The aim of neoadjuvant immunotherapy (nIT) with ICIs is to enhance T cell activity against tumor neoantigens in a preserved primary tumor microenvironment. In the adjuvant setting, ICI therapy may enhance T cell efficacy against neoantigens in a micrometastatic environment in which the primary tumor is no longer present. It is currently unclear whether the microenvironment and metabolism differ between early-stage disease and micrometastatic disease. In a colon cancer liver model, micrometastases exhibited increased T cell infiltration compared with macrometastases [8]. Nonetheless, other metastatic models (e.g., brain metastases from HER2⁺ breast cancer) showed that latent metastases have distinct metabolic fitness from synchronous and metachronous bulk metastases [9, 10]. Nevertheless, better efficacy with the neoadjuvant strategy has recently been shown in phase III trials in non-small cell lung cancer, melanoma, and triple-negative breast cancer [11–14].

One randomized clinical trial of note has completed recruitment in the adjuvant setting (NCT02912559) [15]. This study was initiated in 2017 and is comparing FOLFOX + atezolizumab to FOLFOX alone (n = 700). Recruitment closed in January 2023, and the final results are expected to be reported in April 2025. Although the results are eagerly awaited, because this will be the first randomized data from a neoadjuvant setting, several caveats should be noted. First, the duration (12 months therapy in the atezolizumab arm) compares negatively with the 6-week neoadjuvant period in the NICHE-II trial [16]. Second, the expected toxicity profile will also compare negatively with that of NICHE-II. Finally, even if the experimental arm is superior to the control arm, it would be difficult to match the efficacy of the NICHE-II study, with 100% disease-free survival at 3 years with neoadjuvant therapy with nivolumab-ipilimumab.

We have included in the revision all the published or non-published (but presented at ESMO or ASCO Meetings) with ICI neoadjuvant therapy (see Figure 1). The NICHE-I trial pioneered the evaluation of nIT with nivolumab and ipilimumab in patients with locally advanced colon cancer [17]. The study demonstrated a pCR rate of up to 60% and a favorable safety profile. These findings have prompted further investigations into the safety and efficacy of nIT. Phase II studies consistently demonstrate that nIT in early MSI patients yields significantly higher rates of a pCR (45%–80%) and major pathological response (< 10% residual tumor) (90%–100%), corresponding to Mandard tumor regression grade 1 (complete response) or 2 (near-complete response), compared to 6-week neoadjuvant chemotherapy with FOLFOX, in which only 4% of such patients achieved a pCR and 5% achieved a Mandard tumor regression grade 1 or 2 [18] (see Table 2). However, several questions remain, necessitating further research to obtain robust scientific evidence that could potentially influence standard clinical practice.

First, although radiological staging is well-defined with pelvic magnetic resonance imaging in rectal cancer, the sensitivity and specificity of computed tomography are less clear in colon cancer [19]. This is important because MSI pT1-3N1 tumors, which represent roughly 50% of patients, have an excellent prognosis, with recurrence rates of less than 10%. In contrast, pT4N2 has a poor prognosis and 40% of these patients develop metastasis [20].

Second, it is further unclear whether adjuvant therapy will add clinical value after nIT. Findings from the NICHE-II trial with only 3-week therapy with nivolumab (two cycles) and ipilimumab (one cycle) [16, 21] indicated a 67% pCR and 95% major pathological response and an outstanding 100% disease-free survival rate at 3 years, suggesting that adjuvant therapy will likely not increase the efficacy of nIT.

The remaining question is if monotherapy with anti-PD-1 alone can achieve the same efficacy as combination therapy with anti-CTLA4 [17, 21, 22]. Recently, the NICHE-III trial, involving patients treated with nIT comprising two cycles of nivolumab and one cycle of relatlimab (anti-LAG3), showed similar efficacy (67% pCR) and toxicity as anti-CTLA4 therapy [23]. Most phase II studies suggest that monotherapy with anti-PD-1 is sufficient to achieve a similar pCR (46%–66%) as combination regimens. Only a small phase II randomized study has compared the efficacy of anti-PD-1 alone with that of anti-PD-1 combined with anti-CTLA4. That study showed a significant reduction in the pCR with anti-PD-1 alone (47.7% vs 80%) [22]. The benefit of adding other strategies to anti-PD-1 therapy, such as anti-angiogenic therapy with afatinib [24] and anti-COX2 with celecoxib [25], is currently unknown.

The duration of nIT may also influence outcomes. In the NICHE trials, combination therapies were administered during a 6-week nIT period, whereas other trials involving monotherapy with anti-PD-1 extend the nIT to 12 or even 24 weeks. For instance, a pCR was achieved in 46% and 68% of patients with pembrolizumab alone in 3- or 6-week nIT therapy, respectively, in the IMHOTEP study [26], in 59% of patients with 6-week therapy [27], and in 66% of patients with 24-week therapy [28]. Therefore, it is currently unclear if the number of cycles or the nIT duration improves the pathological response and 3-year disease-free survival. Because of the absence of phase III data compared with conventional therapy (surgery followed by chemotherapy), the results of the ongoing AZUR2 study (NCT05855200), a phase 3 clinical trial comparing nIT with dostarlimab vs standard treatment that is currently enrolling patients, are awaited with interest.

In early rectal cancer, anti-PD-1 alone with dostarlimab improves the strategy of non-operative management (NOM) to almost 100% [29, 30]. Other studies with other anti-PD-1 agents, such as camrelizumab [24] and sintilimab [31], also showed impressive results in NOM. Recently, the AZUR1 study (NCT05723562) has recruited 100 patients with early-stage MSI rectal cancer treated with dostarlimab. The primary objective is a 65%–90% complete clinical response at 12 months. The results are highly expected, particularly because the appropriate radiological evaluation for selecting patients for NOM and nIT is currently unclear.

The clinical characteristics of MSI metastatic patients (only 5% of metastatic patients are MSI) differ in some aspects from those of MSS individuals. Compared with MSS tumors, the primary tumors are typically located on the right side (65%–70% vs 30%–35%), are more prone to the development of BRAF mutations (25%–30% vs 5%–10%), have more peritoneal metastasis (35%–40% vs 10%–15%) and less liver metastasis (35%–40% vs > 70%), and exhibit less abnormally high lactate dehydrogenase levels (5%–10% vs 30%–35%).

In Table 3, we included all the studies that have been published with ICI in advanced MSI tumors. We included phase II trials of first-line therapy with nivolumab-ipilimumab [32] and of pre-treated patients with nivolumab [33], nivolumab-ipilimumab [34, 35], nivolumab-relatlimab [36], avelumab [37, 38], durvalumab [39] and pembrolizumab [40]. Overall, these studies showed a high response rate (33%–65%) and 3-year progression-free rate (38%–60%). Although the results compare favorably with chemotherapy, roughly 25%–40% of patients are ICI refractory and less than 50% remain progression free at 3 years.

Two large phase III trials have compared ICI to standard chemotherapy. The KEYNOTE-177 trial demonstrated superior progression-free survival of pembrolizumab over chemotherapy as frontline treatment for patients with advanced MSI disease [41, 42]. The CheckMate 8HW has been recently published [43]. The dual primary endpoints were progression-free survival, assessed by blinded independent central review, for nivolumab + ipilimumab vs nivolumab across all lines and for nivolumab + ipilimumab vs chemotherapy in the first-line setting in patients with centrally confirmed MSI-H metastatic CRC (mCRC). This study shows the superiority of nivolumab-ipilimumab vs chemotherapy. The comparison between nivolumab-ipilimumab and nivolumab is planned only in patients that received ≥ 2 prior lines of therapy and the study will therefore not clarify if nivolumab-ipilimumab is superior to nivolumab as first-line therapy. Because chemotherapy is currently not an adequate control arm for advanced MSI disease, it is unclear whether nivolumab-ipilimumab should be considered a new standard of care.

Only a few studies have analyzed biomarkers to address ICI therapy efficacy in early MSI disease. In the NICHE-I study, although the tumor mutational burden (TMB) varied widely between MSI (median, 1,438 mutations; range, 648–4,458) and MSS (median, 111 mutations; range, 48–261) tumors, no differences in TMB were found between MSS responders (median, 108; interquartile range, 62–153) and non-responders (median, 117; interquartile range, 93–143). In addition, no major differences were found between MSI and MSS in CD3⁺, FOXP3⁺, PD-L1, and interferon scores, tertiary lymphoid structure presence, and CXCL13 expression. The sole biomarker related to the MSS ICI response was the presence of T cell co-expression of CD8 and PD-1 [17]. In the PICC study, early-stage MSI patients were treated with toripalimab ± celecoxib. The authors found no differences in activity based on TMB, but more responders were seen in the group of patients with HLA-DQA1 and HLA-DQB1 expression and in tumors that showed CD8⁺ T cells with low PD-1 expression [44]. Finally, the same group used single-cell RNA-seq analysis to show that patients who achieved a pCR in post-treatment samples exhibited decreased proinflammatory features, with fewer CD8⁺ TRM-mitotic cells (with exhausted phenotype markers, e.g., PDCD1, LAG3, TIGIT, HAVCR2, TOX, and ENTPD1), CD4⁺ Tregs, proinflammatory IL1B⁺ monocytes, and CCL2 fibroblasts [45]. Although these findings suggest a polarization to an exhausted ICI refractory phenotype after therapy, these studies do not clarify in pre-treated biopsies the intrinsic causes driving this polarization.

There is extensive literature regarding the use of potential biomarkers to assess ICI efficacy in advanced MSI patients (see Table 4). Nonetheless, using our proposed quality biomarker analysis (modified REMARK), none of the studies fulfill all of the strict criteria for their recommended use in clinical practice (Table 1). This means that none of these biomarkers can replace immunohistochemistry (loss of expression of any of four proteins [MLH1, PMS2, MSH2, MSH6]) or determination of mismatch repair deficiency (dMMR) defined by microsatellite analysis (at least two of five microsatellites with different length variations). These methods are cheap, highly reproducible, and, most importantly, have been used prospectively to select patients in clinical trials of MSI patients undergoing ICI. The fact that none of the published MSI predictive biomarkers, are used in clinical practice, reflect the poor credibility of REMARK criteria.

The only study that fulfills at least two modified REMARK criteria was recently published by Quintanilha et al. [46]. They used next-generation sequencing to assess MSI and compared the results with standard methods (immunohistochemistry of dMMR) and showed that next-generation sequencing better discriminates the benefit of ICI therapy.

MSI sensor software analyzes aligned sequencing data to assess available microsatellite regions with sufficient coverage in a tumor/normal tissue pair and thereby identify variations in deletion lengths indicative of MSI. Current evidence does not justify its use in clinical practice [47, 48]. In addition, TMB [49–52], which uses different cut-offs or clonal neoantigen burdens, does not fulfill any of our adapted REMARK criteria [53]. Loss of the expression of beta 2 microglobulin (β2M), which is usually associated with β2M mutations (occurring in 24%–28% of MSI tumors), has been shown pre-clinically [54] to preclude ICI efficacy due to poor major histocompatibility complex (MHC)-I presentation, although the clinical data do not support the pre-clinical findings [55–57]. Importantly, none of these studies fulfill our modified REMARK criteria and therefore cannot be used in clinical practice.

Inflammation has been pre-clinically related to ICI resistance. The inflammation pattern can be measured in different ways. For instance, in blood samples, we can evaluate inflammation by using the ratio between neutrophils/myeloids over lymphocyte infiltration. The main concern of this biomarker is the difficulty in the clear definition of the optimal cut-off. Using our strict REMARK methodology, none of the published studies fulfill the A or B recommendation [58–61] suggest that a close proximity of CD8⁺ T cells expressing PD-1 with CD3, CD8, and CD68 macrophages expressing PD-L1 correlates with ICI efficacy. Finally, transcriptomics has also been used to evaluate the ratio of proliferation to stromal components [62]. Progression-free survival was shorter in tumors treated with ICIs and in stromal-high/proliferation-low tumors. Again, different cut-offs and different ways to evaluate inflammation limit its value as a biomarker. Other authors have focused on pathways that have been described in other tumors as potential biomarkers of ICI resistance. For example, the study by Chida et al. [63] reported that WNT activation, commonly observed in CMS2/3 subtypes, is associated with ICI resistance. Similarly, studies by Sui et al. [64] and Bortolomeazzi et al. [65] further support the link between WNT activation and ICI resistance. Consensus molecular subtype 2, characterized by MYC and WNT expression and reduced immune infiltration [66, 67], has been shown to be increased in CRC metastases, compared with primary tumors [68]. This CMS subtype has been also related to ICI-resistance [69]. Additionally, a transforming growth factor-β (TGF-β) signature, also enriched in colorectal metastases [70] has been correlated with reduced ICI efficacy [71].