2.3.1 Biomarkers for adjuvant therapy protocol

There were discrepancies between the IMpower010 [8, 9] and KEYNOTE-091 [10] trials. An exploratory analysis of the IMpower010 suggested that the DFS benefit was associated with PD-L1 expression (PD-L1 < 1% [HR = 0.97], PD-L1 = 1%-49% [HR = 0.87], PD-L1 ≥ 50% [HR = 0.43]) within the stage II–IIIA population [8]. In contrast, the DFS benefit in the KEYNOTE-091 trial was independent of PD-L1 expression, and adjuvant immunotherapy also conferred a DFS benefit to the PD-L1 < 1% population (HR = 0.78, 95% CI = 0.58-1.03). There was even a negative DFS in the population with tumors expressing PD-L1 on ≥ 50% of cancer cells as a study endpoint [10]. Factors that may have contributed to these differences include the following: (1) Patients registered in the IMpower010 trial received adjuvant chemotherapy, whereas adjuvant chemotherapy was not a requirement in the KEYNOTE-091 trial (14.2% of patients did not receive adjuvant chemotherapy). (2) The proportion of patients with stage IIIA disease was higher in the IMpower010 trial than in the KEYNOTE-091 trial (40.4% vs. 30.0%), whereas stage IB disease was more common in the KEYNOTE-091 trial (14.2% vs. 12.8%). (3) IMpower010 was an open-label trial, whereas KEYNOTE-091 was a randomized, triple-blind trial. (4) The two trials used different assays to determine the PD-L1 expression status (PD-L1 IHC22C3 reagent in KEYNOTE-091 vs. SP142 assay in IMpower010). (5) Finally, in the KEYNOTE-091 trial, the PD-L1 ≥ 50% subpopulation in the placebo group had relatively superior mDFS compared with the other PD-L1 subgroups; this difference may be explained by randomization considerations and is the potential cause of negative results in the PD-L1 ≥ 50% population. Overall, we believe that PD-L1 expression plays a role in predicting the efficacy of adjuvant immunotherapy.

On-treatment ctDNA dynamics also helped to predict the efficacy. The IMpower010 trial showed that postoperative ctDNA non-clearance was associated with worse DFS. However, both the ctDNA-uncleared and ctDNA-cleared stage II-IIIA populations benefitted from adjuvant ICI therapy; only the ctDNA-uncleared, PD-L1 < 1% population did not achieve a significant DFS benefit (HR = 0.88; 95% CI = 0.40-1.91) [20]. A recent trial including patients with stage I-III NSCLC who underwent definitive surgery showed that only 3.2% of patients with longitudinal undetectable MRD developed recurrence, resulting in a negative predictive value of 96.8%; the positive predictive value of longitudinal detectable MRD was 89.1%. Although adjuvant therapy did not improve DFS among patients with undetectable MRD (HR = 2.29, 95% CI = 0.85-6.11), it conferred a survival benefit for patients with detectable MRD (HR = 0.34, 95% CI = 0.12-0.88) [21]. Similarly, the LUNGCA-1 trial (NCT03317080) showed that the postoperative presence of ctDNA-based MRD was associated with disease relapse (HR = 11.1) [22]. Furthermore, recurrence-free survival was higher in MRD-uncleared patients who received adjuvant therapy than in MRD-uncleared patients who did not receive adjuvant therapy (HR = 0.3, P = 0.008). In MRD-negative patients, adjuvant therapy did not prolong recurrence-free survival. Therefore, dynamic monitoring of ctDNA/MRD in the perioperative period demonstrates the potential to predict the risk of recurrence and may indicate a need for adjuvant therapy.

2.5.1 Neoadjuvant mono-immunotherapy

The CheckMate 159 trial (NCT02259621) demonstrated a major pathological response (MPR) rate of 45% and a pathological complete response (pCR) rate of 10% in patients with stage I-IIIA resectable disease after 2 preoperative cycles of nivolumab; recurrence-free survival (RFS) rates at 3, 4, and 5 years were 65%, 60%, and 60%, respectively, and the corresponding OS rates were 85%, 80%, and 80% [49, 50]. The survival benefit reached a plateau after 3 years, reflecting the sustained benefit of neoadjuvant nivolumab immunotherapy. The results of the LCMC3 trial (NCT02927301) also suggested that the application of neoadjuvant atezolizumab monotherapy to stage IB-selective IIIB NSCLC led to a substantially high survival plateau (3-year DFS and OS rates of 72% and 80%, respectively); however, it was associated with relatively low rates of MPR (20.4%) and pCR (6.8%) [51, 52]. Similarly, a trial involving a Chinese population (ChiCTR-OIC-17013726) showed that 2-cycle neoadjuvant sintilimab immunotherapy was associated with an MPR rate of 40.5%, a primary tumor pCR rate of 16.2%, and a lymph node pCR rate of 8.1% in patients with stage IA-IIIB NSCLC; the most pronounced responsiveness was derived from squamous cell carcinoma [53]. The rationale of administering immunotherapy to patients with resectable NSCLC in the neoadjuvant setting was also evaluated. Caushi et al. [54] showed that patients with MPR after neoadjuvant mono-immunotherapy had higher expression levels of genes associated with memory (interleukin-7 receptor [IL7R] and transcription factor 7 [TCF7]) and effector function (granayme K [GZMK]) in mutation-associated neoantigen-specific T cells. Moreover, mutation-associated neoantigen-specific T-cell-related disorder genes (TOX high mobility group box family member 2 [TOX2], cytotoxic T lymphocyte associate protein 4 [CTLA4], hepatitis A virus cellular receptor 2 [HAVCR2], and ectonucleoside triphosphate diphosphohydrolase 1 [ENTPD1]) were significantly higher in the non-MPR population; immune checkpoint and depletion scores tended to be higher. The study provided new insights concerning immune efficacy assessment and efforts to overcome immune resistance.

2.5.2 Neoadjuvant and perioperative chemoimmunotherapy

In the field of neoadjuvant/perioperative chemoimmunotherapy, the first questions are as follows: does the combined therapy produce an improved pathological response, and how does it influence surgical treatment? The NADIM trial (NCT03081689) was a milestone study concerning the feasibility of neoadjuvant chemoimmunotherapy. Although it had a single-arm design, the results showed an MPR rate of 83%, a pCR rate of 63%, and a pathological downstage rate of 90% [55, 56]. The subsequent NADIM II trial (NCT03838159) confirmed that, compared with neoadjuvant chemotherapy, neoadjuvant chemoimmunotherapy in patients with stage IIIA-B NSCLC significantly improved the pCR rate (7% vs. 37%, relative risk = 5.34), MPR rate (14% vs. 52%), and objective response rate (ORR; 48% vs. 74%). Although there was an increase in grade 3-4 adverse events (10% vs. 24%), they were generally manageable and did not affect subsequent surgical treatment. The PD-L1 status had predictive value for the pCR rate (P = 0.008) [57, 58]. Later, as the first phase III analysis of neoadjuvant chemoimmunotherapy, the CheckMate 816 trial (NCT02998528) showed that neoadjuvant chemoimmunotherapy significantly increased the MPR rate (36.9% vs. 8.9%) and pCR rate (24.0% vs. 2.2%, P < 0.001) compared with neoadjuvant chemotherapy, reaching a pCR rate of 23.0% (vs. 0.9%) in patients with stage IIIA disease. The proportion of patients undergoing surgery was higher among patients who received neoadjuvant chemoimmunotherapy than among the control group (83% vs. 75%), but there was no increase in surgical difficulty [59, 60].

The feasibility of neoadjuvant chemoimmunotherapy has been confirmed in four other recent randomized phase III clinical trials using perioperative ICI therapy: AEGEAN (NCT03800134) [61], NEOTORCH (NCT04158440) [62], KEYNOTE-671 (NCT03425643) [63], and CheckMate 77T (NCT04025879) [64]. The MPR rates in these studies were 30.2%-48.5%, and the pCR rates were 17.2%-25.3%. The MPR and pCR rates varied among trials varied, presumably because of differences in the pathological evaluation protocols. The International Association for the Study of Lung Cancer further standardized the pathological evaluation of postoperative resection specimens after neoadjuvant therapy, recommending evaluation of surgical specimens to determine the percentages of surviving tumor, necrosis, and stroma displaying fibrosis and inflammation. Moreover, MPR should be regarded as a reduction in the percentage of active tumors below the optimal threshold value of clinical significance, which is determined by histologic type and prior evidence of a specific treatment regimen. The optimal threshold for MPR was previously defined as 10% for all types of lung cancer, whereas recent studies suggest that the optimal threshold varies according to histologic type (squamous: 10%, adenocarcinoma: 65%). Thus, further validation is needed [65].

In terms of safety, the CheckMate 159 [49] and LCMC3 [51] trials showed that the rates of grade 3 and 4 treatment-related adverse events (TRAEs) were 4.5% and 6.0%, respectively; these were lower than the rate of 33.5% observed in the chemoimmunotherapy arm of the CheckMate 816 trial. However, a meta-analysis showed that although neoadjuvant chemoimmunotherapy increased the incidence of TRAEs compared with immunotherapy monotherapy (73.9% vs. 42.9%), the incidence of severe adverse events did not significantly increase (18% vs. 12.3%, P = 0.32). Moreover, the rates of surgical resection were similar between the two arms (84.4% vs. 89.2%), whereas the incidence of delayed surgery was lower in the neoadjuvant chemoimmunotherapy arm [66]; therefore, the overall adverse effects of the neoadjuvant chemoimmunotherapy strategy were manageable.

Considering that combined modalities improve the MPR rate, the second question regarding neoadjuvant/perioperative chemoimmunotherapy is as follows: what is the optimal duration of inductive chemoimmunotherapy? In addition to ongoing clinical trials concerning the efficacies of various ICIs in neoadjuvant chemoimmunotherapy strategies [67-69], the neoSCORE trial (NCT04459611) is currently exploring the optimal number of neoadjuvant cycles for chemoimmunotherapy. Thus far, this trial has demonstrated numerical superiority in the three-cycle group compared with the two-cycle group (MPR rate: 41.4% vs. 26.9%, P = 0.260; pCR rate: 24.1% vs. 19.2%, P = 0.660) [70]. The optimal number of neoadjuvant therapy cycles is mainly determined by the MPR rate, but there is a logical paradox: if a patient does not undergo surgery, their pathological remission status cannot be determined. Therefore, an accurate method for preoperative evaluation of the pathological result is urgently needed. Indeed, artificial intelligence-based deep learning approaches have emerged to overcome this limitation; one relevant study showed that a combined model integrating clinical features and artificial intelligence-based deep learning scores could effectively predict whether patients with NSCLC achieved MPR after neoadjuvant chemoimmunotherapy (internal validation cohort: area under the curve = 0.77, 95% CI = 0.64-0.89; external validation cohort: area under the curve = 0.75, 95% CI = 0.62-0.87). The study also revealed that the underlying biological basis of deep learning scores may be related to microenvironmental activation of signaling pathways that mediate tumor proliferation and promote antitumor immune cell infiltration [71]. However, the mechanisms underlying neoadjuvant chemoimmunotherapy outcomes have also been reported. Hui et al. [72] scanned the entire immune microenvironment and found a decreased number of intratumoral-activated Tregs in MPR tumors, along with clonal expansion of peripheral cytotoxic CD8⁺ T cells. Another study revealed therapy-promoting expansion of cytotoxic T cells and activation of memory CD8⁺ T cells into an effector phenotype, as well as MPR-associated reduction of neutrophil heterogeneity in the aged CCL3⁺ neutrophil/secreted phosphoprotein 1(+)/tumor-associated macrophage pathway [73].

The third and most critical question regarding neoadjuvant/perioperative chemoimmunotherapy is as follows: can the remarkable pathological response be translated into an improved prognosis? Cross-study comparisons should be performed with caution. However, considering the different study designs and chemotherapy regimens, it is intriguing that the HRs for disease progression, disease recurrence, or death were 0.68 in the CheckMate 816 trial, 0.81 (95% CI = 0.67-0.99) in the IMpower010 trial, and 0.76 (95% CI = 0.63-0.92) in the KEYNOTE-091 trial. These findings indicate that neoadjuvant ICI therapy has a more central role. In support of this perspective, the findings in related studies have suggested that the tumor-draining lymph nodes are important components of the antitumor immunotherapy response; therefore, standard lymph node dissection during surgical resection may affect subsequent efficacy of adjuvant immunotherapy [74-76]. The next priority involves determining whether a perioperative approach would provide benefits beyond either approach alone.

Although positive results were obtained in all perioperative trials, comparable advantages were reported over the CheckMate 816 trial in terms of the 2-year event-free survival (EFS) rate (CheckMate 816: 63.8%; KEYNOTE-671: 61.5%; AEGEAN: 63.3%) and the HR for EFS (CheckMate 816: HR = 0.63 [95% CI = 0.43-0.91]; KEYNOTE-671: HR = 0.59 [95% CI = 0.48-0.72]; AEGEAN: HR = 0.68 [95% CI = 0.53-0.88]; CheckMate 77T: HR = 0.58 [97.36% CI = 0.42-0.81]). Moreover, in the comparison of HR among patients with stage IIIA disease, advantages were almost identical (CheckMate 816: HR = 0.54 [95% CI = 0.37-0.80]; KEYNOTE-671: HR = 0.57 [95% CI = 0.44-0.74]; AEGEAN: HR = 0.57 [95% CI = 0.39-0.83]; NEOTORCH: HR = 0.44 [95% CI = 0.287-0.661]) (Table 5, Figure 1). Very recently, the KEYNOTE-671 firstly reported a statistically significant and clinically important OS improvement (HR 0.72, 95% CI = 0.56-0.93) [77]. However, considering the cost-effectiveness ratio and ICI-related toxicity, more evidence is required to support the use of a perioperative immunotherapy strategy; the final answer to this question will be determined by eventual survival outcomes.

2.5.3 Neoadjuvant dual immunotherapy

The results of the NEOSTAR trial (NCT04230109) suggested that neoadjuvant nivolumab plus ipilimumab improved the rates of MPR (38% vs. 22%) and pCR (29% vs. 9%) compared with nivolumab alone, and the safety profiles were identical (10% vs. 13%); however, more patients in the combination group received nonsurgical treatment (4 vs. 1) [78]. Although dual ICI therapy had better efficacy, this efficacy was offset by the loss of surgical opportunities. PD-L1 expression was correlated with the pathological response. Another study (NCT02259621) evaluating nivolumab plus ipilimumab as a neoadjuvant regimen was discontinued because of the incidence of TRAEs reached 67%, including a 33% incidence of grade ≥ 3 TRAEs [79]. Other dual ICI therapy trials include analyses of volrustomig (anti-PD-1/CTLA4), as well as PD-1/PD-L1 inhibitors combined with oleclumab (anti-CD73), monalizumab (anti-NKG2A) [80], danvatirsen (anti-STAT3) [81], or relatlimab (anti-LAG3) [82]. Considering the limited efficacy in the advanced setting, these novel immune checkpoints require further investigations.

2.5.4 Neoadjuvant radiotherapy combined with immunotherapy

Altorki et al. [83] reported a significantly higher MPR rate with neoadjuvant durvalumab and stereotactic body radiotherapy, compared with durvalumab alone (53.3% vs. 6.7%); the pCR rate in the combined group reached 26.7%. No additional safety concerns were reported (grade 3-4 adverse events: 20% vs. 17%); only one (3%) of 33 patients in the combined group underwent delayed surgery because of adverse reactions. The results of another study suggested that durvalumab combined with neoadjuvant sub-ablative radiation did not prolong the duration of surgery, and no new safety concerns were identified (grade 3-4 adverse events: 38%) [84]. Therefore, neoadjuvant immunotherapy combined with radiotherapy might be a chemotherapy-free alternative for patients with a large tumor load and N2 metastases who can undergo surgery. The ongoing SQUAT study (WJOG 12119L) [85] will provide further insights concerning the benefits of neoadjuvant immunotherapy combined with radiotherapy.

4.1.1 Postoperative adjuvant targeted therapy in resectable EGFR-mutated LA-NSCLC

The BR19 (NCT00049543) [109] and RADIANT (NCT00373425) [110] trials initiated the early exploration of postoperative adjuvant targeted therapy for NSCLC; both trials yielded negative results because of their non-selective design. During the SELECT trial (NCT00462995), adjuvant erlotinib showed an improved 2-year DFS compared with historic EGFR-mutant controls [111]. The ADJUVANT (NCT01405079) and EVIDENCE (NCT02448797) trials confirmed that tyrosine kinase inhibitors (TKIs) had a DFS benefit in the adjuvant setting [112-114], whereas the IMPACT trial (NCT02164513) reported negative results for DFS [115]. Importantly, none of the trials demonstrated an OS benefit, and they revealed poor control of distant and brain metastases [112-114, 116, 117]. On top of the superior PFS and OS outcomes of third-generation EGFR-TKI observed in advanced NSCLC in FLAURA trial [118], the ADAURA trial (NCT02511106) explored the efficacy and safety of osimertinib as adjuvant therapy compared with placebo [119]. The patients received or did not receive postoperative adjuvant chemotherapy and were then randomized, revealing a dual benefit of TKI single-agent targeted therapy and sequential targeted therapy with chemotherapy. The data showed an mDFS of 65.8 vs. 21.9 months (HR = 0.23) in patients with stage II-IIIA disease and an mDFS of 65.8 vs. 28.1 months (HR = 0.27) in patients with stage IB-IIIA disease. For the planned final analysis of OS, the 5-year OS rates were 85% and 73% (HR = 0.49) among patients with stage II to IIIA disease; these rates were 88% and 78% (HR = 0.49) among patients with stage IB to IIIA disease. The results of subgroup analysis suggested that the survival benefit was more pronounced in patients with stage IIIA (HR = 0.37) disease than in patients with stage IB (HR = 0.44) or II (HR = 0.63) disease [120, 121] (Table 7, Figure 1). Analyses of other third-generation TKIs for adjuvant therapy are currently underway (Figure 2).

Although the above-described trials revealed successful use of adjuvant targeted therapy to manage NSCLC, they also revealed some problems (Figure 1). The ADJUVANT study showed a significant convergence of DFS curves after approximately 3 years of follow-up. Similarly, the ADAURA trial suggested rapid convergence of DFS curves after nearly 4 years of follow-up [122]. The phenomenon of rapid relapse after the end of adjuvant therapy suggests that adjuvant targeted therapy can only delay the time to disease recurrence; it cannot cure the disease.

There is no conclusive evidence regarding the optimal duration of adjuvant targeted maintenance. The results of the ICOMPARE trial (NCT02274818) suggested that adjuvant treatment with icotinib for 2 years significantly prolonged the mDFS (HR = 0.512) and improved the 6-year OS rate, compared with 1-year maintenance (74.6% vs. 51.3%) [123]. Adjuvant therapy-related studies of gastrointestinal stromal tumors [124] and breast cancer [125] suggested that further extension of adjuvant therapy can yield better survival benefits. However, patient's compliance decreased with the development of adverse events. Concomitantly, the increased socioethical and economic costs associated with a prolonged duration of targeted therapy should also be considered. There is evidence that ctDNA/MRD is a predictor of postoperative disease recurrence and poor prognosis [21, 22, 126]. Additionally, the time to disease recurrence based on assessment of ctDNA/MRD status was earlier than the time to recurrence based on clinical imagings [19, 127]. These findings support that ctDNA/MRD can be used to identify patients at high risk of recurrence; thus, we suggest that people at high risk of postoperative recurrence should be screened according to ctDNA/MRD status, then administered intensive adjuvant targeted therapy. The ADAURA trial is currently exploring the feasibility of combining the ctDNA/MRD status for adjuvant treatment with osimertinib, and we look forward to publication of the results.

4.1.2 Postoperative adjuvant chemotherapy in resectable EGFR-mutated LA-NSCLC

The findings of retrospective studies have suggested that postoperative adjuvant chemotherapy provides minimal survival benefits to patients with an EGFR mutation [128, 129]. In contrast, the ADAURA trial reported a DFS benefit favoring osimertinib vs. placebo in patients with (HR = 0.16) and without adjuvant chemotherapy (HR = 0.23), regardless of disease stage [130]. Comprehensive genomic profiling of patients in the ADJUVANT trial suggested 5 potential predictive biomarkers for DFS: RB1 alterations (favoring adjuvant chemotherapy) and NKX2-1 copy number gain, CDK4 copy number gain, TP53 exon4/5 missense alterations, and MYC copy number gain (all favoring adjuvant gefitinib) [131]. Overall, although EGFR-TKI plays a determinant role in the adjuvant setting, adjuvant chemotherapy remains the recommendation according to international guidelines (stage IB [tumors of ≥ 4 cm] to IIIA; AJCC Seventh Edition) and for EGFR-mutated NSCLC.

4.1.3 Neoadjuvant targeted therapy in resectable EGFR-mutated LA-NSCLC

The EMERGING trial (NCT01407822, CTONG 1103) explored differences in the efficacy of erlotinib vs. chemotherapy as a neoadjuvant regimen for IIIA (N2) EGFR-mutated NSCLC. There was no statistically significant difference regarding ORR (54.1% vs. 34.3%), and the MPR rate was 9.7% vs. 0%; pCR was not observed in either group. Although there was a benefit in terms of PFS (mPFS: 21.5 vs. 11.4 months, HR = 0.36) and a significant decrease in the incidence of grade 3-4 adverse events (0% vs. 29.4%), OS remained negative (mOS: 42.2 vs. 36.9 months; HR = 0.83) [132, 133]. Another non-comparative phase II trial revealed similar results [134]: the NEOS trial (ChiCTR1800016948) showed an ORR of 71.1%, with an MPR of 11% and pCR of 4%. The downstaging rate was 53.3% when osimertinib was used as a neoadjuvant treatment strategy in patients with EGFR-mutated stage II-IIIB NSCLC [135, 136]. Recently, another phase II single-arm trial testing the efficacy of neoadjuvant osimertinib in EGFR-mutated NSCLC also showed negative results: 15% MPR, 0% pCR, and 44% lymph node downstaging rate [137]. Considering these findings, the efficacy of neoadjuvant TKI therapy is unsatisfactory, despite the use of third-generation EGFR-TKIs (Table 8). Additional trials are urgently needed to explore the roles of combined chemotherapy and TKI therapy, define the optimal TKI inductive duration, and test other neoadjuvant strategies (e.g., chemoimmunotherapy) in the neoadjuvant EGFR-mutated scenario (Figure 2).

4.1.4 Perioperative immunotherapy in resectable EGFR-mutated LA-NSCLC

There is considerable evidence that immunotherapy has minimal effects on EGFR-mutated NSCLC, even after these cancers have become resistant to EGFR-TKIs [138, 139]; there is a risk of hyper-progression in the advanced setting [140]. As mentioned above, EGFR-TKIs have distinct roles at different stages of EGFR-mutated NSCLC. Thus, it is reasonably possible to determine whether perioperative immunotherapy is feasible in the EGFR-positive population. In the advanced disease setting, most perioperative immunotherapy trials excluded patients with EGFR mutations. However, some evidence is available. In the LCMC3 trial, no MPR was observed in patients with known EGFR mutations (n = 7). The CheckMate 816 trial excluded patients with known EGFR mutations or ALK fusions, although the trial protocol did not specify an EGFR/ALK testing requirement. A recent multicenter retrospective study with a limited sample size (n = 40) suggested that neoadjuvant chemoimmunotherapy in resectable driver-positive NSCLC had an ORR of 62.5%, R0 resection rate of 97.4%, MPR of 37.5%, pCR of 12.5%, and mDFS of 28.5 months [141]. The KEYNOTE-671 trial also recruited patients with EGFR-mutated cancer (n = 33); surprisingly, the neoadjuvant chemoimmunotherapy strategy achieved an EFS benefit relative to chemotherapy alone (HR = 0.09; Figure 1). Although potential clinical feasibility was observed, further validation is needed. Before approval, the application of perioperative immunotherapy to resectable oncogene-mutated LA-NSCLC should be used with caution.

In the adjuvant setting, although EGFR mutations showed a DFS benefit in the PEARLS trial, there was no DFS benefit from adjuvant atezolizumab in the IMpower010 trial. Importantly, considering that neoadjuvant TKIs have great potential for improving treatment outcomes, the prioritization of TKIs over adjuvant ICIs in patients with EGFR-mutated NSCLC would never be challenged in the postoperative setting.