If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Tyrosine kinase inhibitors and immune checkpoint inhibitors (ICIs), each requiring testing for precision biomarkers, have recently been approved in the adjuvant setting. We assessed the potential value of multigene testing in early lung adenocarcinoma (LUAD).
Methods
Using a real-world clinicogenomic database linking deidentified electronic health record–derived clinical data to genomic data, we selected patients with LUAD who underwent tissue comprehensive genomic profiling (CGP). Using a probabilistic decision tree, we estimated the cost implications of the avoidance of adjuvant ICI in patients with programmed death-ligand 1–positive (PD-L1+) LUAD and an ALK, ROS1 or RET driver.
Results
The CGP was performed on a specimen collected before advanced disease in 20% (1320 of 6697) of cases and ordered before advanced diagnosis for 12.6% (847 of 6697) of patients. The prevalence of driver alterations in early and advanced-stage specimens was similar, though KRAS mutations were enriched in early disease and drivers including ALK rearrangements in advanced disease. Patients who had CGP results obtained before versus after recurrence had less time between recurrence and the start of any first-line treatment (median 3.6 versus 6 wk, p < 0.001). Through avoidance of ICI in programmed death-ligand 1–positive early LUAD with an ALK, ROS1 or RET driver, we estimated that the universal CGP could reduce expected costs by $1597.23 per patient relative to EGFR single-gene testing.
Conclusions
The CGP can identify driver alterations and accelerate the start of first-line therapy at recurrence. It may also represent a cost-effective approach for avoiding futile adjuvant ICI in patients with drivers that have historically lacked activity with ICI in metastatic disease.
Although approximately 30% of NSCLCs can be completely resected with curative intent, many patients have recurrence typically requiring systemic treatment for metastatic disease.
The treatment paradigm for stage IV NSCLC has changed in recent years as tyrosine kinase inhibitors (TKIs) and immune checkpoint inhibitors (ICIs) have become part of standard of care. Both modalities have dramatically increased overall survival in select biomarker populations of patients. The current National Comprehensive Cancer Network (NCCN) guidelines recommend evaluation of programmed death-ligand 1 (PD-L1) status for all advanced-stage NSCLCs; broad genomic testing to identify oncogenic driver alterations is strongly advised for nonsquamous tissue types and considered for squamous tissue types.
PD-L1 and specifically EGFR mutation testing are now recommended in resectable disease as the Food and Drug Administration (FDA) has recently approved both adjuvant atezolizumab for PD-L1 positive (PD-L1+) greater than or equal to 1% disease and adjuvant osimertinib in patients with classic EGFR driver mutations.
Although smoking cessation programs are credited for the rapid drop in NSCLC incidence, recent analysis of the Surveillance, Epidemiology, and End Results Program national database revealed that mortality from NSCLC decreased even faster than the incidence of the disease.
This drop in mortality corresponded to the timing of approval of targeted therapy. As lung cancer surveillance programs gain more wide acceptance in clinical practice,
more patients are expected to be diagnosed with early stage NSCLC where emerging evidence supports the practice of applying targeted therapy in the perioperative phase.
Though testing for PD-L1 and EGFR is necessary for choosing the optimal adjuvant treatment, multigene testing for other gene alterations in early stage NSCLC may have its own benefits. For example, the NCCN recommends testing for seven other driver alterations besides EGFR mutations in metastatic disease. With comprehensive genomic profiling (CGP), these results would already be available at recurrence to guide treatment selection in a more timely manner, potentially enabling faster start of first-line therapy. Furthermore, ICI therapy in metastatic NSCLC with EGFR, ALK, RET, or ROS1 driver alterations has not improved response rates nor survival in comparison to chemotherapy.
Thus, most immunotherapy clinical trials exclude patients with EGFR mutations and ALK rearrangements. Although EGFR and ALK+ patients were included in the IMpower010 study which evaluated adjuvant atezolizumab and lead to its approval, it was not clear that this small subgroup had any benefit, and adjuvant atezolizumab is not recommended for this population.
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
With this information and given data from the metastatic setting, it can be reasonably inferred that ICI therapy in early stage disease positive for these driver alterations would not improve recurrence-free survival or overall survival and could lead to unnecessary side effects and increased risk of TKI toxicity if disease recurs and further systemic therapy is needed. There is the appropriate concern of the costs of multigene testing, but this must be balanced with the costs of potentially ineffective ICI treatment. Both the potential clinical and financial benefits of CGP in patients with early stage, resected lung adenocarcinoma (LUAD) are explored in this retrospective study.
Materials and Methods
Foundation Medicine CGP
The CGP was performed during routine clinical care (Foundation Medicine, Inc., Cambridge, MA) on tissue specimens from early and advanced-stage LUAD. DNA was extracted from 40 μm of formalin-fixed, paraffin-embedded sections, and CGP was performed on hybridization-captured, adaptor ligation-based libraries to a mean coverage depth of greater than 550 times for 315 or 324 cancer-related genes plus selected introns from genes frequently rearranged in cancer, as previously described.
Tumor mutational burden was calculated by counting the number of synonymous and nonsynonymous mutations across a 0.8 to 1.2 megabase (Mb) region, with computational germline status filtering, and reporting the result as mutations per Mb. This method has been previously validated for accuracy against whole exome sequencing.
PD-L1 expression was determined by immunohistochemistry (IHC) performed on formalin-fixed, paraffin-embedded tissue sections. PD-L1 IHC results were available in 58.2% (3900 of 6697) of total cases. Missing results may be due to real-world practice patterns and changing practice patterns over time. In addition, assays that were only scored by combined positive score or immune cells were excluded. A pathologist determined the percentage of tumor cells with expression (0%–100%) and the intensity of expression (0, 1+, 2+). PD-L1 expression was reported as a continuous variable with the percentage of tumor cells staining with greater than or equal to 1+ intensity. PD-L1 expression was summarized as negative (<1%) or positive (≥1% of tumor cells staining with ≥1+ intensity).
Clinicogenomic Database
This study used real-world data from the Flatiron Health (FH)-Foundation Medicine (FM) NSCLC clinicogenomic database (CGDB), a nationwide deidentified electronic health record (EHR)–derived database which includes patients sequenced at FM who received care within the FH network. The deidentified data originated from approximately 280 U.S. cancer clinics (approximately 800 sites of care). The FH-FM CGDB includes 8378 patients with chart-confirmed NSCLC, with diagnosis of LUAD who received care within the FH network between January 2011 and June 2021. Cohorts included in our analysis were limited to those who had tissue CGP (FoundationOne or FoundationOne CDx) at some point during cancer care. Retrospective longitudinal clinical data were derived from EHR data, comprising patient-level structured and unstructured data, curated by means of technology-enabled abstraction of clinical notes and radiology or pathology reports and linked to CGP data by deidentified, deterministic matching.
Association of patient characteristics and tumor genomics with clinical outcomes among patients with non-small cell lung cancer using a clinicogenomic database.
Patients were excluded if their stage at diagnosis was unavailable, if stage was noted as III without further granularity, or if CGP specimen collection or report date was unavailable. There were 236 cases excluded because advanced diagnosis was noted less than 3 months from initial early diagnosis or a specimen recorded as collected during early disease was from a metastatic site. Early stage disease was classified as stage I, II, or IIIA without documentation of recurrence, and advanced-stage disease was defined as either early stage (I–IIIA) disease that had recurred or progressed or initial diagnosis with stage IIIB or IIIC or IV.
Institutional Review Board Approval
For FM genomic analysis, approval for this study, including a waiver of informed consent and a Health Insurance Portability and Accountability Act (HIPAA) waiver of authorization, was obtained from the Western Institutional Review Board (WIRB) Copernicus Group IRB ( protocol number 20152817). For FH-FM CGDB analysis, IRB approval of the study protocol was obtained before study conduct and included a waiver of informed consent from Copernicus Group IRB.
Statistical Analysis
We evaluated the frequency of known or likely pathogenic genomic alterations in CGP specimens for all LUAD cases and in a PD-L1+ (tumor proportional score [TPS] ≥ 1%) subset. Oncogenes evaluated included ALK rearrangement, BRAF V600E mutation, EGFR mutation (limited to L858R, exon 19 deletion, G719X, L861Q, S768I, and exon 20 insertion), ERBB2 mutation, KRAS mutation, MET amplification and exon 14 skipping alteration, RET rearrangement, ROS1 rearrangement, and NTRK fusion. Frequencies of these driver alterations were compared using Fisher’s exact test, and p values were corrected with the Benjamini-Hochberg false discovery rate method. In patients with disease recurrence who had testing on a specimen collected in the early setting, we compared the time from recurrence with start of systemic first-line treatment for patients with the testing performed and reported before (n = 203) versus after (n = 439) advanced diagnosis using a Cox regression. The number of patients with an EGFR mutation, ALK, ROS1, or RET rearrangement who received matched targeted therapy as first-line treatment and the time from recurrence to start of any first-line therapy were also compared between the two cohorts with Fisher’s exact test. Patients were excluded from these analyses if first-line treatment information was unavailable.
Timing between initial diagnosis of LUAD and CGP biopsy or report was calculated. On the basis of the distribution and anticipated typical clinical practice, cases classified as CGP reported in the early setting were limited to those with CGP specimen collected within 3 months of initial diagnosis and testing results reported within 6 months of initial diagnosis, whereas those in the late setting were those with report after recurrence or advanced initial diagnosis. For genomic analysis, early disease was classified as CGP specimen collected within 3 months of initial diagnosis and before recurrence.
Cost Analysis
The incremental expected cost of CGP multigene testing using FoundationOneCDx compared with EGFR single-gene testing in patients with early stage LUAD was assessed with a probabilistic decision tree. Base-case cost parameters for diagnostic testing, drug administration, and a year of atezolizumab ICI treatment were sourced from the Centers for Medicaid and Medicare Services clinical laboratory fee schedules, physician fee schedules, and ASP drug pricing files (Supplementary Table 1).
The prevalence of PD-L1 1%+ (TPS ≥ 1%) and the prevalence of ALK rearrangements among PD-L1 1%+ patients were direct estimates from the IMpower010 trial.
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
We opted for trial-based prevalence estimates when available rather than using direct estimates from the CGDB due to the potential for a biased sample of patients receiving CGP, many of whom seek out comprehensive testing after negative biomarker results from single-gene tests or limited panels. Because RET and ROS1 prevalence estimates among PD-L1 1%+ patients were not available in the IMpower010 trial, we did use the direct estimates from the CGDB as the basis of those base-case parameter values. We also allowed for imperfect adjuvant therapy adoption (base assumption was 80% of patients considering adjuvant ICI treatment) among the eligible patient population given both the recency of the adjuvant atezolizumab approval and the role of patient preferences. A cost-neutrality threshold analysis, a one-way sensitivity analysis, and a probabilistic sensitivity analysis were conducted (Supplementary Tables 2–4 and Supplementary Figs. 1 and 2) to assess the influence of parameter uncertainty and base-case assumptions. The cost analysis considered only direct diagnostic and therapeutic costs with a 1-year time horizon.
Results
Genomics of Early Disease
A total of 8378 patients were selected with LUAD from the CGDB, of which 6697 had CGP performed on a tissue specimen and were assessable as detailed in the Materials and Methods section. Further details of cohort selection are described in Supplementary Figure 3. Of the patients, 20% (1320 of 6697) had CGP performed on tissue collected before advanced diagnosis. Among this cohort, 89% (1177 of 1320) had CGP performed on a specimen collected within 3 months of initial diagnosis in the absence of any advanced diagnosis, which hereafter we have defined as in the early disease setting, representing 17.6% (1177 of 6697) of the overall cohort (Supplementary Fig. 4A). Furthermore, of the patients, 80.2% (5377 of 6697) had CGP performed on tissue collected in the advanced disease setting. The 143 patients with CGP performed on a specimen collected before any advanced diagnosis but more than 3 months from initial diagnosis were excluded from genomic analysis comparing the biology of early versus advanced disease. PD-L1 IHC results were available for 58.2% of the cases (3900 of 6697). In general, 3506 samples (89.9%) were tested using the Dako 22C3 PD-L1 antibody, 200 (5.1%) were tested using other clones, and for 194 (5.0%), the PD-L1 IHC platform was unknown. Among the patients with CGP testing performed on a specimen collected in the early versus advanced disease setting with PD-L1 IHC available, 54% (362 of 670) and 63% (1999 of 3164), respectively, of the patients were PD-L1+ (TPS ≥ 1%).
NSCLC driver alterations were detected using CGP of specimens collected in the early disease setting (within 3 mo of initial diagnosis). Compared with CGP of specimens collected in advanced disease, the distribution of driver alterations was similar, although some statistically significant differences were observed (Fig. 1 and Supplementary Table 5). KRAS mutations were significantly more common in the early disease setting (41.7% versus 35.5%, adjusted p < 0.01), whereas ALK rearrangements (early versus advanced: 1.8% versus 4.3%, adjusted p < 0.01) and MET amplification (1.4% versus 3.1%, adjusted p < 0.01) had a significantly higher prevalence in advanced disease. When evaluating only patients with PD-L1+ disease, similar results were observed, but only the differences in KRAS mutation and ALK rearrangement frequencies between the two groups were significant (Supplementary Table 5). The distribution of common EGFR and KRAS mutant isoforms was largely consistent in early and advanced disease settings (Fig. 1A-D). Alterations in a limited number of other genes including TP53, CDKN2A, CDKN2B, FGF10, SMARCA4, MYC, RICTOR, and MCL1 were also enriched in advanced disease; however, these enrichments were not significant in the smaller PD-L1+ subset (Fig. 2).
Figure 1Frequencies of driver oncogene alterations detected by CGP in tumor tissue specimens collected during early and advanced stage LUAD. The prevalence of driver gene alterations listed in the National Comprehensive Cancer Network NSCLC guidelines is illustrated for all LUAD cases (A, B) and for the PD-L1+ subset (C, D). B and D illustrate the frequencies of specific EGFR and KRAS alterations. Gene alterations were compared for specimens collected in the early versus advanced disease setting. Early setting is defined as CGP specimen collected within 3 months of initial diagnosis. Amp, amplification; CGP, comprehensive genomic profiling; ex14, exon 14; LUAD, lung adenocarcinoma; Mut, mutation; PD-L1+, programmed death-ligand 1 positive; TPS, tumor proportional score.
Figure 2Enrichment of co-altered genes in early and advanced lung adenocarcinoma. The prevalence of gene alterations was compared for specimens collected in the early versus advanced disease setting. Volcano plots depict all lung adenocarcinoma cases (left) and only PD-L1+ cases (right). Higher frequency alterations are found with larger dot size. The dotted horizontal line is a significance cutoff (FDR = 0.05). p values were corrected with the Benjamini-Hochberg FDR method. Early setting is defined as CGP specimen collected within 3 months of initial diagnosis. CGP, comprehensive genomic profiling; FDR, false discovery rate; PD-L1+, PD-L1+, programmed death-ligand 1 positive.
CGP Testing Patterns and Impact of Early CGP on First-Line Treatment Decisions
In patients with LUAD in the CGDB, 12.6% (847 of 6697) had CGP ordered before advanced diagnosis, for which 9.1% (608 of 6697) had a specimen collected within 3 months and CGP report within 6 months of initial diagnosis, which hereafter we have defined as CGP report in the early disease setting. This included 34%, 28%, and 38% of patients diagnosed with having stage I, II, and IIIA disease, respectively. Furthermore, 87.4% (5850 of 6697) of the patients had CGP testing after advanced diagnosis or recurrence. Testing patterns also continue to evolve over time. The growing ubiquity of CGP in clinical care is reflected in a larger share of the analysis cohort having been tested in recent years, and more of these patients are receiving CGP testing before a diagnosis of advanced cancer (Fig. 3). Patients with CGP testing performed in the early versus advanced disease setting tended to be older (median 69.0 versus 67.0 y old at initial diagnosis), more often female (59.4% versus 55.1%), and treated at an academic center versus in community practice (93.1% versus 88.9%) (Table 1). In patients with CGP testing before advanced diagnosis, CGP was done at a median of 66 days (interquartile range: 35–212 d) after initial diagnosis, and of patients who had recurred by the time of data cutoff, the median time from initial diagnosis to advanced LUAD diagnosis was just under 13 months. Of those with CGP ordered before advanced diagnosis, 73.2% received their CGP results within 6 months of initial diagnosis. The remaining 26.8% received their CGP results more than 6 months after initial diagnosis but before documentation of any recurrence, though we suspect that a subset of these cases could represent testing at suspicion of recurrence (Supplementary Fig. 4B).
Figure 3Dynamics of CGP testing patterns in the early and advanced lung adenocarcinoma disease settings over time. Incidence of CGP testing is increasing over time in both the early and advanced disease settings, and the fraction of cases tested in early disease is increasing over time. Numbers on each bar indicate cases tested. #2021 data are through June 30, 2021, only. Early versus advanced disease designation is based on CGP report date. ∗Early setting refers to patients with CGP report date within 6 months of initial diagnosis with specimen collected within 3 months of initial diagnosis. The "Other before advanced disease" category includes patients with CGP report more than 6 months from initial diagnosis but before advanced diagnosis, or CGP specimen >3 months from initial diagnosis. If a patient had multiple CGP tests performed during their disease course, the first instance is depicted here. CGP, comprehensive genomic profiling.
Table 1Characteristics of Patients With CGP Testing Performed in the Early Versus Advanced Disease Setting
Patient Characteristics
CGP Report in Advanced Disease n = 5850
CGP Report in Early Disease n = 608
p Value
Age at diagnosis, median [IQR]
67.0 [59.0–74.0]
69.0 [62.0–76.0]
<0.001
Community practice
5199 (88.9)
566 (93.1)
0.002
Female
3226 (55.1)
361 (59.4)
0.051
Race
0.030
Asian
192 (3.28)
30 (4.93)
Black or African American
383 (6.55)
29 (4.77)
White
3952 (67.6)
391 (64.3)
Other
846 (14.5)
100 (16.4)
Not documented
477 (8.15)
58 (9.54)
Stage at initial diagnosis
<0.001
Stage I
561 (9.59)
207 (34.0)
Stage II
308 (5.26)
170 (28.0)
Stage IIIA
349 (5.97)
231 (38.0)
Stage IIIB/C
472 (8.07)
0 (0.00)
Stage IV
4160 (71.1)
0 (0.00)
Smoking history
0.102
History of smoking
4595 (78.5)
500 (82.2)
No history of smoking
1241 (21.2)
107 (17.6)
Unknown/not documented
14 (0.24)
1 (0.16)
Note: Early disease setting refers to patients with CGP results reported within 6 months of initial diagnosis, specimen collected within 3 months of initial diagnosis, and before any advanced diagnosis. Advanced disease setting refers to CGP report after advanced diagnosis. Patients with CGP report before any advanced diagnosis but more than 6 months after initial diagnosis or specimen more than 3 months after initial diagnosis are excluded from this analysis. Ages at diagnosis are compared using Kruskal-Wallis test, and frequencies are compared using Fisher’s exact test. All values are n (%) unless otherwise specified.
We evaluated the time from recurrence to the start of first-line systemic therapy in the advanced setting, hypothesizing that patients with CGP ordered any time before recurrence (n = 174) had shorter time to first-line treatment based on existing knowledge of actionable biomarkers compared with patients with CGP performed on specimen collected before recurrence but with the testing ordered after recurrence (n = 370). In patients with CGP before recurrence, the median time to start of first-line therapy was 25 days compared with 42 days for patients with CGP ordered on an existing specimen after recurrence (p < 0.001) (Fig. 4). In the subset of patients from each group with a targetable EGFR, ALK, RET, or ROS1 driver detected, for patients with CGP before recurrence, the median time to start of first-line therapy was 19 days compared with 47 days for patients with CGP ordered on an existing specimen after recurrence (p < 0.001) (Fig. 4). Furthermore, within this subset with EGFR, ALK, RET, or ROS1 drivers, 30 of 39 (77%) with early CGP initiated matched first-line TKI whereas 43 of 65 (66%) with CGP after recurrence received matched first-line TKI (p = 0.3); an additional two patients and one patient, respectively, in each group received only an unspecified clinical study drug.
Figure 4CGP before recurrence is associated with timely delivery of first-line therapy. (A) In patients with LUAD with CGP on samples collected before recurrence, those with CGP results obtained any time before recurrence versus after recurrence had less time from recurrence to start of 1L therapy (median 3.6 versus 6 wk, p < 0.001). A total of 18 patients have record of starting 1L therapy shortly (≤2 wk) before advanced diagnosis. (B) In patients with LUAD with a targetable EGFR, ALK, RET, or ROS1 driver detected and CGP on samples collected before recurrence, those with CGP results obtained any time before recurrence versus after recurrence had less time from recurrence to start of 1L therapy (median 2.7 versus 6.7 wk, p < 0.001). A total of five patients have record of starting 1L therapy shortly (≤2 wk) before advanced diagnosis. 1L, first line; CGP, comprehensive genomic profiling; LUAD, lung adenocarcinoma; Q, quartile.
We also calculated the expected incremental cost of routine CGP testing in early LUAD as an alternative to single-gene EGFR mutation testing, with the assumption that CGP detection of ALK, ROS1, and RET driver rearrangements could identify patients who would most likely not respond to atezolizumab, an ICI approved in the adjuvant setting. This assumption is based on data from studies in the advanced disease setting and the IMpower010 study revealing no clear improvement in survival with ICI in patients with these oncogenic drivers.
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
A probabilistic decision tree was used to estimate the incremental cost of CGP testing (Fig. 5A), where patients with ALK, RET, and ROS1 driver alterations identified on CGP avoid adjuvant atezolizumab. Costs of CGP and EGFR single-gene testing are estimated at $3500 and $324.58, respectively, and treatment with atezolizumab for a year is approximately $160,199.04 (including both drug and administration costs). In the intention-to-treat group in IMpower010, 53.2% of the patients were PD-L1+ (TPS ≥ 1%); 4.8% of IMpower010 PD-L1+ patients also had an ALK driver rearrangement. In the CGDB, among patients with early LUAD with PD-L1+ disease, a RET, ROS1, or ALK rearrangement was detected in 1.4%, 0.8%, and 1.4% of cases, respectively (Supplementary Table 5). Because they were not available from IMpower010, we used the CGDB-derived estimates of ROS1 and RET rearrangements among patients with PD-L1+ early LUAD as our base-case parameters for a total ICI avoidable patient population of 7% based on ALK, RET, and ROS1 status (Supplementary Table 1).
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
Figure 5Cost implications and assumptions for testing paradigms in early stage LUAD. (A) Probabilistic decision tree with alternative testing strategies of single-gene EGFR and PD-L1 IHC compared with multigene testing for at least EGFR, ALK, RET, and ROS1 using CGP. (B) Incremental cost of CGP testing strategy compared with EGFR single-gene testing in early stage LUAD reveals an expected total cost reduction per patient with universal CGP testing compared with EGFR single-gene testing of $1597.23. Cost assumptions are found in Supplementary Table 1. (C) One-way sensitivity analysis. Cost base values and minimum and maximum estimates are described in Supplementary Table 3. +, positive; CGP, comprehensive genomic profiling; ICI, immune checkpoint inhibitor; IHC, immunohistochemistry; LUAD, lung adenocarcinoma; Max, maximum; Min, minimum; PD-L1, programmed death-ligand 1; TPS, tumor proportional score; Tx, treatment.
Using these estimated costs and prevalence estimates, $3175.42 more would be spent on diagnostics per person with CGP instead of just EGFR testing, but $4772.65 could be saved per person on average by avoiding ICI therapy expenses in patients who are both PD-L1 1%+ and ALK, RET, or ROS1+. Universal CGP testing for early stage LUAD could represent an overall cost-saving testing strategy per patient with a total expected incremental cost reduction of $1597.23 (Fig. 5B). These results hold under a wide range of parameter assumptions including a PD-L1+ prevalence as low as 35.4% and CGP testing costing up to $5097.23 (Supplementary Table 2). The most influential parameters in the one-way sensitivity analysis were ALK, RET, and ROS1+ prevalence among PD-L1+ patients when considered together as the total population identified to avoid adjuvant ICI therapy, the ALK+-specific prevalence, the percent of eligible patients for adjuvant ICI therapy who received treatment, and the PD-L1+ prevalence (Fig. 5C and Supplementary Table 3). Across 10,000 parameter draws in the probabilistic sensitivity analysis, CGP was the cost-saving testing strategy in 84% of the simulations (Supplementary Fig. 2).
Discussion
To the best of our knowledge, this is the first study characterizing the use of CGP in patients with early stage LUAD, including the evaluation of potential clinical and financial benefits. Currently, guidelines recommend broad genomic testing only for metastatic NSCLC, but with utilization of a real-world database, a group of patients who underwent CGP in the early disease setting with relatively even distribution of stages I, II, and IIIA diagnosis were selected.
In this population, CGP most often occurred within 3 months of initial diagnosis. This could suggest that there are certain practitioners or practices that are already routinely using CGP in patients with early stage disease, and this trend seems to be increasing over time.
The prevalence of oncogenic drivers was similar between CGP specimens collected in the early versus advanced setting, though KRAS mutations were somewhat more frequent in early disease and other drivers were somewhat more frequent in advanced disease. When limited to only PD-L1+ disease, KRAS and ALK were the only oncogenes that continued to have a significant difference. Although the exact reason for this difference remains elusive, it may be influenced by testing practice patterns or by screening practices by which patients with strong smoking history are more likely to qualify for lung cancer screening whereas patients with EGFR or ALK are more likely to be never smokers and thus more likely to bypass screening and present with advanced disease. These genomic findings highlight the potential utility of KRAS inhibitors in the perioperative setting and should be further explored.
This study raises several potential advantages of CGP in early stage LUAD. First, there are potential advantages of the immediate availability of information regarding targetable genomic alterations at the time of recurrence, which leads to a decrease in time to first-line treatment of approximately 3 weeks. Although it is not clear whether the difference of a few weeks would lead to improved survival, symptomatic patients may be palliated faster, and it may benefit patients from an anxiety and mental health perspective. The practitioner’s time would also be spent more efficiently. These potential benefits could be explored in a small prospective trial. Furthermore, as new assays are developed to monitor molecular residual disease, early CGP tissue testing may provide added value as a baseline for these assays.
Molecular residual disease (MRD) detection with a tissue comprehensive genomic profiling (CGP)-informed personalized monitoring assay: an exploratory analysis of the IMvigor-010 observation arm.
Comprehensive genomic profiling (CGP)-informed personalized molecular residual disease (MRD) detection: an exploratory analysis from the PREDATOR study of metastatic colorectal cancer (mCRC) patients undergoing surgical resection.
There are also an increasing number of precision-targeted therapy trials in early lung cancer that CGP could aid in enrollment; these are summarized in Supplementary Table 4.
The difference in the percentage of patients with CGP specimen collected in early disease treated with appropriate first-line–matched targeted therapy for ALK, RET, or ROS1 between those with CGP testing performed in the early versus advanced setting was not significantly different from a statistical point of view, but there was a positive trend favoring those with earlier testing. Approximately 10% more patients were treated with the guideline-recommended targeted therapy when CGP was already completed at time of recurrence. It is possible that this reflects the ability to initiate therapy more quickly among patients who could otherwise go untreated due to poor performance status or need to initiate therapy rapidly owing to severity of disease.
Cost benefits of universal CGP compared with single-gene EGFR testing could come in the form of avoiding potentially unnecessary adjuvant atezolizumab in patients with ALK, ROS1, or RET driver alterations, an estimated average savings of $1597.23 per patient tested with CGP. Because adjuvant atezolizumab is approved for patients with PD-L1+ disease, the prevalence of PD-L1 1%+ among patients with LUAD is an important parameter to define the total addressable population. We relied on the IMpower010 study-specific estimates of 53.2% PD-L1+ and 4.8% ALK+ to inform our costing analysis, estimates consistent with other trial populations.
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
In our real-world CGDB cohort, 55% of specimens from early LUAD and 63% from advanced LUAD were PD-L1+; however, only 1.4% of early PD-L1+ cases were ALK+, perhaps reflecting a bias toward patients with positive ALK testing by IHC or FISH results not being referred for CGP. In a threshold analysis, we estimated that PD-L1+ prevalence would need to be at least 35.4% and the prevalence of ALK, ROS1 or RET drivers in the PD-L1+ population at least 4.7% in total for CGP to be the cost saving or cost neutral choice.
As noted previously, the cost analysis depends on the assumption that patients with ALK, ROS1, or RET driver alterations would probably not have a survival benefit with adjuvant ICI. This is based on available data in the metastatic setting and from the IMpower010 study, where the few patients with EGFR or ALK alterations did not have a clear improvement in disease-free survival (hazard ratios: 0.99 and 1.04, respectively).
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
EGFR mutations and ALK rearrangements are associated with low response rates to PD-1 pathway blockade in non–small cell lung cancer: a retrospective analysis.
Although this has not yet been proven definitively, and no biomarker guidance for adjuvant ICI is included yet in the NCCN guidelines, several relevant trials are ongoing.
We believe that with the existing and anticipated new data, some physicians may consider avoiding adjuvant atezolizumab in this patient population. For these practitioners in particular, our cost analysis would be relevant. Beyond the costs, there is also the clinical benefit of reduced unnecessary ICI toxicity with the appropriate avoidance of adjuvant ICI. In IMpower010, 52% of the patients receiving atezolizumab experienced an immune-related adverse event compared with just 9% of the patients who got best supportive care and 12% of patients in the atezolizumab arm required treatment with systemic corticosteroids.
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
A phase 1 study of gefitinib combined with durvalumab in EGFR TKI-naive patients with EGFR mutation-positive locally advanced/metastatic non-small-cell lung cancer.
There are multiple limitations that should be noted for this study. First, at this point, adjuvant therapy is recommended for patients with stages II to IIIA NSCLC. There is also a group of high-risk patients with stage IB disease for whom adjuvant therapy can be considered, but for all IA and most IB disease, adjuvant treatment is not used or recommended. These stages were not excluded in the final analysis. Assuming PD-L1 and the driver alteration rates are similar in this population though (as they are similar in both early and advanced diseases), this should not change the final cost benefit results. Second, in March 2022, the FDA approved neoadjuvant nivolumab based on the results of the CHECKMATE-816 trial. Given the newness of this approval, the cost of the avoidance of neoadjuvant nivolumab was not calculated in our study, and there is no consensus in the thoracic oncology community about when to use neoadjuvant or adjuvant ICI (or maybe even both). We do feel that there are patients in whom adjuvant ICI would be used and that this analysis would still be applicable and important for this group. It would be interesting to investigate how CGP could potentially play a role in neoadjuvant selection. Third, this study is biased as all included patients were those who received CGP testing. Finally, clinical information, including receipt of therapy, was captured as documented in the EHR and events occurring outside of the FH network and not documented in the available EHR were not captured. In particular, regarding assessing absolute rates of targeted therapy use in this study, some patients may have received targeted therapy outside the Flatiron network, which would not have been captured, received targeted therapy in later lines, or been treated before therapies currently on-label were approved. In addition, clinical study drugs were masked in this analysis; thus, the patients could have received targeted therapy on a clinical trial and patients with stage IIIB disease may have received curative-intent chemoradiation.
In conclusion, CGP in early stage LUAD can identify EGFR, ALK, ROS1, RET, and other driver alterations and accelerate the start of appropriate first-line TKI at recurrence. CGP could also represent a more cost-effective approach by avoiding potentially futile ICI in patients with driver alterations that have historically lacked activity with ICI in metastatic disease. Additional investigation is needed to confirm that adjuvant ICI should be avoided in these patients, and with the recent approval of neoadjuvant nivolumab, the use of adjuvant atezolizumab may be more limited. Regardless, the recent advances and ongoing trials in perioperative treatment for NSCLC highlight that early stage disease will need a more individualized, nuanced approach. Use of CGP in early stage disease would be helpful in making this process more efficient and effective.
CRediT Authorship Contribution Statement
Bharathi Muthusamy: Conceptualization, Writing—original draft, Writing—review and editing.
Kira Raskina: Formal analysis, Visualization, Data curation, Software, Writing—review and editing.
Katherine Lofgren: Formal analysis, Visualization, Writing—review and editing.
Gerald Li: Formal analysis, Visualization.
Khaled Tolba: Visualization, Writing—review and editing.
Karen Schwed: Writing—review and editing.
Emily Castellanos: Writing—review and editing.
Rischard Huang: Writing—review and editing.
Geoffrey R. Oxnard: Conceptualization, Supervision, Writing—review and editing.
Alexa Schrock: Supervision, Writing—original draft, Visualization, Writing—review and editing.
Nathan Pennell: Supervision, Writing—review and editing.
Adjuvant atezolizumab after adjuvant chemotherapy in resected stage IB-IIIA non-small-cell lung cancer (IMpower010): a randomised, multicentre, open-label, phase 3 trial.
Association of patient characteristics and tumor genomics with clinical outcomes among patients with non-small cell lung cancer using a clinicogenomic database.
Molecular residual disease (MRD) detection with a tissue comprehensive genomic profiling (CGP)-informed personalized monitoring assay: an exploratory analysis of the IMvigor-010 observation arm.
Comprehensive genomic profiling (CGP)-informed personalized molecular residual disease (MRD) detection: an exploratory analysis from the PREDATOR study of metastatic colorectal cancer (mCRC) patients undergoing surgical resection.
EGFR mutations and ALK rearrangements are associated with low response rates to PD-1 pathway blockade in non–small cell lung cancer: a retrospective analysis.
A phase 1 study of gefitinib combined with durvalumab in EGFR TKI-naive patients with EGFR mutation-positive locally advanced/metastatic non-small-cell lung cancer.
Disclosure: Drs. Raskina, Lofgren, Li, Tolba, Huang, Oxnard, and Schrock are employees of Foundation Medicine, Inc., a wholly owned subsidiary of Roche, and have stock ownership in Roche. Drs. Castellanos and Schwed report having employment in Flatiron Health, Inc., an independent subsidiary of the Roche group, and stock ownership in Roche. Dr. Castellanos also reports equity ownership in Flatiron Health, Inc. Dr. Pennell reports consulting/advisory roles with Merck, Pfizer, Eli Lilly/LOXO, Genentech, Amgen, Mirati, G1 Therapeutics, ResistanceBio, Xencor, Janssen, Boehringer Ingelheim, and Sanofi Genzyme. Dr. Muthusamy declares no conflict of interest.
01937-2&id=gr1.jpg){kind=link}
01937-2&id=gr2.jpg){kind=link}
01937-2&id=gr3.jpg){kind=link}
01937-2&id=gr4.jpg){kind=link}
01937-2&id=gr5.jpg){kind=link}