As mentioned above, the definition of resectability is not universal, and practice varies across different centers and according to surgeon expertise. Across the variety of stage III disease however, some tumors are clearly resectable, while others are unequivocally unresectable. Cases that are upfront resectable consist mainly in T3 N1 tumors, and some T4 N0 or N1 disease. On the other side of the spectrum, N3 tumors are considered unresectable, as are most cases of clinically evident multi-station N2 or bulky N2 diseases. In between stands the case of stage IIIA tumors with single station, non-bulky clinically positive N2, for which surgical experience is crucial. Clinical practice varies by country, and guidelines differ on that topic [7]. In these situations, the benefit of surgical resection over CRT remains uncertain. Several trials have compared surgical with radiotherapeutic approaches. The EORTC-08941 [8] and NTOG [9] trials compared surgery vs. RT, after initial neoadjuvant CT for all patients, and showed no differences in OS or EFS. Importantly, in these trials a significant part of the patients attributed to surgery did not undergo complete R0 resection. Moreover, the sequential strategy of CT and RT used in these trials is known to be inferior to concurent CRT. Two additional phase III trials, INT013917 and ESPATUE, evaluated concurrent CRT with or without surgery. The INT0139 trial compared induction CRT followed by surgery vs. CRT followed by further RT [10]. A slight improvement in PFS was observed in favor of surgery (12.8 months vs. 10.5 months) with no difference in OS, a discrepancy potentially attributable to an excess mortality rate after pneumonectomy. In the ESPATUE trial, no OS or PFS difference was observed between induction CT followed by CRT and by surgery, vs. induction CT and CRT alone [11]. However, the trial closed prematurely, resulting in insufficient statistical power for the primary outcome.

These data and the resulting uncertainties underly the paramount importance of discussing these challenging cases in multidisciplinary tumorboard, with thoracic surgeons who are experienced in the management of these patients. In comparison to the 8th edition of the TNM staging system, the 9th edition, recently presented at the World Conference on Lung Cancer, redefined the mediastinal N stages, distinguishing between single ipsilateral N2a and multi-station ipsilateral N2b, following the observation of different prognosis in these subgroups [12]. However, the treatment choice will not be based solely on this new TNM staging, and resectability, should be determined at a multidisciplinary team discussion.

For patients presenting with unresectable stage III NSCLC, the mainstay of management is definitive concurrent CRT, with a platinum-based CT doublet. Several trials and meta-analyses demonstrated the superiority of CRT over sequential CT and RT [13] with a five-year OS of 16%. The preferred CT regimens in this situation are cisplatin-etoposide [14] and carboplatin-paclitaxel [15], as well as cisplatin-pemetrexed for non-squamous histology [16].

The management of unresectable NSCLC has changed since the results of the PACIFIC trial. This double-blind, randomized, placebo-controlled trial evaluated the efficacy and safety of 12 months of adjuvant durvalumab, an immune-checkpoint inhibitor (ICI) targeting PD-L1, after definitive CRT with a platinum-based regimen, for unresectable NSCLC. The co-primary endpoints were PFS and OS, and the secondary endpoints included PFS at 12 months and 18 months, objective response rate (ORR), duration of response and safety, among others. Seven hundred and thirteen patients were randomized 2:1, to receive durvalumab 10 mg/kg every 2 weeks for up to 12 months, or placebo [17]. The five-year survival outcomes, published in 2021, with a median follow-up of 34.2 months, demonstrated a sustained benefit with a median OS of 47.5 months in the durvalumab arm vs. 29.1 months [HR: 0.72, 95% confidence interval (CI): 0.59–0.89] in the placebo arm and a median PFS of 16.9 months in the immunotherapy arm vs. 5.6 months (HR: 0.55, 95% CI: 0.45–0.68). This benefit was observed in the majority of prespecified subgroups with the notable exception of the EGFR positive subgroup (HR: 0.84, 95% CI: 0.40–1.75) [18]. However, these exploratory analyzes and the small size of this group require in-depth analyzes to confirm this trend. The ORR was better with durvalumab than placebo (28.4% vs. 16%, P < 0.001). Since these results, the PACIFIC regimen became standard of care for unresectable stage III NSCLC [19]. Like in advanced NSCLC, the optimal duration of durvalumab is unknown [20].

Oncogene addicted lung adenocarcinoma represents a subset of NSCLC patients that have worse response rates to ICI [21]. EGFR represents the most frequent oncogene in NSCLC. There are limited data of specific treatment for EGFR mutated patients in this setting. The LAURA trial, which is in progress (NCT03521154), is a randomized phase III, double-blind, placebo-controlled, that is currently assessing the efficacy and safety for osimertinib, a third-generation irreversible oral EGFR-tyrosine kinase inhibitor, as a maintenance for unresectable NSCLC after CRT, with exon 19 deletion (ex19del) or Leu858Ar (L858R) EGFR mutations [22]. Patients are randomized 2:1 to either osimertinib 80 mg once a day or placebo, until disease progression. The primary endpoint is PFS and the secondary endpoints are among others: central nervous system PFS and cumulative incidence at 12 months and 24 months, OS, PFS by mutation status, response/disease control rate and safety. The results are expected at the end of 2023.

For stage III NSCLC that are deemed resectable upfront, the standard of care has long been surgery followed by adjuvant CT with a cisplatin doublet [23]. The commonly used regimens are cisplatin-vinorelbine, etoposide, docetaxel, pemetrexed or gemcitabine. Adjuvant CT increases OS by 5.4% at 5 years in NSCLC, for tumor of more than 4 cm or with nodal disease [24]. Recently, the major advance in the treatment landscape of resectable NSCLC is the incorporation of ICI with anti-programmed cell death 1 (anti-PD-1)/PD-L1 in the treatment sequence of these patients.

Results from the Keynote-091/PEARLS trial showed that adjuvant pembrolizumab improved DFS compared to a placebo for patients with resected stage IB–III NSCLC: DFS of 53.6 months vs. 42.0 months (HR: 0.76, 95% CI: 0.63–0.91, P = 0.0014) [25]. The IMpower010 trial showed comparable results for atezolizumab, which conferred a DFS benefit vs. observation for stage II–IIIA PD-L1 positive tumors: (HR: 0.66, 95% CI: 0.50–0.88, P = 0.0039) [26]. The latest update of IMpower010 also revealed an OS advantage in the PD-L1 high subgroup exclusively. There are differences between Keynote-91 and IMpower010. First, Keynote-091 compared ICI to placebo while IMpower010 had an open-label design. Second, all patients in IMpower010 received adjuvant CT before receiving ICIs, while in Keynote-091 adjuvant CT was at the investigator’s discretion. Third, IMpower010 clearly showed a direct association between PD-L1 expression and the benefit of ICI, while Keynote-091 did not find a benefit among patients with PD-L1 > 50%, a finding that contradicts the known predictive value of PD-L1 in metastatic stages and a growing body of data suggesting a similar predictive role of PD-L1 in early-stage NSCLC. These trials led to the approval of atezolizumab for patients with PD-L1 > 50% by the European Medicines Agency (EMA), United Kingdom, and Canada.

Recently, the results of several randomized phase II and III trials of neoadjuvant and pre-operative ICI, combined with CT for resectable NSCLC, have been presented or published [27]. The theoretical advantage of a neoadjuvant approach stands in an enhance antitumor immune response as tumor antigens and lymph nodes are still in place [28, 29].

The first phase III data came from the CheckMate-816 trial, which compared three neoadjuvant cycles of nivolumab and CT with CT alone [30]. This is the only phase III trial that did not include an adjuvant phase. Nivolumab improved the coprimary endpoints of EFS, with an (HR: 0.63, 95% CI: 0.49–0.93), and pCR, at 24.0% (95% CI: 18.0–31.0) vs. 2.2% (95% CI: 0.6–5.6). CheckMate-77T, on the other hand, evaluated the use of peri-operative ICI in combination with CT. Treatment consisted in 4 cycles of nivolumab-CT or placebo-CT, followed by surgery and adjuvant treatment of nivolumab 480 mg every four weeks or placebo for one year in total [31]. The primary endpoint of EFS was met, with a median EFS not reached vs. 18.4 months, and a HR for EFS of 0.58 (97.36% CI: 0.42–0.81; P = 0.00025). The Keynote-671 trial compared perioperative pembrolizumab and CT to neoadjuvant CT alone in resectable stage II–IIIB NSCLC, according to the TNM American Joint Committee on Cancer (AJCC) 8th edition [32, 33]. There were 4 cycles of neoadjuvant CT with placebo or pembrolizumab, followed by 13 cycles of adjuvant placebo or pembrolizumab. After a median follow-up of 36.6 months, EFS was improved in the intervention arm with (HR: 0.59, 95% CI: 0.48–0.72). MPR, defined as 10% or less of viable tumor cells was improved at 30.2% vs. 11.1%, as was pCR at 18.1% vs. 4%. Most importantly, the Keynote-671 trial was the first phase III trial to demonstrate a survival benefit in favor of perioperative ICI with an OS of 71.3% vs. 64.0% at three years (HR: 0.72, 95% CI: 0.56–0.93; P = 0.00517). The pembrolizumab-combination led to slightly more adverse events, which did not impair the surgical resection rates in the intervention group, compared to the control arm. The phase III AGEAN trial evaluated the combination of durvalumab-CT vs. placebo-CT for four cycles in the preoperative setting, followed by adjuvant durvalumab or placebo for one year [34, 35]. The durvalumab combination improved EFS and pCR compared to placebo: EFS (HR: 0.68, 95% CI: 0.53–0.88, P = 0.003902), and pCR: 17.2% vs. 4.3%. Finally, the NEOTORCH phase III trial compared the association of toripalimab with CT to placebo-CT for three pre-operative and one post-operative cycle, followed by one year of toripalimab or placebo[36]. The primary endpoint, EFS among stage III patients, was met with a (HR: 0.40, 95% CI: 0.277–0.565, P < 0.0001). There was also higher MPR and pCR in the toripalimab arm: 48.5% vs. 8.4% and 24.8% vs. 1.0%, respectively.

These phases III studies did not only include stage III patients [37]. In subgroup analyses, stage III patients derived equal or greater benefit from immunotherapy than stage II patients. Two phase II studies specifically looked at resectable stage III diseases. First, the NADIM-2 trial randomized patients between nivolumab-CT and CT alone [38]. Patients with R0 resection in the intervention group received 6 months of adjuvant nivolumab. Nivolumab improved pCR at 37% vs. 7% in the control group. The secondary endpoints of PFS and OS were also significantly better with nivolumab. The Chinese TD-FOREKNOW phase II trial, comparing neoadjuvant camrelizumab-CT to CT alone, showed comparable results, with a pCR rate of 32.6% (95% CI: 19.1–48.5%) with camrelizumab vs. 8.9% (95% CI: 2.5–21.2%) for the control group [39].

Several questions remain regarding neoadjuvant ICI in NSCLC. First, the role of PD-L1 expression requires clarification. As in other setting in NSCLC, PD-L1 positive patients seem to derive more benefit from ICI than PD-L1 negative patients. For instance, in an exploratory subgroup analysis of CheckMate-816, no EFS benefit was seen in PD-L1 negative patients from the a ddition of nivolumab to CT, with a HR for EFS of 0.81 (95% CI: 0.48–1.36) [40]. The EMA approved neoadjuvant nivolumab exclusively for patients with PD-L1 positive disease based on this analysis. Second, the question of the optimal CT backbone remains uncertain. At present, it is difficult to establish whether the efficacy of ICI is impacted by a specific CT regimen. Third, the role of the adjuvant ICI after neoadjuvant therapy remains open for debate. While cross-comparing trials should be done with caution, the EFS rate at 2 years seems similar in the Keynote-671 and CheckMate-816 trials. To specifically address the role of the adjuvant therapy, further studies will be necessary. An interesting approach would be to randomize patients based on post-operative biomarkers such as circulating tumor DNA. Finally, a major limitation of these phase III trials is post-protocol drug access. In CheckMate-816 and Keynote-671, more than a third of patients in the control group who relapsed and required new systemic treatment did not receive ICI, which is standard of care in this situation.

Regarding oncogene-addicted NSCLC and the efficacy of ICI, some patients with EGFR or ALK were included in the phase III trials [32, 34]. As previously mentioned, the value of ICI for these patients is uncertain and the subgroup analyses from these trials have too few patients to draw any conclusions. There are greater expectations for targeted therapies in this setting. Osimertinib has recently been approved for the adjuvant treatment of NSCLC harboring classical EGFR mutations following the results of the ADAURA trial [41]. Compared to placebo, three years of adjuvant osimertinib improved DFS in stage IB–IIIA patients according to the AJCC 7th edition, with (HR: 0.20, 99% CI: 0.14–0.30, P < 0.001). The benefit for stage IIIA patients is greatest, with a HR for DFS of 0.12 (95% CI: 0.07–0.20). The OS results have recently been published, showing an improvement in survival in the stage II–IIIA population, with an OS of 88% in the osimertinib group vs. 78% in the placebo at five years (HR: 0.49, 95% CI: 0.33–0.73, P < 0.001) [42]. A major limitation in the interpretation of these OS results is the low cross-over rate to osimertinib at progression for patients in the control group, with only about 40% of patients with progressive disease in the control group receiving osimertinib, even though it represents the standard of care in this situation [43]. The potential role of osimertinib in the neoadjuvant setting is currently being explored in the phase III NeoADAURA for patients with resectable stage II to IIIB NSCLC harboring classical EGFR mutations.

The use of adjuvant alectinib, an ALK inhibitor, for NSCLC harboring ALK rearrangement has recently been reported in the ALINA trial [44]. The trial evaluated two-years of adjuvant alectinib compared to standard adjuvant CT for resected ALK positive NSCLC and showed a clear DFS benefit for alectinib in stage II to IIIA, with (HR: 0.24, 95% CI: 0.13–0.45). Several questions remained, however, such as if these impressive results will translate in an OS benefit such as in the ADAURA study; if adjuvant CT should still be considered for some patients in addition to alectinib; and if other ALK inhibitor such as lorlatinib could even do better in that setting [44].

As seen with EGFR, such trials have the potential to reshape the therapeutic sequence in other oncogenic alterations. As data mature, it may be difficult not to extrapolate the benefit to rarer oncogenic drivers for which organizing early-stage phase III randomized trials are unrealistic.

As highlighted by these new data on perioperative ICI and targeted therapies for stage III disease, the determination of resectability remains the cornerstone of the management of these patients, guiding towards a strategy of surgery or CRT. Recent data of pre- and perioperative ICI and the relatively high rate of MPR or pCR could challenge the definition of resectability, and question if stage IIIB disease should be directed toward surgery in priority. However, if the high rates of R0 resections of stage IIIB disease observed in trials involving highly experienced thoracic surgeons [38] can be extrapolated in real-life setting remains debated. Moreover, in all the phases III perioperative ICI trials, about 20% of patients were not able to undergo surgery, even if all included patients were judged resectable upfront. Thus, the rate of patients with R0 would certainly be much lower if patients with borderline or unresectable disease were included in a perioperative and resection strategy in real-life settings. Having to redirect patients with incomplete resection towards a salvage strategy could impose potential higher risks of toxicities, and possibly worst outcome than an upfront CRT strategy followed by durvalumab. Therefore, in situations where resectability is uncertain upfront, induction chemoimmunotherapy does not yet constitute a strategy that should be adopted unconditionally, at the present time.