Prognostic Role of KRAS G12C Mutation in Non-Small Cell Lung Cancer: A Systematic Review and Meta-Analysis

KRAS G12C mutation (mKRAS G12C) is the most frequent KRAS point mutation in non-small cell lung cancer (NSCLC) and has been proven to be a predictive biomarker for direct KRAS G12C inhibitors in advanced solid cancers. We sought to determine the prognostic significance of mKRAS G12C in patients with NSCLC using the meta-analytic approach. A protocol is registered at the International Prospective Register for systematic reviews (CRD42022345868). PubMed, EMBASE, The Cochrane Library, and Clinicaltrials.gov.in were searched for prospective or retrospective studies reporting survival data for tumors with mKRAS G12C compared with either other KRAS mutations or wild-type KRAS (KRAS-WT). The hazard ratios (HRs) for overall survival (OS) or Disease-free survival (DFS) of tumors were pooled according to fixed or random-effects models. Sixteen studies enrolling 10,153 participants were included in the final analysis. mKRAS G12C tumors had poor OS [HR, 1.42; 95% CI, 1.10–1.84, p = 0.007] but similar DFS [HR 2.36, 95% CI 0.64–8.16] compared to KRAS-WT tumors. Compared to other KRAS mutations, mKRAS G12C tumors had poor DFS [HR, 1.49; 95% CI, 1.07–2.09, p < 0.0001] but similar OS [HR, 1.03; 95% CI, 0.84–1.26]. Compared to other KRAS mutations, high PD-L1 expression (>50%) [OR 1.37 95% CI 1.11–1.70, p = 0.004] was associated with mKRAS G12C tumors. mKRAS G12C is a promising prognostic factor for patients with NSCLC, negatively impacting survival. Prevailing significant heterogeneity and selection bias might reduce the validity of these findings. Concomitant high PD-L1 expression in these tumors opens doors for exciting therapeutic potential.


Introduction
Approximately 236,000 new lung cancer cases are expected to be diagnosed in the United States in 2022, contributing to 12.3% of all new cancer cases and 21.4% of all cancerrelated deaths [1]. Of these, 85% are non-small cell lung Cancer (NSCLC), and 75% of NSCLC cases present at either the advanced or relapsed stage [2]. Kirsten rat sarcoma viral oncogene homolog (KRAS) is one of the most common oncogenic drivers in NSCLC, seen in over 30% of lung adenocarcinomas (LUAD), depending on ethnicity and tumor stage and associated with smoking and female patients [3][4][5]. Notwithstanding this, KRAS testing is not included in the routine genomic panel for NSCLC, probably due to its less well-established efficacy in daily clinical practice based on current evidence [The European Society for Medical Oncology (ESMO) Scale for Clinical Actionability of Molecular Targets

Search Strategy and Study Selection
Preferred Reporting Items for Systematic Review and Meta-analysis (PRISMA) guidelines were followed to conduct this study [28] (Table S1). The protocol was registered in the International Prospective Register of Systematic Reviews (PROSPERO identifier: CRD42022345868). A systematic and comprehensive search was conducted by two reviewers (DW and PH) of PubMed, Embase, Cochrane Library, and ClinicalTrials.gov from inception until April 2023. The following combinations of keywords and Medical Subject Headings (MeSH)/EMTREE terms were used: 'KRAS OR Kirsten rat sarcoma viral homolog', 'mutation* OR mutated', and 'lung* OR pulmonary', 'cancer* OR tumor* or tumour* OR carcinom* OR neoplas* or malignan*' and 'prognos* or survival or recurren* or mortality or predict* or outcome* or death'. The search was restricted to articles published in English.
Inclusion criteria for eligible studies were defined as follows: (c) Studies reporting mixed data for KRAS G12C and other G12 substitutions; (d) Conference abstracts and preprint articles; (e) Narrative and systematic reviews, meta-analyses, expert opinions, editorial letters, case reports, and case series with a sample size of fewer than five individuals.
Studies exported from the databases were deduplicated, and unique studies were screened based on title, keyword, and abstract information. Subsequently, eligible studies were screened for full-text information, and data extracted from the studies met the eligibility criteria.

Data Extraction and Quality Assessment
Two reviewers independently retrieved the following data: authors, patient source, study design, histology, stage, mutation analysis methods, the total number of patients and patients with KRAS G12C mutation, and median follow-up duration in patients with KRAS WT, mKRAS G12C, other KRAS mutations. In addition, statistics extraction for timeto-event analyses (OS and DFS) were Hazard Ratio (HR) and 95% Confidence Interval (CI), median survival, and p-values. The OS and DFS were defined per the definition in each primary study. In case of a lack of reported survival outcomes, data were reconstructed from Kaplan-Meier curves using Tierney's or Parmar's methods [29,30]. We also assessed whether the survival outcomes were adjusted for clinicopathological covariates through univariate or multivariate analysis. We would utilize the latter if the author reported univariate and multivariate survival analysis results. Items were listed as "not available" (NA) when data from any of the categories mentioned above was unavailable".
The publication that provided the most recent or informative data for studies with multiple publications was selected. Discrepancies between reviewers were resolved by consensus or involving a third investigator. Survival outcomes were compared between KRAS G12C vs. other KRAS mutations or KRAS G12C vs. KRAS-WT. In studies where survival data for individual point mutations were provided, we compared KRAS G12C with KRAS G12D mutation. The risk of bias assessment was done using the Newcastle-Ottawa Scale (NOS) for observational studies [31].

Statistical Methods
The pooled HR evaluated the prognostic role of KRAS G12C for OS and DFS using the generic inverse variance method. The statistical heterogeneity within studies was tested with the Cochrane Q test and measured using I 2 indices. If HRs were found to have mild (I 2 ≤ 30%) to moderate (I 2 = 30-60%) heterogeneity, a fixed-effects model was used. In case of significant heterogeneity (I 2 > 60%), a random-effects model was used. By convention, an observed HR > 1 implies worse survival for the group with KRAS G12C mutations. The impact of KRAS status on survival was considered statistically significant if the 95% CI did not overlap with 1. Statistical significance was set at p < 0.05, and all tests were two-sided. Sensitivity analyses were conducted using the leave-one-out method to determine the undue influence of an individual study on the summary estimate or heterogeneity.
We conducted subgroup analyses for outcomes with ≥5 studies and >1 study in each subgroup [32]. The studies were stratified according to ethnicity (Asian/non-Asian), testing methodology (next-generation sequencing (NGS)/polymerase-chain-reaction (PCR)), and adjustment for clinical covariates (HR derived from multivariate (MV)/univariate or recreated from survival curves (UV)). In order to decrease the likelihood of chance differences arising from multiple testing in the subgroup analyses, we used 99% CI for the study estimates and 95% CI for the summary estimate.
Pooled Odds Ratio (OR) with 95% CI for binary variables and standardized mean difference for continuous variables were generated to investigate the relationship of mutant KRAS G12C tumors with PD-L1 expression status (<1%, 1-50%, >50%) and co-occurring mutations (TP53, STK11). Publication bias was assessed by Begg's funnel plots and Egger's test [33]. In case of publication bias, the trim-and-fill method was used to determine an adjusted pooled estimate [34]. Primary analyses were performed using the Review Manager version 5.4 (The Cochrane Collaboration, Copenhagen, Denmark), and publication bias evaluation was performed in the JASP software (JASP 0.16, the JASP team) [35].
likelihood of chance differences arising from multiple testing in the subgroup analyses, we used 99% CI for the study estimates and 95% CI for the summary estimate.
Publication bias was assessed by Begg's funnel plots and Egger's test [33]. In case of publication bias, the trim-and-fill method was used to determine an adjusted pooled estimate [34]. Primary analyses were performed using the Review Manager version 5.4 (The Cochrane Collaboration, Copenhagen, Denmark), and publication bias evaluation was performed in the JASP software (JASP 0.16, the JASP team) [35].

Quality of Studies
The Newcastle-Ottawa Scale indicated that nine studies had a low risk of bias, and the remainder had a moderate risk of bias (Tables 1 and S2). The overall quality of studies was moderate to high, with an average score of 7 (range 5-9). A few studies failed to report the duration and adequacy of follow-up. In a few studies, comparability between the two groups could not be ascertained due to the lack of adjustment of confounding variables that were likely to affect the survival outcomes.
The summary HR for OS showed no statistically significant survival difference between patients with KRAS G12C and non-KRAS G12C mutations [HR 1.03, 95% CI: 0.84-1.26, p = 0.79], though between-study heterogeneity was high (I 2 = 68%, p < 0.0001) (Figure 2A Studies that compared KRAS G12C to KRAS G12D mutated tumors also showed similar outcomes [HR, 0.93, 95% CI, 0.67-12.9; p = 0.66] ( Figure S1). Sensitivity analysis did not identify any undue influence of individual studies on effect size or heterogeneity. An asymmetrical right-skewed funnel plot ( Figure S2) and significant Egger's test (p = 0.013) suggested the presence of publication bias. The trim-and-fill method led to the The summary HR for DFS showed that patients with KRAS G12C mutations had a higher risk of relapse compared to patients with other KRAS mutations [HR 1.49 95% CI 1.07-2.09, p < 0.0001], and significant heterogeneity was observed [I 2 = 68%, p = 0.02] ( Figure 2B). However, studies that compared KRAS G12C to KRAS G12D mutated tumors found a non-significant outcome [HR, 1.36, 95% CI, 0.59-3.15; p = 0.48] ( Figure S9). Leaveone-out analysis revealed that none of the studies contributed to heterogeneity; however, the exclusion of any of the following studies led to a non-significant outcome: Finn et al.

Secondary Outcomes
We found that the proportion of patients with tumors expressing PD-L1 > 50% [OR 1.

Secondary Outcomes
We found that the proportion of patients with tumors expressing PD-L1 > 50% [OR 1.37 95% CI 1.

Discussion
Small molecule inhibitors of KRAS G12C mutant proteins finally broke the curse of the "undruggable" status of KRAS mutation based on the promising results of CodeBreak 100 and KRYSTAL-1 trials [25,52,53]. There is renewed interest in evaluating long-term oncological outcomes in patients with mKRAS G12C tumors [54]. Our systematic review and meta-analysis of sixteen retrospective studies comprising 10,153 patients showed that the mKRAS-G12C predicts poor survival in patients with NSCLC. The key findings of our study were as follows: First, compared to patients with other KRAS mutations, mKRAS-G12C predicted poor DFS. Second, mKRAS-G12C tumors were at a higher risk of all-cause mortality than KRAS-WT tumors despite similar DFS. Third, mKRAS G12C tumors were associated with high PD-L1 expression (>50%) compared with other KRAS mutations.

Discussion
Small molecule inhibitors of KRAS G12C mutant proteins finally broke the curse of the "undruggable" status of KRAS mutation based on the promising results of CodeBreak 100 and KRYSTAL-1 trials [25,52,53]. There is renewed interest in evaluating long-term oncological outcomes in patients with mKRAS G12C tumors [54]. Our systematic review and meta-analysis of sixteen retrospective studies comprising 10,153 patients showed that the mKRAS-G12C predicts poor survival in patients with NSCLC. The key findings of our study were as follows: First, compared to patients with other KRAS mutations, mKRAS-G12C predicted poor DFS. Second, mKRAS-G12C tumors were at a higher risk of all-cause mortality than KRAS-WT tumors despite similar DFS. Third, mKRAS G12C tumors were associated with high PD-L1 expression (>50%) compared with other KRAS mutations. Finally, TP53 and STK11 were not associated with either mKRAS G12C or other KRAS mutant tumors. Overall, KRAS G12C mutated tumors harbor poor prognosis; however, considerable between-study heterogeneity existed in most of these analyses.
In prior meta-analyses, the overall link between mKRAS and survival in NSCLC patients is relatively weak and primarily restricted to advanced disease [10,11]. Different mutation testing methodologies, variable patient selection criteria, and lack of stratification across stages further muddled the conclusions [55]. Furthermore, the downstream signaling of KRAS mutation subtypes uniquely alters tumor biology and thus may influence distinct clinical behavior that may not be apparent in examining KRAS mutations in toto [17,56]. Our analyses built on these lacunae by reporting the survival outcomes for the most common KRAS mutation subtype in NSCLC and showed consistent results across ethnicities and testing methodologies.
Surgically resected NSCLC are at a higher risk of disease relapse if associated with the mKRAS G12C than other KRAS mutations [47,49]. In contrast, tumors with mKRAS G12C and other KRAS mutations showed comparable OS rates in our study. In a study on earlystage NSCLC (n = 179), mKRAS G12C was associated with poor DFS (p = 0.006) compared with KRAS-WT; however, the statistical strength of this analysis dropped significantly for stage I patients, suggesting the influence of pathological stage on DFS [49]. Notwithstanding this, KRAS G12C inhibitors showed clinical efficacy in eradicating micrometastases and enhanced anti-tumor activity when combined with targeted agents [23]. Based on these results and our findings, adjuvant or neoadjuvant KRAS G12C inhibitor monotherapy or in combination with targeted agents in patients with early-stage mKRAS G12C tumors may open the door for exciting therapeutic paradigms.
We found that the frequency of mKRAS-G12C in the NSCLC and mKRAS cohorts reflected the real-world data that were previously reported, suggesting a representative population and thus more likely to reflect prognosis in clinical practice [69,70]. Acknowledging these findings cautiously is warranted. Most analyses involved high heterogeneity, which could not be attributed to an individual study and was presumably due to vast differences in patient characteristics. Publication bias was observed despite following expanded search criteria to include studies with either non-significant or negative results. The trim-and-fill method performs poorly in the presence of substantial between-study heterogeneity [71]. Approximately 49.7% of total patients had an advanced stage. Effect estimates from seven studies were unadjusted, which may influence the summary estimates, particularly by stage and systemic therapy. Lastly, this meta-analysis relied on the summary estimates from observational studies rather than individual patient data, which is considered a statistically superior method.
Trials like ADAURA and PACIFIC have provided sufficient evidence for the role of targeted agents in early-stage NSCLC. In addition, ANVIL (NCT02595944), PEARLS/Keynote-091 (NCT02504372), and IMpower010 (NCT02486718) trials are currently exploring the role of immune checkpoint inhibitors in the adjuvant setting [72,73]. However, the current literature lacks data to determine the prognostic and predictive role of KRAS G12C mutation in early-stage NSCLC; thus, further investigations are needed.
Our findings contribute to the evolving landscape of KRAS mutations in patients with NSCLC. Extrapolating these findings, the following recommendations may be made for clinical practice: 1) Inclusion of KRAS status, especially KRAS G12C mutation, as a routine test using a comprehensive molecular gene panel. Similarly, future studies should focus on 1) Evaluation of KRAS G12C inhibitors-anti-PD-1/PD-L1 inhibitor combination therapy in advanced NSCLC; 2) evaluation of adjuvant KRAS G12C inhibitor in surgically resected NSCLC with mKRAS G12C; and 3) assessment of the impact of KRAS G12C-comutations including TP53, STK11, and KEAP1, on prognosis.

Conclusions
Our meta-analysis on NSCLC found that tumors with mKRAS G12C were associated with worse DFS than tumors with other KRAS mutations and worse OS than tumors with KRAS-WT. The presence of significant heterogeneity and publication bias collectively undermines the validity of these findings. However, these outcomes reflect real-world prognoses and may be utilized in clinical practice to stratify high-risk patients and provide more effective therapeutic strategies for patients with NSCLC.  Figure S11. Association of PD-L1 expression 1-49% with KRAS G12C, Figure S12. Association of TP53 mutation with KRAS G12C, Figure S13. Association of STK11 mutation with KRAS G12C; Table S1: PRISMA checklist, Table S2: Newcastle-Ottawa Scale for stdy quality assessment. Funding: The authors would like to thank the FHU OncoAge, "Ligue Départementale 06 de Lutte contre le Cancer", and "Conseii Départemental des Alpes Maritimes" for supporting the fee related to the publication of this manuscript.
Institutional Review Board Statement: Not applicable.

Informed Consent Statement: Not applicable.
Data Availability Statement: The datasets generated during and/or analyzed during the current study are available in the Open Science Framework repository at 10.17605/OSF.IO/AUF2T without third-party permission.