|Home | About | Journals | Submit | Contact Us | Français|
Interindividual variation in genetic background may influence the response to chemotherapy and overall survival for patients with advanced-stage non–small cell lung cancer (NSCLC).
To identify genetic variants associated with poor overall survival in these patients, we conducted a genome-wide scan of 307 260 single-nucleotide polymorphisms (SNPs) in 327 advanced-stage NSCLC patients who received platinum-based chemotherapy with or without radiation at the University of Texas MD Anderson Cancer Center (the discovery population). A fast-track replication was performed for 315 patients from the Mayo Clinic followed by a second validation at the University of Pittsburgh in 420 patients enrolled in the Spanish Lung Cancer Group PLATAX clinical trial. A pooled analysis combining the Mayo Clinic and PLATAX populations or all three populations was also used to validate the results. We assessed the association of each SNP with overall survival by multivariable Cox proportional hazard regression analysis. All statistical tests were two-sided.
SNP rs1878022 in the chemokine-like receptor 1 (CMKLR1) was statistically significantly associated with poor overall survival in the MD Anderson discovery population (hazard ratio [HR] of death = 1.59, 95% confidence interval [CI] = 1.32 to 1.92, P = 1.42 × 10−6), in the PLATAX clinical trial (HR of death = 1.23, 95% CI = 1.00 to 1.51, P = .05), in the pooled Mayo Clinic and PLATAX validation (HR of death = 1.22, 95% CI = 1.06 to 1.40, P = .005), and in pooled analysis of all three populations (HR of death = 1.33, 95% CI = 1.19 to 1.48, P = 5.13 × 10−7). Carrying a variant genotype of rs10937823 was associated with decreased overall survival (HR of death = 1.82, 95% CI = 1.42 to 2.33, P = 1.73 × 10−6) in the pooled MD Anderson and Mayo Clinic populations but not in the PLATAX trial patient population (HR of death = 0.96, 95% CI = 0.69 to 1.35).
These results have the potential to contribute to the future development of personalized chemotherapy treatments for individual NSCLC patients.
Although platinum-based chemotherapy is the main treatment for non–small cell lung cancer (NSCLC), most patients either do not respond to the therapy or develop resistance. Variation among patients might account for at least some of the variability in response and response duration.
A genome-wide scan for genetic variants associated with decreased overall survival was performed in a population of NSCLC patients who received platinum-based chemotherapy with or without radiation. Variants highly statistically significantly associated with decreased overall survival in the initial discovery population were validated in two independent patient populations as well as in pooled analyses.
A genetic variation in the chemokine-like receptor 1 (CMKLR1) gene was statistically significantly associated with decreased overall survival in the three individual populations as well as in pooled analyses.
NSCLC patients carrying a genetic variation in CMKLR1 may not respond to platinum-based chemotherapy. This study demonstrates an analytical approach to identify genetic factors associated with clinical outcomes that could be used to develop refined treatment strategies for these patients.
There are no previous reports of CMKLR1 or its protein being associated with lung cancer. Further studies are necessary to confirm a role and elucidate a mechanism for this G-protein-coupled receptor in NSCLC and resistance to platinum-based chemotherapeutic agents.
From the Editors
Lung cancer causes approximately 28% of cancer-related deaths per year in the United States and has remained the leading cause of all cancer deaths for the past decade with a 5-year survival rate of 15% (1). Non–small cell lung cancer (NSCLC) accounts for more than 80% of all lung cancers and is often diagnosed at an advanced stage. Disease stage and performance status are two of the most important clinical factors used to determine prognosis and guide treatment options for NSCLC patients. Platinum-based chemotherapy is the main treatment option for advanced-stage NSCLC patients and has demonstrated improved overall survival (2,3). Unfortunately, response to platinum-based chemotherapy varies among patients with similar clinical characteristics (4). Therefore, identification of biomarkers that can better predict a patient’s clinical outcome might prove helpful in guiding the physician in the selection of an optimal treatment regimen.
Germline genetic variations, such as single-nucleotide polymorphisms (SNPs), have attracted much attention as potential predictors of overall survival for NSCLC patients treated with platinum-based chemotherapy (2,5–8). A majority of these studies have applied a candidate gene approach requiring a priori knowledge of SNPs and gene function. For example, studies focused on the association between genetic variants in DNA repair pathway genes and clinical outcome have assumed that suboptimal DNA repair capacity influences individual responses to chemotherapy and overall survival. However, the results have often been conflicting and difficult to replicate (8).
Recently, several genome-wide association studies have been completed with the aim of identifying genetic variants influencing the risk of lung cancer (9–13). The availability of the data from the patient population of our previous genome-wide association studies for lung cancer risk together with data from additional patients selected from our ongoing epidemiological lung cancer study at MD Anderson Cancer Center provided us with the opportunity to comprehensively examine genome-wide genetic data to identify common genetic variants associated with overall survival for NSCLC. To construct a relatively homogeneous treatment regimen, this study was restricted to patients who did not receive surgery and were treated with platinum-based chemotherapy with or without radiation. We implemented a three-stage study design with a fast-track replication for 60 selected SNPs in an independent NSCLC patient cohort at Mayo Clinic followed by further validation for seven SNPs in NSCLC patients who received cisplatin and docetaxel as part of the PLATAX clinical trial.
All participants were selected from 1235 newly diagnosed histologically confirmed lung cancer patients. Of these 1235 patients, 1154 patients from our recent genome-wide association studies of lung cancer risk (9) and an additional 81 lung cancer patients from an ongoing epidemiological lung cancer study at the University of Texas MD Anderson Cancer Center were included in the analysis. All patients were recruited from January 1, 1995, to December 30, 2007, and had a median lag time of 34 days between the date of initial diagnosis and the start of chemotherapy. To be eligible for this study, patients had to be white ever-smokers with stage III or IV NSCLC without surgery or other prior therapy and treated at MD Anderson Cancer Center with first-line platinum-based chemotherapy with or without radiotherapy. In total, 327 NSCLC patients were selected for inclusion in the discovery phase. We also performed a separate analysis on 213 NSCLC patients with incurable stage IIIB (wet) and IV disease. All study participants gave written informed consent, and the study was approved by the Institutional Review Board of the University of Texas MD Anderson Cancer Center.
Of the 945 lung cancer patients who received platinum-based chemotherapy recruited from the Mayo Clinic (14), we applied the same eligibility criteria as was used in the discovery step and identified 315 NSCLC patients who met the criteria outlined above. Patients were recruited from January 1, 1997, to December 30, 2008, and had a median lag time of 25 days between the date of initial diagnosis and the start of chemotherapy. All study participants gave written informed consent, and the study was approved by the Institutional Review Board of the Mayo Foundation.
Specimens from 420 NSCLC stage III and IV patients who received chemotherapy (cisplatin and docetaxel) within the pharmacogenomic, open-label, single-arm multicenter PLATAX trial were included in the validation analysis. Lag time for each patient was less than 1 month. Specimen collection was performed as part of a collaborative study under a Grupo Espanol de Cancer de Pulmon (Spanish Lung Cancer Group) Clinical Research Ethics Committee–approved protocol involving multiple institutions in Spain (a complete list of participating Spanish hospitals is found in the Notes). All patients provided written informed consent and de-identified specimens, and patient data were sent to the University of Pittsburgh (Pittsburgh, PA) for genotyping analyses under a separately approved Institutional Review Board protocol.
Genotypes were generated using Illumina’s HumanHap300 BeadChip (San Diego, CA) for the 1154 patients included in our previous genome-wide association studies of lung cancer risk (9). We carried out genotyping for an additional 81 patients from MD Anderson Cancer Center using Illumina’s HumanHap317 BeadChip. The analysis focused on 307 260 SNPs that were included in both chips and passed quality control filters including call rate of at least 95% and minor allele frequencies of at least 0.01. All 1235 specimens had call rates greater than 95%, reported gender consistent with X chromosome heterozygosity, and passed additional quality control tests for duplicate sample detection and outlier sample detection implemented in the software package PLINK (version 1.03, http://pngu.mgh.harvard.edu/~purcell/plink/) (15). Genotyping for the Mayo Clinic validation population was performed in the Mayo Clinic Genomic Shared Resources facilities using the SNPstream genotyping platform (Beckman Coulter, Fullerton, CA) and TaqMan assays (Applied Biosystems, Foster City, CA) according to the manufacturer’s instructions. Genotyping was performed at the University of Pittsburgh using a custom Illumina GoldenGate panel (Illumina) and TaqMan assays following standard protocols.
We assessed three genetic models of inheritance (dominant, recessive, and additive) using the discovery dataset for each SNP by multivariable Cox proportional hazard regression analysis. Adjustments for age (continuous), sex (male or female), pack-years (continuous), clinical stage (IIIA, IIIB [dry], IIIB [wet], or IV), and pretreatment performance status (0, 1, or 2–4) were made with the use of STATA software (version 10; STATA Corporation, College Station, TX). The model with the smallest P value was used to measure the statistical significance of the association between each SNP and overall survival for the genome-wide SNP data. Only the dominant model was considered when the rare homozygous genotype was less than 5% in both living and deceased patients. Overall survival time was defined as the time from the date of chemotherapy start to the date of death or the date of last follow-up, whichever came first. Kaplan–Meier curves and log-rank tests were used to calculate the survival difference associated with individual genotypes. All statistical tests were two-sided.
All patients from the discovery population were used to determine population substructure using the software packages PLINK and EIGENSTRAT (version 3.0, http://genepath.med.harvard.edu/~reich/Software.htm) (16). The 1235 patients were classified into 27 strata on the basis of genetic similarity using PLINK clustering analysis. When restricted to the 327 patients analyzed in the discovery phase, 22 strata were identified. Analysis using EIGENSTRAT software and principle component analysis identified the five top eigenvalues from the 1235 available eigenvalues. Multivariable Cox regression analyses were used to allow for potential population substructure by two separate analyses (one with the PLINK identified strata and the other with the five eigenvectors from EIGENSTRAT).
We divided the data equally using two approaches—by randomly splitting the data into training and test sets and by assigning alternating samples into training and test sets according to the date of sample collection. For SNPs showing statistically significant differences in patients with stage III and IV disease, a SNP was selected for validation in the Mayo Clinic validation population if it was statistically significantly associated with overall survival in the training and testing sets generated by both approaches or if it was among the top 20 statistically significant SNPs identified in the training and testing sets generated by either approach. For SNPs statistically significantly associated with overall survival by analysis of the subgroup of specimens from patients with stage IIIB (wet) and IV disease, a SNP was selected for validation if it was one of the top 20 SNPs with a statistically significant association with overall survival in both the training and testing sets, generated by either approach (Supplementary Table 1, available online).
The analysis of specimens from the Mayo Clinic validation studies was performed using multivariable Cox regression analysis with adjustment variables identical to those of the initial scan. Seven SNPs showing consistent association with overall survival, resulting in the corresponding hazard ratios in the MD Anderson and Mayo Clinic studies, were selected for further replication in the PLATAX clinical trial samples (Supplementary Table 1, available online). Multivariable Cox regression analysis with adjustments for age, sex, clinical stage, and pretreatment performance status (same as above definitions) was used to investigate an association between a SNP and overall survival. Smoking history was not available for these patients, so the data were not adjusted for this variable. To summarize results for the discovery set and the two validation studies, we performed pooled analysis to obtain the summary hazard ratio and 95% confidence interval (CI). We tested the proportional-hazards assumption on the basis of Schoenfeld residuals. Data from MD Anderson, Mayo Clinic, and the PLATAX clinical trial satisfied the proportionality assumption with P = .21, P = .48, and P = .15, respectively, when rs1878022 and all covariates were included in the analysis.
Haploview software (version 4.1, http://www.broad.mit.edu/mpg/haploview/) was used to determine pair-wise linkage disequilibrium structure across the genomic region under study (17). To impute SNPs in the region containing susceptibility loci, we used MACH software (http://www.sph.umich.edu/csg/abecasis/MaCH) and haplotype information for the multimarker tags from the International HapMap Project release 22, human genome build 36 (www.hapmap.org).
We constructed receiver operating characteristic curves and calculated the area under the curve (AUC) to evaluate the specificity and sensitivity of predicting 1-year survival for the pooled dataset by clinical and epidemiological variables and by the combination of clinical, epidemiological, and genetic variables. We used 1000 bootstrapping samples to compute a 95% bias-corrected confidence interval for the difference of the AUCs between the two models to determine the statistical significance of adding genetic markers to the model.
The initial genome-wide scan was performed for 307 260 SNPs that passed strict quality control measures. In the discovery phase, we observed 31 751 SNPs that met our selection criteria (P < .05; among them, 20 with P < 10−5 and one with P < 10−6) for the analysis focused on stage III and IV NSCLC patients (Figure 1, A). When the analysis was restricted to only stage IIIB (wet) and IV patients, 30 258 SNPs (P < .05; among them, 15 with P < 10−5 and three with P < 10−6) were identified (Figure 1, B). In a sensitivity analysis, we performed two separate analyses (one with clusters from PLINK and the other with eigenvectors from EIGENSTRAT) to assess the potential effect of population substructure using multivariable Cox regression analysis. The data indicated that the patient population is relatively ethnically homogeneous and the observed associations were not driven by potential population substructure (Supplementary Table 2, available online).
To determine which associations identified in the discovery phase were robust, we identified 60 SNPs (Supplementary Table 1, available online) to perform a fast-track validation study using an independent NSCLC patient cohort from the Mayo Clinic following the same eligibility criteria as the discovery population (Table 1). As summarized in Table 2, under the additive model, rs1878022 was statistically significantly associated with poor overall survival in the MD Anderson discovery population (hazard ratio [HR] of death = 1.59, 95% CI = 1.32 to 1.92, P = 1.42 × 10−6), and the difference in survival time was dependent on the number of variant alleles carried by the patient (P = 0.001) (Figure 2). This association reached borderline statistical significance in the Mayo Clinic validation population (HR of death = 1.16, 95% CI = 0.95 to 1.41, P = .15) but was associated with survival time differences in patients with different genotypes (homozygous common genotype [TT] vs heterozygous genotype [TC] vs homozygous variant genotype [CC], P = .04) (Figure 2).
Another candidate SNP, rs10937823, was associated with statistically significantly poorer survival in the MD Anderson discovery dataset (HR of death = 2.40, 95% CI = 1.67 to 3.43, P = 1.80 × 10−6), the Mayo Clinic validation dataset (HR of death = 1.45, 95% CI = 1.02 to 2.08, P = .04), and the combined MD Anderson and Mayo Clinic dataset (HR of death = 1.82, 95% CI = 1.42 to 2.33, P = 1.73 × 10−6). In the combined dataset, the median survival time for individuals with the common homozygous genotype (16.05 months) was statistically significantly longer than the mean survival time for those carrying the variant-containing genotypes (10.72 months, P = 6.76 × 10−5).
In a sensitivity analysis, we repeated the analysis for rs1878022 by restricting to all patients who had chemotherapy treatment before 2004, 2005, and 2006 (ie, those patients who had at least 5, 4, and 3 years of follow-up). The associations between overall survival and rs1878022 were very similar to the overall analysis, and for the Mayo study, we observed stronger association with poor overall survival when the follow-up was longer. We also assessed the effect on overall survival of rs1878022 in patients who received radiotherapy as part of their treatment regimen. In both the MD Anderson and Mayo Clinic populations, the results were similar between the two treatment groups (MD Anderson: chemotherapy and radiotherapy, HR of death = 1.50, 95% CI = 1.16 to 1.94, vs chemotherapy only, HR of death = 1.64, 95% CI = 1.22 to 2.20; Mayo Clinic: chemotherapy and radiotherapy, HR of death = 1.17, 95% CI = 0.90 to 1.52, vs chemotherapy only, HR of death = 1.23, 95% CI = 0.89 to1.69) with overlapping 95% confidence intervals.
To further provide evidence that these genetic loci are associated with poor overall survival in patients receiving platinum-based chemotherapy, we analyzed rs1878022 and rs10937823 in patients enrolled in the PLATAX clinical trial. Under the dominant model, rs10937823 was non-statistically significantly associated with poor overall survival in the PLATAX validation population (HR of death = 0.96, 95% CI = 0.69 to 1.35, P = .84). Also, rs1878022 was validated in the PLATAX population (HR of death = 1.23, 95% CI = 1.00 to 1.51, P = .05) (Table 2). However, a statistically significant difference in median survival times (log-rank P = .04) was evident based on the number of variant alleles. Patients with the common genotype had a mean survival time of 10.76 months compared with 8.03 and 8.91 months for patients carrying the heterozygous genotype and the variant genotype, respectively (Figure 2). This association was stronger in a pooled analysis of the two validation datasets (HR of death = 1.22, 95% CI = 1.06 to 1.40, P = .005) and was highly statistically significant overall using data from all three studies (HR of death = 1.33, 95% CI = 1.19 to 1.48, P = 5.13 × 10−7).
SNP rs1878022 is located on chromosome 12q23.3 within the intron of the chemokine-like receptor 1 gene (CMKLR1). Haplotype analysis of the genomic region surrounding rs1878022 indicated rs1878022 was not in high linkage disequilibrium with any neighboring SNPs, thus rs1878022 is not located in any major haplotype blocks within this region of the chromosome (Figure 3). Imputation of HapMap SNPs within this genomic region confirmed these findings, and no other SNPs demonstrated a statistically significant association with poor overall survival in NSCLC patients as calculated using this method.
We created two prediction models based on 1-year survival for the pooled dataset to investigate the clinical relevance of rs1878022. The AUC based on the clinical and epidemiological variables (age, sex, clinical stage, and pretreatment performance status) was 69.1%, demonstrating reasonable predictive power. The addition of the single SNP to the model increased the AUC to 70.5%. Results based on 1000 bootstrapping samples showed that the distribution of the difference of the AUCs has a 95% bias-corrected confidence interval of 0.4% to 3.4%, indicating a statistically significant improvement in prediction of decreased overall survival in NSCLC patients after adding rs1878022 to the model.
To better understand the genetic factors modulating response to platinum-based therapy in stage III and IV patients, we used a three-stage analytical approach that took advantage of two of the largest lung NSCLC patient cohorts in the United States and patients from a large multicenter clinical trial. Our data identified a common genetic polymorphism in CMKLR1 that is statistically significantly associated with decreased overall survival. An additional candidate SNP was also identified that was non-statistically significantly associated with an increased risk of poor outcome in our study populations but may reach statistical significance after further analyses in a larger patient population. This is the first genome-wide study, to our knowledge, taking an unbiased noncandidate gene–driven approach to investigate the genetic factors influencing overall survival for advanced-stage NSCLC patients receiving platinum-based chemotherapy.
Previous candidate gene studies have indicated that genetic variation in genes within the platinum drug action pathway may be associated with response to chemotherapy (18). However, the observed effect of each individual SNP is modest with a stronger association in combined analyses of multiple risk alleles (8). Linkage studies using lymphoblastoid cell lines from pedigrees of Caucasian individuals have indicated that approximately 47% of the variation in susceptibility to cisplatin-induced cytotoxicity is because of heritable factors (19). These results indicate that there are unidentified genetic factors playing a major role in determining patient responses to platinum-based chemotherapeutics. Genome-wide association studies of cisplatin cytotoxicity in lymphoblastoid cell lines have also been performed and identified several novel associations between genetic variation and cytotoxicity (20,21). Similar to the findings of our study, the variants identified in these previous reports were located within genes not previously considered as candidates for chemotherapy response.
The validated SNP rs1878022 is an intronic variant located within CMKLR1 on human chromosome 12. This gene encodes for a seven transmembrane G-protein coupled receptor (also known as the ChemR23 protein) that is involved in several cellular pathways, including inflammation, adipogenesis, and osteoblastogenesis (22–25). Although primarily expressed on dendritic cells and macrophages, high CMKLR1 protein expression has been previously reported in lung tissue (25). Binding to this receptor by its ligands chemerin and resolvin E1 has been shown to activate multiple downstream signaling pathways including phosphoinositide-3 kinase/AKT, mitogen-activated protein kinase (MAPK), and extracellular signal-related kinase 1 and 2 (ERK1/2) (24–27), and more recently, endothelial cell expression and angiogenesis (27). Expression of CMKLR1 was found to be higher in normal mucosa of patients with esophageal squamous dysplasia patients whose disease had progressed (28). However, to date, no studies have been published reporting an association between either CMKLR1 or its protein and lung cancer. Further studies are needed to elucidate the function of CMKLR1 and its protein specifically within the context of lung cancer and to determine how genetic variation within this gene modulates these functions resulting in differences in advanced-stage NSCLC patients’ responses to platinum-based chemotherapeutics.
The SNP rs10937823 was statistically associated with decreased overall survival in the MD Anderson and Mayo Clinic patient populations, but not associated with decreased overall survival in patients enrolled in the PLATAX clinical trial. This genetic variant is located within an intron, a gene encoding for the poorly characterized SORCS2 protein. SORCS2 contains a VPS10 domain that is known to have a role in intracellular trafficking and lysosomal processing (29). Other members of this gene family with VPS10 domains, such as sortilin (SORT1), are known to be responsible for both trafficking and regulation of neurotrophin signaling (30). However, SORCS2 is mainly located on the cell surface and is not believed to have a major role in cellular trafficking (31), in part because its VPS10 domain differs from that of other members of the family (32). SORCS2 is highly expressed in the brain and central nervous system during development (33,34), and protein expression has been detected in the developing and adult lung (32,34). Although the exact function of this receptor is unclear, several gene expression studies have hypothesized a role for SORCS2 in breast cancer clinical outcome (35) and lymphatic metastasis after treatment for oral carcinoma (36). Also, SORCS2 gene expression was strongly related to sensitivity to chemotherapeutics in human gastric cancer cell lines in vitro (37). Our data indicate that rs10937823 may be associated with overall survival in NSCLC patients. Additional studies are necessary to determine the function of SORCS2 in NSCLC and further establish rs10937823 as a candidate SNP for poor overall survival in this patient population.
Somatic alterations in tumor tissues are immensely useful in characterizing the molecular changes that occur during cancer development and may aid in selecting the appropriate targeted therapy for those patients with specific mutations (38). However, platinum-based chemotherapy is often used to treat patients for whom surgery is not an option because of disease severity or other preexisting conditions, making it difficult to screen for somatic alterations if tumor specimens are limited. Because of these characteristics, it may be difficult to screen for somatic alterations due to lack of tumor availability and underscores the need to identify germline genetic variations that can be used as prognostic and predictive markers. Furthermore, platinum-based chemotherapy is also used in the adjuvant setting before surgery. Germline SNPs are stable markers that do not vary with disease severity and can be screened for using DNA isolated from a blood sample before treatment. Identification of these common genetic polymorphisms that can reliably predict response to chemotherapy has the potential to guide personalized therapeutic prediction.
Our study has two limitations that are major concerns inherent of any genome-wide approach—multiple comparisons and the presence of false-positive findings. To address these issues, we used a three-stage study design in multiple independent populations using a progressively smaller panel of candidate SNPs. The study design allows for validation of any findings from the large discovery population that is more prone to false-positive findings, SNPs that remain statistically significantly associated with decreased overall survival across all study populations have a high likelihood of being true-positive findings, thus reducing the need for strict multiple comparison correction. Future studies to replicate our data in other patient populations with similar treatment regimens are necessary to confirm the conclusions of our study.
Because of the widespread use of platinum-based chemotherapy to treat advanced-staged NSCLC, it is difficult to establish SNPs as pharmacogenetic markers specific for response to treatment or more general prognostic markers. This distinction would only be possible with a parallel analysis in a platinum-based chemotherapy-naive patient population. Nevertheless, by focusing on a select homogenous population of NSCLC patients, we are able to determine that a statistically significant increase in risk of death is present in patients with specific genetic variants when treated with platinum-based chemotherapy. There was only a moderate statistically significant increase in risk of death associated with either of the two candidate SNPs. However, the combinatorial effect of multiple SNPs and potential unidentified gene–gene interactions may result in a stronger association between genetic variants and risk of death in NSCLC. This has been previously demonstrated for risk models developed for prostate and breast cancers (39,40). Together with epidemiology, demographic, clinical, and other genetic risk factors, our findings have the potential to influence personalized treatment by identifying patients who would respond best to a specific treatment regimen.
Advanced-stage NSCLC patients are commonly treated with platinum-based chemotherapy. Although this course of treatment has been shown to increase overall survival, many patients develop resistance or dose-limiting side effects to these agents and even with therapy, the 1-year survival rate is 29% (41). To the best of our knowledge, this is the first genome-wide study to assess the genetic factors influencing overall survival for advanced-stage NSCLC patients receiving platinum-based chemotherapy. This study provides additional biomarkers that can be integrated with known epidemiological, clinical, and genetic risk factors to potentially identify patients who are more likely to respond to chemotherapy, thereby helping the physician develop individualized treatment regimens.
The study was supported in part by National Institutes of Health (R01CA111646 and P50CA070907 to X.W., R01CA55769 to M.R.S., R01CA121197 to C.I.A., R01CA084354 and R01CA115857 to P.Y., P50CA090440 and 5U01CA114771 to M.R.).
M. Romkes and P. Yang contributed equally to this work.
The participating institutions from Spain include Hospital Doce de Octubre, Madrid; Xeral Cies de Vigo, Vigo; Catalan Institute of Oncology, Badalona; Clinica Sagrado Corazon, Sevilla; Hospital General Yague, Burgos; Hospital La Princesa, Madrid; Catalan Institute of Oncology, Girona; Hospital de Leon, Leon; Hospital General de Valencia, Valencia; Hospital Arnau de Vilanova, Valencia; Catalan Insitute of Oncology, Bellvitge; Hospital Clinico San Carlos, Madrid; Hospital General de Asturias, Oviedo; Hospital Alcoy, Alicante; Autonomous University of Madrid, Madrid.
The sponsors had no role in the design or conduct of the study; collection, management, analysis, or interpretation of the data; and preparation, review, or approval of the article.