Effectiveness of crizotinib in patients with ROS1-positive non-small-cell lung cancer: real-world evidence in Japan

The ROS1 proto-oncogene encodes a member of the tyrosine kinase insulin receptor family (ROS1) with substantial homology to anaplastic lymphoma kinase (ALK) [1–3]; however, the physiological role of wild-type ROS1 is as yet unclear [1,2]. Chromosomal rearrangements can lead to fusion of the ROS1 tyrosine kinase domain with one of multiple partner proteins [1,3]. These constitutively active ROS1 fusion kinases are oncogenic drivers of cell proliferation, migration and survival via upregulation of the PI3K/AKT/mTOR, JAK/STAT and MAPK/ERK signaling pathways [1,3,4]. The expression of ROS1 fusion proteins has been detected in multiple malignancies and tumor types, including in melanomas, glioblastomas, angiosarcomas and ovarian, breast, gastric and colorectal cancers [1,2]. ROS1 fusions are also observed in 1 to 2% of non-small-cell lung cancer (NSCLC) cases [3–7], with an estimated 10,000 to 15,000 patients diagnosed with ROS1-positive NSCLC each year globally [7]. In East Asian patients, the proportion of NSCLCs (lung adenocarcinomas) harboring ROS1 fusions is 2–3% [8].

Crizotinib is a potent, oral tyrosine kinase inhibitor (TKI) that targets the ALK, MET and ROS1 kinases [9–11] and was first approved for patients with ALK-positive metastatic NSCLC (in 2011 by the US FDA and in 2012 by the European Commission), with subsequent approval for the ROS1-positive indication in 2016 [12–16]. In Japan, crizotinib was approved by the Pharmaceuticals and Medical Devices Agency for the treatment of ALK-positive NSCLC in 2012 and for ROS1-positive NSCLC in 2017 [17,18].

The approval in ROS1-positive NSCLC was granted based on the results from two clinical trials: an expansion cohort of the phase I PROFILE 1001 study (NCT00585195), which evaluated 50 patients with ROS1-positive metastatic NSCLC [3,5,19], and OO12-01, a phase II study of 127 East Asian patients, of whom 26 were Japanese (NCT01945021) [20,21]. Crizotinib was highly active in patients with ROS1-positive NSCLC, and efficacy results were consistent between the two studies. Objective response rates and median progression-free survival (PFS) were 72% (95% CI: 58–84%) and 19.2 months (95% CI: 14.4 months–not reached) in PROFILE 1001 [3] and 72% (95% CI: 63–79%) and 16 months (95% CI: 13–24%) in OO12-01 [20]. Median overall survival was 51.4 months (95% CI: 29.3 months–not reached) in PROFILE 1001 [5] and 32.5 months (95% CI: 32.5 months–not reached; data considered immature at time of publication) in OO12-01 [20]. Moreover, objective response rates in OO12-01 were similar regardless of the number of lines of prior therapy in the metastatic setting (<2 vs ≥2) [20]. The safety profile of crizotinib in both studies was consistent with those of previous reports [3,5,20].

However, given the low incidence of ROS1 fusions in NSCLC (1–2% overall [3–7]; 2–3% in East Asian patients with lung adenocarcinomas [8]), real-world data on the clinical outcomes of Japanese patients with ROS1-positive NSCLC treated with crizotinib are limited. To date, real-world evidence has been published from only a single study of pretreated patients by one institute [18]. Moreover, in clinical practice, treatment statuses and post-crizotinib therapies are unclear for Japanese patients with ROS1-positive NSCLC. Therefore, clarification on the use of crizotinib post-approval in Japan is of great value.

This study aimed to understand the duration of treatment (DoT) and treatment pattern in Japanese patients with NSCLC tested for ROS1 who were administered crizotinib in order to evaluate therapeutic effectiveness and treatment status in the real-world setting.

Materials & methods

Study design

This descriptive, retrospective, observational study analyzed claims data for patients who were tested for ROS1 NSCLC and had a prescription order for crizotinib. ROS1 fusions were identified using the OncoGuide AmoyDx ROS1 Gene Fusions Detection Kit (Amoy Diagnostics Co., Ltd, Xiamen, China), the approved companion diagnostic test for use with crizotinib.

The patient claims data were obtained from the Japanese Medical Data Vision (MDV) database between June 2017 and March 2021. The MDV database contains data from more than 40 million patients in Japan, from over 460 hospitals, with coverage of approximately 26% of diagnosis procedure combination hospitals [22]. Available data from the MDV database used in this study included patient demographics, diagnoses, inpatient and outpatient visits, medications and laboratory test results; however, no data were available for Eastern Cooperative Oncology Group performance status (ECOG PS) or the presence or absence of brain metastases.

The primary end point was to describe the DoT and treatment pattern of crizotinib monotherapy in Japanese patients who were tested for ROS1 NSCLC. Primary outcome measures included the DoT with crizotinib in any line of treatment, in the first-line setting, and in the second-line or later settings. The secondary end point was to investigate the associations among treatment sequence, duration, patient demographics and clinical characteristics. Secondary outcome measures included information on post-crizotinib treatments and the association between pre-crizotinib treatments and crizotinib DoT.

Institutional review board and independent ethics committee approvals were not required due to the nature of the study, which analyzed secondary data from the MDV database. Informed patient consent was not required due to the anonymized, structured data used in the study.

Eligibility criteria

Inclusion criteria were evidence of visiting healthcare facilities that contribute data to the MDV database; enrollment into the MDV database between June 2017 and March 2021; and confirmed diagnosis for lung cancer (International Classification of Diseases, Tenth Revision [ICD-10] code C34). Patients' first prescription order for crizotinib had to be after 1 June 2017, based on the May 2017 approval of crizotinib in ROS1-positive NSCLC; date of first prescription was used as the index date. Also, patients had to have been tested for ROS1 prior to initiating crizotinib treatment. Conversely, patients treated with crizotinib were excluded from the study if they had received alectinib or ceritinib. They were also excluded if they were tested with an ALK test only or if they had been tested with an ALK test within 2 weeks of a ROS1 test; this criterion was applied because it was not possible to distinguish which of the ALK or ROS1 tests were positive when crizotinib was administered due to the test results not being available in the MDV database.

Assessments & statistical analyses

The full analysis set consisted of all patients with NSCLC who were tested for ROS1, initiated crizotinib treatment after the testing date, and met the selection criteria. DoT was calculated from the index date until the last prescription of crizotinib. For assessment of treatment patterns, the DoT for crizotinib was calculated from the index date to day 1 of the prescription date of the next line of treatment. The median DoT was estimated using the Kaplan–Meier method as the number of months from the index date until discontinuation of treatment for any documented reason. Patients were defined as receiving first-line crizotinib if they had no prior therapy recorded in the MDV and had crizotinib prescribed as first-line treatment. Patients were defined as receiving second- or later-line crizotinib if they were prescribed crizotinib after a prior drug prescription. Patients without an event (events were defined as switching from crizotinib treatment, stopping crizotinib treatment, or addition of a treatment to crizotinib) were censored at the last follow-up visit or at the end of the follow-up period. Patient numbers, percentages and descriptive statistics (mean, standard deviation, median, minimum and maximum) were reported for continuous data, while frequency (n, %) was reported for categorical variables. Treatment patterns were visualized using swimmer plots.

Results

Patients

The screening process for eligible patients in this study is shown in Figure 1. At data cutoff (March 2021), the MDV database had data on 24,233,737 patients. Of these, 292,873 had a diagnosis of lung cancer (according to ICD-10 code C34), of whom 419 had received their first prescription for crizotinib after 1 June 2017. Of these, 148 patients had had their tumors tested for ROS1 fusions. After identifying these patients, further screening was then carried out to exclude patients who had received only the ALK test or had received the ALK and ROS1 tests within 2 weeks of each other, as well as patients who had received treatment with alectinib or ceritinib, leaving a total of 58 eligible patients. Patient characteristics are summarized in Table 1: 67% were female, the median age was 67 years (range: 32.0–81.0), and the majority of patients (n = 35, 60%) had never smoked. The median body mass index was 21.3 kg/m² (range: 14.8–29.7).

Crizotinib DoT

When crizotinib was given in the first-line setting, the median DoT was 13.0 months (95% CI: 4.4–32.0 months; Figure 2A) compared with 14.0 months (95% CI: 4.6–22.2 months) in the second-line setting (Figure 2B). The median DoT in any line was 12.9 months.

Treatment post-crizotinib

27 patients (46.6%) were treated with crizotinib in the first-line setting. 13 patients continued to receive crizotinib, seven patients had stopped treatment with crizotinib and seven were receiving post-crizotinib treatment. For these seven patients, post-crizotinib treatments included chemotherapy (cyclophosphamide, epirubicin, paclitaxel, docetaxel, cisplatin, carboplatin, pemetrexed and/or gemcitabine), cancer immunotherapy (CIT; pembrolizumab), and anti–vascular endothelial growth factor receptor (VEGF/R) therapy (ramucirumab or bevacizumab), as well as further treatment with crizotinib (Figure 3A). The duration of post–first-line crizotinib treatment ranged from 0.9 to 22.4 months.

31 patients (53.4%) were treated with crizotinib in the second- or later-line setting. Nine continued to receive crizotinib, 16 had stopped treatment with crizotinib and six were receiving post-crizotinib treatment, which included chemotherapy (carboplatin, pemetrexed, paclitaxel, docetaxel, vinorelbine, and/or gemcitabine), CIT (nivolumab), anti–VEGF/R therapy (bevacizumab) and further treatment with crizotinib (Figure 3B). The duration of post–second- or later-line crizotinib treatment ranged from 2.2 to 15.3 months.

Correlation between prior treatment type & crizotinib DoT

Of the 31 patients treated with crizotinib in the second- or later-line setting, the type of prior treatments received were as follows: chemotherapy (n = 8; 26%), CIT (n = 10; 32%), anti–VEGF/R + chemotherapy (n = 12; 39%) and other therapy (n = 1; 3%).

As shown in Figure 4A-C, there did not appear to be any association between the type of prior treatment received (CIT based; anti–VEGF/R + chemotherapy; or chemotherapy) or the duration of prior therapy and the DoT of crizotinib; however, no formal statistical tests were performed, and population sizes were quite small. One patient who received prior treatment with anti–VEGF/R therapy, CIT and chemotherapy was excluded from this analysis.

Discussion

Crizotinib is strongly recommended by the Japanese Lung Cancer Society guidelines for the first-line treatment of ROS1-positive metastatic NSCLC [23] based on efficacy and safety data from four phase II clinical trials and one phase I clinical trial, although there are variations in the efficacy reported in these studies [3,5,20,24–26]. However, available data for crizotinib in Japanese patients is limited to a subpopulation from a single clinical trial [20] and from a study of real-world evidence from a single institute in Japan [18].

This analysis of real-world, post-approval data from the Japanese MDV database provides further evidence of the effectiveness of crizotinib monotherapy treatment for ROS1-positive NSCLC, including a median DoT of 12.9 months, regardless of the line of crizotinib treatment. Indeed, at the data cutoff, some patients continued to receive crizotinib treatment (in any line) with an ongoing treatment duration of >33 months.

The median DoT of crizotinib in this study (~13 months for the first-line setting) was comparable with median PFS previously reported in clinical or real-world studies of Japanese patients (10.0–15.9 months) [18,20] but shorter than median PFS reported in European or global clinical studies (19.2–22.8 months) [3,24,25], whereas real-world data from patients in China showed some variation in efficacy (median PFS, 11.0–23.0 months) [27–29]. Differences in the duration of crizotinib treatment between this study and others may be due to variations in efficacy and/or due to the large proportion of patients censored in this analysis (in the first-line setting, 13 of 27 patients were censored [48%]). However, comparisons between studies should be made with caution given the differences in study design, patient populations, and baseline characteristics such as age. Previous studies have also focused primarily on patients with ECOG PS 0–1 [3,18,20,24,25,28,29]; however, ECOG PS data were not collected in this study. Real-world data for agents targeting ROS1-positive NSCLC are scarce; an integrated analysis of three clinical studies of entrectinib in ROS1-positive locally advanced or metastatic NSCLC reported a median PFS of 15.7 months [30]; however, there are no reports to date of real-world data for entrectinib. Moreover, the DoT of crizotinib was comparable between patients treated in the first and second lines and was not affected by duration of prior therapies, complementing the results from the OO12-01 study that showed a clinical benefit of crizotinib irrespective of the number of prior lines of therapy [20].

Previous studies have shown that patients with programmed death-ligand 1-positive advanced NSCLC that is also ROS1-positive are less likely to derive benefit from CIT [31–33]. Moreover, sequential CIT and crizotinib treatment has been shown to be associated with a significantly increased risk of hepatotoxicity [34]. Therefore, testing for ROS1 fusions at initial diagnosis is important to ensure that these patients are treated with the most appropriate therapy upfront, such as ROS1-targeted TKIs. In this study, several patients received single-agent CIT prior to crizotinib, with varying durations of CIT pretreatment. Although the number of patients was low, no association or trend was observed between the type of prior treatment and the duration of treatment following crizotinib. Moreover, to date, no real-world data from other countries have been published on the effects of prior therapies on crizotinib DoT.

Patients who were treated with chemotherapy post-crizotinib did derive some benefit, consistent with studies of the efficacy of chemotherapy in ROS1-positive NSCLC [27,35,36]. In this study, post-crizotinib treatments mostly consisted of CIT, chemotherapy alone, or chemotherapy in combination with anti–VEGF/R therapies, and usage did not seem to differ according to the line of treatment. Treatment with crizotinib in the first-, second- or later-line settings did not appear to affect the duration of subsequent therapies.

These data also highlight the variation in treatment sequencing in Japanese medical institutions for patients tested for ROS1 NSCLC. Of the patients included in the study, only half were treated with crizotinib in the first-line setting despite recommendations from clinical guidelines for crizotinib to be used as a first-line treatment [23]. Prior therapies for patients who were administered crizotinib in the second- or later-line setting were varied and included CIT and chemotherapy alone or in combination with anti–VEGF/R therapy, reflecting the diverse treatment landscape in Japan.

The limitations of this study were that patients whose tumors were tested for ROS1 and were subsequently treated with crizotinib were considered to be ROS1-positive for the purposes of this analysis; however, the actual results of the ROS1 tests were unknown in this study because the results were not available as structured data in the MDV database. Furthermore, only the OncoGuide AmoyDx ROS1 Gene Fusions Detection Kit was permitted in the study protocol; alternative testing kits such as the Oncomine DX Target Test (Thermo Fisher Scientific, MA, USA) were not included, thus potentially excluding some patients from this study. Other limitations include the fact that the calculation of DoT and Kaplan–Meier events was dependent on the order of treatment, and therefore DoT may have been underestimated compared with the actual treatment duration; also, due to the nature of the database, it was not possible to account for periods of dose reduction or drug withdrawal due to adverse events. Additionally, as survival time was not analyzed, the results of this study are difficult to contextualize with respect to those of clinical trials. Finally, the generalizability of these findings is limited by the fact that only 58 of 419 patients receiving crizotinib were included in this analysis, and that this study did not include entrectinib as a post-crizotinib treatment due to the timing of entrectinib's approval for ROS1-positive NSCLC in Japan (February 2020) [37].

Conclusion

In conclusion, this study reports valuable real-world data on the clinical use of crizotinib monotherapy and treatment sequencing for Japanese patients tested for ROS1 NSCLC and further complements the effectiveness of crizotinib in ROS1-positive NSCLC seen previously in clinical-trial and real-world settings.