- Email: [email protected]
ROS1 Gene Rearrangements Are Associated With an Elevated Risk of Peridiagnosis Thromboembolic Events
ORIGINAL ARTICLE
ROS1 Gene Rearrangements Are Associated With an Elevated Risk of Peridiagnosis Thromboembolic Events Terry L. Ng, MD,a,* Derek E. Smith, MS,b Rao Mushtaq, MD,a Tejas Patil, MD,a Anastasios Dimou, MD,a Shuo Yang, MD,c Qian Liu, MD,c Xuefei Li, PhD,c Caicun Zhou, MD, PhD,c Robert T. Jones, BS,d Megan M. Tu, PhD,e Flora Yan, BS,f I. Alex Bowman, MD,f Stephen V. Liu, MD,g Siera Newkirk, BS,h Joshua Bauml, MD,h Robert C. Doebele, MD, PhD,a Dara L. Aisner, MD, PhD,i Dexiang Gao, PhD,b Shengxiang Ren, MD, PhD,c D. Ross Camidge, MD, PhDa a
Division of Medical Oncology, University of Colorado, Aurora, Colorado Department of Pediatrics, Cancer Center Biostatistics Core, University of Colorado and Children’s Hospital Colorado, Aurora, Colorado c Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai, China d Department of Pharmacology, University of Colorado, Aurora, Colorado e Department of Surgery, University of Colorado, Denver, and University of Colorado Comprehensive Cancer Center, Aurora, Colorado f University of Texas at Southwestern, Dallas, Texas g Georgetown University, Washington, DC h Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania i University of Colorado, Department of Pathology, Aurora, Colorado b
Received 7 October 2018; revised 2 December 2018; accepted 2 December 2018 Available online - 10 December 2018
ABSTRACT Introduction: This study aims to determine whether advanced ROS1 gene-rearranged NSCLC (ROS1þ NSCLC) has a higher than expected thromboembolic event (TEE) rate. Methods: Venous and arterial TEEs within ±365 days of diagnosis of ROS1þ, ALKþ, EGFRþ, or KRASþ advanced NSCLC at five academic centers in the United States and China were captured (October 2002–April 2018). The primary endpoint was incidence of TEE in ROS1þ compared to anaplastic lymphoma kinase (ALK)þ, EGFRþ, and KRASþ NSCLC within ±90 days of diagnosis. Logistic regression was used to assess if the odds of TEE differed among oncogene drivers. Results: Eligible data from 95 ROS1þ, 193 ALKþ, 300 EGFRþ, and 152 KRASþ NSCLC patients were analyzed. The incidence rate of TEE was 34.7% (n ¼ 33), 22.3% (n ¼ 43), 13.7% (n ¼ 41), and 18.4% (n ¼ 28), respectively. In univariate analysis, the odds of a TEE in ROS1þ NSCLC were higher than ALKþ, EGFRþ, and KRASþ cohorts. In multivariable analysis, the odds of a TEE were significantly higher for ROS1þ compared to EGFRþ and KRASþ cohorts, the odds ratio (OR) was 2.44, with a 95% confidence interval of 1.31–4.57 (p ¼ 0.005), and OR: 2.62, with a 95% confidence interval of 1.26–5.46 (p ¼ 0.01), respectively.
Although numerically superior, the odds for a TEE with ROS1þ compared to ALKþ was not statistically significant (OR: 1.45, p ¼ 0.229). Overall survival was not significantly *Corresponding author. Disclosure: Dr. Ng has received personal fees from Takeda Oncology/Ariad and Boehringer Ingelheim. Dr. Mushtaq has stock ownership in Abbott Laboratories, Amgen Inc., Bristol Myers-Squibb, Celgene Corp., Edwards Lifesciences, Gilead Sciences, Inc., Johnson & Johnson, and Medtronic PLC. Dr. S. Liu has received grants from Astra Zeneca, Bayer, Blueprint, Clovis, Corvus, Esanex, Genentech/Roche, Lilly, Pfizer, and Threshold; and has received personal fees from Astra Zeneca, Bristol Myers-Squibb, Celgene, Genentech/Roche, Heron, Lilly, Pfizer, Regeneron, Taiho, and Takeda. Dr. Bauml has received grants from Merck, Carevive Systems, Novartis, Bayer, Janssen, Astra Zeneca, and Takeda; and has received personal fees from Merck, Janssen, Astra Zeneca, Takeda, Bristol MyersSquibb, Celgene, Genentech, Guardant Health, and Boehringer Ingelheim. Dr. Doebele has received grants from Ignyta and Loxo Oncology; has received personal fees from Pfizer, Trovagene, Ariad, Astra Zeneca, Guardant, Ignyta, Spectrum, Takeda, Genentech/Roche, Bayer, and Rain Therapeutics; has a patent with Abbott Molecular on NTRK1 diagnostics; and has a patent pending with Rain Therapeutics. Dr. Aisner has received personal fees from AbbVie, Bristol Myers-Squibb, Genentech, and Bayer Oncology. The remaining authors declare no conflict of interest. Address for correspondence: Terry L. Ng, MD, Division of Medical Oncology, University of Colorado, 12801 E. 17th Ave, 8122, 8th Floor, Aurora, Colorado 80045. E-mail: [email protected] ª 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. ISSN: 1556-0864 https://doi.org/10.1016/j.jtho.2018.12.001
Journal of Thoracic Oncology
Vol.
-
No.
-: ---
2 Ng et al
different in patients with or without TEE within ±90 days of diagnosis in the overall study cohort or within each molecular group. Conclusions: The risk of peridiagnostic TEEs is significantly elevated in patients with advanced ROSþ NSCLC compared to EGFRþ and KRASþ cases. TEE risk may be similarly elevated in ALKþ NSCLC. 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. Keywords: Lung cancer; ROS1; ALK; Oncogene driver; Thrombosis
Introduction Cancer-associated thrombosis is a common condition, although the risk of thromboembolic disease depends on cancer type, stage, treatment, and patient-related factors.1,2 Past studies have shown that cancer patients with thrombotic events have a worse prognosis than cancer patients without thrombotic complications.3-8 Importantly, thrombotic complications are the second leading cause of death in cancer patients.5 Registration studies comparing cancer patients with matched cases from the general population report up to a 20-fold increase in thromboembolic risk in lung cancer, with adenocarcinoma at higher risk than squamous cell carcinoma.9,10 Although the treatment of advanced NSCLC has changed drastically due to the discovery of oncogene drivers, the rate of thromboembolic events (TEEs) in advanced ROS1þ lung cancer has never been reported. Based on an a priori clinical observation from one of the authors (D.R.C.), we hypothesized that the rate of TEEs in this molecular cohort may be higher than in other common molecular drivers including anaplastic lymphoma kinase (ALK) gene rearrangements (ALKþ), EGFR mutations (EGFRþ), and KRAS mutations (KRASþ). The primary objective of this multicenter, retrospective cohort study was to determine the incidence of TEEs (both venous and arterial) in ROS1þ lung cancer relative to ALKþ, EGFRþ, and KRASþ lung cancer. The secondary objectives were to explore other clinical and pathological risk factors that may predict TEEs, and survival outcomes across and within molecular driver subgroups with and without TEE.
Materials and Methods Study Population All patients with unresectable, locally advanced or metastatic (advanced) ROS1þ, ALKþ, EGFRþ, and KRASþ lung cancer from each site were identified in a sequential fashion through a search of our respective electronic medical records from October 2002 to April
Journal of Thoracic Oncology
Vol.
-
No.
-
2018. ROS1þ, ALKþ, EGFRþ, and KRASþ lung cancers were chosen because they are the most commonly tested and well-established oncogene drivers, including a representative population of smoking- and nonsmokingrelated lung cancer. A separate cohort of wild-type patients (i.e., NSCLC that are ROS1, ALK, EGFR, and KRAS wild-type) was not included because these patients can still be driven by other molecular oncogenes including MNNG HOS transforming gene (MET), BRAF, erb-b2 receptor tyrosine kinase 2 (HER2), SLIT and NTRK like family member 1 (NTRK), and ret proto-oncogene (RET) rather than a truly homogeneous separate population. Inclusion of these less common molecular subsets would not allow for a statistically powered analysis. Advanced NSCLC was defined as stage IIIB or greater (according to the seventh edition of the American Joint Committee on Cancer TNM staging system) at initial diagnosis or recurrent metastatic disease.11 Venous TEE (deep venous thrombosis [DVT] or pulmonary embolism [PE]) and arterial TEE (myocardial infarction [MI] or transient ischemic attack [TIA] / cerebrovascular accident [CVA]) within ±365 days of advanced lung cancer diagnosis were identified by retrospective review of the medical record including diagnostic imaging results, medication history, key term searches, and review of progress notes and outside medical records where available. Each participant’s last follow-up date and living status (dead or alive) were captured for survival analysis. Overall survival (OS) was defined as the time from the date of advanced lung cancer diagnosis to the date of last follow-up or death. This study was a joint effort between the University of Colorado, Shanghai Pulmonary Hospital, Georgetown University Hospital, University of Texas Southwestern, and University of Pennsylvania. All participating sites received institutional review board approval to collect clinicopathologic data on their patients.
Patient Demographics and Risk Factors for TEEs Baseline demographics (age, sex, race, performance status, albumin), information on lung cancer diagnosis and treatment (histology, American Joint Committee on Cancer seventh edition staging, sites of metastases, firstline systemic treatment [chemotherapy versus targeted therapy versus other (i.e., immunotherapy or clinical trial)]), and molecular information regarding the participant’s lung cancer diagnosis were collected.11 Each molecular cohort was stratified by their gene fusion partner for the ROS1þ (CD74 versus non-CD74 or analysis by individual fusion partners) and ALKþ (EML4 versus non-EML4 or analysis by individual gene fusion partners) cohorts or by their mutation subtype for the EGFRþ (exon 19 del versus L858R versus other) and
---
2019
KRASþ (transversion mutations [G12A/G12C/G12V/ G13C] versus transition mutations [G13D / G12D/ G12S] versus individual mutation) cohorts. For patients with a documented TEE, Response Evaluation Criteria in Solid Tumors v.1.1 response based on the most recent scan relative to the TEE was also captured (untreated or progressive disease were grouped to reflect active disease, whereas stable disease and partial/complete response were grouped to reflect controlled disease). Other clinical and laboratory variables that may influence TEE risk were stratified into categories for univariate and multivariable analysis to preserve statistical power and feasibility of data collection. This includes variables in the Khorana risk score, a validated score to stratify risk of developing venous TEE in an outpatient cancer population. It consists of five variables (lung cancer [1 point]; pre-chemotherapy platelet count 350 10⁹/L [1 point], hemoglobin < 10g/dL [1 point], pre-chemotherapy leukocyte count 11 109/L [1 point], and body mass index [BMI] 35kg/m2 [1 point]), whereby the 2.5-month rate of developing a venous TEE is 0.3 to 0.8% in the low risk group (0 points), 1.8% to 2.0% in the intermediate risk group (1 to 2 points), and 6.7% to 7.1% in the high-risk group (3 points).
Study Endpoints The primary endpoint was the development of a TEE within ±90 days of advanced lung cancer diagnosis, with the aim to compare differences in risk between the ROS1þ cohort and the ALKþ, EGFRþ, and KRASþ cohorts. Our goal was to assess whether the risk of TEE on developing advanced lung cancer was different among the molecular groups with the hypothesis that ROS1þ would have the highest risk. TEEs were limited to within ±90 days of diagnosis to reduce potential confounding factors other than a new diagnosis of NSCLC with longer follow-up and to minimize the risk of death as a competing event. Individuals with less than 90 days of follow-up and without a TEE event were excluded from the analysis. The secondary endpoints were clinical variables other than molecular group that were associated with TEEs, and OS according to the presence or absence of TEE within ±90 days of advanced lung cancer diagnosis for the overall study cohort and for each molecular driver cohort.
Association of Clinical and Molecular Variables With TEE Logistic regression was used to assess the association between TEEs and various clinical variables. In the event of multiple TEEs within ±90 days, only the first record was assessed. Both univariate and multivariable logistic
ROS1 and Elevated Risk of Thromboembolic Events
3
regression analyses were performed and odds ratios (ORs), 95% confidence interval (CI), and p values were computed. Fisher’s exact test determined whether the proportion of TEEs was different across fusion or mutation subtype categories for a given molecular driver group. All significance testing was conducted using an alpha of 0.05. Variables missing for more than 50 records were assessed univariately but were not included in the multivariable model. Categorical variables with low variability, defined as less than 10 records in one or more categories, were excluded from the univariate and multivariable analyses due to their inability to provide stable estimates. Analyses were conducted using R (V 3.5.0) and SAS (V 9.4) software.
Long-Term Survival Outcomes Associated With TEE Across Molecular Cohorts Kaplan-Meier and log-rank tests were used to examine and test for differences in OS for those that experienced a TEE, and for those that did not, both in the overall study population, and separately for each molecular group (i.e., ROS1þ, ALKþ, EGFRþ, and KRASþ). Survival rates across the four molecular driver groups were also compared. A Cox proportional hazard model was used to assess survival between the subgroup of patients with and without a TEE in the overall study cohort, and in each molecular driver group after adjusting for differences in stage at diagnosis (IIIB/IV versus
Results Baseline Patient Characteristics A total of 910 patients were included in this study. Of those, 13 were excluded because they did not have unresectable or advanced lung cancer at the last followup date or did not have one of the predefined molecular drivers (ROS1/ALK/EGFR/KRAS), or both. Additionally, 63 subjects were excluded because they were missing survival status data and/or their date of last follow-up. Lastly, 94 patients with less than 90 days of follow-up and who did not develop a TEE were excluded, of which, 14.4% (16 of 111) were ROS1þ, 8.1% (17 of 210) were ALKþ, 7.4% (24 of 324) were EGFRþ, and 19.6% (37 of 189) were KRASþ patients. In total, 740 patients were available for the final analysis, comprised of 95 ROS1þ, 193 ALKþ, 300 EGFRþ, and 152 KRASþ lung cancer patients. A comparison of baseline characteristics of all analyzable patients across the molecular subgroups is summarized in Table 1.
4 Ng et al
Journal of Thoracic Oncology
Vol.
-
No.
-
Table 1. Baseline Clinical Characteristics of the Molecular Driver Cohorts
Mean age at diagnosis (SD), y Sex (%) Female Male Race (%) White Asian Other Patient geographical location (%) China USA ECOG (%) <2 2 Albumin (%) <3.5 3.5 Smoking status (%) Never smoked (100 cigarettes in lifetime) Former smoker Current smoker Smoking, pack years (%) <10 10-20 >20 Histology (%) Adenocarcinoma Adeno-squamous Squamous Large cell NSCLC not otherwise specified Stage at diagnosis (%) I II IIIA IIIB IVA IVB Number of metastatic sites (%) 0 1-2 3 Metastatic site Brain (%) No Yes Liver (%) No Yes Lung (%) No Yes Adrenal (%) No Yes Soft tissue (%) No Yes
ROS1 (n ¼ 95)
ALK (n ¼ 193)
EGFR (n ¼ 300)
KRAS (n ¼ 152)
53.83 (11.76)
52.07 (11.77)
60.73 (10.82)
64.48 (9.93)
51 (53.7) 44 (46.3)
101 (52.3) 92 (47.7)
186 (62.0) 114 (38.0)
79 (52.0) 73 (48.0)
31 (32.6) 59 (62.1) 5 (5.3)
98 (50.8) 70 (36.3) 25 (13.0)
175 (58.3) 75 (25.0) 50 (16.7)
85 (55.9) 47 (30.9) 20 (13.2)
52 (54.7) 43 (45.3)
64 (33.2) 129 (66.8)
48 (16.0) 252 (84.0)
43 (28.3) 109 (71.7)
24 (26.4) 67 (73.6)
46 (27.2) 123 (72.8)
80 (31.0) 178 (69.0)
25 (18.9) 107 (81.1)
21 (24.7) 64 (75.3)
37 (28.0) 95 (72.0)
50 (25.6) 145 (74.4)
32 (26.9) 87 (73.1)
73 (77.7) 10 (10.6) 11 (11.7)
149 (77.2) 26 (13.5) 18 (9.3)
200 (67.1) 82 (27.5) 16 (5.4)
24 (15.9) 92 (60.9) 35 (23.2)
7 (35.0) 4 (20.0) 9 (45.0)
11 (26.8) 19 (46.3) 11 (26.8)
37 (40.7) 27 (29.7) 27 (29.7)
11 (9.2) 23 (19.2) 86 (71.7)
89 (93.7) 2 (2.1) 0 (0.0) 0 (0.0) 4 (4.2)
189 (98.4) 0 (0.0) 0 (0.0) 1 (0.5) 2 (1.0)
292 (97.7) 1 (0.3) 3 (1.0) 1 (0.3) 2 (0.7)
145 (95.4) 1 (0.7) 0 (0.0) 0 (0.0) 6 (3.9)
2 (2.1) 1 (1.1) 5 (5.3) 17 (18.1) 25 (26.6) 44 (46.8)
6 (3.1) 2 (1.0) 12 (6.2) 16 (8.3) 59 (30.7) 97 (50.5)
20 (6.7) 10 (3.3) 12 (4.0) 6 (2.0) 67 (22.4) 184 (61.5)
11 (7.3) 9 (6.0) 9 (6.0) 14 (9.3) 32 (21.2) 76 (50.3)
1 (1.1) 44 (48.4) 46 (50.5)
0 (0.0) 82 (43.4) 107 (56.6)
1 (0.3) 130 (45.0) 158 (54.7)
2 (1.4) 82 (56.9) 60 (41.7)
69 (74.2) 24 (25.8)
139 (72.4) 53 (27.6)
199 (67.7) 95 (32.3)
113 (76.4) 35 (23.6)
83 (88.3) 11 (11.7)
160 (82.9) 33 (17.1)
242 (81.2) 56 (18.8)
137 (90.7) 14 (9.3)
46 (48.9) 48 (51.1)
79 (40.9) 114 (59.1)
91 (30.5) 207 (69.5)
58 (38.4) 93 (61.6)
85 (90.4) 9 (9.6)
177 (91.7) 16 (8.3)
277 (93.9) 18 (6.1)
138 (91.4) 13 (8.6)
92 (97.9) 2 (2.1)
187 (96.9) 6 (3.1)
286 (96.6) 10 (3.4)
143 (94.7) 8 (5.3) (continued)
---
ROS1 and Elevated Risk of Thromboembolic Events
2019
5
Table 1. Continued
Bone (%) No Yes Pleural (%) No Yes Lymph node (%) No Yes Other (%) No Yes Systemic therapy Chemotherapy Targeted Therapy Other
ROS1 (n ¼ 95)
ALK (n ¼ 193)
EGFR (n ¼ 300)
KRAS (n ¼ 152)
65 (69.1) 29 (30.9)
120 (62.2) 73 (37.8)
145 (48.8) 152 (51.2)
98 (64.9) 53 (35.1)
57 (61.3) 36 (38.7)
100 (52.1) 92 (47.9)
189 (63.9) 107 (36.1)
112 (75.2) 37 (24.8)
10 (10.8) 83 (89.2)
40 (20.9) 151 (79.1)
131 (44.4) 164 (55.6)
50 (33.3) 100 (66.7)
89 (95.7) 4 (4.3)
165 (85.9) 27 (14.1)
277 (93.3) 20 (6.7)
139 (92.7) 11 (7.3)
50 (57.5) 37 (42.5) 0 (0.0)
97 (50.8) 87 (45.5) 7 (3.7)
89 (30.5) 189 (64.7) 14 (4.8)
125 (94.7) 3 (2.3) 4 (3.0)
ECOG, Eastern Cooperative Oncology Group.
TEEs Across Molecular Subgroups Within 90 Days of Diagnosis Of the eligible patients (N ¼ 740), the incidence rate of TEE within ±90 days of advanced lung cancer was 34.7% (33 of 95), 22.3% (43 of 193), 13.7% (41 of 300), and 18.4% (28 of 152) for the ROS1þ, ALKþ, EGFRþ, and KRAS cohorts, respectively. Death as a competing risk within 90 days of advanced lung cancer diagnosis was negligible (10 of 740 [1.35%]) (Supplementary Fig. 1A). In univariate analysis, the odds of TEE occurring in the ROS1þ cohort was significantly higher than in the ALKþ (OR: 1.86, 95% CI: 1.08–3.19, p ¼ 0.025), EGFRþ (OR: 3.36, 95% CI: 1.96–5.57, p < 0.001), and KRASþ (OR: 2.36, 95% CI: 1.31–4.27, p ¼ 0.004) cohorts. In multivariable analysis, the odds of TEE in the ROS1þ cohort remained significantly higher than in the EGFRþ (OR: 2.44, 95% CI: 1.31–4.57, p ¼ 0.005) and KRASþ (OR: 2.62, 95% CI: 1.26–5.46, p ¼ 0.01) cohorts. However, although numerically higher, the difference compared to the ALKþ cohort was no longer statistically significant (OR: 1.45, 95% CI: 0.79–2.64, p ¼ 0.229) (Table 2). The cumulative incidence over time of TEEs within ±90 days of advanced lung cancer diagnosis for each molecular group is shown in Figure 1. Although most
of the TEEs were detected after the diagnosis of advanced lung cancer, there was no significant difference in the proportion of symptomatic and asymptomatic TEEs before and after diagnosis (p ¼ 0.13). The proportion of subjects with venous TEE (DVT/PE) versus arterial TEE (MI/TIA/CVA) as their first event within ±90 days of diagnosis was 57.6% (19 of 33) and 42.4% (14 of 33) for the ROS1þ cohort, 55.8% (24 of 43) and 44.2% (19 of 43) for the ALKþ cohort, 58.5% (24 of 41) and 41.5% (17 of 41) for the EGFRþ cohort, and 39.3% (11 of 28) and 60.7% (17 of 28) for the KRASþ cohort, respectively. This, and the proportion of subjects with DVT, PE, MI, and TIA/CVA separately as their first event were not different across molecular driver groups (p ¼ 0.39 and p ¼ 0.49, respectively) (Supplementary Table 1). The median time (interquartile range [IQR]) to TEE from the time of diagnosis for ROS1þ/ALKþ/ EGFRþ/ KRASþ was 0 days (IQR: -6.0 to 7.0), 0 days (IQR: -3.5 to 2), 2 days (IQR: -5.0 to 14.0), and 4.5 days (IQR: -1.3 to 17.5), respectively.
Association of Other Clinical Variables With TEE Within 90 Days of Diagnosis Several clinical variables were excluded from the analyses due to their low variability (less than 10
Table 2. Logistic Regression Analysis of Molecular Driver Groups and Thromboembolic Events Univariate analysis OR Molecular Driver ROS1 vs. ALK ROS1 vs. EGFR ROS1 vs. KRAS
1.86 3.36 2.36
Lower 95% CI 1.08 1.96 1.31
OR, odds ratio; CI, confidence interval.
Multivariable analysis Upper 95% CI
p Value
3.19 5.75 4.27
<0.001 0.025 <0.001 0.004
OR 1.45 2.44 2.62
Lower 95% CI 0.79 1.31 1.26
Upper 95% CI
p Value
2.64 4.57 5.46
0.017 0.229 0.005 0.01
6 Ng et al
Figure 1. Cumulative incidence of thromboembolic events within ±90 days of advanced lung cancer diagnosis according to molecular driver group. Time 0 is when advanced lung cancer was diagnosed.
records) in the dataset, including congestive heart failure, chronic kidney disease, stage at diagnosis (IIIB þ versus
Journal of Thoracic Oncology
Vol.
-
No.
-
platelet count were excluded from multivariable analysis because they were not available for more than 10% of patients. As a result, the final dataset used in the multivariable model was reduced from 740 to 694 patients. Although age was not significant in the univariate setting, the odds of a TEE increased by 4% for every 1-year increase in age at the time of diagnosis (p < 0.001) in multivariable analysis. Hypertension also emerged in multivariable analysis as being significantly associated with TEE. Among the different metastatic sites that could have been involved at advanced stage diagnosis, only the presence of lymph node metastasis was found to be significantly related to increased odds of a TEE in the multivariable model (OR: 2.29; 95% CI: 1.31–4.01; p ¼ 0.004) (Table 3). Importantly, there was no significant difference across the molecular groups in the proportion of patients with active disease (untreated disease or progressive disease) versus those with controlled disease (stable disease or complete response/partial response) in cases where TEE was diagnosed (Supplementary Table 3). Also, of the patients who received systemic anticancer therapy within 90 days of diagnosis, the proportion of patients that received a vascular endothelial growth factor–related monoclonal antibody was 5.7% (5 of 87) for ROS1, 8.4% (16 of 191) for ALK, 6.5% (19 of 292) for EGFR, and 12.1% (16 of 132) for KRAS (p ¼ 0.23). Among patients with information on anticoagulation within 30 days before diagnosis of venous TEE, the proportion that received anticoagulation was 0% (0 of 19) for ROS1, 4.2% (1 of 24) for ALK, 4.2% (1 of 24) for EGFR, and 11.1% (1 of 9) for KRAS (p ¼ 0.51). Among patients with information on antiplatelet therapy within 30 days before diagnosis of arterial TEE, the proportion that received antiplatelet therapy was 0% (0 of 14) for ROS1, 0% (0 of 19) for ALK, 5.9% (1 of 17) for EGFR, 0% (0 of 17) for KRAS (p ¼ 0.72).
Subtypes of Gene Fusions or Mutations and TEE Within 90 Days of Diagnosis Among available data on gene fusion or mutation subtype, EGFR was the only molecular group with significantly different TEE rates across its molecular subtype categories (p ¼ 0.018), with the largest proportion of TEEs in the exon 19 deletion subtype (44.7% [17 of 38]) versus L858R (23.7% [9 of 38]) or other EGFR mutations (31.6% [12 of 38]) (Supplementary Table 4). Supplementary Table 5 shows the proportion of patients with and without TEEs across individual gene fusions or mutations within each molecular group.
---
ROS1 and Elevated Risk of Thromboembolic Events
2019
7
Table 3. Logistic Regression Assessing Clinical Association With Thromboembolic Events Clinical Variable
OR
Lower 95% CI
Upper 95% CI
p Value
OR
Lower 95% CI
Upper 95% CI
p Value
Age at diagnosis, y Sex Males vs. females Race Asian vs. White Other vs. White ECOG performance 2 vs. <2 Albumin (g/dL) 3.5 vs. <3.5 Smoking status Former vs. never Current vs. never Smoking pack years 10-20 vs. <10 >20 vs. <10 Number of sites <3 vs. 3þ Brain site Yes vs. no Liver site Yes vs. no Lung site Yes vs. no Bone site Yes vs. no Pleural site Yes vs. no Lymph node site Yes vs. no Other site Yes vs. no WBC > 11 109 /L Yes vs. no HGB < 10 g/dL Yes vs. no PLT count 350 109 /L Yes vs. no Diabetes Yes vs. no Hypertension Yes vs. no Khorana score 1 to 5 Geographic location USA vs. China
1.01
1
1.03
0.08
1.04
1.02
1.06
<0.001
1.5
1.04
2.16
1
0.65
1.54
0.982 0.246
2.81 1.12
1.9 0.58
4.18 2.04
0.029 0.001 <0.001 0.723
1.57 1.15
0.93 0.58
2.65 2.29
1.87
1.18
3.09
0.011
—
—
—
—
0.42
0.27
0.65
—
—
—
0.58 2.61
0.36 1.56
0.92 4.32
<0.001 <0.001 0.025 <0.001 0.732
0.89 2.19
0.49 1.16
1.59 4.13
0.98 1.25
0.41 0.6
2.34 2.73
— —
— —
— —
— 0.032 0.682 0.016 — — —
1.12
0.77
1.62
0.552
—
—
—
—
0.81
0.53
1.22
0.33
0.93
0.58
1.5
0.76
1.12
0.67
1.8
0.663
1.43
0.81
2.51
0.22
0.62
0.43
0.89
0.01
0.69
0.44
1.1
0.123
0.93
0.64
1.34
0.686
0.99
0.64
1.52
0.959
1.45
1
2.1
0.052
1.43
0.94
2.17
0.093
3.24
2.02
5.44
<0.001
2.29
1.31
4.01
0.004
1.09
0.55
2.01
0.789
1.17
0.58
2.36
0.671
1.76
1.01
2.98
0.041
—
—
—
—
2.96
1.23
6.92
0.012
—
—
—
—
1.87
1.04
3.25
0.03
—
—
—
—
0.99
0.46
1.95
0.971
1.7
0.75
3.85
0.204
0.7
0.44
1.08
0.114
0.52
0.31
0.89
0.017
1.62
1.18
2.22
0.002
—
—
—
—
0.29
0.2
0.42
<0.001
—
—
—
—
OR, odds ratio; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; WBC, white blood cell; HGB, hemoglobin; PLT, platelet.
OS Across and Within Molecular Cohorts With and Without TEEs Median duration of follow-up across all patients was 19.9 months, which varied among the molecular groups (ROS1þ 10.8 months, ALKþ 21.1 months, EGFRþ 21.2 months, and KRASþ 6.1 months). Median duration of follow-up for patients enrolled from institutions in the United States and in China was 25.3
months and 11.6 months, respectively. When the overall study cohort and each molecular group were stratified by the presence or absence of TEE within ±90 days of advanced lung cancer diagnosis, there was no evidence that OS was significantly different between those subgroups (Figs. 2A-E). After adjusting for stage, age at diagnosis, sex, race, smoking status, and molecular driver group, the survival between the
8 Ng et al
Journal of Thoracic Oncology
Vol.
-
No.
-
Figure 2. Kaplan Meier graphs of survival. (A-E) Survival in the presence or absence of TEE within 90 days of advanced lung cancer diagnosis, both for the entire cohort and for each molecular driver group. (F) Overall survival across the different molecular driver groups. The log-rank test was used to assess for differences in survival. TEE, thromboembolic events; CI, confidence interval.
subgroup with versus without TEE was not significantly different (hazard ratio: 1.35, 95% CI: 0.89–2.05, p ¼ 0.152). OS among the molecular driver groups was found to differ significantly using the log-rank test (p < 0.0001), with KRAS displaying the poorest survival (Fig. 2F).
Discussion This retrospective cohort study examined the odds of developing TEEs associated with ROS1þ lung cancer compared with other common molecular driver cohorts. Our results show that the odds of developing TEEs in ROS1þ lung cancer were significantly higher than in
---
2019
EGFRþ and KRASþ lung cancer. Although the numerical difference in TEE incidence between ROS1þ and ALKþ lung cancer (34.7% versus 22.3%) was significant in univariable analysis, this difference was not present in multivariable analysis (p ¼ 0.229). Two prior retrospective cohort studies of 264 and 99 ALKþ lung cancer patients reported a TEE incidence of 30% and 36%, respectively, much higher than historically reported rates in the general lung cancer population.12,13 Although our study reported a shorter period of follow-up for TEEs, it also reproduced the observation that ALKþ patients are at increased risk of developing TEEs. Among the other clinical variables, although ECOG status greater than or equal to 2, current/former smoker, pleural and lymph node metastases, a higher Khorana score, (including WBC >11 109/L, platelet count 350, and HGB <10 g/dL), Asian race and/or patients from China, and albumin less than 3.5 g/dL increased the odds of TEE in univariate analysis, only lymph node metastases and smoking status continued to be associated with increased odds of a TEE in multivariable analysis. Older age and hypertension emerged in multivariable analysis as being significantly associated with TEE. Despite prior studies showing an association between cancer-associated thrombosis and worse survival, our study did not show worse survival in the subgroups that developed a TEE, both in the overall study cohort and across molecular driver groups.6,12 Possible reasons for this different observation include insufficient duration of follow-up and potentially insufficient sample size. Considering the prolonged disease control feasible from targeted therapies among ROSþ, ALKþ, and EGFRþ patients, it is also possible that subsequent TEE propensity and its associated risks on survival may be abrogated by highly active anti-cancer therapy. High TEE rates in ROS1þ and ALKþ lung cancer raises the possibility of an etiological mechanism of prothrombotic risk associated with certain molecular drivers. Because ROS1þ and ALKþ are gene fusions that are phylogenetically related, clotting risk may be mediated by specific signaling emanating from these oncogenic abnormalities. Although elevated TEE rate in both groups may be related to treatment with a common drug (e.g., crizotinib), TEEs in both groups occurred primarily at the time of diagnosis, suggesting that commencement of therapy was unlikely a relevant factor. However, this question is worthy of exploration in future projects. This is the largest evaluation of TEEs in ROS1þ lung cancer patients to date, and the only study that we are aware of to simultaneously compare the odds of TEE across multiple molecular driver groups. Directly
ROS1 and Elevated Risk of Thromboembolic Events
9
comparing TEE risk between ROS1þ, ALKþ, EGFRþ, and KRASþ patients minimized sampling bias seen in cross-trial comparisons or when comparing study results against historical control cohorts. We controlled for other clinicopathologic risk factors that may have confounded TEE risk in this study by incorporating multivariable logistic regression. A ROS1/ALK/EGFR/ KRAS-wild type cohort was not included because this is, by definition, not a uniform population. Limitations of this study other than the retrospective study design include the imbalance in the number of patients included from China and the United States in each molecular cohort. However, race was highly correlated to country of enrollment (86% Asians from China and 100% Caucasians from the United States), and was not significantly associated with TEE in multivariable analysis. More than 10% of records were missing information on albumin concentration, ECOG score, BMI, Khorana score, WBC count, HGB concentration, and platelet count, and we were unable to incorporate them into multivariable analysis. There were fewer than 10 records in at least one of the comparator subgroups for congestive heart failure, chronic kidney disease, stage at diagnosis (IIIBþ versus
Acknowledgments Drs. Camidge, Doebele, and Aisner were supported in part by the University of Colorado Lung Cancer Specialized Program of Research Excellence (P50CA058187). Mr. Smith and Dr. Gao were supported by the Colorado Cancer Center Support Grant (P30CA046934). This study was not influenced by any of the funding bodies.
Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of the Journal of Thoracic Oncology at www.jto.org and at http://doi. org/10.1016/j.jtho.2018.12.001.
10 Ng et al
Journal of Thoracic Oncology
References 1. Timp JF, Braekkan SK, Versteeg HH, et al. Epidemiology of cancer-associated venous thrombosis. Blood. 2013;122:1712–1723. 2. Gade IL, Brækkan SK, Næss IA, et al. The impact of initial cancer stage on the incidence of venous thromboembolism: the Scandinavian Thrombosis and Cancer (STAC) Cohort. J Thromb Haemost. 2017;15:1567–1575. 3. Braekkan SK, Borch KH, Mathiesen EB, et al. Body height and risk of venous thromboembolism: the Tromsø Study. Am J Epidemiol. 2010;171:1109–1115. 4. Chew HK, Wun T, Harvey D, et al. Incidence of venous thromboembolism and its effect on survival among patients with common cancers. Arch Intern Med. 2006;166:458–464. 5. Khorana AA, Francis CW, Culakova E, et al. Thromboembolism is a leading cause of death in cancer patients receiving outpatient chemotherapy. J Thromb Haemost. 2007;5:632–634. 6. Sørensen HT, Mellemkjaer L, Olsen JH, et al. Prognosis of cancers associated with venous thromboembolism. N Engl J Med. 2000;343:1846–1850. 7. Horlander KT, Mannino DM, Leeper KV. Pulmonary embolism mortality in the United States, 1979-1998: an
8.
9.
10.
11.
12.
13.
Vol.
-
No.
-
analysis using multiple-cause mortality data. Arch Intern Med. 2003;163:1711–1717. Ogren M, Bergqvist D, Wåhlander K, et al. Trousseau’s syndrome — what is the evidence? A population-based autopsy study. Thromb Haemost. 2006;95:541–545. Blom JW, Osanto S, Rosendaal FR. The risk of a venous thrombotic event in lung cancer patients: higher risk for adenocarcinoma than squamous cell carcinoma. J Thromb Haemost. 2004;2:1760–1765. Blom JW, Doggen CJM, Osanto S, et al. Malignancies, prothrombotic mutations, and the risk of venous thrombosis. JAMA. 2005;293:715–722. American Joint Committee on Cancer. AJCC Cancer Staging Manual. 7th ed. Chicago, Illinois: American Joint Committee on Cancer; 2010. Zugazagoitia J, Biosca M, Grau JF, et al. Incidence, predictors and prognostic significance of thromboembolic events in patients with advanced ALK-rearranged NSCLCs. Presented at IASLC 18th World Conference on Lung Cancer. October 15-18, 2017; Yokohama, Japan. Zer A, Moskovitz M, Hwang DM, et al. ALK-rearranged non–small-cell lung cancer is associated with a high rate of venous thromboembolism. Clin Lung Cancer. 2017;18:156–161.
ROS1 Gene Rearrangements Are Associated With an Elevated Risk of Peridiagnosis Thromboembolic Events Terry L. Ng, MD,a,* Derek E. Smith, MS,b Rao Mushtaq, MD,a Tejas Patil, MD,a Anastasios Dimou, MD,a Shuo Yang, MD,c Qian Liu, MD,c Xuefei Li, PhD,c Caicun Zhou, MD, PhD,c Robert T. Jones, BS,d Megan M. Tu, PhD,e Flora Yan, BS,f I. Alex Bowman, MD,f Stephen V. Liu, MD,g Siera Newkirk, BS,h Joshua Bauml, MD,h Robert C. Doebele, MD, PhD,a Dara L. Aisner, MD, PhD,i Dexiang Gao, PhD,b Shengxiang Ren, MD, PhD,c D. Ross Camidge, MD, PhDa a
Division of Medical Oncology, University of Colorado, Aurora, Colorado Department of Pediatrics, Cancer Center Biostatistics Core, University of Colorado and Children’s Hospital Colorado, Aurora, Colorado c Shanghai Pulmonary Hospital, Tongji University School of Medicine, Shanghai, China d Department of Pharmacology, University of Colorado, Aurora, Colorado e Department of Surgery, University of Colorado, Denver, and University of Colorado Comprehensive Cancer Center, Aurora, Colorado f University of Texas at Southwestern, Dallas, Texas g Georgetown University, Washington, DC h Perelman School of Medicine, University of Pennsylvania, Philadelphia, Pennsylvania i University of Colorado, Department of Pathology, Aurora, Colorado b
Received 7 October 2018; revised 2 December 2018; accepted 2 December 2018 Available online - 10 December 2018
ABSTRACT Introduction: This study aims to determine whether advanced ROS1 gene-rearranged NSCLC (ROS1þ NSCLC) has a higher than expected thromboembolic event (TEE) rate. Methods: Venous and arterial TEEs within ±365 days of diagnosis of ROS1þ, ALKþ, EGFRþ, or KRASþ advanced NSCLC at five academic centers in the United States and China were captured (October 2002–April 2018). The primary endpoint was incidence of TEE in ROS1þ compared to anaplastic lymphoma kinase (ALK)þ, EGFRþ, and KRASþ NSCLC within ±90 days of diagnosis. Logistic regression was used to assess if the odds of TEE differed among oncogene drivers. Results: Eligible data from 95 ROS1þ, 193 ALKþ, 300 EGFRþ, and 152 KRASþ NSCLC patients were analyzed. The incidence rate of TEE was 34.7% (n ¼ 33), 22.3% (n ¼ 43), 13.7% (n ¼ 41), and 18.4% (n ¼ 28), respectively. In univariate analysis, the odds of a TEE in ROS1þ NSCLC were higher than ALKþ, EGFRþ, and KRASþ cohorts. In multivariable analysis, the odds of a TEE were significantly higher for ROS1þ compared to EGFRþ and KRASþ cohorts, the odds ratio (OR) was 2.44, with a 95% confidence interval of 1.31–4.57 (p ¼ 0.005), and OR: 2.62, with a 95% confidence interval of 1.26–5.46 (p ¼ 0.01), respectively.
Although numerically superior, the odds for a TEE with ROS1þ compared to ALKþ was not statistically significant (OR: 1.45, p ¼ 0.229). Overall survival was not significantly *Corresponding author. Disclosure: Dr. Ng has received personal fees from Takeda Oncology/Ariad and Boehringer Ingelheim. Dr. Mushtaq has stock ownership in Abbott Laboratories, Amgen Inc., Bristol Myers-Squibb, Celgene Corp., Edwards Lifesciences, Gilead Sciences, Inc., Johnson & Johnson, and Medtronic PLC. Dr. S. Liu has received grants from Astra Zeneca, Bayer, Blueprint, Clovis, Corvus, Esanex, Genentech/Roche, Lilly, Pfizer, and Threshold; and has received personal fees from Astra Zeneca, Bristol Myers-Squibb, Celgene, Genentech/Roche, Heron, Lilly, Pfizer, Regeneron, Taiho, and Takeda. Dr. Bauml has received grants from Merck, Carevive Systems, Novartis, Bayer, Janssen, Astra Zeneca, and Takeda; and has received personal fees from Merck, Janssen, Astra Zeneca, Takeda, Bristol MyersSquibb, Celgene, Genentech, Guardant Health, and Boehringer Ingelheim. Dr. Doebele has received grants from Ignyta and Loxo Oncology; has received personal fees from Pfizer, Trovagene, Ariad, Astra Zeneca, Guardant, Ignyta, Spectrum, Takeda, Genentech/Roche, Bayer, and Rain Therapeutics; has a patent with Abbott Molecular on NTRK1 diagnostics; and has a patent pending with Rain Therapeutics. Dr. Aisner has received personal fees from AbbVie, Bristol Myers-Squibb, Genentech, and Bayer Oncology. The remaining authors declare no conflict of interest. Address for correspondence: Terry L. Ng, MD, Division of Medical Oncology, University of Colorado, 12801 E. 17th Ave, 8122, 8th Floor, Aurora, Colorado 80045. E-mail: [email protected] ª 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. ISSN: 1556-0864 https://doi.org/10.1016/j.jtho.2018.12.001
Journal of Thoracic Oncology
Vol.
-
No.
-: ---
2 Ng et al
different in patients with or without TEE within ±90 days of diagnosis in the overall study cohort or within each molecular group. Conclusions: The risk of peridiagnostic TEEs is significantly elevated in patients with advanced ROSþ NSCLC compared to EGFRþ and KRASþ cases. TEE risk may be similarly elevated in ALKþ NSCLC. 2018 International Association for the Study of Lung Cancer. Published by Elsevier Inc. All rights reserved. Keywords: Lung cancer; ROS1; ALK; Oncogene driver; Thrombosis
Introduction Cancer-associated thrombosis is a common condition, although the risk of thromboembolic disease depends on cancer type, stage, treatment, and patient-related factors.1,2 Past studies have shown that cancer patients with thrombotic events have a worse prognosis than cancer patients without thrombotic complications.3-8 Importantly, thrombotic complications are the second leading cause of death in cancer patients.5 Registration studies comparing cancer patients with matched cases from the general population report up to a 20-fold increase in thromboembolic risk in lung cancer, with adenocarcinoma at higher risk than squamous cell carcinoma.9,10 Although the treatment of advanced NSCLC has changed drastically due to the discovery of oncogene drivers, the rate of thromboembolic events (TEEs) in advanced ROS1þ lung cancer has never been reported. Based on an a priori clinical observation from one of the authors (D.R.C.), we hypothesized that the rate of TEEs in this molecular cohort may be higher than in other common molecular drivers including anaplastic lymphoma kinase (ALK) gene rearrangements (ALKþ), EGFR mutations (EGFRþ), and KRAS mutations (KRASþ). The primary objective of this multicenter, retrospective cohort study was to determine the incidence of TEEs (both venous and arterial) in ROS1þ lung cancer relative to ALKþ, EGFRþ, and KRASþ lung cancer. The secondary objectives were to explore other clinical and pathological risk factors that may predict TEEs, and survival outcomes across and within molecular driver subgroups with and without TEE.
Materials and Methods Study Population All patients with unresectable, locally advanced or metastatic (advanced) ROS1þ, ALKþ, EGFRþ, and KRASþ lung cancer from each site were identified in a sequential fashion through a search of our respective electronic medical records from October 2002 to April
Journal of Thoracic Oncology
Vol.
-
No.
-
2018. ROS1þ, ALKþ, EGFRþ, and KRASþ lung cancers were chosen because they are the most commonly tested and well-established oncogene drivers, including a representative population of smoking- and nonsmokingrelated lung cancer. A separate cohort of wild-type patients (i.e., NSCLC that are ROS1, ALK, EGFR, and KRAS wild-type) was not included because these patients can still be driven by other molecular oncogenes including MNNG HOS transforming gene (MET), BRAF, erb-b2 receptor tyrosine kinase 2 (HER2), SLIT and NTRK like family member 1 (NTRK), and ret proto-oncogene (RET) rather than a truly homogeneous separate population. Inclusion of these less common molecular subsets would not allow for a statistically powered analysis. Advanced NSCLC was defined as stage IIIB or greater (according to the seventh edition of the American Joint Committee on Cancer TNM staging system) at initial diagnosis or recurrent metastatic disease.11 Venous TEE (deep venous thrombosis [DVT] or pulmonary embolism [PE]) and arterial TEE (myocardial infarction [MI] or transient ischemic attack [TIA] / cerebrovascular accident [CVA]) within ±365 days of advanced lung cancer diagnosis were identified by retrospective review of the medical record including diagnostic imaging results, medication history, key term searches, and review of progress notes and outside medical records where available. Each participant’s last follow-up date and living status (dead or alive) were captured for survival analysis. Overall survival (OS) was defined as the time from the date of advanced lung cancer diagnosis to the date of last follow-up or death. This study was a joint effort between the University of Colorado, Shanghai Pulmonary Hospital, Georgetown University Hospital, University of Texas Southwestern, and University of Pennsylvania. All participating sites received institutional review board approval to collect clinicopathologic data on their patients.
Patient Demographics and Risk Factors for TEEs Baseline demographics (age, sex, race, performance status, albumin), information on lung cancer diagnosis and treatment (histology, American Joint Committee on Cancer seventh edition staging, sites of metastases, firstline systemic treatment [chemotherapy versus targeted therapy versus other (i.e., immunotherapy or clinical trial)]), and molecular information regarding the participant’s lung cancer diagnosis were collected.11 Each molecular cohort was stratified by their gene fusion partner for the ROS1þ (CD74 versus non-CD74 or analysis by individual fusion partners) and ALKþ (EML4 versus non-EML4 or analysis by individual gene fusion partners) cohorts or by their mutation subtype for the EGFRþ (exon 19 del versus L858R versus other) and
---
2019
KRASþ (transversion mutations [G12A/G12C/G12V/ G13C] versus transition mutations [G13D / G12D/ G12S] versus individual mutation) cohorts. For patients with a documented TEE, Response Evaluation Criteria in Solid Tumors v.1.1 response based on the most recent scan relative to the TEE was also captured (untreated or progressive disease were grouped to reflect active disease, whereas stable disease and partial/complete response were grouped to reflect controlled disease). Other clinical and laboratory variables that may influence TEE risk were stratified into categories for univariate and multivariable analysis to preserve statistical power and feasibility of data collection. This includes variables in the Khorana risk score, a validated score to stratify risk of developing venous TEE in an outpatient cancer population. It consists of five variables (lung cancer [1 point]; pre-chemotherapy platelet count 350 10⁹/L [1 point], hemoglobin < 10g/dL [1 point], pre-chemotherapy leukocyte count 11 109/L [1 point], and body mass index [BMI] 35kg/m2 [1 point]), whereby the 2.5-month rate of developing a venous TEE is 0.3 to 0.8% in the low risk group (0 points), 1.8% to 2.0% in the intermediate risk group (1 to 2 points), and 6.7% to 7.1% in the high-risk group (3 points).
Study Endpoints The primary endpoint was the development of a TEE within ±90 days of advanced lung cancer diagnosis, with the aim to compare differences in risk between the ROS1þ cohort and the ALKþ, EGFRþ, and KRASþ cohorts. Our goal was to assess whether the risk of TEE on developing advanced lung cancer was different among the molecular groups with the hypothesis that ROS1þ would have the highest risk. TEEs were limited to within ±90 days of diagnosis to reduce potential confounding factors other than a new diagnosis of NSCLC with longer follow-up and to minimize the risk of death as a competing event. Individuals with less than 90 days of follow-up and without a TEE event were excluded from the analysis. The secondary endpoints were clinical variables other than molecular group that were associated with TEEs, and OS according to the presence or absence of TEE within ±90 days of advanced lung cancer diagnosis for the overall study cohort and for each molecular driver cohort.
Association of Clinical and Molecular Variables With TEE Logistic regression was used to assess the association between TEEs and various clinical variables. In the event of multiple TEEs within ±90 days, only the first record was assessed. Both univariate and multivariable logistic
ROS1 and Elevated Risk of Thromboembolic Events
3
regression analyses were performed and odds ratios (ORs), 95% confidence interval (CI), and p values were computed. Fisher’s exact test determined whether the proportion of TEEs was different across fusion or mutation subtype categories for a given molecular driver group. All significance testing was conducted using an alpha of 0.05. Variables missing for more than 50 records were assessed univariately but were not included in the multivariable model. Categorical variables with low variability, defined as less than 10 records in one or more categories, were excluded from the univariate and multivariable analyses due to their inability to provide stable estimates. Analyses were conducted using R (V 3.5.0) and SAS (V 9.4) software.
Long-Term Survival Outcomes Associated With TEE Across Molecular Cohorts Kaplan-Meier and log-rank tests were used to examine and test for differences in OS for those that experienced a TEE, and for those that did not, both in the overall study population, and separately for each molecular group (i.e., ROS1þ, ALKþ, EGFRþ, and KRASþ). Survival rates across the four molecular driver groups were also compared. A Cox proportional hazard model was used to assess survival between the subgroup of patients with and without a TEE in the overall study cohort, and in each molecular driver group after adjusting for differences in stage at diagnosis (IIIB/IV versus
Results Baseline Patient Characteristics A total of 910 patients were included in this study. Of those, 13 were excluded because they did not have unresectable or advanced lung cancer at the last followup date or did not have one of the predefined molecular drivers (ROS1/ALK/EGFR/KRAS), or both. Additionally, 63 subjects were excluded because they were missing survival status data and/or their date of last follow-up. Lastly, 94 patients with less than 90 days of follow-up and who did not develop a TEE were excluded, of which, 14.4% (16 of 111) were ROS1þ, 8.1% (17 of 210) were ALKþ, 7.4% (24 of 324) were EGFRþ, and 19.6% (37 of 189) were KRASþ patients. In total, 740 patients were available for the final analysis, comprised of 95 ROS1þ, 193 ALKþ, 300 EGFRþ, and 152 KRASþ lung cancer patients. A comparison of baseline characteristics of all analyzable patients across the molecular subgroups is summarized in Table 1.
4 Ng et al
Journal of Thoracic Oncology
Vol.
-
No.
-
Table 1. Baseline Clinical Characteristics of the Molecular Driver Cohorts
Mean age at diagnosis (SD), y Sex (%) Female Male Race (%) White Asian Other Patient geographical location (%) China USA ECOG (%) <2 2 Albumin (%) <3.5 3.5 Smoking status (%) Never smoked (100 cigarettes in lifetime) Former smoker Current smoker Smoking, pack years (%) <10 10-20 >20 Histology (%) Adenocarcinoma Adeno-squamous Squamous Large cell NSCLC not otherwise specified Stage at diagnosis (%) I II IIIA IIIB IVA IVB Number of metastatic sites (%) 0 1-2 3 Metastatic site Brain (%) No Yes Liver (%) No Yes Lung (%) No Yes Adrenal (%) No Yes Soft tissue (%) No Yes
ROS1 (n ¼ 95)
ALK (n ¼ 193)
EGFR (n ¼ 300)
KRAS (n ¼ 152)
53.83 (11.76)
52.07 (11.77)
60.73 (10.82)
64.48 (9.93)
51 (53.7) 44 (46.3)
101 (52.3) 92 (47.7)
186 (62.0) 114 (38.0)
79 (52.0) 73 (48.0)
31 (32.6) 59 (62.1) 5 (5.3)
98 (50.8) 70 (36.3) 25 (13.0)
175 (58.3) 75 (25.0) 50 (16.7)
85 (55.9) 47 (30.9) 20 (13.2)
52 (54.7) 43 (45.3)
64 (33.2) 129 (66.8)
48 (16.0) 252 (84.0)
43 (28.3) 109 (71.7)
24 (26.4) 67 (73.6)
46 (27.2) 123 (72.8)
80 (31.0) 178 (69.0)
25 (18.9) 107 (81.1)
21 (24.7) 64 (75.3)
37 (28.0) 95 (72.0)
50 (25.6) 145 (74.4)
32 (26.9) 87 (73.1)
73 (77.7) 10 (10.6) 11 (11.7)
149 (77.2) 26 (13.5) 18 (9.3)
200 (67.1) 82 (27.5) 16 (5.4)
24 (15.9) 92 (60.9) 35 (23.2)
7 (35.0) 4 (20.0) 9 (45.0)
11 (26.8) 19 (46.3) 11 (26.8)
37 (40.7) 27 (29.7) 27 (29.7)
11 (9.2) 23 (19.2) 86 (71.7)
89 (93.7) 2 (2.1) 0 (0.0) 0 (0.0) 4 (4.2)
189 (98.4) 0 (0.0) 0 (0.0) 1 (0.5) 2 (1.0)
292 (97.7) 1 (0.3) 3 (1.0) 1 (0.3) 2 (0.7)
145 (95.4) 1 (0.7) 0 (0.0) 0 (0.0) 6 (3.9)
2 (2.1) 1 (1.1) 5 (5.3) 17 (18.1) 25 (26.6) 44 (46.8)
6 (3.1) 2 (1.0) 12 (6.2) 16 (8.3) 59 (30.7) 97 (50.5)
20 (6.7) 10 (3.3) 12 (4.0) 6 (2.0) 67 (22.4) 184 (61.5)
11 (7.3) 9 (6.0) 9 (6.0) 14 (9.3) 32 (21.2) 76 (50.3)
1 (1.1) 44 (48.4) 46 (50.5)
0 (0.0) 82 (43.4) 107 (56.6)
1 (0.3) 130 (45.0) 158 (54.7)
2 (1.4) 82 (56.9) 60 (41.7)
69 (74.2) 24 (25.8)
139 (72.4) 53 (27.6)
199 (67.7) 95 (32.3)
113 (76.4) 35 (23.6)
83 (88.3) 11 (11.7)
160 (82.9) 33 (17.1)
242 (81.2) 56 (18.8)
137 (90.7) 14 (9.3)
46 (48.9) 48 (51.1)
79 (40.9) 114 (59.1)
91 (30.5) 207 (69.5)
58 (38.4) 93 (61.6)
85 (90.4) 9 (9.6)
177 (91.7) 16 (8.3)
277 (93.9) 18 (6.1)
138 (91.4) 13 (8.6)
92 (97.9) 2 (2.1)
187 (96.9) 6 (3.1)
286 (96.6) 10 (3.4)
143 (94.7) 8 (5.3) (continued)
---
ROS1 and Elevated Risk of Thromboembolic Events
2019
5
Table 1. Continued
Bone (%) No Yes Pleural (%) No Yes Lymph node (%) No Yes Other (%) No Yes Systemic therapy Chemotherapy Targeted Therapy Other
ROS1 (n ¼ 95)
ALK (n ¼ 193)
EGFR (n ¼ 300)
KRAS (n ¼ 152)
65 (69.1) 29 (30.9)
120 (62.2) 73 (37.8)
145 (48.8) 152 (51.2)
98 (64.9) 53 (35.1)
57 (61.3) 36 (38.7)
100 (52.1) 92 (47.9)
189 (63.9) 107 (36.1)
112 (75.2) 37 (24.8)
10 (10.8) 83 (89.2)
40 (20.9) 151 (79.1)
131 (44.4) 164 (55.6)
50 (33.3) 100 (66.7)
89 (95.7) 4 (4.3)
165 (85.9) 27 (14.1)
277 (93.3) 20 (6.7)
139 (92.7) 11 (7.3)
50 (57.5) 37 (42.5) 0 (0.0)
97 (50.8) 87 (45.5) 7 (3.7)
89 (30.5) 189 (64.7) 14 (4.8)
125 (94.7) 3 (2.3) 4 (3.0)
ECOG, Eastern Cooperative Oncology Group.
TEEs Across Molecular Subgroups Within 90 Days of Diagnosis Of the eligible patients (N ¼ 740), the incidence rate of TEE within ±90 days of advanced lung cancer was 34.7% (33 of 95), 22.3% (43 of 193), 13.7% (41 of 300), and 18.4% (28 of 152) for the ROS1þ, ALKþ, EGFRþ, and KRAS cohorts, respectively. Death as a competing risk within 90 days of advanced lung cancer diagnosis was negligible (10 of 740 [1.35%]) (Supplementary Fig. 1A). In univariate analysis, the odds of TEE occurring in the ROS1þ cohort was significantly higher than in the ALKþ (OR: 1.86, 95% CI: 1.08–3.19, p ¼ 0.025), EGFRþ (OR: 3.36, 95% CI: 1.96–5.57, p < 0.001), and KRASþ (OR: 2.36, 95% CI: 1.31–4.27, p ¼ 0.004) cohorts. In multivariable analysis, the odds of TEE in the ROS1þ cohort remained significantly higher than in the EGFRþ (OR: 2.44, 95% CI: 1.31–4.57, p ¼ 0.005) and KRASþ (OR: 2.62, 95% CI: 1.26–5.46, p ¼ 0.01) cohorts. However, although numerically higher, the difference compared to the ALKþ cohort was no longer statistically significant (OR: 1.45, 95% CI: 0.79–2.64, p ¼ 0.229) (Table 2). The cumulative incidence over time of TEEs within ±90 days of advanced lung cancer diagnosis for each molecular group is shown in Figure 1. Although most
of the TEEs were detected after the diagnosis of advanced lung cancer, there was no significant difference in the proportion of symptomatic and asymptomatic TEEs before and after diagnosis (p ¼ 0.13). The proportion of subjects with venous TEE (DVT/PE) versus arterial TEE (MI/TIA/CVA) as their first event within ±90 days of diagnosis was 57.6% (19 of 33) and 42.4% (14 of 33) for the ROS1þ cohort, 55.8% (24 of 43) and 44.2% (19 of 43) for the ALKþ cohort, 58.5% (24 of 41) and 41.5% (17 of 41) for the EGFRþ cohort, and 39.3% (11 of 28) and 60.7% (17 of 28) for the KRASþ cohort, respectively. This, and the proportion of subjects with DVT, PE, MI, and TIA/CVA separately as their first event were not different across molecular driver groups (p ¼ 0.39 and p ¼ 0.49, respectively) (Supplementary Table 1). The median time (interquartile range [IQR]) to TEE from the time of diagnosis for ROS1þ/ALKþ/ EGFRþ/ KRASþ was 0 days (IQR: -6.0 to 7.0), 0 days (IQR: -3.5 to 2), 2 days (IQR: -5.0 to 14.0), and 4.5 days (IQR: -1.3 to 17.5), respectively.
Association of Other Clinical Variables With TEE Within 90 Days of Diagnosis Several clinical variables were excluded from the analyses due to their low variability (less than 10
Table 2. Logistic Regression Analysis of Molecular Driver Groups and Thromboembolic Events Univariate analysis OR Molecular Driver ROS1 vs. ALK ROS1 vs. EGFR ROS1 vs. KRAS
1.86 3.36 2.36
Lower 95% CI 1.08 1.96 1.31
OR, odds ratio; CI, confidence interval.
Multivariable analysis Upper 95% CI
p Value
3.19 5.75 4.27
<0.001 0.025 <0.001 0.004
OR 1.45 2.44 2.62
Lower 95% CI 0.79 1.31 1.26
Upper 95% CI
p Value
2.64 4.57 5.46
0.017 0.229 0.005 0.01
6 Ng et al
Figure 1. Cumulative incidence of thromboembolic events within ±90 days of advanced lung cancer diagnosis according to molecular driver group. Time 0 is when advanced lung cancer was diagnosed.
records) in the dataset, including congestive heart failure, chronic kidney disease, stage at diagnosis (IIIB þ versus
Journal of Thoracic Oncology
Vol.
-
No.
-
platelet count were excluded from multivariable analysis because they were not available for more than 10% of patients. As a result, the final dataset used in the multivariable model was reduced from 740 to 694 patients. Although age was not significant in the univariate setting, the odds of a TEE increased by 4% for every 1-year increase in age at the time of diagnosis (p < 0.001) in multivariable analysis. Hypertension also emerged in multivariable analysis as being significantly associated with TEE. Among the different metastatic sites that could have been involved at advanced stage diagnosis, only the presence of lymph node metastasis was found to be significantly related to increased odds of a TEE in the multivariable model (OR: 2.29; 95% CI: 1.31–4.01; p ¼ 0.004) (Table 3). Importantly, there was no significant difference across the molecular groups in the proportion of patients with active disease (untreated disease or progressive disease) versus those with controlled disease (stable disease or complete response/partial response) in cases where TEE was diagnosed (Supplementary Table 3). Also, of the patients who received systemic anticancer therapy within 90 days of diagnosis, the proportion of patients that received a vascular endothelial growth factor–related monoclonal antibody was 5.7% (5 of 87) for ROS1, 8.4% (16 of 191) for ALK, 6.5% (19 of 292) for EGFR, and 12.1% (16 of 132) for KRAS (p ¼ 0.23). Among patients with information on anticoagulation within 30 days before diagnosis of venous TEE, the proportion that received anticoagulation was 0% (0 of 19) for ROS1, 4.2% (1 of 24) for ALK, 4.2% (1 of 24) for EGFR, and 11.1% (1 of 9) for KRAS (p ¼ 0.51). Among patients with information on antiplatelet therapy within 30 days before diagnosis of arterial TEE, the proportion that received antiplatelet therapy was 0% (0 of 14) for ROS1, 0% (0 of 19) for ALK, 5.9% (1 of 17) for EGFR, 0% (0 of 17) for KRAS (p ¼ 0.72).
Subtypes of Gene Fusions or Mutations and TEE Within 90 Days of Diagnosis Among available data on gene fusion or mutation subtype, EGFR was the only molecular group with significantly different TEE rates across its molecular subtype categories (p ¼ 0.018), with the largest proportion of TEEs in the exon 19 deletion subtype (44.7% [17 of 38]) versus L858R (23.7% [9 of 38]) or other EGFR mutations (31.6% [12 of 38]) (Supplementary Table 4). Supplementary Table 5 shows the proportion of patients with and without TEEs across individual gene fusions or mutations within each molecular group.
---
ROS1 and Elevated Risk of Thromboembolic Events
2019
7
Table 3. Logistic Regression Assessing Clinical Association With Thromboembolic Events Clinical Variable
OR
Lower 95% CI
Upper 95% CI
p Value
OR
Lower 95% CI
Upper 95% CI
p Value
Age at diagnosis, y Sex Males vs. females Race Asian vs. White Other vs. White ECOG performance 2 vs. <2 Albumin (g/dL) 3.5 vs. <3.5 Smoking status Former vs. never Current vs. never Smoking pack years 10-20 vs. <10 >20 vs. <10 Number of sites <3 vs. 3þ Brain site Yes vs. no Liver site Yes vs. no Lung site Yes vs. no Bone site Yes vs. no Pleural site Yes vs. no Lymph node site Yes vs. no Other site Yes vs. no WBC > 11 109 /L Yes vs. no HGB < 10 g/dL Yes vs. no PLT count 350 109 /L Yes vs. no Diabetes Yes vs. no Hypertension Yes vs. no Khorana score 1 to 5 Geographic location USA vs. China
1.01
1
1.03
0.08
1.04
1.02
1.06
<0.001
1.5
1.04
2.16
1
0.65
1.54
0.982 0.246
2.81 1.12
1.9 0.58
4.18 2.04
0.029 0.001 <0.001 0.723
1.57 1.15
0.93 0.58
2.65 2.29
1.87
1.18
3.09
0.011
—
—
—
—
0.42
0.27
0.65
—
—
—
0.58 2.61
0.36 1.56
0.92 4.32
<0.001 <0.001 0.025 <0.001 0.732
0.89 2.19
0.49 1.16
1.59 4.13
0.98 1.25
0.41 0.6
2.34 2.73
— —
— —
— —
— 0.032 0.682 0.016 — — —
1.12
0.77
1.62
0.552
—
—
—
—
0.81
0.53
1.22
0.33
0.93
0.58
1.5
0.76
1.12
0.67
1.8
0.663
1.43
0.81
2.51
0.22
0.62
0.43
0.89
0.01
0.69
0.44
1.1
0.123
0.93
0.64
1.34
0.686
0.99
0.64
1.52
0.959
1.45
1
2.1
0.052
1.43
0.94
2.17
0.093
3.24
2.02
5.44
<0.001
2.29
1.31
4.01
0.004
1.09
0.55
2.01
0.789
1.17
0.58
2.36
0.671
1.76
1.01
2.98
0.041
—
—
—
—
2.96
1.23
6.92
0.012
—
—
—
—
1.87
1.04
3.25
0.03
—
—
—
—
0.99
0.46
1.95
0.971
1.7
0.75
3.85
0.204
0.7
0.44
1.08
0.114
0.52
0.31
0.89
0.017
1.62
1.18
2.22
0.002
—
—
—
—
0.29
0.2
0.42
<0.001
—
—
—
—
OR, odds ratio; CI, confidence interval; ECOG, Eastern Cooperative Oncology Group; WBC, white blood cell; HGB, hemoglobin; PLT, platelet.
OS Across and Within Molecular Cohorts With and Without TEEs Median duration of follow-up across all patients was 19.9 months, which varied among the molecular groups (ROS1þ 10.8 months, ALKþ 21.1 months, EGFRþ 21.2 months, and KRASþ 6.1 months). Median duration of follow-up for patients enrolled from institutions in the United States and in China was 25.3
months and 11.6 months, respectively. When the overall study cohort and each molecular group were stratified by the presence or absence of TEE within ±90 days of advanced lung cancer diagnosis, there was no evidence that OS was significantly different between those subgroups (Figs. 2A-E). After adjusting for stage, age at diagnosis, sex, race, smoking status, and molecular driver group, the survival between the
8 Ng et al
Journal of Thoracic Oncology
Vol.
-
No.
-
Figure 2. Kaplan Meier graphs of survival. (A-E) Survival in the presence or absence of TEE within 90 days of advanced lung cancer diagnosis, both for the entire cohort and for each molecular driver group. (F) Overall survival across the different molecular driver groups. The log-rank test was used to assess for differences in survival. TEE, thromboembolic events; CI, confidence interval.
subgroup with versus without TEE was not significantly different (hazard ratio: 1.35, 95% CI: 0.89–2.05, p ¼ 0.152). OS among the molecular driver groups was found to differ significantly using the log-rank test (p < 0.0001), with KRAS displaying the poorest survival (Fig. 2F).
Discussion This retrospective cohort study examined the odds of developing TEEs associated with ROS1þ lung cancer compared with other common molecular driver cohorts. Our results show that the odds of developing TEEs in ROS1þ lung cancer were significantly higher than in
---
2019
EGFRþ and KRASþ lung cancer. Although the numerical difference in TEE incidence between ROS1þ and ALKþ lung cancer (34.7% versus 22.3%) was significant in univariable analysis, this difference was not present in multivariable analysis (p ¼ 0.229). Two prior retrospective cohort studies of 264 and 99 ALKþ lung cancer patients reported a TEE incidence of 30% and 36%, respectively, much higher than historically reported rates in the general lung cancer population.12,13 Although our study reported a shorter period of follow-up for TEEs, it also reproduced the observation that ALKþ patients are at increased risk of developing TEEs. Among the other clinical variables, although ECOG status greater than or equal to 2, current/former smoker, pleural and lymph node metastases, a higher Khorana score, (including WBC >11 109/L, platelet count 350, and HGB <10 g/dL), Asian race and/or patients from China, and albumin less than 3.5 g/dL increased the odds of TEE in univariate analysis, only lymph node metastases and smoking status continued to be associated with increased odds of a TEE in multivariable analysis. Older age and hypertension emerged in multivariable analysis as being significantly associated with TEE. Despite prior studies showing an association between cancer-associated thrombosis and worse survival, our study did not show worse survival in the subgroups that developed a TEE, both in the overall study cohort and across molecular driver groups.6,12 Possible reasons for this different observation include insufficient duration of follow-up and potentially insufficient sample size. Considering the prolonged disease control feasible from targeted therapies among ROSþ, ALKþ, and EGFRþ patients, it is also possible that subsequent TEE propensity and its associated risks on survival may be abrogated by highly active anti-cancer therapy. High TEE rates in ROS1þ and ALKþ lung cancer raises the possibility of an etiological mechanism of prothrombotic risk associated with certain molecular drivers. Because ROS1þ and ALKþ are gene fusions that are phylogenetically related, clotting risk may be mediated by specific signaling emanating from these oncogenic abnormalities. Although elevated TEE rate in both groups may be related to treatment with a common drug (e.g., crizotinib), TEEs in both groups occurred primarily at the time of diagnosis, suggesting that commencement of therapy was unlikely a relevant factor. However, this question is worthy of exploration in future projects. This is the largest evaluation of TEEs in ROS1þ lung cancer patients to date, and the only study that we are aware of to simultaneously compare the odds of TEE across multiple molecular driver groups. Directly
ROS1 and Elevated Risk of Thromboembolic Events
9
comparing TEE risk between ROS1þ, ALKþ, EGFRþ, and KRASþ patients minimized sampling bias seen in cross-trial comparisons or when comparing study results against historical control cohorts. We controlled for other clinicopathologic risk factors that may have confounded TEE risk in this study by incorporating multivariable logistic regression. A ROS1/ALK/EGFR/ KRAS-wild type cohort was not included because this is, by definition, not a uniform population. Limitations of this study other than the retrospective study design include the imbalance in the number of patients included from China and the United States in each molecular cohort. However, race was highly correlated to country of enrollment (86% Asians from China and 100% Caucasians from the United States), and was not significantly associated with TEE in multivariable analysis. More than 10% of records were missing information on albumin concentration, ECOG score, BMI, Khorana score, WBC count, HGB concentration, and platelet count, and we were unable to incorporate them into multivariable analysis. There were fewer than 10 records in at least one of the comparator subgroups for congestive heart failure, chronic kidney disease, stage at diagnosis (IIIBþ versus
Acknowledgments Drs. Camidge, Doebele, and Aisner were supported in part by the University of Colorado Lung Cancer Specialized Program of Research Excellence (P50CA058187). Mr. Smith and Dr. Gao were supported by the Colorado Cancer Center Support Grant (P30CA046934). This study was not influenced by any of the funding bodies.
Supplementary Data Note: To access the supplementary material accompanying this article, visit the online version of the Journal of Thoracic Oncology at www.jto.org and at http://doi. org/10.1016/j.jtho.2018.12.001.
10 Ng et al
Journal of Thoracic Oncology
References 1. Timp JF, Braekkan SK, Versteeg HH, et al. Epidemiology of cancer-associated venous thrombosis. Blood. 2013;122:1712–1723. 2. Gade IL, Brækkan SK, Næss IA, et al. The impact of initial cancer stage on the incidence of venous thromboembolism: the Scandinavian Thrombosis and Cancer (STAC) Cohort. J Thromb Haemost. 2017;15:1567–1575. 3. Braekkan SK, Borch KH, Mathiesen EB, et al. Body height and risk of venous thromboembolism: the Tromsø Study. Am J Epidemiol. 2010;171:1109–1115. 4. Chew HK, Wun T, Harvey D, et al. Incidence of venous thromboembolism and its effect on survival among patients with common cancers. Arch Intern Med. 2006;166:458–464. 5. Khorana AA, Francis CW, Culakova E, et al. Thromboembolism is a leading cause of death in cancer patients receiving outpatient chemotherapy. J Thromb Haemost. 2007;5:632–634. 6. Sørensen HT, Mellemkjaer L, Olsen JH, et al. Prognosis of cancers associated with venous thromboembolism. N Engl J Med. 2000;343:1846–1850. 7. Horlander KT, Mannino DM, Leeper KV. Pulmonary embolism mortality in the United States, 1979-1998: an
8.
9.
10.
11.
12.
13.
Vol.
-
No.
-
analysis using multiple-cause mortality data. Arch Intern Med. 2003;163:1711–1717. Ogren M, Bergqvist D, Wåhlander K, et al. Trousseau’s syndrome — what is the evidence? A population-based autopsy study. Thromb Haemost. 2006;95:541–545. Blom JW, Osanto S, Rosendaal FR. The risk of a venous thrombotic event in lung cancer patients: higher risk for adenocarcinoma than squamous cell carcinoma. J Thromb Haemost. 2004;2:1760–1765. Blom JW, Doggen CJM, Osanto S, et al. Malignancies, prothrombotic mutations, and the risk of venous thrombosis. JAMA. 2005;293:715–722. American Joint Committee on Cancer. AJCC Cancer Staging Manual. 7th ed. Chicago, Illinois: American Joint Committee on Cancer; 2010. Zugazagoitia J, Biosca M, Grau JF, et al. Incidence, predictors and prognostic significance of thromboembolic events in patients with advanced ALK-rearranged NSCLCs. Presented at IASLC 18th World Conference on Lung Cancer. October 15-18, 2017; Yokohama, Japan. Zer A, Moskovitz M, Hwang DM, et al. ALK-rearranged non–small-cell lung cancer is associated with a high rate of venous thromboembolism. Clin Lung Cancer. 2017;18:156–161.