), esophageal SCC (nivolumab), and SCC with elevated PD-L1 expression (pembrolizumab) [5,6]. Despite studies demonstrating the vital role of sequence of treatments in clinical outcomes of gastric cancer [26,27], there is no consensus regarding the preferred treatment regimen in the 2L and 3L+ therapy settings in the USA [26].

Studies in Asia have demonstrated that patients treated with RAM + PAC after failing an ICI treatment had better outcomes versus patients treated with RAM + PAC therapy who were ICI-treatment naive [28,29]. Another study by Kawai et al. demonstrated that 3L RAM monotherapy has similar efficacy and toxicity as that of 2L RAM monotherapy in patients with advanced gastric cancer [30]. In another study, differences in disease characteristics and management of advanced gastric cancer by geography were reported [31]. There are limited prospective clinical trial data and real-world studies available on the sequence and outcomes of RAM and ICI treatments in the USA. Furthermore, there are no real-world data available for potential use of ICI in 1L therapy, including subsequent lines of therapy and outcomes. Therefore, real-world evidence on utilization of RAM before or after ICI initiation in patients with advanced gastroesophageal cancer could contribute to improving evidence-based clinical decision making in the US. This study aimed to fill this evidence gap by describing the sequences of RAM relative to ICI in USA patients receiving treatment for advanced gastroesophageal cancer, including RAM by line of therapy and by regimen, as part of their routine cancer care.

Patients & methods

Study design & data source

This retrospective, observational study used the Advanced Gastric/Esophageal Cancer Cohort from the USA nationwide Flatiron Health electronic health record (EHR)-derived de-identified database, which is a longitudinal database, comprising patient-level structured and unstructured data, curated via technology-enabled abstraction [32,33]. Regimens and lines of therapy in the database were derived using rules based on clinical expert input. The study was conducted using data on patients diagnosed with advanced gastric, GEJ or esophageal cancer from 1 January 2011 through end of follow-up (last visit dates) through 30 June 2020. During the study period, the de-identified data originated from approximately 280 cancer clinics (∼800 sites of care).

The study followed the ethical principles of the Declaration of Helsinki, Good Pharmacoepidemiology Practices, and applicable laws, and regulations in the USA. These research activities are covered in Flatiron Health's parent protocol which was reviewed and approved by the Copernicus Group Institutional Review Board (IRB). However, this general protocol does not cover targeted research questions or complex statistical analyses. All data were de-identified to protect patients' confidentiality in compliance with Health Insurance Portability and Accountability Act regulations. Informed consent from patients was not required.

Patient population

The database included patients diagnosed with advanced gastric, GEJ or esophageal cancer on or after 1 January 2011 and who have ≥2 visits on different days documented in the EHR during that same period. The patients were identified based on the International Classification of Diseases (ICD). For gastric cancer (ICD-9 151.x or ICD-10 C16.x), patients should have had advanced disease at diagnosis with either distant recurrence or a second locoregional recurrence, or a first locoregional recurrence that was not completely resected, or no surgical resection of the primary tumor. For GEJ or esophageal cancer (ICD-9 150.x or ICD-10 C15.x), patients should have had distant metastases at diagnosis or distant recurrence, any locoregional recurrence, or no surgical resection of the primary tumor.

Study-specific inclusion criteria were: i) initiating 1L therapy for advanced gastric, GEJ or esophageal cancer at age ≥18 years; ii) receiving RAM in any line of therapy on or after 21 April 2014 (initial regulatory approval for RAM) and iii) had structured activity within 90 days of advanced diagnosis date in the EHR.

Assessments

We described the treatment sequence with RAM in patients with advanced gastroesophageal cancer as four mutually exclusive patient groups based on the use of the first treatment in the treatment sequence: 1) RAM then ICI, 2) ICI then RAM, 3) RAM with ICI in the same line of therapy and 4) RAM without ICI in any line of therapy. In addition, we described the use of RAM in the sequences by line of therapy and by regimen, patient and disease characteristics, the use of prior therapies by line of therapy (regimens), and the use of subsequent therapies by line of therapy (percentage who receive, regimens, number of subsequent lines). Time on RAM within treatment sequences (duration of treatment) by regimen and by line of therapy was also assessed.

Patient groups

Patients in RAM without ICI group were further classified based on the line of therapy: 1L, 2L, 3L and any other line of therapy. Patients were also classified based on RAM use in these sequences by line of therapy (as 1L, 2L, 3L, fourth line or higher [4L+] therapy when either RAM or ICI was used for the first time), by regimen (RAM + PAC, monotherapy, with other agents), and primary tumor site (gastric, GEJ, esophageal). Line of therapy was based on Flatiron Health oncologist-defined rules with a revision so that substitution of any biosimilar agents did not advance the line of therapy.

Statistical analysis

Data for the study cohort and subgroups were summarized descriptively. Continuous variables were summarized using mean (standard deviation, SD), median (interquartile range, [IQR]), and range (minimum and maximum). Categorical variables were summarized by patient count and percentage for each category including missing/unknown as a special category. Sankey diagrams were generated to visualize treatment sequences.

Kaplan–Meier method was used to assess time to event variables, and medians (interquartile range, IQR) were reported. Time on RAM for a line of therapy was calculated using Kaplan–Meier method and was defined as time from the first date of RAM administration to the last date of administration. The time at risk for therapy discontinuation for both discontinued and censored patients was calculated from the first date of the therapy administration to last date of the therapy administration. A discontinuation event was defined by the presence of a subsequent line of therapy, patient death, or any gap of ≥42 days with no RAM treatment before the patient's last activity date; otherwise, the patient was censored on the last administration date for RAM.

All summaries and analyses were performed using the Instant Health Data (IHD) software (Panalgo, MA, USA) and R, version 3.2.1 (R Foundation for Statistical Computing, Vienna, Austria).

Results

Demographics & clinical characteristics

A total of 10,138 patients comprised the Flatiron Health Advanced Gastric/Esophageal Cancer Cohort, of which, 1117 patients were included in this study, Supplementary Figure 1. Of the 1117 patients, 74.9% were males, 66.0% were Caucasians, mean (SD) age was 63.2 (11.2) years at initiation of 1L therapy and mean (SD) age was 64.0 (11.2) years at initiation of the 1L therapy containing RAM. The majority of the patients (91.6%) in the study were from community oncology settings, with the remainder from the academic setting. Among the patients, 217 received RAM and ICI: RAM before ICI (n = 148, 68.2%), ICI then RAM (n = 50, 23.0%), RAM and ICI in same line of therapy (n = 19, 8.8%). ICI was not administered in any line of therapy for the remaining 900 patients.

Table 1 presents the baseline and clinical characteristics for each group. Among the patients from different groups, patients in the ICI then RAM group were the oldest (mean [SD] age 65.0 [10.2] years) at 1L initiation and had the highest proportion of patients aged ≥65 years (58.0%). A similar pattern was observed for age at RAM initiation; patients in the ICI then RAM group were older on average (mean [SD] age 66.3 [10.2] years), and patients in RAM with ICI group had the highest proportion of patients aged ≥65 years (68.4%). In all the groups, the majority of the patients had stage IV disease at initial diagnosis and had Eastern Cooperative Oncology Group (ECOG) status of 0 or 1 at 1L therapy initiation. At RAM initiation, RAM then ICI and ICI then RAM groups had >50% patients with ECOG PS of 0 or 1.

Treatment characteristics

The proportion of patients who received RAM in 2L therapy was the highest among those who received RAM then ICI and lowest in those who received ICI then RAM (RAM then ICI: 61.5%; ICI then RAM: 6.0%; RAM with ICI: 26.3%; RAM without ICI: 52.9%). RAM without ICI was administered primarily in 2L (52.9%) and 3L (22.9%) therapies. Irrespective of treatment sequence, RAM + PAC combination was the most common RAM regimen (72.0%) compared with RAM monotherapy (17.0%) and RAM + Other therapy (11.0%). In patients whose 1L initiated from 2011 to 2016, 13.6% received both RAM and ICI (any sequence). After both RAM and ICI were approved, 27.3% of patients who received 1L therapy between 2011 and 2017 received both RAM and ICI.

Median time to initiate RAM after 1L was the longest in the ICI then RAM group (12.2 months IQR 8.9–19.4 months). For each group, baseline treatment characteristics including the patients treated with RAM-containing therapy, line of therapy containing first RAM, and number of lines of therapy are included in Table 2.

Treatment patterns

RAM then ICI Group

Figure 1 presents the treatment sequences in the 148 patients who received RAM then ICI. All patients received at least two lines of therapy, with 137 receiving 3L therapy, 89 receiving 4L therapy and 42 receiving 5L therapy. In this group, 70.3% (n = 104) of patients received ICI immediately after RAM. The most common treatment received was fluoropyrimidine + platinum in 1L therapy (n = 67/148, 45.3%), RAM + taxanes in 2L therapy (n = 72/148, 48.6%), and ICI monotherapy in 3L therapy (n = 61/137, 44.5%).

Irrespective of the sequence, the most common treatment regimens were FOLFOX in 1L (n = 43/148, 29.1%), RAM + PAC in 2L (n = 68/148, 46.0%), and pembrolizumab in 3L (n = 44/137, 32.1%), 4L (n = 25/89, 28.1%), and 5L (n = 20/42, 47.6%) therapies. Other prominent treatment regimens were carboplatin + PAC (n = 14, 9.5%), and PAC + RAM (n = 13, 8.8%) in 1L therapy; FOLFOX (n = 14, 9.5%) and pembrolizumab (n = 8, 54.%) in 2L therapy; and PAC + RAM (n = 19, 13.9%) and FOLFIRI (n = 18, 13.1%) in 3L therapy. All other regimens in each line of therapy are presented in Table 3.

ICI then RAM Group

Figure 2 presents the treatment sequences in the 50 patients who received ICI then RAM. All 50 patients received at least two lines of therapy, with 47 receiving 3L therapy, 27 receiving 4L therapy, and 17 receiving 5L therapy. A total of 84.0% (n = 42) patients received ICI immediately prior to RAM. The most common treatments received in first three lines were fluoropyrimidine + platinum in 1L (n = 26, 52.0%), ICI monotherapy in 2L (n = 27, 54.0%), and RAM + taxane in 3L (n = 19, 40.4%) therapies.

In the ICI then RAM group, the most common treatment regimens across lines of therapy were FOLFOX (n = 19/50, 38.0%) in 1L, pembrolizumab (n = 19/50, 38.0%) in 2L, and RAM + PAC (n = 18/47, 38.3%) in 3L. Other prominent treatment regimens for the group included carboplatin + PAC (n = 7/50, 14.0%), and FOLFOX + trastuzumab (n = 6/50, 12.0%) in 1L therapy; nivolumab (n = 7/50, 14.0%) and FOLFOX (n = 4/50, 8.0%) in 2L therapy; and pembrolizumab (n = 11/47, 23.4%) and FOLFIRI (n = 3/47, 6.4%) in 3L therapy (Table 4).

RAM with ICI Group

RAM + ICI was mostly received in 1L therapy (n = 8/19, 42.1%), and 2L therapy (n = 6/13, 46.2%). In 3L therapy, 28.6% (n = 2/7) patients each received FOLFIRI/irinotecan or RAM + ICI. Treatment sequences are presented in Supplementary Figure 2.

Among the treatment regimens in different lines of therapy, the most common regimens were FOLFOX (n = 3/19, 15.8%) in 1L therapy and RAM + PAC + pembrolizumab (n = 4/13, 30.8%) in 2L therapy. Overall, 42.1% patients (n = 8) received 3L therapy, with all patients receiving different treatment regimens, Table 5.

RAM without ICI Group

Supplementary Figure 3 presents the treatment sequences for the 900 patients receiving RAM without ICI. Of these patients, 800 received 2L, 463 received 3L, 179 received 4L, and 78 received 5L therapy. The most common treatments received were fluoropyrimidine + platinum (n = 375/900, 41.7%) in 1L and RAM + taxane (n = 369/800, 46.1%) in 2L.

Irrespective of treatment sequence, the most common treatment regimen in the RAM without ICI group was FOLFOX in 1L (n = 269/900, 29.9%). RAM + PAC was the most common regimen in 2L (n = 346/800, 43.3%), 3L (n = 137/463, 29.6%), 4L (n = 32/179, 17.9%) and 5L (n = 15/78, 19.2%). Supplementary Tables 1–3 present the treatment regimens for the RAM without ICI group based on the line of therapy in which RAM was received.

Time on RAM in 2L & 3L therapy

When RAM was administered in combination with PAC, median duration of RAM therapy did not exceed 3.0 months. Patients who received RAM + PAC in 2L therapy in the RAM then ICI group had a similar duration of therapy compared with those who received 2L RAM + PAC therapy in the RAM without ICI group (3.0 months [IQR, 1.7–5.1 months] vs 2.3 months [IQR 1.1–5.0 months]). Patients who received RAM + PAC in 3L therapy in the RAM then ICI group had a similar median duration of therapy compared with those who received 3L RAM + PAC therapy in the ICI then RAM and the RAM without ICI groups. For patients who received RAM monotherapy in 2L therapy, the duration of RAM therapy in RAM then ICI group was similar compared with those who received 2L RAM monotherapy in the RAM without ICI group. For 3L therapy, duration of RAM monotherapy in RAM then ICI group was similar to ICI then RAM and RAM without ICI groups, Table 6.

Discussion

To the best of our knowledge, this is the first study describing the sequence of RAM and ICI therapies in patients with advanced or metastatic gastroesophageal cancer from the USA. The analysis is based on a large USA multi-institutional and geographically diverse dataset. The most common RAM-containing regimen in this study was RAM + PAC and the most common ICI regimen was pembrolizumab monotherapy. Whether RAM was used before or after ICI, treatment most commonly initiated with 1L fluoropyrimidine + platinum therapy, with 1L FOLFOX as the most common regimen.

In this descriptive analysis, a greater number of patients received RAM then ICI (n = 148) compared with ICI then RAM (n = 50), indicating a greater use of 2L RAM followed by 3L ICI than 2L ICI followed by 3L RAM for the time period of April 2014 to June 2020. While some patients received RAM (most commonly with PAC) in 3L therapy, the most common use of RAM was with PAC in 2L therapy. In addition, RAM monotherapy was used more commonly in later lines of therapy. The treatments administered in 2L therapy were aligned with the NCCN guidelines for gastric and esophageal/GEJ cancers, which also recommend use of RAM + PAC (preferred regimen) or RAM monotherapy (other recommended regimen) in gastric cancer (category 1) and the use of RAM + PAC for adenocarcinoma (preferred regimen: category 1 for GEJ and category 2A for esophageal carcinoma) or RAM monotherapy other recommended regimen: category 1 for GEJ and category 2A for esophageal carcinoma [5,6]. During the study period, NCCN guidelines recommended the use of ICIs for gastroesophageal adenocarcinoma for later lines of therapy, except in cases of MSI-H or TMB high tumors [5,34]. This study did not appreciably include patients with SCC histology as RAM is approved only for gastric/GEJ adenocarcinoma [7]. A systematic review of the treatment regimens used in advanced gastric/GEJ adenocarcinoma synthesized the evidence from 70 publications on clinical studies to provide insights on evidence-based treatment sequence. The review determined the optimal treatments in each line by HER2 expression status and reported that for patients with HER2-negative status, 2L RAM + PAC and 3L ICI therapy was considered as the optimal treatment sequence. However, patients with HER2-negative status and CPS ≥5 or ≥10 were considered eligible for ICI therapy in 1L [26]. The systematic review was limited by the fact that it included only a descriptive cross-trial comparison of the randomized clinical trials and comprised evidence available only at the time of conducting the review (2009–2019) [26], which was before the US FDA approval of 1L ICIs and their incorporation into the NCCN guidelines [8,9,16,23].

Other studies have demonstrated that exposure to ICI therapies before RAM plus taxane therapy improved the objective response rate (ORR), progression-free survival (PFS) and overall survival (OS) in patients with advanced gastroesophageal adenocarcinoma. A previous study has shown that tumor response to RAM-containing 2L therapy is an independent predictor of OS with ICI in 3L therapy. With 3L ICI, OS was significantly longer in patients who had achieved response with 2L RAM-containing regimen versus patients without response to 2L RAM-containing regimen (p < 0.001) [35]. In a retrospective study of advanced gastroesophageal adenocarcinoma, investigators compared outcomes of 19 patients who had received RAM + PAC after ICI therapy with 68 patients who had received RAM + PAC without preceding ICI therapy. Compared with the ICI-naive group, patients who received RAM + PAC after ICI had better effectiveness outcomes [28]. Another retrospective analysis included 233 patients with advanced gastric cancer: 67 exposed to ICI before RAM and 166 ICI naive. Patients exposed to ICI before RAM + taxane therapy had significantly better effectiveness outcomes compared with the ICI naive group [29]. Better outcomes in patients exposed to ICI therapy before RAM + taxanes was attributed to Tregs, tumor associated macrophages, and enhanced PD-1 inhibitor activity due to persistence of PD-1 blocking antibodies [28,29]. These studies suggest potential enhanced effectiveness for RAM + taxane after prior ICI exposure which could be a sequential option for patients who receive 1L ICI-containing regimens. The current study was not designed to compare effectiveness of sequences; tumor assessments required for ORR and PFS analyses are not included in the database, and the definitions we used to categorize treatment sequences would not be suitable to use in comparative OS analyses. While the sequence of therapy could affect the duration on RAM in each group, no important differences in duration were observed across groups. The median time on RAM + PAC therapy in the RAM then ICI group was similar to the ICI then RAM and RAM without ICI groups in 2L and 3L therapies. Similar observations were made for RAM monotherapy in the three groups for 2L and 3L therapies. Future research is needed to understand the optimal treatment sequence for the patients with advanced gastroesophageal cancer in the USA. As the opportunity to evaluate sequences is limited in the clinical trial setting, real-world studies such as this are a potential source of evidence.

Our study has several strengths. These included a relatively large, nationwide, multi-institution study design, and timeframe of the study which minimized the potential COVID impact on cancer treatment. The majority of the cohort received a 1L platinum-based regimen, representative of the standard of care for advanced gastroesophageal cancers. There are a few limitations to our research. First, the derived lines of therapy may not accurately represent all the treatment regimens and sequences. Second, the lack of ICI biomarker data in the Flatiron Health database at the time of this analysis limits the characterization of ICI use. Third, although we did not analyze surgical history, the majority of patients were initially diagnosed with advanced disease at baseline. A previous study utilizing the same database reported resection in <15% patients, thus, limiting any impact of perioperative chemotherapy [36]. Fourth, the results are descriptive with no formal comparisons between groups. Last, the study period does not include the recent introduction of 1L ICI, so the data may not reflect the most current treatment paradigm. However, until real-world data reflecting 1L ICI are available, data from this study suggest RAM effectiveness regardless of sequence.

Conclusion

In real-world routine practice in the USA during April 2014 through June 2020, patients with advanced gastric, esophageal, or GEJ cancer who received both RAM and ICI typically received RAM before an ICI, with RAM most commonly administered in 2L therapy. Among the different RAM-containing regimens, RAM + PAC was the most common RAM regimen irrespective of line of therapy or the sequence of administration of RAM, i.e., RAM before or after ICI. The study also showed that time on RAM was similar irrespective of the sequence. With approval of ICIs as 1L therapy, the treatment sequences are likely to evolve.