Background
Non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancers [
1]. NSCLC is a complex disease consisting of numerous molecular subtypes [
2]. Biomarker testing of NSCLC has a direct impact on treatment decisions, and screening tumors for actionable driver mutations is essential before initiating treatment [
3]. In general, recommended treatment for patients include targeted therapies for patients with advanced NSCLC (aNSCLC) bearing alterations in driver oncogenes (
EGFR,
ALK,
ROS1, BRAF, NTRK, RET, or
MET), cancer immunotherapy (CIT) with immune checkpoint inhibitors alone or in combination with chemotherapy [
3].
Mutations in
Kirsten Rat Sarcoma Viral Oncogene Homolog (
KRAS) are prevalent in NSCLC, occurring in approximately 25 to 40% of patients (≈5–10% in Asian patients) [
4‐
7]. In particular, the
KRAS glycine 12 to cysteine (
G12C) activating mutation has the highest prevalence (≈40% of all
KRAS mutations in NSCLC) [
8,
9]. Tumors that harbor
KRAS mutations are usually among the most aggressive and refractory to treatment [
8]. Recently, sotorasib, a mutation-specific inhibitor of KRAS G12C has received marketing authorization in
KRAS G12C-mutated aNSCLC patients who have received at least one prior systemic therapy [
10,
11]. While KRAS G12C inhibitors provide an exciting new second-line or later (2L+) treatment option for patients with
KRAS G12C–positive tumors, the standard-of-care therapies for first-line (1L)
KRAS G12C-mutated aNSCLC patients currently remain CIT targeting programmed cell death protein 1 (PD-1) or programmed cell death 1 ligand 1 (PD-L1) alone or combined with chemotherapy.
Patients with tumors that harbor
KRAS mutations have shown a longer median overall survival (OS) when treated with CIT alone vs chemotherapy alone (28 months (95% CI: 23-NR) vs 11 months (95% CI: 7–25)) [
10] or when treated with combination therapy (CIT and chemotherapy) vs chemotherapy alone (21 months (95% CI: 16-NR) vs 14 months (95% CI: 8-NR)) [
10]. While patients with
KRAS G12C–positive cancer may benefit from treatment with CIT alone or in combination with chemotherapy [
12,
13], numbers of patients with
KRAS G12C-positive cancer were low in these previous studies, and it remains unclear how the benefit compares with that seen in patients with
KRAS wild-type (WT) cancer. Furthermore, somatic genomic alterations in serine/threonine kinase 11
(STK11) or kelch like ECH associated protein 1
(KEAP1) commonly co-occur with
KRAS mutations in NSCLC (25–30%) [
14], and co-mutation of
KRAS with
STK11 or
KEAP1 is associated with significantly worse survival [
14‐
16]. It is unknown whether co-occurring mutations affect prognosis and whether differential responses to treatment and consequent effects on survival outcomes exist in these patient populations. This study used real-world data to assess the effect of
KRAS G12C mutational status on OS in patients with aNSCLC with or without
STK11 and/or
KEAP1 co-mutations who received CIT, chemotherapy, or both in the 1L and 2L using real-world data.
Methods
Data source
This study used the nationwide (US-based) deidentified Flatiron Health-Foundation Medicine NSCLC clinico-genomic database (FH-FMI CGDB). Retrospective longitudinal clinical data were derived from electronic health record (EHR) data, comprising patient-level structured and unstructured data, curated via technology-enabled abstraction, and were linked to genomic data derived from FMI comprehensive genomic profiling (CGP) tests in the CGDB by de-identified, deterministic matching [
17]. Genomic alterations were identified via CGP of > 300 cancer-related genes on FMI’s next-generation sequencing (NGS) test (FMI sequencing platform[s]: FoundationOne®CDx, FoundationOne®, FoundationOne®Liquid, or FoundationOne®Liquid CDx) [
18‐
20]. Both liquid and solid assays were used in this study. As liquid assays may not detect alterations if shedding of circulating DNA is low, only solid assays were used to define WT
KRAS,
KEAP1, and
STK11.
Patient population
Eligibility criteria included: (1) aged ≥18 years with aNSCLC newly diagnosed between January 1, 2011, and March 31, 2020; (2) had structured activity within 90 days after aNSCLC diagnosis and had 6 months of follow-up after treatment initiation; (3) did not have functional or likely functional driver alterations (short variants, copy number alterations, or fusions) in EGFR, ALK, ROS1, BRAF, ERBB2, MET, or RET; (4) received treatment with CIT (eg, immune checkpoint inhibitors) alone, combination CIT and chemotherapy, or chemotherapy alone and did not receive targeted therapies for driver mutations or nonapproved monotherapy CIT for aNSCLC; (5) displayed no evidence of being diagnosed with other cancers in the database; and (6) had ≥1 definitive (ie, positive or negative) molecular test, including a test for the KRAS gene, with results before or after 1L treatment initiation. If a patient had several specimen collections, the closest to initiation of 1L treatment was used. Patients were categorized by KRAS mutational status (G12C vs WT). The KRAS WT group excluded patients with any KRAS alteration.
Outcomes and analysis
The primary outcome was OS in patients with KRAS WT or KRAS G12C–positive aNSCLC, sorted by the following factors: (1) treatment line (1L or 2L); (2) treatment ([a] CIT alone or in combination with other CIT or with nonchemotherapy; [b] combination CIT and chemotherapy; or [c] chemotherapy alone); 3) presence of mutated STK11 (mSTK11) and/or mutated KEAP1 (mKEAP1) vs STK11 WT and KEAP1 WT, including any functional status (unknown, likely and known). OS is defined as time from an index date to the date of death for individual patients who have died. Patients without a death date are censored at the last evidence of them being alive, e.g., structured activity in the database.
All statistical analyses were performed with R. Kaplan-Meier (KM) curves, associated medians, and 95% confidence intervals were estimated for survival outcomes. Cox proportional hazards models adjusted for baseline demographics and clinical characteristics (age, sex, race, cancer type [de novo or recurrent], PD-L1 status, any metastasis, tumor mutational burden, histology, and 1L treatment [for 2L analysis only]) were used to analyze the effect of mutational status on OS in patients receiving CIT alone, combination CIT and chemotherapy, or chemotherapy alone. A separate category within a variable was created for the missing values. Adjustments to account for left truncation and immortal bias were applied to the KM analysis and the Cox regression model.
Discussion
This retrospective real-world study used an EHR-linked CGDB to assess OS in patients with KRAS WT and KRAS G12C–positive aNSCLC by treatment line, treatment type, and co-mutations in STK11 and/or KEAP1.
Patients with aNSCLC with
KRAS G12C–positive tumors were found to have comparable OS relative to patients with
KRAS WT tumors when receiving combination CIT and chemotherapy or chemotherapy alone in the 1L or 2L; this finding is consistent with majority of the previous literatures using other datasets [
12,
13,
21,
22].
Although not statistically significant, potentially due to limited patient counts, patients with
KRAS G12C–positive cancer treated with 1L CIT showed a trend toward longer survival vs patients with
KRAS WT cancer. Similar results were observed in patients with high (≥50%) PD-L1 expression level. These results with significantly larger sample size are consistent with previous reports of improved survival with immune checkpoint inhibitors in
KRAS-mutant aNSCLC [
12]. Consistent with the findings of this study,
KRAS mutations are associated with increased PD-L1 expression in patients with aNSCLC [
23] and contribute to immunosuppression [
24]. A recent study [
25] also found a strong association between mutated
KRAS and immune biomarkers linked to response to immune checkpoint inhibition [
26]. Together these results suggest that patients with aNSCLC whose tumors harbor
KRAS mutations may have particularly favorable outcomes with CIT [
26], which is distinct from
EGFR-mutated NSCLC [
26,
27]. Also, since the single agent KRAS G12C inhibitors appear to have a lower overall response rate compared to the target-specific tyrosine kinase inhibitors [
26,
28‐
30], it is possible that the KRAS G12C inhibitors need to be combined with CIT moving into 1L. Further analyses are needed to evaluate
KRAS mutational status in patients with aNSCLC treated with CIT alone and to explore the effect of different
KRAS variants (e.g.,
G12D and
G12V) on survival relative to
G12C.
In this study, patients with aNSCLC and concurrent double mutations in
STK11 and
KEAP1 had significantly shorter OS vs patients with
STK11 WT and
KEAP1 WT receiving any type of 1L or 2L therapy regardless of
KRAS G12C mutational status. This is consistent with recent retrospective analyses conducted using an EHR-linked CGDB that found that mutations in
STK11 and
KEAP1 in aNSCLC were associated with worse outcomes (shorter progression-free survival and OS) in patients treated with anti–PD-L1/PD-1 therapies and platinum-based chemotherapy [
31]. Our results build on these previous findings and further underscore the poor prognosis and high unmet medical need that exists in patients with aNSCLC with co-occurring mutations in
STK11 and
KEAP1, with either
KRAS WT cancer or
KRAS G12C–positive cancer. If
KRAS G12C-positive patients with
STK11 co-mutations are sensitive to the KRAS G12C inhibitors, this may serve as the rationale of expediting the exploration of single-agent KRAS G12C inhibitors in certain biomarker-selected patient sub-population in 1L.
Initial studies of KRAS G12C inhibitors excluded patients with active brain metastasis [
11]. Our study showed that
KRAS G12C-positive patients have a higher prevalence of brain metastasis as compared to patients with
KRAS WT tumors. It is therefore critical to evaluate whether KRAS G12C inhibitors can be beneficial for patients with brain metastasis and brain-penetration may be a key consideration for future generations of KRAS G12C inhibitors.
This study has several limitations. CGDB data are generated from real-world clinical practice; thus, some data may have been miscoded or may be subject to errors encountered in an oncology clinic. The data do not capture information about patients’ history or treatment outside of the specific cancer care site, which may lead to underreporting or missing data. Limited data exist from patients attending or beginning treatment elsewhere, such as academic medical centers, as the CGDB largely reflects community oncology treatment information. The study population was comprised of patients who received treatment in the United States, and the results may not be generalizable to patients treated globally. For specimen collection, patients were categorized as having
KRAS G12C or
KRAS WT cancer based on the closest specimen collection to index date (1L or 2L treatment initiation). With this approach, patients may have developed a
KRAS mutation after receiving therapy; however, sensitivity analyses were conducted to evaluate selection bias. Analyses were rerun in patients with a specimen collection date within 90 days of the index date (1L or 2L treatment initiation), and results were found to be consistent. We categorized patients with
KRAS G12C,
KEAP1, and
STK11 mutational status using both solid and liquid assays; patients with cancer that was categorized as WT were assigned based on solid assays only. To avoid selection bias, we further restricted and subsequently reperformed our analyses of patients categorized using solid assays only. The results remained consistent; however, to ensure a larger sample size, patients whose mutational status was determined using liquid assays were included in the final analysis. Due to the entry selection rules, the CGDB is inherently left truncated. For inclusion in the cohort, patients were required to have undergone NGS testing by FMI, and, therefore, must have been alive until the date of the NGS test. The analyses were adjusted for left truncation to control for this potential immortal bias. Eastern Cooperative Oncology Group performance statuses were missing in 14 to 23% of patients and not adjusted for due to this high missingness. A higher proportion of patients with
KRAS G12C–positive tumors had non-squamous histology than patients with
KRAS WT tumors. Although squamous, NOS vs non-squamous histology was adjusted for, no further details such as adenocarcinoma was available in the CGDB. This study included patients with locally advanced (stage IIIB and IIIC) and metastatic (stage IV) diseases, which could have also progressed from an initial diagnosis at early stages (stage I, II and IIIA). Although stage at initial diagnosis was not adjusted for, cancer type (de novo vs recurrent) and presence of metastasis at baseline were included in the model to adjust for heterogeneity in patient population. Finally, this study included a high percentage of patients with missing PD-L1 data; PD-L1 status was unknown for > 50% of patients in each group. Despite these limitations, these findings highlight the importance of evaluating genomic alterations in clinical practice to better understand how selection of treatment and therapy type affect survival in patients with tumors bearing genomic alterations. This study offers a unique design advantage to the published literature to date. Other studies have assessed the effect of
STK11 and/or
KEAP1 mutations either in patients whose tumors harbor
KRAS mutations only, or in a mixed patient population whose tumors could harbor either
KRAS WT or
KRAS mutations [
32‐
35]. This study evaluated survival benefit with different 1L and 2L therapies in patients with
KRAS G12C or WT cancer with or without
STK11 and/or
KEAP1 mutations, providing additional insights into these frequently co-occurring mutations. Further analyses with a larger sample size evaluating the interplay between
KRAS and
STK11 and/or
KEAP1 by select treatment types are warranted.
Results from this study may inform personalized treatment for patients with
KRAS-mutated NSCLC, as certain combinations of mutations in
KRAS and other genes may generate biological diversity that may respond to tailored treatment [
36]. Together, these results may enable personalized care and help optimize patient outcomes.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.