Impact of ALK variants on brain metastasis and treatment response in advanced NSCLC patients with oncogenic ALK fusion.

BACKGROUND: To investigate the impact of ALK variants on the features of brain metastases (BM), the outcome of chemotherapy and targeted therapy using crizotinib, as well as the progression pattern in patients with ALK fusion.
METHODS: Patients with ALK fusion were retrospectively collected from January 2013 to July 2017 in Shanghai Pulmonary Hospital. ALK rearrangements were identified via ARMS-PCR. ALK variants were identified via Sanger Sequencing.
RESULTS: A total of 135 patients and 41 with brain metastasis were identified. Radiological features showed that the patients with ALK variant 1 had a larger BM size compared with patients with ALK non-variant 1 (median tumor size: 16.89 vs. 11.01 mm, P=0.031). Similar time to treatment failure (TTF) was observed in patients with ALK variant 1 and non-variant 1 who received first-line crizotinib (median TTF: 15.7 vs. 13.8 months, HR =0.75, P=0.34). Patients with ALK variant 1 who had baseline BM had significantly shorter TTF than non-variant 1 with baseline BM when treated with first-line crizotinib (median TTF: 9.1 vs. 14.9 months, HR =2.68, P=0.037). In patients treated with chemotherapy, ALK variant 1 was associated with inferior TTF (median TTF: 5.6 vs. 8.1 months, HR =1.66, P=0.039). Progression pattern was similar between ALK variant 1 and non-variant 1.
CONCLUSIONS: Patients with ALK variant 1 and baseline BM had inferior TTF on first-line crizotinib treatment and presented with more aggressive radiological features. Patients with ALK non-variant 1 had better clinical outcome on first-line chemotherapy. 2020 Translational Lung Cancer Research. All rights reserved.

Introduction

Anaplastic lymphoma kinase (ALK) gene rearrangements were first discovered in 2007 and account for 2–7% of non-small cell lung cancer (NSCLC) (1,2). Thereafter, several studies have established the role of crizotinib as standard of care in later line, 2nd line and first line setting of advanced NSCLC patients with ALK fusion (3-7). However, nearly all patients receiving crizotinib inevitably progress, with the majority of them presenting with disease progression in the central nervous system (CNS) (8). Several 2nd or 3rd generation ALK inhibitors such as ceritinib (9), alectinib (10,11), brigatinib (12) and lorlatinib (13) with better CNS penetration, were developed and have demonstrated superior efficacy over crizotinib for both extracranial and intracranial disease (14,15). Besides ALK inhibitors, pemetrexed-based chemotherapy is potentially more efficacious than other chemotherapy against lung cancer with ALK fusion (16). However, response to these treatments is heterogeneous and precisive strategies are needed to further improve the outcome of patients with ALK fusion.

Echinoderm microtubule-associated protein-like 4 (EML4) is the most common fusion partner to ALK. To date, 15 EML4-ALK variants have been identified. The most common are variant 1 [v1; exon 13 of EML4 fused to exon 20 of ALK (E13;A20)] and variant 3a/b [exon 6a/b of EML4 fused to exon 20 of ALK (E6a/b;A20)] (17,18). More and more evidences suggest that ALK fusion variants may have biologic and clinical implications in ALK-positive lung cancer (19-26). For example, Yoshida et al. found crizotinib had better efficacy in patients with EML4-ALK variant 1 than those with non-variant 1 (21). Furthermore, Lin et al. revealed that variant 3a/b were associated with G1202R, an ALK mutation that confers resistance to crizotinib and all 2nd generation ALK TKIs (e.g., ceritinib, alectinib and brigatinib), which suggested the link between ALK fusion variant and clinical outcome (25). Thus, analyzing ALK variants might help provide new insight in the era of precision medicine.

ALK gene rearranged lung cancer have a high incidence of BM and exploring the characteristics including ALK variants in ALK+ patients with baseline BM might help guide therapeutic strategy (8,16,27). Here, in order to have a better understanding of clinical impact of ALK variants on BM, we comprehensively investigated the association of ALK variants with the feature of brain metastasis, as well as the efficacy of crizotinib and chemotherapy in 135 Chinese advanced NSCLC patients with ALK fusions.

We present the following article in accordance with the STROBE Reporting Checklist (available at http://dx.doi.org/10.21037/tlcr-19-346).

Methods

The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013) and was approved by the Ethics Committee of Shanghai Pulmonary Hospital, Tongji University School of Medicine (No. K18-089-1). Because of the retrospective nature of the research, the requirement for informed consent was waived.

Study population and study design

As shown in , we enrolled 235 patients with positive ALK-rearrangement from January 2013 to July 2017 in Shanghai Pulmonary Hospital. In total, 41 patients had confirmed BM at baseline. Among them, 8 patients had brain MRI data but were lack of evaluation of treatment response. This was due to either to be enrolled in double-blind clinical trials, or no further anti-cancer therapy, or lost follow-up. Therefore, 33 patients with baseline BM were included as a subgroup to compare time to treatment failure (TTF) and objective response rate (ORR) between variant 1 vs. non-variant 1 in both crizotinib-treated and chemo-treated settings. In addition, 6 out of the 41 patients had their BM confirmed in other hospitals but not ours. Therefore, only the remaining 35 patients with both the baseline BM and available MRI data were included as an exploratory subgroup to compare (I) radiological features; (II) neurological symptoms; (III) the graded prognostic assessment (GPA) score between variant 1 vs. non-variant 1.

Figure 1

Summary of study design and measurement of PTBE. (A) Flow chart of the study design; (B,C) PTBE is the difference in maximum diameter of tumor in T2-weighted MRI image (right: bidirectional arrow labeled “b”) and maximum diameter of tumor in T1-weighted MRI image (left: bidirectional arrow labeled “a”).

Among 235 patients, 100 patients were excluded owing to either ineligible variant detection results (PCR product was not abundant to perform Sanger Sequencing), missing information on first-line lung cancer treatment (the patient made treatment decision at a peripheral hospital), or lack of survival data. Finally, 135 patients with either crizotinib or chemotherapy as their first-line treatment were included to compare (I) TTF and ORR; (II) progression pattern between variant 1 vs. non-variant 1 in both crizotinib-treated and chemo-treated groups.

Data reporting and acquisition

TTF was defined as the time from the start of crizotinib treatment or chemotherapy to the date of treatment discontinuation for any reason (including switching treatment strategy, death, whichever occurred first). ORR was calculated as percentage of patients with evaluated complete response (CR) or partial response (PR).

To investigate the progression pattern, patients enrolled in this study routinely got chest CT scan every 2 months, brain MRI every 6 months, bone scan every 6 months and abdominal ultrasound every 6 months. Any radiological examination was applied whenever related symptoms occurred.

Radiological features of BM

The brain images obtained by MRI were collected. All scans were performed on the same 1.5 Tesla MRI machine (United Imaging Health-care, Shanghai). Three variables were measured: (I) number of BMs; (II) size of BM; (III) sum of BM size; (IV) size of peritumoral brain edema (PTBE). For multiple brain tumors, the BM size was defined using the longest diameter of the largest lesion in one dimension on T1-weighed images. PTBE was defined as the subtraction of diameter measured on T1-weighed imaging from that measured on corresponding T2-weighed images (). If no edema was observed, the size of PTBE was defined as zero (). The size of the tumor and PTBE were measured by two experienced oncologists and one radiologist, and averaged eventually.

Figure S1

Tumor located in right lateral ventricle on T1-weighted image (A) compared to T2-weighted image demonstrates absence of PTBE (B).

Graded prognostic assessment (GPA)

The GPA is a validated assessment scale for lung cancer patients with BM (28). Four prognostic factors comprise the GPA: Karnofsky performance score (KPS) (0: KPS <70, 0.5: KPS =70–90, 1.0: KPS >90), age (0: age >60, 0.5: age =50–60, 1: age <50), presence of extracranial metastasis (ECM) (0: ECM yes or 1: ECM no), and the number of BM (0: >3 BM, 0.5: 2–3 BM,1: 1 BM). The GPA score showed good prognostic discrimination, and depending on the GPA score, lung cancer patients with BM in this study had a median overall survival (OS) ranging from 3.02 to 14.78 months (29).

Detection of ALK rearrangements and variants

The Amplification Refractory Mutation System (ARMS) was adopted to detect ALK fusion using the Human AmoyDx EML4-ALK Fusion Gene Diagnostic Kit (Amoy Diagnostics Co, Ltd). ALK variants were detected via Sanger sequencing of PCR product. The detailed procedures were performed as previously described (30-35).

Statistical analysis

All data were analyzed by using the Statistical Package for Social Science (SPSS) software (version 23.0 for Mac) and GraphPad Prism software (Version 6 for Mac). Differences in baseline clinical characteristics between groups (variant 1 vs. non-variant 1) were analyzed using Chi-square test or Fisher’s exact test. The Kaplan-Meier method was used to analyze the survival probability and log-rank test was used to calculate the significance of differences. Cox proportional hazard model was applied for the univariate and multivariate analyses to calculate the hazard rations (HR) and 95% confidence intervals (95% CI). P values in this article were two-sided and considered statistically significant when less than 0.05.

Results

Distribution of ALK variants and the association with clinical characteristics

In total, 135 patients with known ALK variant and medical record were analyzed. As shown in , the most frequent was ALK variant 1 (E13;A20) in 61 patients (45.2%), followed by variant 3a/b (E6;A20/E6ins33;A20) in 51 patients (37.8%) and variant 2 (E20;A20) in 17 patients (12.6%). Other variants included 3 patients with variant V5(5’) (E18;A20), 2 with variant 5a (E2;A20) and one with variant 8a/b (E17;A20) (Baseline characteristics of patients listed in ). In the overall study cohort, median age was 53.4 (28 to 80) years old and 48.9% (66/135) were male, 73.3% (99/135) were never-smokers, 91.9% (124/135) had lung adenocarcinoma, and 8.1% (11/135) had recurrence after initial diagnosis of early stage disease. As first-line therapy, 51.9% (70/135) received chemotherapy and 48.1% (65/135) received crizotinib. Comparing the variant 1 and non-variant 1 cohorts, there were no significant differences observed with regard to age (P=0.500), gender (P=0.148), ECOG PS (P=0.183), histology (P=0.110), and smoking status (P=0.498). In total, 33/135(24.4%) patients had baseline BM before treatment. In the overall study cohort, 15/33 (45.4%) patients with BM, 4/14 (28.6%) patients with bone metastasis, and 6/12 (50%) patients with liver metastasis were variant 1. In our study cohort, there was no statistical significance in baseline incidence of metastasis between variant 1 and non-variant 1 (BM: P=0.971, bone metastasis: P=0.187, liver metastasis: P=0.726).

Figure 2

Overview of ALK variants and characteristics of radiological features in ALK positive patients with BM. (A) The distribution of ALK fusion variants; (B) the radiological features (BM size, PTBE) of brain tumor size in group of ALK-variant 1 and non-variant 1. Each blue bar represents an individual patients with variant 1 and red bar represents an individual with non-variant 1. The length of the bar represents either BM size (left extended) or PTBE size (right extended). The followed table summarized the characteristics related to BM in variant 1 and non-variant 1.

Table 1

Clinical characteristics of included patients

Characteristic	Total (N=135) (%)	Variant 1 (N=61) (%)	Non-variant 1 (N=74) (%)	P value
Median age, years (range)	53.4 (28–80)	52.7 (29–80)	53.9 (28–80)	0.500
Gender				0.148
Male	66 (48.9)	34 (55.7)	32 (43.2)
Female	69 (51.1)	27 (44.3)	42 (56.8)
ECOG PS				0.183
0–1	125 (92.6)	59 (96.7)	66 (89.2)
2–3	10 (7.4)	2 (3.3)	8 (10.8)
Histology				0.110
Adenocarcinoma	124 (91.9)	53 (86.9)	71 (95.9)
Non-adenocarcinoma	11 (8.1)	8 (13.2)	3 (4.1)
Smoking status				0.498
Current/former smokers	36 (26.7)	18 (29.5)	18 (24.3)
Never-smokers	99 (73.3)	43 (70.5)	56 (75.7)
Stage				1.000
Unresected IIIB–IV	124 (91.9)	56 (91.8)	68 (91.9)
Postoperative recurrent	11 (8.1)	5 (8.2)	6 (8.1)
First-line treatment strategy				0.827
Chemotherapy	70 (51.9)	31 (50.8)	39 (52.7)
Crizotinib	65 (48.1)	30 (49.2)	35 (47.3)
Crizotinib treatment line				0.899
First	65	30	35
Second	13	8	5
≥Third	11	4	7
Baseline brain metastasis				0.971
Yes	33 (24.4)	15 (24.6)	18 (24.3)
No	102 (75.6)	46 (75.4)	56 (75.7)
Baseline bone metastasis				0.187
Yes	14 (10.3)	4 (6.6)	10 (13.5)
No	120 (88.9)	57 (93.4)	64 (86.5)
Baseline liver metastasis				0.726
Yes	12 (8.9)	6 (9.8)	6 (8.1)
No	123 (91.1)	55 (90.2)	68 (91.9)

Radiological features associated with ALK variants in patients with baseline BM

In patients with BM, there was no statistical significance in clinical characteristics between variant 1 and non-variant 1 (). We further examined the radiological features according to ALK variants (proportion of patients with ≤3 BM, symptomatic BM at baseline, BM size, sum of BM size and PTBE) in patients with baseline BM (N=35) (). There was no significant difference in the proportion of patients with three BMs or less [11 (78.6%) patients in variant 1 subgroup and 16 (76.2%) patients in non-variant 1 subgroup]. Overall, 51.4% (18/35) ALK+ patients with baseline BM suffered from neurological symptoms, defined as headache, dizzy, nausea, vomiting, and etc. Of those cases, 64.3% (9/14) of patients with ALK variant 1 had neurological symptoms related to BMs compared to 42.8% (9/21) in the non-variant 1 group (P=0.21). In addition, the patients with ALK variant 1 had larger BM size (median BM size: 16.89 vs. 11.01 mm, P=0.031) and numerically higher CNS burden (30.32 vs. 21.98 mm, P=0.272) together with wider PTBE (17.94 vs. 12.80 mm, P=0.447) compared to patients with ALK non-variant 1. No significant difference was observed in constituent ratio of GPA score between two groups (P=0.934).

Table S1

Clinical characteristics of included patients with BM

Characteristic	Total (N=33) (%)	Variant 1 (N=15) (%)	Non-variant 1 (N=18) (%)	P value
Median age, years (range)	53.2 (32–73)	54.5 (32–73)	52.2 (35–69)	0.520
Gender				0.126
Male	15 (45.5)	9 (60.0)	6 (33.3)
Female	18 (54.5)	6 (40.0)	12 (66.7)
Histology				0.868
Adenocarcinoma	30 (90.9)	13 (86.7)	17 (94.4)
Non-adenocarcinoma	3 (9.1)	2 (13.3)	1 (5.6)
Smoking status				0.748
Current/former smokers	9 (27.3)	5 (33.3)	4 (22.2)
Never-smokers	24 (72.7)	10 (66.7)
Stage				1.000
Unresected IIIB-IV	31 (93.9)	14 (93.3)	17 (94.4)
Postoperative recurrent	2 (6.1)	1 (6.7)	1 (5.6)
First-line treatment strategy				0.126
Chemotherapy	15 (45.5)	9 (60.0)	6 (33.3)
Crizotinib	18 (54.5)	6 (40.0)	12 (66.7)

Impact of ALK-variants on efficacy of crizotinib

Patients who received crizotinib as first-line treatment strategy had longer TTF than those who had chemotherapy (median TTF: 13.8 vs. 6.3 months, HR =0.43, P<0.01, ). There was no significant TTF difference between variant 1 and non-variant 1 in the subgroup treated with first-line crizotinib (). Both ORR and the DCR were not significantly different between variant 1 and non-variant 1 in the crizotinib-treated group (). Univariate and multivariate analysis showed that only ECOG PS and patient sex had a significant impact on TTF. None of the other variables including ALK variant were associated with TTF ().

Figure 3

Time to treatment failure (TTF) in (A) overall population; (B) patients treated with crizotinib as the first-line therapy; (C) patients with baseline brain metastasis; (D) patients with baseline brain metastasis and treated with crizotinib as first-line therapy; (E) patients without baseline brain metastasis; (F) patients without baseline brain metastasis and treated with crizotinib as first-line therapy.

Table S2

The objective response (ORR) and disease control rate (DCR) in group of variant 1 and non-variant 1 treated with crizotinib

RECIST	Total				First-line				Non-first line
	Variant 1 (n=42)	Non-variant 1 (n=47)	P value		Variant 1 (n=30)	Non-variant 1 (n=35)	P value		Variant 1 (n=12)	Non-variant 1 (n=12)	P value
PR	28	28			21	23			7	5
SD	12	13			7	8			5	5
PD	2	6			2	4			0	2
ORR (%)	66.7	59.6	0.489		70	65.7	0.713		58.3	41.7	0.414
DCR (%)	95.2	87.2	0.344		93.3	88.6	0.817		100	83.3	0.478

Table 2

Univariate and multivariate analyses of clinical parameters on TTF in patients with ALK rearrangement-positive NSCLC treated with crizotinib

Factor	Univariate analysis				Multivariate analysis
	HR (log rank)	95% CI	P value		HR (log rank)	95% CI	P value
Sex (male/female)	2.279	1.197–4.336	0.012		3.455	1.302–9.166	0.013
Age (≥65/<65)	1.474	0.648–3.354	0.355		2.119	0.800–5.612	0.131
Smoking (never/ever)	0.635	0.338–1.193	0.158		1.703	0.649–4.470	0.280
Stage (IIIB–IV/postoperative recurrent)	0.556	0.196–1.579	0.271		0.716	0.191–2.688	0.620
ECOG PS (2–3/0–1)	3.416	1.177–9.916	0.024		3.733	1.108–12.581	0.034
Histology type (adenocarcinoma/non-adenocarcinoma	0.712	0.215–2.023	0.524		0.456	0.141–1.478	0.191
Brain metastasis (yes/no)	1.570	0.826–2.982	0.168		1.939	0.940–3.997	0.073
Bone metastasis (yes/no)	0.917	0.216–3.892	0.906		1.545	0.339–7.041	0.574
Liver metastasis (yes/no)	1.65	0.498–5.465	0.412		2.225	0.588–8.421	0.239
Variant (non-variant 1 vs. variant 1)	1.349	0.722–2.521	0.348		1.452	0.682–3.092	0.334

Patients with baseline BM had a median TTF of 12.7 months treated with crizotinib, compared to 7.1 months in the chemotherapy-treated group (P=0.017, ). Patients with variant 1 and baseline BM had a significantly shorter TTF than non-variant 1 on first-line crizotinib (HR =2.68, P=0.037, ). Regardless of first-line treatment strategy, patients with variant 1 and baseline BM had significantly lower ORR than patients with non-variant 1 (20% vs. 66.7%, P=0.007) and numerically lower in crizotinib-treated group (50.0% vs. 75.0%, P=0.596, ). In patients without baseline metastases, patients treated with crizotinib had median TTF of 15.7 months and no statistically significant difference in TTF was observed between variant 1 and non-variant 1 (median TTF: 17.0 vs. 12.8 months, HR 0.61, P=0.19) ().

Figure S2

The objective response rate (ORR) in ALK+ patients with baseline brain metastasis.

Impact of ALK-variants on efficacy of chemotherapy

A total of 70 patients received chemotherapy as their first-line treatment strategy. Patients with ALK variant 1 had a significantly inferior TTF compared to those with ALK non-variant 1 (median TTF: 5.6 vs. 8.1 months, HR =1.66, P=0.039) (). In the subgroup receiving pemetrexed-based chemotherapy, patients with variant 1 had a TTF of 6.3 months, whereas patients with non-variant 1 had a TTF of 9.2 months (P=0.094) (). Otherwise, ORR and DCR were similar between variant 1 and non-variant 1 cohorts receiving first-line chemotherapy (). In the ALK variant 1 cohort, patients receiving 1st line chemotherapy (N=31) had numerically inferior TTF compared to non-variant 1, regardless of whether they had BM at baseline, but the difference was not statistically significant (P=0.12 and P=0.15, respectively, ). Other than ECOG PS, only non-variant 1 status was significantly associated with a longer TTF in both univariate analysis (vs. variant 1, HR =0.579, P=0.043) and multivariate analysis (vs. variant 1, HR =0.470, P=0.009) ().

Figure 4

Time to treatment failure (TTF) outcomes in ALK variant 1 versus non-variant 1 in patients treated with (A) any first-line chemotherapy; (B) first-line pemetrexed-based chemotherapy; (C) patients with baseline brain metastasis and treated with first-line chemotherapy; (D) patients without baseline brain metastasis and treated with first-line chemotherapy.

Table S3

ORR and DCR in group of variant 1 and non-variant 1 treated with chemotherapy

RECIST	Variant 1 (n=31)	Non-variant 1 (n=39)	P value
PR	6	14
SD	15	18
PD	10	7
ORR (%)	19.4	35.9	0.128
DCR (%)	67.7	82.1	0.165

Table 3

Univariate and multivariate analyses of clinical parameters on TTF in patients with ALK rearrangement-positive NSCLC treated with chemotherapy

Factor	Univariate analysis				Multivariate analysis
	HR (log rank)	95% CI	P value		HR (log rank)	95% CI	P value
Sex (male/female)	0.699	0.417–1.170	0.173		0.579	0.303–1.108	0.099
Age (≥65/<65)	0.821	0.371–1.817	0.627		0.753	0.314–1.810	0.527
Smoking (never/ever)	1.031	0.541–1.967	0.925		1.069	0.465–2.458	0.875
Stage (IIIB-IV/postoperative recurrent)	1.931	0.863–4.319	0.109		1.535	0.662–3.561	0.318
ECOG PS (2–3/0–1)	2.728	1.128–6.602	0.026		2.748	1.008–7.492	0.048
Histology type (adenocarcinoma/non-ade	0.524	0.159–1.719	0.286		0.610	0.138–2.695	0.515
Brain metastasis (yes/no)	0.969	0.523–1.795	0.920		0.807	0.410–1.588	0.535
Bone metastasis (yes/no)	1.582	0.793–3.158	0.193		1.992	0.898–4.420	0.09
Liver metastasis (yes/no)	1.582	0.793–3.158	0.193		1.413	0.405–4.925	0.587
Variant (non-variant 1 vs. variant 1)	0.579	0.342–0.982	0.043		0.470	0.268–0.825	0.009

Progression pattern according to variants

Patterns of disease progression were separated into five categories: CNS progression, liver progression, bone progression, intrathoracic progression and other extra-thoracic progression. Other extra-thoracic progression was defined as metastases occurring in the adrenal glands, spleen or extra-pulmonary lymph nodes. Overall, no significant difference in progression pattern was observed in variant 1 and non-variant 1 (Figure S3A). At the time of analysis, 55 patients who were once treated with crizotinib and 61 patients treated with chemotherapy exhibited progression of disease (PD), including 56 patients with ALK variant 1 (28 once received crizotinib; 28 received chemotherapy) and 60 patients with ALK non-variant 1 (first-line treatment: 27 received crizotinib; 33 received chemotherapy). However, there were no significant differences in the patients with intra-thoracic and extra-thoracic recurrence between the variant 1 and non–variant 1 group (, Figure S3B).

Table S4

Progression pattern between variant 1 and non-variant 1 in face of different treatment strategies

Progression pattern	Crizotinib (n=55)			Chemotherapy (n=61)
	Variant 1 (n=28)	Non-variant 1 (n=27)		Variant 1 (n=28)	Non-variant 1 (n=33)
CNS relapse	13 (46.5)	9 (33.3)		2 (7.1)	8 (24.2)
Bone progression	3 (10.7)	0 (0)		3 (10.7)	8 (24.2)
Liver progression	1 (3.6)	0 (0)		1 (3.6)	2 (6.1)
Thoracic progression	15 (53.6)	16 (59.3)		21 (75.0)	24 (72.7)
Other extrathoracic progression	1 (3.6)	4 (14.8)		0 (0)	2 (6.3)

Discussion

This real-world study collected the largest cohort of ALK positive patients and comprehensively investigated the association of ALK variants with brain metastases together with clinical outcomes on first-line crizotinib and chemotherapy. We observed more aggressive radiological features (larger BM size and numerically larger area of PTBE) in those with ALK variant 1 than non-variant 1. We found a numerically longer but not statistically significant difference of TTF in patients with ALK variant 1 and non-variant 1 in crizotinib-treated group (median TTF: 15.7 vs. 13.8 months, P=0.75), while subgroup analysis showed that patients with ALK-variant 1 and baseline BM had significantly shorter TTF (P=0.037). Additionally, patients with ALK non-variant 1 had longer TTF in the chemotherapy-treated group than patients with variant 1 (median TTF: 8.1 vs. 5.4 months, P=0.039; multivariate analysis: P=0.009).

As far as we know, this is the first report on the association of ALK variants with the radiologic features of BM. We found that the size of baseline metastasis in variant 1 group was larger than in non-variant 1 group (median TS: 16.89 vs. 11.01 mm, P=0.031) with numerically higher CNS burden and wider edema range. Additionally, patients with ALK variant 1 and baseline BM had a higher frequency (64.3%) of neurologic symptoms related to BM than patients with ALK-non variant 1 (42.8%). Previous studies demonstrated that the size of BM was positively associated with the thickness of PTBE and both factors were associated with worse survival outcome (36,37). Although statistical significance regarding the size of PTBE was not reached owing to the limited sample size, it is suggested that patients with ALK-variant 1 and baseline BM likely have more aggressive BM radiographically and might be related to the different efficacy of crizotinib in this study. Our subgroup analysis showed that patients with baseline BM and ALK-variant 1 had a significantly worse TTF than those with non-variant 1 (9.1 vs. 14.9 months, P=0.037) after the treatment of crizotinib.

The frequency of BM in treatment-naïve ALK+ NSCLC ranges from 20–40%. The clinical outcome of ALK+ patients with baseline BM receiving crizotinib varied across different studies (PFS range, 7.0–21.2 months) (6,7,38-40). It is conceivable that different ALK variant might explain the heterogeneity of treatment response in ALK patients with baseline BM. Although previous investigations including our current study indicated that variant 1 might have better crizotinib response than non-variant 1 but lack of significance (22,23,25), Yoshida, et al. did report a significantly longer PFS to crizotinib in patients with variant 1 (n=19) versus non-variant 1 (n=16) (median PFS: 11.0 vs. 4.2 months, P<0.05) (21). The inconsistency might be due to lower percentage of patients with baseline BM in Yoshida’s study (20%) comparing to other studies including ours (ranging from 26% to 44.4%) (22,23,25).

The most frequent ALK variants were variant 1 (45.2%) and variant 3a/b (35.8%), which was consistent with the findings from previous publications (). In consistent with previous studies (20-25), similar baseline clinical characteristics between different ALK variants were found in our study. Previous studies including ours found that the level of thymidylate synthetase (TS) was lower in ALK-positive tumors (41) and correlated with enhanced sensitivity to pemetrexed than ALK-wild type tumors (16,42). This present study found that ALK-non variant 1 had statistically significant longer TTF than ALK variant 1 in the chemotherapy-treated group (median TTF: 8.1 vs. 5.6 months, P=0.039) and possibly longer TTF than ALK variant 1 in the pemetrexed-treated group (median TTF: 9.2 vs. 6.3 months, P=0.094), although statistical significance for the latter comparison was not shown. In consistent with worse outcomes seen in ALK variant 1 patients treated with pemetrexed-based chemotherapy, an in-vitro study showed that the IC50 to pemetrexed was higher in ALK variant 1 cell line, H3122 (68.7±21.1 nM) than in an ALK variant 3a/b cell line, H2228 (33.0±10.7 nM) (42). Therefore, we hypothesized that in ALK-positive group with low TS expression, the delicate difference in level of TS might not be the critical factor that determined the efficacy to pemetrexed when we stratified the ALK-positive patients into different ALK variants.

Table S5

Previous studies about the impact of ALK variants on treatment response

Author	Year	Study design	Methods for genotyping ALK variants	Proportion of ALK variants	Results	Reference
Heuckmann	2012	In vitro experiments conducted to analyze the cytotoxic efficacy of crizotinib on different ALK variants by making Ba/F3 cell-line model	–	–	EML4-ALK variant 2 was the most sensitive variant to the crizotinib in vitro	(19)
Lei	2015	A retrospective study conducted on 61 patients treated with crizotinib	rapid amplification of cDNA ends (RACE)-coupled PCR		No significant difference existed among three groups	(20)
Yoshida	2016	A retrospective study conducted on 35 patients treated with crizotinib	RT-PCR		Varaint 1 had better PFS than non-variant 1 (median PFS: 11 vs. 4.2 months, P<0.05)	(21)
Cha	2016	A retrospective study conducted on 52 patients who have once treated with ALK inhibitors or chemotherapy	PNA-mediated qPCR assay		Variant 1 had superior PFS than other variants in patients once treated with pemetrexed	(22)
					No difference observed according to ALK variants treated with ALK inhibitors
Woo	2016	A retrospective study conducted on 54 patients once treated with ALK inhibitors together with experients in vitro by making Ba/F3 and Beas-2B cell lines	PNA-mediated qPCR assay		2-year PFS had no difference among variant 1/2/others or variant 3 treated with crizotinib (P=0.108)	(23)
					Variant 3a/variant 5a were resistant to ALK inhibitors in vitro
Li	2017	A retrospective study conducted on 35 patients treated with crizotinib	targeted NGS		Variant 2 had prolonged PFS (P=0.021)	(24)
					Variant 3a/b and non-variant 3a/b had no PFS difference
Lin	2018	A retrospective study conducted on 129 patients treated ALK inhibitors including 55 patients received crizotinib as first-line therapy	targeted NGS		Similar PFS was observed while comparing patients with variant 1 or varaint 3 treated with crizotinib (P=0.163)	(25)
Mitiushkina	2018	A retrospective study conducted on 64 patients treated ALK inhibitors	RT-PCR		Similar PFS was observed while comparing variant 1 and non-variant 1 (P=0.604)	(26)

This study has several limitations that need to be acknowledged. Firstly, although we analyzed the largest cohorts of patients harboring ALK aberrations, the sample size in the cohort of patients with BM was relatively small. Only 33 patients were enrolled for the analysis of survival outcome and 35 patients included for the analysis of radiological features. Secondly, it is a single-institution study. Thirdly, the impact of second and third generation ALK-TKIs with improved CNS penetration on clinical outcomes in different ALK variants were not examined in this study. Additionally, given the nature of retrospective study, many confounding factors might have impact on drawing a convincing conclusion. Finally, due to a limited sample size of some of the less common variant subtypes, we had to analyze non-variant 1 as a conglomerate cohort. Therefore, the results need to be interpreted with caution.

Conclusions

In summary, our findings provide a possible explanation for the discrepancy between studies on crizotinib efficacy among ALK variants—that ALK variant 1 with BM had inferior TTF than non-variant 1 with BM due to more aggressive and frequent BMs. Moreover, ALK non-variant 1 had longer TTF than variant 1 in the chemotherapy-treated group, which further strengthen the need to explore additional treatment strategies based on ALK variant in patients with ALK fusion.

Progression pattern in overall population stratified by (A) ALK variant, (B) treatment strategy (crizotinib versus chemotherapy).

The article’s supplementary files as