A Prognostic Assay to Identify Patients at High Risk of Mortality Despite Small, Node-Negative Lung Tumors

To the Editor: Low-dose computed tomography screening¹ may increase diagnoses of T1a node-negative non–small-cell lung cancers (NSCLC). One-quarter of these patients die within 5 years.² Maximizing the benefit of screening requires a reliable method to identify patients with high mortality risk. A molecular prognostic assay has been clinically validated for nonsquamous NSCLC, but performance of the assay was not studied in small node-negative tumors.³

A total of 1439 patients who had undergone resection of nonsquamous NSCLC in either the Kaiser Permanente Northern California system between 1998 and 2005 or at 1 of 3 institutions from the China Clinical Trials Consortium between 2000 and 2008 were enrolled in 2 original validation studies using consecutive sampling (follow-up end date: May 31, 2011).³ All patients with node-negative tumors of less than 2 cm from the Kaiser system (n = 155 patients) and the China Consortium (n = 114 patients) were included in this study.

The prognostic test measures the expression of 14 genes using quantitative polymerase chain reaction on RNA extracted from formalin-fixed paraffin-embedded specimens, and assigns patients to low-, intermediate-, and high-risk groups based on clinically validated cutoff values for a calculated risk score.³ Five-year survival was the primary end point. Vital status was obtained from multiple redundant sources including medical records, direct patient or family contact, the Kaiser Permanente Cancer Registry, California Death Records, and the Social Security Death Master File.

Kaplan-Meier analysis and the log-rank test for trend were used to evaluate the association between risk category and mortality. Cox proportional hazards models were used to evaluate predictors of 5-year mortality; proportional hazards assumptions were validated using Schoenfeld residuals. Wald and nested likelihood ratio tests were used to evaluate significance. Time-dependent area under the receiver operating characteristic curves were used to evaluate the improvement in risk prediction vs conventional staging. A 2-sided α level of .05 was considered statistically significant. Analyses were conducted using R version 2.12.2 (R Foundation for Statistical Computing) and Stata version 11 (StataCorp).

Waiver of informed consent was approved by the institutional review boards of the University of California, San Francisco, and Kaiser Permanente Northern California Division of Research; written informed consent was obtained from Chinese patients at the time of surgery, and the study was conducted in compliance with institutional review board policies at the Chinese institutions.

The mean (SD) age of patients was 62.4 (10.4) years; median follow-up among survivors was 74.4 months (interquartile range, 44.5-104 months), and 5-year mortality was 28.6%. Ninety-two patients (34.2%) were identified as high risk by the prognostic assay; survival was significantly different among the high-risk group (52.3%; 95% CI, 41.1%-62.4%), the intermediate-risk group (69.1%; 95% CI, 56.8%-78.6%), and the low-risk group (83.0%; 95% CI, 72.8%-89.7%) (P < .001; Figure).

Significant differences in survival also were observed among patients with the 26 tumors that were 1.0 cm or smaller in the low-risk group (100%; 95% CI not calculated), the intermediate-risk group (76.2%; 95% CI, 33.2%-93.5%), and the high-risk group (33.3%; 95% CI, 4.6%-67.6%) (P < .001). High-risk categorization was a significant predictor of mortality (Table) and remained significant when analyzed independently in the Kaiser cohort (hazard ratio, 3.31 [95% CI, 1.51-7.24]; P = .001) and in the Chinese cohort (hazard ratio, 3.33 [95% CI, 1.00-11.08]; P = .046). Adding risk category to a tumor size cutoff of 1.5 cm increased the C statistic from 0.57 to 0.68 (P < .001 by Wilcoxon rank sum test).

Current practice suggests consideration of adjuvant chemotherapy in high-risk stage I tumors,⁴ but no guidelines aid in the identification of high-risk T1a tumors. The early mortality of 25% in patients with T1a tumors indicates a high rate of undetected metastasis, and adjuvant chemotherapy is increasingly effective as the risk of metastasis and mortality increases.⁵ Trials that have relied on conventional staging, however, failed to demonstrate benefit in stage Ia disease,⁵ while a retrospective analysis in patients with early-stage tumors suggested that high-risk patients identified by prognostic gene signatures benefit from adjuvant chemotherapy.⁶

Limitations of the present study include subgroup analysis of a previously blinded validation study and a small number of tumors that were smaller than 1 cm. Despite these limitations, these data suggest the potential clinical utility of a new prognostic assay in the postoperative management of node-negative T1a disease. The identification of high-risk patients may further maximize the benefit of early detection of T1a node-negative tumors through low-dose computed tomography screening.

Author Contributions: Dr Mann had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Kratz, He, Jablons, Mann.

Acquisition of data: Van Den Eeden, He, Jablons, Mann.

Analysis and interpretation of data: Kratz, Van Den Eeden, He, Jablons, Mann.

Drafting of the manuscript: Kratz, Mann.

Critical revision of the manuscript for important intellectual content: Kratz, Van Den Eeden, He, Jablons, Mann.

Statistical analysis: Kratz, Mann.

Obtained funding: Mann.

Administrative, technical, or material support: Kratz, Van Den Eeden, He, Mann.

Study supervision: Jablons, Mann.

Funding/Support: Funded by private endowments to the University of California, San Francisco, Thoracic Oncology Laboratory and by Pinpoint Genomics Inc.

Role of the Sponsors: The University of California, San Francisco, Thoracic Oncology Laboratory funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript. Pinpoint Genomics collected data for the study, but otherwise had no role in the design and conduct of the study; management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript.