Comparative efficacy and safety of first-line treatments for advanced non-small cell lung cancer with ALK-rearranged: a meta-analysis of clinical trials

Background Whereas there are many pharmacological interventions prescribed for patients with advanced anaplastic lymphoma kinase (ALK)- rearranged non-small cell lung cancer (NSCLC), comparative data between novel generation ALK-tyrosine kinase inhibitors (TKIs) remain scant. Here, we indirectly compared the efficacy and safety of first-line systemic therapeutic options used for the treatment of ALK-rearranged NSCLC. Methods We included all phase 2 and 3 randomised controlled trials (RCTs) comparing any two or three treatment options. Eligible studies reported at least one of the following outcomes: progression free survival (PFS), overall survival (OS), objective response rate (ORR), or adverse events of grade 3 or higher (Grade ≥ 3 AEs). Subgroup analysis was conducted according to central nervous system (CNS) metastases. Results A total of 9 RCTs consisting of 2484 patients with 8 treatment options were included in the systematic review. Our analysis showed that alectinib (300 mg and 600 mg), brigatinib, lorlatinib and ensartinib yielded the most favorable PFS. Whereas there was no significant OS or ORR difference among the ALK-TKIs. According to Bayesian ranking profiles, lorlatinib, alectinib 600 mg and alectinib 300 mg had the best PFS (63.7%), OS (35.9%) and ORR (37%), respectively. On the other hand, ceritinib showed the highest rate of severe adverse events (60%). Conclusion Our analysis indicated that alectinib and lorlatinib might be associated with the best therapeutic efficacy in first-line treatment for major population of advanced NSCLC patients with ALK-rearrangement. However, since there is little comparative evidence on the treatment options, there is need for relative trials to fully determine the best treatment options as well as the rapidly evolving treatment landscape. Supplementary Information The online version contains supplementary material available at 10.1186/s12885-021-08977-0.


Introduction
Lung cancer is the second most commonly diagnosed cancer and the leading cause of cancer death globally in 2020, with a 5-year survival rate of only 10 to 20% [1], and non-small cell lung cancer (NSCLC) accounts for about 85% of overall reported cases [2]. The anaplastic lymphoma-kinase (ALK)-rearrangements are detectable in approximately 2-7% of patients with NSCLC, especially those who are light/never-smokers and younger patients [3][4][5].
Currently multiple generation ALK-tyrosine kinase inhibitors (TKIs) have been developed (including crizotinib (first generation); alectinib, ceritinib, brigatinib, and ensartinib (second generation); and lorlatinib (third generation)), and most of these TKIs have been established as standard first line treatments [6,7]. However, despite an initial response to ALK-TKIs, acquired resistance inevitably develops and the way to overcome it is an open challenge [8]. Furthermore, patients harboring an ALK -rearrangements are particularly prone to central nervous system (CNS) metastasis, therefore, the ideal treatment could differ in patients stratified by CNS metastasis [9].
Several randomised controlled trials (RCTs) using only the direct comparison model have been conducted for conclusive evidence about the comparative efficacy and safety of first line treatments for patients with advanced ALK-rearranged NSCLC [10][11][12][13][14]. However, they have been unable to address the aforementioned problems, especially the second and third-generation TKIs. To provide additional evidence to guide treatment choices for ALK -rearranged NSCLC patients, we conducted a Bayesian network meta-analysis to comprehensively integrate all available direct and indirect evidence with a well-designed and comparative synthesis.

Methods
To conduct this network meta-analysis, we followed the preferred reporting items for systematic reviews and meta-analyses (PRISMA) extension statement for network meta-analysis (Additional files Table S1). The protocol of this study had been registered in the International Prospective Register of Systematic Reviews (PROSPERO), under the registration number of CRD42020173238 [15,16].

Data sources and searches
We searched PubMed, Embase, Cochrane Library, and ClinicalTrials.gov databases to find relevant articles up to 10 Sep 2021 in all languages. Then, to include complete and updated outcomes, abstracts of ongoing RCTs on NSCLC from several of the most important international conferences (American Society of Clinical Oncology, European Society of Medical Oncology, European Cancer Conference, and World Conference on Lung Cancer) from 2016 to 2021 were inspected. We used the following key words:"non-small-cell lung cancer", "NSCLC", "non-small-cell lung carcinoma", "treatment", "ALK","TKIs", "randomized controlled trial", and "clinical trial". The detailed search strategy is presented in Additional files Table S2.

Study selection
Two independent investigators (MHC and LYF) screened relevant records to identify all the relevant RCTs, and any disagreements were resolved via consensus. We included phase II/III RCTs involving adult patients with histologically or cytologically confirmed advanced (stage III/IV/recurrent) NSCLC with ALKrearrangement, comparing any two or more first line treatments. Studies that compared different doses of one ALK-TKI were also included.
As the current standard of first-line therapy for patients with advanced ALK -rearranged NSCLC is treatment with ALK-TKIs, the comparator of this metaanalysis is dominated by ALK-TKIs, and the comparison between chemotherapy were not included in our analyses. Unpublished data and case reports were also excluded from the analysis.

Data extraction
Two independent investigators (DKL and ZWJ) extracted relevant data parameters. Disagreements were resolved by discussion with the section partner. The following data parameters were extracted: study ID, first author, year of publication, region, RCT design, number of participants in each arm, total number of patients, patients in safety analysis, patient characteristics, interventions, and outcomes. Survival data assessed by an independent review facility were extracted to avoid potential assessment bias by the investigators.

Risk of bias assessment
The risk of bias of the included original studies was assessed using a modified version of the Cochrane Collaboration's Risk-of Bias Tool [17]. The following domains were assessed: (1) sequence generation; (2) allocation concealment; (3) blinding of participants, personnel, and outcome assessors; (4) incomplete outcome data; (5) selective outcome reporting; and (6) other potential threats to validity. Each aspect was assigned an assessment index associated with the risk of bias classified as yes, no, or unclear.

Clinical outcomes
The prespecified outcomes were progression free survival (PFS), overall survival (OS), objective response rate (ORR), and adverse events of grade 3 or higher (Grade ≥ 3 AEs). PFS and OS were analyzed as a survival outcome and reported as Hazard Rate (HR) with an associated 95%CrI. ORR and Grade ≥ 3 AEs were analyzed as a binary outcome and reported as odds ratio (OR) with an associated 95%CrI.

Statistical analyses
The network meta-analyses were performed with a Bayesian hierarchical random effects model using GeMTC (version 0.14.3) and R (version 3.5.3; Package gemtc) [18,19]. Model fit was assessed using deviance information criteria (DIC) and random effects standard deviation [20]. A three-chain or four-chain model with non-informative priors was run with an adaptation phase of 40,000 iterations followed by 200,000 model iterations with 20,000 burn-ins for each chain (the thinning interval was 10) [21,22]. Model convergence was estimated by the Brooks-Gelman-Rubin diagnostic and Potential Scale Reduction Factor (PSRF). In addition, the software can calculate the probability that each intervention is rated as the best (or worst), second, third, etc. Based on these data, we use GraphPad Prism (version 7.0) [23] to showed them in ranking plots.
Subgroup analysis was conducted according to CNS metastases. In order to measure the consistency of the effect size (OR and HR), we performed pairwise metaanalyses to assess heterogeneity between the studies using the Q test and I 2 statistic in Stata (version 14.0) [24]. Since the number of included studies is less than 10, we did not perform publication bias assessment. Sensitivity analyses were performed to assess the robustness and reliability of the results within each network metaanalysis.

Network meta-analysis for advanced ALK-rearranged NSCLC
The network meta-analysis included 9 studies for PFS, OS, ORR and Grade ≥ 3 AEs (Fig. 2).

Adverse events
For adverse events of grade 3 or higher, different toxicity spectrums were interrogated for individual ALK-TKIs. There was no significant difference between the various treatment options in Grade ≥ 3 AEs (Fig. 5). According to Bayesian ranking profiles, ceritinib was most likely to be associated with the highest (60%) cause of in Grade ≥ 3 AEs, followed by lorlatinib (18%). On the other hand, alectinib 300 mg had the highest probability (59%) of being the safest intervention, followed by alectinib 600 mg (22%) (Fig. 4).
In case of mild AEs (Grade1 and Grade2 AEs), no significant differences among the various treatment options. Based on Bayesian ranking profiles, alectinib 300 mg most likely leads to mild AEs (73%), followed by alectinib 600 mg (17%)(Additional files Fig. S2). This result is negatively associated with Grade ≥ 3 AEs.

Assessment of heterogeneity and inconsistency
Forest plots for four feasible pairwise comparisons showed low heterogeneity. However, there was high (75.0%) heterogeneity between alectinib 600 mg and crizotinib, for Grade ≥ 3 AEs (Additional files Fig. S6 and Fig. S7).

Sensitivity analysis
The sensitive analysis was performed by excluding the J-ALEX study, the results were stable and were similar to main analysis (Additional files Fig. S8 and Fig. S9).

Principal findings
In this systematic review and network meta-analysis, we comprehensively summarized the comparative effectiveness and safety of multiple first line treatment options for patients with advanced ALK-rearranged NSCLC. Our data showed that alectinib (300 mg and 600 mg) and lorlatinib had favorable effectiveness with tolerable adverse effects. On the other hand, toxicity profiles showed that ceritinib resulted in the highest rate of severe adverse events. The efficacy of crizotinib as the first-line treatment for patients with advanced ALK-rearranged NSCLC was proved in PROFILE 1014 [6,25], however, novelgeneration ALK-TKIs in the first-line setting have shown improved PFS versus crizotinib [10, 11, 13, 26-31, 33, 34]. This might be because of its poor brain penetrance and lower intracranial response [35]. Previous studies have shown that crizotinib, a P-glycoprotein (P-gp) substrate, is easily excreted from the blood-brain barrier (BBB) by active efflux with a cerebrospinal fluid concentration of only 0.26% of plasma concentration, compromising the attainment of effective drug concentrations at the CNS, moreover, it has a very low cerebrospinal fluid (CSF) penetration of 0·26% and CSF: IC50 of 0·03 [36,37]. The second generation ALK-TKIs are non-P-gp substrates, have lower binding to transport proteins which prevent the drug from being excreted from the brain. Therefore, while its BBB penetration rate is significantly increased, with a penetration rate of 63 to 94%, the probability of CNS progression is reduced [38,39]. Related basic experiments also demonstrated that the CSF penetration and CSF:IC50 value of the second generation ALK-TKIs were significantly higher than the first generation ALK-TKIs, with an improved drug specificity and efficacy [40][41][42].
There is a lack of head-to-head comparative evidence between the novel-generation ALK-TIKs. Consistent with our findings, previous studies have associated alectinib and lorlatinib with favorable effectiveness [10, 26, 29-31, 33, 34]. The median PFS of alectinib was more  Fig. 4 Bayesian ranking profiles of comparable treatments on efficacy for patients with advanced ALK-rearranged, non-small cell lung cancer. The profiles indicate the probability of each comparable treatment being ranked from first to last on progression free survival, overall survival, objective response rate, and grade ≥ 3 adverse events. Ale 300 mg, alectinib 300 mg; Ale 600 mg, alectinib 600 mg; Bri, brigatinib; Cer, ceritinib; Chemo, chemotherapy; Cri, crizotinib; Ens, ensartini; Lor, lorlatinib than 30 months in two large phase 3 RCTs (ALEX and J-ALEX) [26,30], while that of brigatinib and ceritinib was 24 months and 16.6 months respectively [12,27]. AS for OS, median OS was not reached with alectinib 600 mg versus 57.4 months with crizotinib (HR 0.67, 95%CrI 0.46-0.98) in the ALEX study [26]. Although OS data were immature, trends toward improvement will likely sustain and the detail reasons could be concluded as follows. First, the updated data from the ALEX study showed that the 5-year OS rate of alectinib 600 mg exceeded 62.5% while 45.5% for crizotinib, the difference is clinically meaningful [26]. Second, a series of first-line treatment studies on EGFR-TKIs have shown that longterm PFS benefits may be translated into OS, when the OS data is immature [43,44]. Therefore, PFS is an important indicator in predicting overall survival. Lastly, alectinib may increase overall survival because of the better control of the intracranial lesions. In contrast, the  final OS analysis of J-ALEX indicated that OS HR was 1.03 (95%CI 0.67-1.58) and 5-year OS rate in the alectinib 300 mg and crizotinib arm were 60.85 and 64.11%, respectively [31]. Alectinib 300 mg did not demonstrate a significant overall survival benefit versus crizotinib. The acceptance of cross-over in clinical trial might be causable for this difference between ALEX and J-ALEX. The data showed that 83.7% of the patients in the crizotinib arm received alectinib 300 mg as subsequent therapy when follow-up was 42.4 months, and 78.8% in the final OS analysis. On the other hand, ALEX study and J-ALEX study used different doses of alectinib, this may be one possible reason for this difference in OS, but this needs to be further confirmed.
In addition, alectinib has a broad-spectrum of safety medications. Our results reveal no obvious differences in terms of severe AEs rates and mild AEs rates in 600 mg and 300 mg. Similarly, Gainor et al. increased the dose of alectinib to 900 mg in a case report, and did not observe any drug-related adverse reactions, but mild constipation [45]. This increased drug dosages led to an increase in the intracranial drug concentration of alectinib, while the other second generation ALK-TIKs had limited increase of intracranial drug concentration due to adverse reaction. Thus, this explains why alectinib has may have yielded better control of CNS metastases, as a first-line treatment.
Similarly, lorlatinib provides improved control of CNS disease, 71% of the lorlatinib arm patients had an intracranial complete response in CROWN study [34]. Zou et al. also showed that lorlatinib induced superior regression of intracranial EML4-ALK tumors and prolonged survival in mice, compared with crizotinib. Moreover, the enhanced activity of lorlatinib on brain metastasis model was attributed to its ability to cross the BBB [46,47]. Besides, lorlatinib remain the antitumor activity for all known single ALK resistance mutations [48,49]. These factors might be the basis for the marked efficacy of lorlatinib as first-line therapy.
Regarding safety, our studies demonstrate that ceritinib 750 mg are expected to be at the highest risk of developing severe adverse effects. In the ASCEND-1 and ASCEND-4 study [12,50], administration of ceritinib at a dose of 750 mg led to 56-80% grade 1-2 gastrointestinal AEs, and about 5% grade 3 gastrointestinal AEs, which was consistent with our finding. However, we only incorporated ASCEND-4 study that was 750 mg, analysis of other doses was lacking. In the ASCEND-8 study, the researchers adjusted the ceritinib dosage to explored whether it can effectively reduce the gastrointestinal side effects [51]. They evaluated the efficacy and safety of ceretinib 450 mg/600 mg compared with 750 mg in ALK-positive advanced NSCLC patients. The results showed that compared with the 750-mg fasted dose, the frequency of dose reduction and withdrawal of ceritinib 450-mg was significantly lower, the median dose intensity was higher, and the frequency and severity of gastrointestinal adverse reactions are significantly reduced. These ASCEND-8 results were similar to the data from real-world studies (RWS) [52].

Strengths and limitations
We conducted a meta-analysis to compare the safety and adverse events of all ALK-TKIs approved by FDA. The results showed that alectinib was the safest secondgeneration TKI. Besides, SAE occurred in 40% of the patients treated with ceritinib and brigatinib [53]. Another Network meta-analysis on WCLC2019 indirectly compared the efficacy of the second-generation ALK inhibitors, and showed that ceritinib (450 mg taken with food) has similar effect as alectinib [54]. Compared with other previous meta-analyses investigating treatments for patients with advanced ALK-rearranged NSCLC, our network meta-analysis only included RCTs, reducing the impact of heterogeneity by study design.
However, most of the trials included in our analysis had immature OS data, without median OS. The limited data on the OS suggest that it might cause heterogeneity when taken as an endpoint to evaluate each personal treatment's realistic effect. Besides, most of the included studies are mainly clinical trials, posing potential publication and selection bias. Thirdly, only the data on PFS was included in the subgroup analysis, while the data on AEs, ORR, and OS was lacking. As a final note, patients were not stratified according to factors like race, which might modify treatment benefits, appropriate dose and efficacy of ALK-TKIs in the Asian population may differ from the Western population, Subsequent studies should investigate the relative treatment efficacy according to these clinical characteristics.

Implications
By synthesizing all the evidence in the RCTs, this review provides clinicians a reference source to evaluate strengths and weaknesses associated with all the promising therapeutic options. Due to its promising effectiveness and safety, alectinib and lorlatinib were recommended as the first-line treatment option for advanced NSCLC patients with ALK-rearrangement. Therefore, these findings can help answer questions on whether there were differences in the efficacy and safety of different doses of alectinib compared with the other ALK-TKIs, in controlling CNS metastasis. Besides, the review shows whether ensartinib and brigatinib can challenge the use of alectinib as the first-line treatment.
We, however, recommend that there should be more head-to-head RCTs between the second-generation and third-generation ALK-TKIs. After ALK-TKI drug resistance, it is necessary to interrogate how to carry out reasonable sequential treatment, optimize the whole medication arrangement and achieve the longest survival time. Moreover, it is also necessary to continue exploring the mechanism of resistance to provide guidance for clinical applications in basic research.

Conclusions
These results indicated that alectinib and lorlatinib might be associated with the best therapeutic efficacy in first-line treatment for major population of advanced NSCLC patients with ALK-rearrangement.