STK11 Inactivation Predicts Rapid Recurrence in Inoperable Early-Stage Non–Small-Cell Lung Cancer

Abstract

PURPOSE

Molecular factors predicting relapse in early-stage non–small-cell lung cancer (ES-NSCLC) are poorly understood, especially in inoperable patients receiving radiotherapy (RT). In this study, we compared the genomic profiles of inoperable and operable ES-NSCLC.

MATERIALS AND METHODS

This retrospective study included 53 patients with nonsquamous ES-NSCLC (stage I-II) treated at a single institution (University of Chicago) with surgery (ie, operable; n = 30) or RT (ie, inoperable; n = 23) who underwent tumor genomic profiling. A second cohort of ES-NSCLC treated with RT (Stanford, n = 39) was included to power clinical analyses. Prognostic gene alterations were identified and correlated with clinical variables. The primary clinical end point was the correlation of prognostic genes with the cumulative incidence of relapse, disease-free survival, and overall survival (OS) in a pooled RT cohort from the two institutions (N = 62).

RESULTS

Although the surgery cohort exhibited lower rates of relapse, the RT cohort was highly enriched for somatic STK11 mutations (43% v 6.7%). Receiving supplemental oxygen (odds ratio [OR] = 5.5), 20+ pack-years of tobacco smoking (OR = 6.1), and Black race (OR = 4.3) were associated with increased frequency of STK11 mutations. In the pooled RT cohort (N = 62), STK11 mutation was strongly associated with inferior oncologic outcomes: 2-year incidence of relapse was 62% versus 20% and 2-year OS was 52% versus 85%, remaining independently prognostic on multivariable analyses (relapse: subdistribution hazard ratio = 4.0, P = .0041; disease-free survival: hazard ratio, 6.8, P = .0002; OS: hazard ratio, 6.0, P = .022). STK11 mutations were predominantly associated with distant failure, rather than local.

CONCLUSION

In this cohort of ES-NSCLC, STK11 inactivation was associated with poor oncologic outcomes after RT and demonstrated a novel association with clinical hypoxia, which may underlie its correlation with medical inoperability. Further validation in larger cohorts and investigation of effective adjuvant systemic therapies may be warranted.

INTRODUCTION

Lung cancer is the most common cause of cancer-related death.¹ Although most patients with non–small-cell lung cancer (NSCLC) were diagnosed with advanced disease in past decades, the recent implementation of risk-stratified, low-dose screening computed tomography on a population-wide level has resulted in more patients being diagnosed with early-stage disease amenable to curative local therapies.^2,3 Medically fit patients are most commonly treated with surgical tumor resection, whereas those deemed medically inoperable generally undergo stereotactic body radiotherapy (SBRT).⁴

CONTEXT

• Key Objective

• To examine the prognostic significance of genomic alterations in early-stage (stage I to II) non–small-cell lung cancer, including a novel cohort of patients receiving definitive radiotherapy.

• Knowledge Generated

• STK11 mutations were more prevalent in patients receiving definitive radiotherapy, compared with those receiving surgery. Increased frequencies of STK11 mutations were observed in patients receiving supplemental oxygen, with 20+ pack-years of tobacco smoking history, or of Black race. In a unique cohort of patients receiving definitive RT, STK11 inactivation was associated with rapid distant (rather than local) recurrence and worse overall survival.

• Relevance

• Since STK11 inactivation may be associated with resistance to immune checkpoint blockade, investigation of effective adjuvant systemic therapies may be warranted in this setting. Furthermore, the novel association with chronic hypoxia suggests that hypoxia may apply a selective pressure that promotes inactivation of STK11, such that medically inoperable patients are more prone to develop STK11-mutant tumors.

Although both surgical resection and SBRT provide an excellent likelihood for local tumor control, up to a quarter of patients experience distant metastatic failure, which generally results in death.^5-8 To date, clinical and molecular factors that predict relapse are poorly understood, and the role of additional therapies remains uncertain.⁹ Adjuvant cytotoxic chemotherapies with cisplatin-based regimens do not offer clear benefit for stage I tumors (≤ 4 cm in size) and may even lead to a worse overall survival (OS) for stage IA disease.¹⁰ The addition of immunotherapy and targeted therapies (for tumors with actionable mutations) improves disease-free survival (DFS) in early-stage resected NSCLC.^11-14 Comprehensive genomic profiling may help identify tumors susceptible to specific systemic therapies; however, existing studies have been primarily limited to surgically resected tumors.^15,16 Comparable studies in inoperable patients receiving definitive SBRT are lacking.

In this study, we characterized the genomic landscape of early-stage NSCLC (stage I and II) and present a comparative, next-generation sequencing (NGS) analysis of patients receiving surgery and patients receiving radiotherapy (RT). We identified gene variants associated with relapse and correlated prognostic genes with various clinical factors. Importantly, we propose that prognostic genes with differential prevalence in the surgery and RT groups may be correlated with underlying clinical features associated with operability.

MATERIALS AND METHODS

Study Design and Patients

The results for this retrospective cohort study were reported following the Reporting Recommendations for Tumor Marker Prognostic Studies guidelines.¹⁷ Comparative analyses of molecular profiles were performed in 53 patients diagnosed with early-stage (stage I and II, node-negative) nonsquamous NSCLC and treated at our institution (University of Chicago Medical Center, Chicago, IL) with either surgery (n = 30) or hypofractionated/stereotactic body RT (n = 23) who underwent NGS on primary tumor tissue between 2015 and 2020. NGS was performed using OncoPlus, a validated 1212-gene proprietary hybrid capture genomic sequencing assay (Data Supplement).¹⁸ Analyses were restricted to nonsquamous histologies because patients with squamous histologies did not routinely undergo molecular profiling during the study period. Demographic and clinical data were recorded with review of the electronic medical record. This study was reviewed and approved by the institutional review board. Because of the study's retrospective nature, a waiver for informed consent was provided.

To further power the association of prognostic gene variants with clinical factors and oncologic outcomes, primary analyses included a cohort of patients (n = 39) receiving RT with curative intent for early-stage nonsquamous NSCLC at Stanford Cancer Center (Stanford, CA).¹⁹ In the Stanford cohort, mutation status was available for five genes (KRAS, TP53, STK11, CDKN2A, and EGFR). For correlation of gene variants with clinical factors, a total pooled cohort of all patients was analyzed (N = 92). For correlation of gene variants with clinical outcomes in the RT group, analyses were limited to the pooled cohort of patients receiving RT at University of Chicago and Stanford (N = 62). Using a pooled cohort augmented external validity of our findings through avoiding bias from subsets of the population specific to a given medical center. Bidirectional institutional review board approval was obtained to share data between institutions.

Genomic Analyses

Processing of sequencing reads and determination of tumor mutational burden and programmed death-ligand 1 expression is described in the Data Supplement.^18,20-22 To screen potentially prognostic variants, pathogenic mutations in each gene were correlated with time to relapse (Data Supplement). To correct for multiple comparisons, q-values were computed using the Benjamini-Hochberg method with the goal of keeping the false discovery rate (FDR) < 0.05. Of the genes with q-value < 0.05, only one gene (STK11) met our threshold of > 10% pathogenic mutation prevalence for correlation with clinical outcomes. STK11 mutations analyzed in this study were predicted to be pathogenic, inactivating mutations.

Clinical Outcomes and Statistical Analyses

Because there was a significant imbalance in the prevalence of STK11 mutations in the RT and surgery groups, it was hypothesized that somatic STK11 mutations were correlated with underlying clinical features associated with operability. To investigate the association of various demographic and clinical variables with the presence of STK11 mutations in the sequenced tumors, univariable logistic regression was performed in the total pooled cohort (N = 92) including sex, age, race and ethnicity, Eastern Cooperative Oncology Group (ECOG) performance status, body mass index, diabetes mellitus status, supplemental oxygen, pack-years of tobacco smoking, forced expiratory volume in 1 second (FEV1; % predicted), diffusing capacity of the lung for carbon monoxide (DCLO; % predicted), and stage. Details regarding rationale for covariate inclusion are described in the Data Supplement.^23-28 Multivariable logistic regression was performed using backward selection (P < .10). The FEV1 % predicted (not assessed in 18 patients) and DLCO % predicted (not assessed in 31 patients) were excluded from multivariable analyses, and only univariable analyses were reported because they were not assessed in all patients.

The primary clinical analyses included the association of STK11 mutations with time to relapse, DFS, and OS in the pooled inoperable cohort receiving RT (N = 62). Primary clinical analyses were limited to the pooled RT cohort because STK11 mutations were only present in two patients receiving surgery. All relapses were classified as local, lobar, regional (hilar or mediastinal lymph nodes), or distant (including contralateral lung and outside of the thorax). Time to relapse was analyzed using the cumulative incidence function (with death as a competing risk) using the Fine-Gray model for univariable and multivariable analyses. DFS and OS were analyzed using the Kaplan-Meier method and the log-rank P values were reported (P < .05 deemed statistically significant). To account for demographic and clinical covariates, multivariable Cox proportional hazards models were computed. Detailed definition of clinical end points and statistical analyses are described in the Data Supplement.^29,30

RESULTS

Patient Characteristics and Outcomes by Treatment Type

From September 21, 2015, to April 13, 2020, 53 patients with nonsquamous early-stage non–small-cell lung cancer were treated definitively with either hypofractionated/stereotactic body RT (n = 23) or surgery (n = 30) and underwent NGS of tumor tissue at University of Chicago (median follow-up, 23.6 mo). Details regarding staging workup and RT regimens are described in the Data Supplement.³¹ All surgical patients underwent mediastinal lymph node dissection. Overall, patients receiving RT exhibited adverse clinical features associated with worse OS, including worse performance status, lower FEV1, increased utilization of supplemental oxygen, and greater pack-years of tobacco smoking (Table 1).

TABLE 1.

Patient Characteristics in University of Chicago Cohort by Treatment Type

Compared with patients who underwent surgery, patients receiving RT exhibited increased relapse, primarily driven by high rates of distant relapse as opposed to local failure (Fig 1). At 2 years, the cumulative incidence of relapse was 16.6% (95% CI, 5.1% to 33.9%) for surgery and 39.1% (95% CI, 19.9% to 58.0%) for RT, respectively (Fine-Gray P = .070). The 2-year cumulative incidence of local failure was 0% for surgery and 4.8% for RT, while the 2-year distant failure was 8.6% and 26.1%, respectively. At 2 years, the OS was 93.1% (95% CI, 75.1% to 98.2%) for surgery and 65.9% (95% CI, 41.1% to 82.2%) for RT (log-rank P = .22). Despite favorable local control in the RT group, the risk of distant relapse was significantly higher than expected, compared with our institutional surgical cohort and published outcomes for operable patients receiving SBRT.²⁹ In this context, we investigated whether there were biological differences in early-stage NSCLC between operable patients receiving surgery and inoperable patients receiving RT that could explain the differences in disease relapse.

FIG 1.

Clinical outcomes in the University of Chicago cohort in the medically inoperable group (patients receiving RT) and medically operable group (patients receiving surgery): (A) freedom from relapse and (B) overall survival. RT, radiotherapy.

Genomic Landscape of Early-Stage NSCLC by Treatment Type

The genomic landscape of early-stage NSCLC in the University of Chicago cohort (N = 53) is summarized in Figure 2. Pathogenic variants were detected in 60 genes and were most commonly present in KRAS (62%), TP53 (42%), STK11 (23%), CDKN2A (19%), and EGFR (15%). To screen for prognostic variants in the combined University of Chicago cohort, the freedom from relapse was analyzed via the Kaplan-Meier method for each of the 60 variants. Variants were considered potentially prognostic if their log-rank q-values were < 0.05 (ie, FDR < 0.05). These are listed in the Data Supplement. Of five potentially prognostic variants, only STK11 variants were present with sufficient frequency (n = 12, 23%) for further clinical analyses. STK11 also remained prognostic when screening variants by cumulative incidence of relapse with death as a competing risk (q = 0.035). Notably, STK11 mutations were present in only 2 (6.7%) patients receiving surgery and 10 (43%) patients receiving RT. STK11/KRAS comutations were present in 7 (13%), 1 (3.3%), and 6 (26%) patients in the overall, surgery, and RT cohorts, respectively. Neither tumor mutational burden nor programmed death-ligand 1 TPS varied with STK11 status (Data Supplement). In summary, with comprehensive molecular profiling of early-stage NSCLC, STK11 variants were associated with relapse and were present in approximately one quarter of patients, but almost exclusively limited to medically inoperable patients.

FIG 2.

Genomic landscape of early-stage NSCLC in the medically inoperable group (patients receiving RT) and medically operable group (patients receiving surgery). (A) OncoPrint demonstrating top 15 recurrently altered genes in each group. All gene alterations are reported (including nonpathogenic variants); however, frequencies to the left of the plot represent pathogenic alteration prevalence. (B) Frequencies of most common gene alterations in each cohort, counting only pathogenic variants. ECOG, Eastern Cooperative Oncology Group; PD-L1, programmed death-ligand 1; TMB, tumor mutation burden.

Clinical Features Associated With STK11 Variants

The associations of various demographic and clinical factors with STK11 variants were investigated in the combined University of Chicago and Stanford cohorts. Univariable associations are displayed in the Data Supplement, demonstrating that treatment with RT, treatment at University of Chicago, Black race, ECOG > 1, FEV1 % predicted, supplemental oxygen, and 20+ pack-years of tobacco smoking history were associated with increased frequency of STK11 mutations (Data Supplement). Multivariable logistic regression for presence of an STK11 mutation was performed with backward selection (P < .1), including sex, age, Black race, ECOG > 1, diabetes mellitus status, body mass index, supplemental oxygen receipt, 20+ pack-years of tobacco smoking history, and stage. The final model is displayed in Table 2. Supplemental oxygen (odds ratio [OR] 5.5, P = .027), 20+ pack-years of tobacco smoking history (OR 6.1, P = .03), and Black race (OR 4.3, P = .024) were all associated with increased frequency of STK11 variants in the corresponding tumors. Finally, these clinical markers of hypoxia and demographic factors did not appear to be correlated with each other (Data Supplement).

TABLE 2.

Multivariable Logistic Regression After Backward Selection (P < .1) for Presence of STK11 Variants in the Profiled Tumors

STK11 Mutation Associated With Relapse in Patients Receiving RT

We analyzed the impact of STK11 mutations on oncologic outcomes in the pooled cohort of patients receiving RT from University of Chicago (n = 23) and Stanford (n = 39), comprising 62 total patients (median follow-up, 18.9 mo). Analysis was limited to the RT cohort because only two patients in the surgery cohort exhibited STK11 mutations. Overall, STK11 mutations were more common in the University of Chicago cohort (43% v 8%, P = .002). Other demographic and clinical variables for each cohort were summarized in the Data Supplement.

In the pooled RT cohort, STK11 variants were strongly associated with time to relapse, DFS, and OS (Fig 3). The 2-year cumulative incidence of relapse was 20.0% (95% CI, 9.3% to 33.7%) for STK11 wild-type and 61.5% (95% CI, 30.8% to 81.8%) for STK11-mutant tumors (Fine-Gray P = .004). Likewise, the 2-year DFS was 71.2% (95% CI, 53.5% to 83.2%) for STK11 wild-type and 30.8% (95% CI, 9.5% to 55.4%) for STK11-mutant tumors (log-rank P = .0003). The 2-year OS was 85.4% (95% CI, 67.7% to 93.8%) for STK11 wild-type and 52.2% (95% CI, 18.9% to 77.6%) for STK11-mutant (log-rank P = .006). Within the individual institutional cohorts, STK11 variants remained associated with relapse (Data Supplement). When analyzing each of the factors associated with STK11 mutations (ie, supplemental oxygen status, 20+ pack-years of tobacco smoking history, and Black race) by mutation status, STK11 mutations remained associated with significantly worse time to relapse, although statistical comparisons were not feasible because of small subgroup sample sizes (Data Supplement).

FIG 3.

Clinical outcomes in the pooled University of Chicago and Stanford medically inoperable cohort by STK11 mutation status: (A) freedom from relapse, (B) disease-free survival, and (C) overall survival.

Multivariable regression analyses were performed to evaluate the effect of STK11 variants while controlling for multiple covariates, using a Fine-Gray model for time to relapse and Cox models for DFS and OS. Finalized models after backward selection (P < .1) are displayed in Table 3, while the full models are shown in the Data Supplement. STK11 variants exhibited a subdistribution hazard ratio (SHR) of 4.0 (95% CI, 1.5 to 10.2; P = .0041) for time to relapse. STK11 variants demonstrated a hazard ratio (HR) of 6.8 (95% CI, 2.5 to 18.3; P = .0002) for DFS, and an HR of 6.0 (95% CI, 1.3 to 27.8; P = .022) for OS. Thus, STK11 mutations remained strongly associated with increased relapse and decreased survival on multivariable analyses.

TABLE 3.

Multivariable Fine-Gray Model After Backward Selection (P < .1) for Time to Relapse (Death Modeled as Competing Risk) and Cox Proportional Hazards Models After Backward Selection (P < .1) for DFS and OS

Furthermore, STK11 mutations were predominantly associated with a distant pattern of failure, rather than local failure (Data Supplement). At 2 years, the cumulative incidence of distant metastasis was 12.1% for STK11 wild-type and 40.4% STK11-mutant tumors (Fine-Gray P = .030). By contrast, the 2-year cumulative incidence of local failure was 5.3% for STK11 wild-type and 9.3% for STK11-mutant (Fine-Gray P = .68). In the STK11-mutant group, the single local failure presented with concomitant regional nodal failure as well. There was no association between RT dose and local failure in the pooled RT cohort (Data Supplement). Taken together, in a pooled inoperable cohort receiving definitive RT, STK11 mutations distinguished a biologically aggressive subset of early-stage NSCLC that was associated with significantly worse OS, DFS, and cumulative incidence of relapse, with the primary mode of failure being distant recurrence.

DISCUSSION

In this study, the genomic landscapes of early-stage NSCLC (stage I to II) treated with surgery and RT were compared. Compared with the surgery cohort, patients undergoing RT demonstrated unique enrichment for STK11 mutations (43% v 6.7%) that defines a poor prognostic subgroup associated with rapid distant progression. Somatic STK11 mutations are more prevalent in patients on supplemental oxygen, reporting 20+ pack-years of tobacco smoking history, and identifying as Black race. The association of STK11 mutations with clinical hypoxia raises the novel hypothesis that the underlying clinical factors associated with medical inoperability may promote the emergence of STK11 mutations.

STK11 (also known as LKB1) encodes a serine/threonine-protein kinase with potent and pleiotropic tumor suppressor functions.²⁵ STK11 is one of the most frequently mutated genes in lung adenocarcinoma, and STK11/KRAS comutations in particular define a subset of advanced lung adenocarcinomas exhibiting resistance to immune checkpoint blockade.^32-34 Although these prior studies have characterized STK11 mutations in either advanced/locally advanced or early-stage operable disease, this study uniquely characterizes its frequency and significance in early-stage disease receiving RT.

In the response to cellular stress, STK11 plays a key role in activating the cGAS/STING pathway, an essential mediator of antitumor immunity. Therefore, mutational inactivation of STK11 promotes an immune evasive phenotype.^35,36 Recent literature demonstrates the importance of immune evasion in early NSCLC carcinogenesis and immune-mediated pruning in the evolution of lung cancer metastasis, which suggests that immune-related mechanisms may explain the adverse outcome in STK11-mutant NSCLC.^37,38 Although effective therapies can be difficult to develop for loss-of-function mutations in tumor suppressor genes, multiple drugs with preclinical efficacy have been identified and may warrant further study in clinical trials.²⁵ These include biguanides (eg, metformin), mTOR inhibitors (eg, rapamycin), and combinatorial regimens targeting mTOR, PI3K, and MEK.^39-41 Finally, since STK11/KRAS comutations promote resistance to immunotherapy, the durable clinical benefit observed for sotorasib for KRAS p.G12C-mutant tumors (including those with STK11 comutations) represents a promising therapeutic approach.⁴²

A key finding of our analysis is that STK11 mutations are associated with the use of supplemental oxygen (OR 5.5), 20+ pack-years of tobacco smoking history (OR 6.1), and Black race (OR 4.3). The strong association with clinical markers of hypoxia may point to distinct biological mechanisms underlying NSCLC in these cohorts. Recent studies highlight the role of hypoxia in promoting an immunosuppressive microenvironment through the induction of M2 macrophage polarization-related genes and the recruitment of myeloid-derived suppressor cells.⁴³ Concomitantly, there is increasing appreciation of the role of STK11 in the cellular response to hypoxia. Preclinical data have shown that STK11 inactivation promotes metabolic reprogramming via HIF-1a, which is the key transcriptional regulator of the cellular response to hypoxia.⁴⁴ Furthermore, the severity of underlying lung disease is correlated with the expression of HIF-1a in the non-neoplastic lung tissue of patients with lung cancer. Taken together, these data suggest that chronic hypoxia may apply a selective pressure that promotes inactivation of STK11, such that medically inoperable patients are more prone to develop STK11-mutant tumors.

The causes of racial disparities in lung cancer outcomes are certainly multifactorial, heavily influenced by inequities in access to screening, receipt of guideline-concordant treatment, and structural racism.^45-48 It is possible, however, that an elevated frequency of STK11-mutant tumors in Black patients contributes to these disparate lung cancer outcomes. We observed a significant association between Black race and somatic STK11 mutations. This result is consistent with published whole-exome data demonstrating that STK11 mutations are more frequent in tumors from African Americans (25% for adenocarcinoma and 8% for squamous cell carcinoma) relative to tumors from European Americans (13% and 1%, respectively).²⁷ This association appears to be functionally relevant, since relapse rates in Black versus non-Black patients are not significantly different after accounting for STK11 status (Data Supplement).

The identification of a poor prognostic subgroup in early-stage inoperable NSCLC defined by STK11 mutations highlights a subset of patients who may benefit from additional systemic therapies. Despite decades of investigation into predictive biomarkers of recurrence in early-stage, operable NSCLC, including promising epigenetic markers, patient selection for adjuvant chemotherapy remains largely determined on the basis of tumor size.^49,50 Immunotherapy is poised to become standard of care for resectable lung cancer, with adjuvant atezolizumab (IMpower010), neoadjuvant nivolumab (CheckMate 816), and adjuvant pembrolizumab (PEARLS/KEYNOTE-091) improving DFS/event-free survival.^11-13 Adjuvant osimertinib improved DFS for resected EGFR-mutant NSCLC in ADAURA.¹⁴ For patients receiving SBRT, ongoing trials, such as PACIFIC-4 (ClinicalTrials.gov identifier: NCT03833154), are investigating the benefit of adjuvant immune checkpoint blockade (eg, durvalumab). The identification of STK11 mutations as a poor prognostic factor is especially important in this context, considering the association with immunotherapy resistance in the metastatic setting.³⁴ Further studies are warranted to determine the level of benefit for patients with STK11 mutations in the early-stage setting.

The unique enrichment of STK11 mutations in the RT cohort and the association of STK11 mutations with features that correlate with medical inoperability (eg, supplemental oxygen, pack-years of tobacco smoking history, and FEV1 % predicted) may contribute to the variation in distant failure rates and oncologic outcomes observed after SBRT. Trials of medically inoperable patients, including Radiation Therapy Oncology Group (RTOG) 0236, RTOG 0813, and RTOG 0915, exhibit distant metastasis rates of approximately 20%-40%, despite all trials requiring staging positron emission tomography.^6,51,52 By contrast, distant metastases occurred in 3%-12% in trials of medically operable patients (ie, RTOG 0618, pooled analysis of STARS and ROSEL, and revised STARS).^29,53,54 Comparisons across trials are difficult, especially with possible differences in invasive nodal staging, but differential underlying frequencies of STK11 mutations may have partially contributed to these outcomes.

STK11 mutations were primarily associated with a distant pattern of failure, as opposed to local, and probably do not predict a decrease in the therapeutic efficacy of RT. However, it is worth acknowledging that diagnosing local failure after lung SBRT can be challenging radiographically, because of post-treatment changes and resultant fibrosis.^55,56 In a retrospective analysis of stage III patients, STK11 mutations were associated with increased locoregional recurrence, with in vitro experiments suggesting that STK11-mediated radioresistance was KEAP1/NRF2-dependent.⁵⁷ KEAP1/NFE2L2 mutations have previously been associated with high rates of local recurrence after RT but not surgery.¹⁹ Unfortunately, KEAP1 mutations were not profiled by our OncoPlus panel in the University of Chicago cohort. Murine models further support that STK11/LKB1-deficient tumors demonstrate more frequent metastases.⁵⁸ We hypothesize that STK11 mutations portend early metastatic dissemination and distant relapse.

Limitations of this study include its retrospective design and the small fraction of early-stage patients who undergo tumor molecular profiling, even at high-volume cancer centers. To address this, a pooled RT cohort was analyzed across two institutions, University of Chicago and Stanford. Although STK11 mutations were highly prognostic even in multivariable analyses and efforts were made to include various clinical covariates that influence the risk of relapse, unmeasured confounding variables, including those related to possible differences in patient selection between institutions, may have contributed to the poor oncologic outcomes observed in patients with STK11 mutations. Finally, because of limited sample size, it was not possible to analyze the prognostic significance of STK11 mutations in the surgical cohort (n = 2) and the significance of STK11/KRAS comutations in the pooled RT cohort (n = 7). Nonetheless, STK11/KRAS comutations were numerically associated with increased risk of relapse in the pooled RT cohort (57.1% v 25.6% at 2 years).

In conclusion, in the growing setting of early-stage, medically inoperable NSCLC receiving RT, to our knowledge, we present the first evidence for a prognostic and potentially targetable biomarker in the form of STK11 inactivation. Although prospective, multicenter validation is warranted, if the importance of STK11 inactivation in predicting poor outcomes in patients with early-stage NSCLC is confirmed, clinical trials of evolving adjuvant therapies targeting relevant pathways would be of significant interest.

APPENDIX

FIG A1.

(A) Freedom from relapse in patients receiving supplemental oxygen or not, stratified by STK11 mutation status. (B) Freedom from relapse in patients with 20+ pack-years smoking history or not, stratified by STK11 mutation status. (C) Freedom from relapse in patients of Black race or not, stratified by STK11 mutation status. O2(–): not receiving supplemental oxygen. O2(+): receiving supplemental oxygen. ≥ 20 PY(+): 20+ pack-years smoking history. ≥ 20 PY(–): less than 20 pack-years smoking history. STK11(–): STK11 wild-type. STK11(+): STK11-mutant.

FIG A2.

Patterns of progression in the pooled University of Chicago and Stanford medically inoperable cohort by STK11 mutation status: (A) freedom from local failure and (B) freedom from distant metastasis.

TABLE A1.

Genes Associated With Freedom From Relapse Using Kaplan-Meier Method (FDR < 0.05)

TABLE A2.

Univariable Associations Between Various Demographic and Clinical Factors and the Presence of STK11 Variants in the Profiled Tumors

TABLE A3.

Patient Characteristics for Pooled University of Chicago and Stanford Cohort of Early-Stage NSCLC Receiving Definitive RT

TABLE A4.

Full Multivariable Fine-Gray Model for Time to Relapse (Death Modeled as Competing Risk) and Full Multivariable Cox Proportional Hazards Models for DFS and OS

DATA SHARING STATEMENT

Deidentified individual participant data, including clinical and genomic data, for the University of Chicago cohort can be made available upon request.

AUTHOR CONTRIBUTIONS

Conception and design: Stanley I. Gutiontov, Daniel R. Gomez, Christine M. Bestvina, Maximilian Diehn, Sean P. Pitroda

Financial support: Maximilian Diehn

Provision of study materials or patients: Daniel R. Gomez, Christine M. Bestvina, Mark K. Ferguson, Jessica S. Donington, Maximilian Diehn

Collection and assembly of data: Rohan R. Katipally, Liam F. Spurr, Stanley I. Gutiontov, William Tyler Turchan, Aditya Juloori, Michael S. Binkley, Alice L. Jiang, Sherin J. Rouhani, Carolina Soto Chervin, Jeremy P. Segal, Victor Ng, Billy W. Loo, Christine M. Bestvina, Jessica S. Donington, Maximilian Diehn, Sean P. Pitroda

Data analysis and interpretation: Rohan R. Katipally, Liam F. Spurr, Stanley I. Gutiontov, William Tyler Turchan, Philip Connell, Renuka Malik, Michael S. Binkley, Pankhuri Wanjari, Billy W. Loo, Daniel R. Gomez, Christine M. Bestvina, Everett E. Vokes, Mark K. Ferguson, Jessica S. Donington, Sean P. Pitroda

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = immediate family member, Inst = my institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).