Overall survival with crizotinib and next-generation ALK inhibitors in ALK-positive non-small-cell lung cancer (IFCT-1302 CLINALK): a French nationwide cohort retrospective study

Overall survival (OS) with the anaplastic lymphoma kinase (ALK) inhibitor (ALKi) crizotinib in a large population of unselected patients with ALK-positive non-small-cell lung cancer (NSCLC) is not documented. We sought to assess OS with crizotinib in unselected ALK-positive NSCLC patients and whether post-progression systemic treatments affect survival outcomes. ALK -positive NSCLC patients receiving crizotinib in French expanded access programs or as approved drug were enrolled. We collected clinical and survival data, RECIST-defined progressive disease (PD) and post-PD systemic treatment efficacy. We performed multivariable analysis of OS from crizotinib initiation and PD under crizotinib. At time of analysis, 209 (65.7%) of the 318 included patients had died. Median OS with crizotinib was 16.6 months. The line of crizotinib therapy did not impact survival outcomes. Of the 263 patients with PD, 105 received best supportive care, 74 subsequent drugs other than next-generation ALKi and 84 next-generation ALKi. Next-generation ALKi treatment correlated with better survival outcomes in multivariate analysis. These patients had a median post-PD survival of 25.0 months and median OS from metastatic disease diagnosis of 89.6 months. Unselected ALK-positive NSCLC patients achieve good survival outcomes with crizotinib therapy. Next-generation ALKi may provide survival improvement after PD under crizotinib.


INTRODUCTION
The anaplastic lymphoma kinase (ALK) gene is rearranged in approximately 5% of non-small-cell lung cancer (NSCLC) cases, leading to constitutive activation of the ALK tyrosine kinase domain and tumorigenesis [1,2]. Crizotinib is an inhibitor of ALK kinase activity that has demonstrated its superiority over conventional chemotherapy in advanced ALK-positive NSCLC [3,4]. Crizotinib was compared to standard first-line and second-line chemotherapy in two randomized Phase III trials (PROFILE 1014 andPROFILE 1007), achieving higher response rates and a significantly longer median progression-free survival (PFS) [3,4]. It is now approved worldwide for treating advanced ALK-positive NSCLC.
The estimation of overall survival (OS) with crizotinib has not yet been fully documented. A retrospective analysis comparing 30 crizotinib-treated ALK-positive NSCLC patients to 23 crizotinib-naïve ones reported longer OS in the former (1-year OS: 70% vs. 44%; 2-year OS: 55% vs. 12%, respectively) [5]. In the PROFILE 1007 trial comparing crizotinib to pemetrexed or docetaxel as second-line following platinum-based regimen failure, an updated survival analysis showed a median OS with crizotinib at 21.7 months but identified no difference in OS between the crizotinib arm and chemotherapy arm, probably due to a cross-over in the chemotherapy arm [6]. Another limitation comes from the restrictive inclusion criteria implemented in clinical trials, meaning any benefit observed in a selected population might not reflect that in daily practice.
Most patients experience progressive disease (PD) within one year of crizotinib initiation [3,4]. Locallyablative treatment could extend disease control with crizotinib in oligoprogressive disease [7,8], and continuing crizotinib beyond PD (CBPD) might favourably impact survival outcomes [9]. Anecdotal clinical response to crizotinib rechallenge has been reported [10,11]. The efficacy of conventional chemotherapy is debatable as very little clinical data is available after crizotinib failure [12].
The IFCT-1302 CLINALK study sought to evaluate OS under crizotinib in a large unselected population of ALK-positive NSCLC patients. We also aimed to examine the different systemic treatments' effect on survival following crizotinib and whether next-generation ALKis improve survival outcomes in this setting.

Patient characteristics
In total, 318 patients were selected for data collection and analysis ( Figure 1). Of them, 214 met the inclusion criteria in the French crizotinib expanded access program (EAP) database. Following EAP discontinuation, 104 additional ALK-positive patients treated with crizotinib as second-line approved drug were enrolled. Table 1 presents the patients' characteristics on initiating crizotinib. Half were male, the median age was under 60 years old, and the majority had never smoked and presented with adenocarcinomas. One third exhibited brain metastasis, and 77.3% PS 0 or 1.

Crizotinib efficacy
Disease progression was observed in 284 patients (89.3%) by the time of data cut-off. The median PFS under crizotinib was

Effect on overall survival of systemic treatments following progression on crizotinib
Disease progression on crizotinib was documented in 284 patients. To avoid an immortal time bias and the inclusion in survival analysis of patient who could not receive subsequent treatments after crizotinib, the 21 patients who died under crizotinib were not considered for the survival analysis of systemic treatments following progression on crizotinib. Finally, a population of 263 with documented progressive disease was considered.
The sites of progression are provided in Supplementary Table 1, including 99 cases (37.6%) of cerebral progression, regardless of the extra-cerebral status. There were 136 (51.7%) oligoprogressive diseases, 60 (44.1%) of which occurred in the brain.
CBPD was documented in 86 patients (32.7%). The median duration of CBPD was 6.6 months (range: 0.7-35.2). The baseline characteristics and progression patterns of CBPD and non-CBPD patients are provided in Table 3. Patients under 65 years, with PFS on crizotinib ≥median, cerebral progression, and oligoprogression were more common in the CBPD population. This population was more commonly treated with next-generation ALKis following disease progression. The median OS from the first crizotinib dose was significantly longer in CBPD patients than in non-CBPDs (32.2 months [95% CI: 25.4-NR] versus ; log rank p <0.0001), as was median post-PD survival ] versus 4.0 months [95% CI: 3.0-5.6]; log rank p <0.0001).
The drugs administered as first-line and secondline post-crizotinib are shown in Table 4 (full details in Supplementary Table 2). The next-generation ALKis administered were ceritinib for 57 patients, alectinib for 19, ceritinib then alectinib for five, ceritinib then lorlatinib in one, and alectinib then ceritinib in two. Crizotinib was rechallenged in nine (3.4%). Chemotherapy was combined with ALKis in eight. Twenty patients received unmonitored subsequent systemic treatment after second- line post-PD: 10 in the group receiving next-generation ALKis and 10 in the group receiving subsequent drugs other than next-generation ALKi. In order to analyse the effect different systemic treatments have on survival following progression on crizotinib, the patients were separated into three groups: best supportive care (BSC) only (n=105, 40%), subsequent drugs other than next-generation ALKis (n=74, 28.1%), and next-generation ALKis (n=84, 31.9%). The characteristics at baseline and time of disease progression are provided in Table 5. Patients receiving BSC only were older, had poorer PS, were heavily pretreated, and more frequently exhibited PFS on crizotinib <median than those receiving subsequent drugs. Patients receiving next-generation ALKis were younger, more frequently exhibited PFS on crizotinib ≥median, more frequently received CBPD, and more frequently exhibited cerebral progression than patients receiving subsequent drugs other than next-generation ALKis.
Median post-PD survival was 6.5 months (95% CI: 5.3-9.8). Multivariable Cox regression (Table 6) revealed that next-generation ALKis and CBPD were associated with improved survival from the first crizotinib dose, as well as improved post-PD survival, as were baseline PS 0-1, PFS with crizotinib ≥median, cerebral progression, and oligoprogression. Subsequent treatment other than next-generation ALKis did not improve survival. BSC only was associated with worse survival from the first crizotinib dose and worse post-PD survival.
For patients receiving next-generation ALKis, median OS from the first crizotinib dose was not reached ( Figure 3A). The 1-and 3-year survival rates from the  Figure 3B). The median OS from diagnosis of metastatic disease was 89.6 months (95% CI: 53.5-not reached) for patients receiving next-generation ALKi and significantly longer than either those receiving subsequent drugs other than next-generation ALKi (28.2 months [95% CI: 22.1-33.0]) or those receiving BSC only ) (log rank P<0.001) ( Figure 3C).

DISCUSSION
Our findings provide a robust OS estimation for patients receiving crizotinib for advanced ALK-positive NSCLC primarily pretreated with first-line platinumbased regimens. Additionally, our analysis of systemic treatments following disease progression on crizotinib suggests that next-generation ALKis substantially prolong survival after crizotinib failure in comparison with other treatment strategies.
We report a median OS of 16.6 months after initiation of crizotinib, which is slightly shorter than the previous estimation of 21.7 months reported by the PROFILE 1007 trial evaluating crizotinib in the second-line setting. This could be the result of selecting patients in clinical trials compared to all-comer patients treated in daily practice. Our patients were older, less commonly women and non-smokers, with poorer PS than those enrolled in PROFILE 1007 [4]. In 40.9% of cases, patients received more than one systemic treatment before crizotinib and were less able to receive subsequent treatments, though the line of crizotinib treatment did not impact survival outcomes on crizotinib in our study. Almost all were of non-Asian ethnicity, compared to Table 4: Drugs used in first-line and second-line post-disease progression on crizotinib in patients receiving subsequent drugs other than next-generation ALKis and patients receiving next-generation ALKis

First-line post-PD on crizotinib Second-line post-PD on crizotinib
No next-generation ALKis n=74
CBPD was associated with remarkably similar survival outcomes to those previously reported by Ou et al. [9]. CLINALK and Ou et al. studies reported median OS with crizotinib of 32.2 and 29.6 months in the CBPD population and 11.2 and 10.8 months in non-CBPD, respectively [9]. The similar survival benefit we observed with CBPD in two independent cohorts, and persistence of this benefit following adjustment for the different systemic treatment strategies initiated after disease progression on crizotinib, confirm the validity of this approach, previously legitimised by the lack of effective and well-tolerated drugs available for strategies following crizotinib failure. Next-generation ALKis have emerged as the preferred treatment in this setting due to their ability to overcome Table 6: Cox multivariate analysis* of survival from the first crizotinib dose and post-progressive disease survival in the population of patients with documented progressive disease on crizotinib (n=263)

Variable
Tested Reference <0.0001 PD = progressive disease; HR = hazard ratio; CI = confidence interval; PS = performance status; PFS = progression-free survival; BSC = best supportive care; ALKi = ALK inhibitor; NS = not significant. *The full Cox univariate and multivariate proportional hazard ratio analysis is provided in Supplementary Table 3 and  Supplementary Table 4. crizotinib resistance. In our study, they deeply impacted survival outcomes when given after disease progression on crizotinib, whereas other systemic treatments did not and, as expected, BSC only was associated with worse survival. Given its retrospective nature, this survival analysis had several limitations. We included patients receiving crizotinib at different times in their disease course, in second-or third-line in 72.7% of the cases. The survival analysis focused on three non-randomized and unmatched groups of patients according to systemic treatment received after disease progression on crizotinib. As a result, the characteristics of these three groups are partially imbalanced, with biases in patient selection potentially causing the improved survival observed in those receiving CBPD or next-generation ALKis. Alternatively, this could also be due to specific tumour biology and high sensitivity to ALK inhibition, rather than a direct effect of treatment strategies. Patients received next-generation ALKis as clinical trial participants, all with good prognosis. Finally, locally-ablative treatments in cases of oligoprogressive disease were not recorded, potentially causing a partial bias in our results.

Multivariable analysis
Nevertheless, we provide an estimation of survival benefit with next-generation ALKis adjusted for potential confounding factors, including patterns of progression on crizotinib and CBPD, compared to a population treated with systemic treatments other than next-generation ALKis or BSC only. A recently published retrospective analysis of 73 patients treated with crizotinib then ceritinib reported a 49.4-month OS from diagnosis of metastatic disease, though did not provide a comparator population [20]. Two smaller studies with 11 and 13 patients treated with crizotinib then reported good survival outcomes [21,22]. In our study, the 84 patients receiving next-generation ALKis after crizotinib achieved an 89.6-month OS from diagnosis of metastatic disease, with a 59.2% 3-year survival from the first crizotinib dose and median post-PD survival of 25.0 months. The survival rates reported here in a large population treated with crizotinib then next-generation ALKis could represent an interesting benchmark for ongoing clinical trials assessing how best to sequence the available ALKis. The ALEX trial (NCT02075840) is comparing crizotinib and alectinib in first-line setting with no crossover at time of progression. In contrast, the NCI ALK Master Protocol (NCT02465060) will compare in first-line a standard treatment arm with crizotinib to several next-generation survival post-progressive disease on crizotinib for the 84 patients receiving next-generation ALK inhibitors after progression on crizotinib, compared with the 74 patients receiving subsequent treatments other than next-generation ALK inhibitors and the 105 patients receiving best supportive care only. C. Overall survival from the diagnosis of metastatic disease in the 84 patients receiving next-generation ALK inhibitors after progression on crizotinib.
ALKis and evaluate different sequential strategies by incorporating systematic crossover in each treatment arm.
In conclusion, the prolonged survival observed in crizotinib clinical trials in ALK-positive NSCLC can also be observed in less selective patient populations treated in routine practice. The remarkable median survival from diagnosis of metastatic disease of 89.6 months reported in patients treated with next-generation ALKis after crizotinib emphasizes the importance of accelerated access to diagnostic tools and targeted therapy in molecularly-defined populations. While this study was unable to resolve the question of how best to sequence ALKis, it could provide a rationale that supports the use of sequential ALKis rather than non-targeted treatment after crizotinib failure, along with offering an historical benchmark for the ongoing clinical trials in this setting.

Study population and procedures
This retrospective study included only patients with diagnosis of ALK-rearranged NSCLC determined by fluorescent in situ hybridization (FISH, performed on a routine basis at certified molecular genetics French National Cancer Institute [INCa] platforms using a certified break-apart FISH assay), with advanced/ metastatic NSCLC, aged ≥18 years, not enrolled in a crizotinib trial, having received at least 7 days of crizotinib treatment. All received 250mg oral crizotinib twice daily at initiation.
The French crizotinib expanded access program (EAP) enrolled 313 patients exhibiting any ALK-positive tumours from November 18 th 2010 to October 23 th 2012. The EAP database was provided by Pfizer. Of the 117 identified investigational centres, 80 agreed to participate. After EAP discontinuation, we enrolled patients receiving second-line crizotinib as approved drug until December 31 th 2013 at participating centres.
Data and survival follow-up were extracted from medical records by investigators in each centre and documented in a standard case report form. Database is held by the French Collaborative Thoracic Intergroup (IFCT) that ensured the quality of the data collected by monitoring the centres via periodic visits of IFCT clinical research associates. Medical monitoring was performed by two co-authors (MD, DMS). The source documents proving the collected data's integrity are filed at the investigational centre.

Definitions and study endpoints
The sites where PD manifested were reported. Oligoprogressive disease was defined as progression in only one site. CBPD was defined as continuing crizotinib for over 21 days following RECIST-defined PD and best response to crizotinib other than PD. First-line and secondline drugs following crizotinib failure and corresponding response according to RECIST 1.1. were monitored. Crizotinib rechallenge was defined as crizotinib initiation following at least one systemic therapy following PD under crizotinib [28].
The primary end-point was OS measured from the date of first crizotinib dose. Secondary endpoints included: objective response rate (ORR) according to RECIST 1.1, evaluated by investigators; disease control rate (DCR); PFS, according to RECIST 1.1.; OS from PD under crizotinib (post-PD survival); OS from diagnosis of metastatic disease.

Study oversight
This non-interventional study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines, approved by a national ethics committee, French Advisory Committee on Information Processing in Material Research in the Field of Health, and France's national data protection authority (CNIL). All participating departments approved the study protocol. All included patients still alive received information from their referring physician.

Statistical analysis
Variable characteristics were compared with the chi-squared or Fisher's exact tests for qualitative variables and Student's t-test or ANOVA for quantitative variables. The Kaplan-Meier method was used to estimate all OS endpoints. We estimated hazard ratios (HRs) and 95% confidence intervals (CIs) using a Cox model. Univariate Cox models were applied to select the most promising prognostic variables (threshold p=0.20). A multivariate Cox model was then applied using a backwards procedure to adjust for potential confounders. OS was defined as the date of first crizotinib dose to death or final follow-up. Post-PD survival was defined as the date of RECISTdefined PD under crizotinib to death or final follow-up. The cut-off for survival analysis was July 31 st 2015. All statistical tests were two-sided, and a p value <0.05 was deemed statistically significant. All analyses were performed using SAS software, Version 9.3 (SAS Institute).

ACKNOWLEDGMENTS
We thank all the investigators and their staff, including Samir Abdiche from Centre Hospitalier (CH) de Libourne; Pascal Assouline from CH de Longjumeau; Fabrice Barlési from Assistance Publique -Hôpitaux de Marseille; Patricia Barre from CH de Cahors; Olivier Bernard from Clinique Calabet; Céline Bremeault from