Background

Microsatellite instability-high (MSI-H) subtypes make up approximately 15–20% of all gastrointestinal (GI) cancers. Due to the deficiency of mismatch repair (dMMR), these tumors are characterized by high levels of tumor mutation burden and abundant tumor-infiltrating lymphocytes [1,2,3]. Researches have shown that immune checkpoint inhibitors (ICIs) could be used as standard therapy of unresectable or metastatic dMMR/MSI-H gastric cancer (GC) and colorectal cancer (CRC) [4,5,6,7,8,9,10].

Taken into consideration the promising efficacy of ICIs in the palliative setting, the role of ICIs has also been investigated in perioperative settings. Overall, the use of neoadjuvant ICIs is safe and has little impact on sequent surgery [11,12,13,14,15]. However, these data all come from PD1 antibody with or without cytotoxic T lymphocyte-associated antigen-4 (CTLA4) antibody in prospective clinical trials, and could not reflect the clinical practice in the real world, where chemotherapy might be combined with ICIs. And also, the safety profile with adjuvant ICIs has not been explored.

Regarding to the efficacy, some small sample studies have shown that neoadjuvant ICIs could lead to relatively high pathologically complete response (pCR) rates [11,12,13,14,15]. However, it remains unclear whether the high pCR rates can translate into subsequent survival benefits. As for adjuvant setting, patients with dMMR/MSI-H gastric cancer (GC) seemed to not benefit from adjuvant chemotherapy [16]. And around 20–25% of patients with stage III dMMR/MSI-H colorectal cancer (CRC) suffered from recurrence even after adjuvant chemotherapy [17]. It is widely discussed whether adjuvant ICIs should be applied in patients not treated with preoperative ICIs. In China, a small proportion of patients decided to receive adjuvant ICIs with or without chemotherapy after being fully informed of the potential pros and cons in a patient-doctor shared decision-making process. However, no study reported the survival data of adjuvant ICIs in dMMR/MSI-H GI cancers so far.

In order to investigate the long-term survivals of neoadjuvant and adjuvant ICIs in dMMR/MSI-H GC and CRC, we retrospectively collected and analyzed the efficacy data from 85 patients enrolled in a prospective observational study (NCT04640103) and other 87 patients retrieved from medical systems from 17 centers.

Methods

Patients Selection and study design

Patients from a prospective observational real-world study (NCT04640103) and patients derived from medical systems from 17 centers were combined together for retrospective efficacy analysis. Main inclusion criteria were: (1) age ≥ 18 years old, (2) confirmed by pathology as gastric cancer (including esophagogastric junction adenocarcinoma) or colorectal adenocarcinoma, (3) dMMR by immunohistochemistry or MSI-H confirmed by polymerase chain reaction (PCR) or next generation sequencing (NGS), and (4) received at least one dose of ICIs with or without chemotherapy or targeted drugs in neoadjuvant or adjuvant setting. (5) patients with stage IV disease were also included if all tumor lesions were assessed as resectable (6) patients treated with neoadjuvant ICIs with or without adjuvant therapy were included in the neoadjuvant cohort (NAC), and those treated with adjuvant ICIs and no any neoadjuvant therapy were included in the adjuvant cohort (AC). Patients were excluded if detailed medical records or follow-up were not available. Patients with proficient mismatch repair (pMMR) or unknown MMR status but MSI-H were included, while those with dMMR and microsatellite stability (MSS) were excluded.

This study was approved by the ethics committee of Beijing Cancer Hospital. All patients had signed informed consent for data collection.

Assessments and outcomes

The NAC patients were evaluated by imaging examination every 6–8 weeks after the initial of treatment (including chest, abdominal and pelvic CT, as well as pelvic and liver MRI if necessary). The time for operation and adjuvant therapy were decided by doctors in charge. Patients were followed up according to local guidelines. The criterion for terminating follow-up is death for any reason.

The duration of neoadjuvant therapy was defied as the time from the first dose of neoadjuvant treatment to surgery or clinical complete response (cCR) (for patients who achieved cCR did not receive surgery) or the last dose (for patients who did not achieve cCR and did not receive surgery), and the duration of adjuvant therapy was defied as the time from the first dose of adjuvant treatment to the last dose. Efficacy outcomes included best response (for NAC), pCR rate (for patients receiving surgery in NAC), disease-free survival (DFS), overall survival (OS). DFS was defied as the time from surgery or cCR until death or disease recurrence. OS was defied as the time from the first dose of neoadjuvant therapy (for NAC) or surgery (for AC) until death or the date the patient was last known to be alive.

Statistical analysis

All data were processed using SPSS 26.0 statistical software. Kaplan Meier curves was created by GraphPad Prism 9.5.1. Continuous variables were represented by a median (range). The classified variables were tested by chi-square test or Fisher exact test. A binary logistic regression model was performed to identify independent prognostic factors of pCR in NAC. The difference was statistically significant with P < 0.05. Survival rates were estimated by life table analysis.

Results

A total of 181 patients were screened in this study, with 89 from our prospective observational trial from January 2020 to February 2024 and 92 from 17 centers from September 2017 to July 2023 (Fig. 1), and finally 172 patients were included in the study, with 124 in NAC and 48 in AC.

Fig. 1
figure 1

Trial profile, Bold and Gray refers to prospective population and retrospective population, respectively. CRC: colorectal cancer; GC: gastric cancer; ICI: immune checkpoint inhibitor

Full size image

Neoadjuvant cohort

Patient characteristics

On April 10, 2024, at data cut-off, the median follow-up time was 29.4 months (4.3 to 79.4 months) for the CRC and 33.0 months (3.2 to 68.9 months) for GC.

The median age for CRC was 45.5 years (ranging from 19 to 71 years), younger than that for GC (median 65.5 years, ranging from 36 to 81 years). More CRC tumors (77.2%) were defined as MSI-H by PCR or NGS than GC. There was no significant difference in other clinical features between CRC and GC (Table 1). As for treatment regimen, most patients, especially those with GC, received anti-PD1 antibody combined with chemo- or targeted therapy as neoadjuvant treatment (Supplementary Tables 1 and 2). The median duration of neoadjuvant treatment was 4 months (ranging from 1 to 25 months) and 3 months (ranging from 1 to 25 months) for CRC and GC, respectively. Observation was the most preferred strategy after surgery, followed by anti-PD1 antibody monotherapy (Supplementary Tables 1 and 2). The median duration of adjuvant treatment was 4 months (ranging from 1 to 26 months) and 4 months (ranging from 1 to 12 months) for CRC and GC, respectively.

Table 1 Clinical characteristics of Neoadjuvant Cohort and Adjuvant Cohort
Full size table

Efficacy

All the 124 patients underwent regular imaging evaluations during the neoadjuvant period. Among them, there were two patients with dual primary colorectal cancer (rectal cancer and right-sided colon cancer). They both receive radical resection of all tumors, and therefore, the tumor assessments (radiological and pathological) were recorded separately. The cCR rates were 20.0% and 3.6%, respectively, in CRC and GC. The objective response rates were 75.7% and 55.4%, respectively, in CRC and GC (Table 2).

Table 2 Radiographic best response in neoadjuvant cohort
Full size table

Ninety-nine patients underwent primary tumor resection. The pCR rate was 73.4% and 47.7% for CRC and GC, respectively (Table 3). In multivariate logistic regression analysis and univariate analysis of caner type, gender, age, clinical TNM stage, clinical T stage, neoadjuvant therapy regimen and duration of neoadjuvant therapy, cancer type was the only factor independently significantly associated with pCR rate was found. (Table 3). Notably, the pCR rate in patients with stage IV disease was similar to those without metastasis. And T4 was not a negative predictor of pCR in CRC and GC (Table 3). Twenty-five patients did not receive surgery, including 18 who refused surgery, stopped treatment (10 cCR, 5 partial response [PR], 3 stable disease [SD]) and were in close follow-up and 7 still receiving treatment.

Table 3 Chi-square test and multivariate logistic analysis of factors related to pathologic characteristics of surgical patients in neoadjuvant cohort
Full size table

The 2-year DFS (2y-DFS) and 3-year DFS (3y-DFS) rates were 96% (95% CI 90–100%) and 96% (95% CI 90–100%) for CRC, and 88% (78–98%) and 84% (72–96%) for GC (Fig. 2A). The 2-years OS (2y-OS) and 3-year OS (3y-OS) rates were 100% and 100% for CRC, and 96% (90–100%) and 93% (85–100%) for GC (Fig. 2B). Although there were no statistically significant differences, patients who achieved pCR tended to have longer DFS and OS than those did not achieve, especially for GC (Supplementary Figs. 1 and 2). For patients with clinical stage IV, the OS and DFS are numerically inferior to those with non-stage IV (Supplementary Figs. 1 and 3).

Fig. 2
figure 2

Kaplan Meier curves of disease-free survival (DFS) and overall survival (OS) in neoadjuvant cohort (NAC) and adjuvant cohort (AC). A, DFS of colorectal cancer (CRC) (n = 62) and gastric cancer (GC) (n = 46) in NAC, analyzed among patients undergoing surgery or achieving clinical complete response. B, OS of CRC (n = 68) and GC (n = 56) in NAC, analyzed among all patients. C and D, DFS (C) and OS (D) of CRC (n = 22) and GC (n = 26) in AC, analyzed among all patients

Full size image

Adjuvant group

Patient characteristics

Forty-eight patients were enrolled finally. At data cut-off, the median follow-up time were 35.5 months (11.0–53.2 months) for CRC and 40.4 months (9.9–69.3 months) for GC. The median age for CRC was 42.5 years (ranging from 18 to 71 years), younger than that for GC (median 59.5 years, ranging from 39 to 69 years). CRC patients were more likely to suffer from Lynch syndrome than GC patients (Table 1). Most of patients (> 80%) received anti-PD1 antibody monotherapy as adjuvant treatment (Supplementary Tables 1 and 2).

Efficacy

The 2y-DFS and 3y-DFS rates were 94% (84–100%) and 94% (84–100%) for CRC, and 92% (82–100%) and 92% (82–100%) for GC (Fig. 2C). The 2y-OS and 3y-OS rates were 100% and 100% for CRC, and 96% (88–100%) and 96% (88–100%) for GC (Fig. 2D).

Patients with relapse

In NAC, two CRC patients and five GC patients had tumor relapse after radical resection, and among patients who did not receive surgery, only one GC patient (case 11) had tumor progression 45 months later after stopping ICIs and died of COVID-19 (Supplementary Table 3). Case 3 had a pMMR GC tumor which was confirmed as MSI-H with tumor mutation burden (TMB) of 58 Muts/Mb using biopsy sample from the primary tumor. The pathological stage was ypT3N0 with TRG of 3 after ICI-based neoadjuvant treatment. Immunohistochemistry and NGS were performed again using the surgery sample and it was found that the tumor turned to be pMMR and MSS with TMB of 8 Muts/Mb. Similarly, the primary tumor of case 7 (GC) was dMMR/MSI-H at initial diagnosis, but turned to pMMR after treatment. Case 4 (mixed adeno-neuroendocrine carcinoma of stomach) had only negative MSH2 expression and no PCR or NGS was performed to validate the MSI status. Case 5 (GC) progressed on neoadjuvant nivolumab plus chemotherapy and received salvage radical resection, and liver metastasis occurred soon after surgery. In AC, three patients (1 CRC and 2 GC) had tumor recurrence or metastasis, and one GC patients died of cancer. Among the seven patients with distant metastases, four received dual blockade of PD1 and CTLA4 combined with or without chemo- and targeted drugs, with three PR and one PD (Supplementary Table 3).

Discussion

To our best knowledge, this is the first study reporting the promising long-term survivals of neoadjuvant ICIs and adjuvant ICIs in dMMR/MSI-H CRC and GC with the longest follow-up period.

More and more efficacy data supported the use of neoadjuvant ICIs in patients with dMMR/MSI-H CRC and GC. However, long-term survival benefits remained unclear. With a median follow-up of 17.2 months in CRC patients treated with neoadjuvant PD1 antibody plus other treatment, the 2y-DFS rate was 100% in a retrospective study [18]. For GC patients treated with dual blockade of PD1 and CTLA4, no relapse was observed with a median follow-up of 13.4–14.9 months in NEONIPIGA study [15]. INFINITY study showed that disease recurred in 2 of 18 GC patients treated with dual blockade of PD1 and CTLA4 with a follow-up of 13.4 months (9.7–14.2 moths)[14]. In this study, the follow-up time was 29.4 and 33.0 months for CRC and GC respectively, much longer compared with other reports, and our data showed 3y-DFS were 96% and 84%, respectively. Notably, a total of 30 (23.8%) patients with stage IV disease were included and received neoadjuvant ICIs in our study. Although the long-term survivals of these patients were numerally lower than that of those with non-IV disease, the DFS and OS were exciting. These results indicated that neoadjuvant ICIs could bring promising long survival benefits in non-metastatic as well as resectable metastatic CRC and GC with dMMR/MSI-H.

In this study, patients with non-pCR seemed to have poorer (but not significantly) DFS and OS in NAC, especially in GC. And in a more recent publication, no local regrowth or distant metastasis was observed in 24 patients with clinical complete response with a median follow-up time of 29.1 months [19]. Therefore, it is reasonable to assume that pCR was associated with better survivals, and further study should be conducted to increase the pCR rate. However, there was no convincing data regarding the most appropriate strategy associated with the highest pCR rate,

For CRC, patients with non-metastatic dMMR colon cancer receiving around 6 weeks of dual blockade of PD1 and CTLA4 and 8 weeks of dual blockade of PD1 and LAG3, respectively in the NICHE2 and NICHE3 studies, and the pCR rates were 67% (72/107) and 79%, respectively. The role of PD1 antibody monotherapy had been explored. In the PICC study, 11 out of 17 (65%) patients treated with 3 months of toripalimab achieved pCR. Creck et al. reported that all the 12 patients achieved cCR after 6 months of dostarlimab. Considering all the results above, it seemed that 6 months of PD1 antibody monotherapy might be preferred to achieve the greatest possibility of CR in CRC, which was supported by another prospective study, which demonstrated that the median time to reach a clinical complete response was 5·2 months [13, 20]. In GC, 3 months of PD1 antibody plus CTLA4 antibody was the only regimen investigated with pCR rate of 58.6–60% [14, 15]. PD1 antibody monotherapy was also used [21]. Interestingly, PD1 antibody plus chemotherapy was the most used neoadjuvant regimen in most retrospective studies related to this field [22,23,24], which was similar in our study, where 47.4% of CRC patients and 70.5% of GC patients received PD1 blockade combined with chemotherapy. The combination with chemotherapy might be not necessary because neoadjuvant chemotherapy was not associated with better prognosis compared with surgery alone in GC [25] and were likely to result in poor response in CRC [26], which was further supported by our data that treatment regimen was not associated with pCR rate. In clinical practice, the decision to add chemotherapy was likely driven by concerns about disease progression with PD1 antibody monotherapy and the high cost and toxicity associated with dual PD1 and CTLA4 blockade. Prospective clinical trials are warrant to explore the most suitable neoadjuvant strategy, in resectable dMMR/MSI-H CRC and GC, in particular to investigate the role of chemotherapy added to ICIs and the treatment duration.

The optimal adjuvant treatment is also controversial for patients with dMMR/MSI-H CRC or GC who did not receive neoadjuvant ICIs. For CRC, chemotherapy of fluoropyridine plus oxaliplatin had been proven to increase OS or DFS in Stage III or pT4N0 CRC [17, 27]. However, the survival outcomes were still not satisfying. Our results firstly reported 3y-DFS of 94% in CRC after adjuvant anti-PD1 antibody (81.8% treated without chemotherapy) with a relatively long follow-up (median 35.5 months), much higher than historical data (3-DFS 75–80% for stage III CRC [17]). These data highlighted the potential benefits of ICIs in adjuvant setting. Clinical trials are needed to confirm the superiority of anti-PD1 antibody with or without chemotherapy over chemotherapy alone.

Regrading GC, dMMR/MSI-H status might predict lack of benefit in DFS and OS of adjuvant chemotherapy in a multinational, individual-patient-data meta-analysis [16]. Moreover, most retrospective studies did not demonstrate a significant improvement in survivals with the implementation of adjuvant chemotherapy [28]. However, no alternative treatment was proposed for this population. Our study firstly reported the 3y-DFS of 92% after adjuvant ICIs in a relatively late-stage disease (65.4% stage III and 11.5% stage IV), which was numerally higher than that with surgery only or postoperative chemotherapy (~ 80%) [16]. This study might provide a potential treatment option for dMMR/MSI-H GC, in spite of the small sample and mixed therapies used. The ongoing clinical trials will give us more data about the necessity of chemotherapy and the value of ICIs in the adjuvant setting (NCT05236972, NCT05468138, NCT04969029).

Few relapses had been reported in patients with dMMR/MSI receiving perioperative ICIs [13, 14, 18], and the reasons for relapse remain largely unknown. Although our study observed limited cases of relapse, they may offer insights into the underlying mechanisms. Two GC patients (cases 3 and 7) exhibited different MMR/MSI statuses between pre-treatment and post-treatment samples. This could be attributed to MMR/MSI heterogeneity, where tumor cells with dMMR/MSI-H were eliminated after ICIs and only pMMR/MSS cells remained. We suspect that the recurrent tumors were also pMMR/MSS; however, these two patients declined re-biopsy. One patient (case 4) had negative MSH2 expression and positive MSH6 expression, which is uncommon as methylation or mutations in MSH2 are typically associated with IHC loss of both MSH2 and MSH6 [29], raising suspicion that the tumor was not MSI-H; however, no PCR or NGS tests were performed. Additionally, one patient (case 5) underwent salvage surgery after progression on ICIs developed liver metastasis soon after surgery. Therefore, MMR/MSI heterogeneity, potential wrong assessment of MMR/MSI and resistance to neoadjuvant ICIs might be associated with tumor relapse. Seven patients with relapse had distant metastases and 4 of them received dual blockade PD1 and CTLA4 with or without chemo- and targeted therapy, with 3 PR and 1 PD. These findings supported that rechallenge with ICIs plus other drugs was effective, consistent with our previous retrospective study [30]. We are conducting a clinical trial to investigate the efficacy of intensive ICIs plus anti-VEGF treatment beyond progression on monotherapy PD1 antibody (NCT06099821).

This study has some limitations. As a retrospective study, the heterogeneity among patients, stages, treatment regimen and duration is the major limitation of our study; however, the application of ICIs in neoadjuvant and/or adjuvant period did improve the prognosis of dMMR/MSI-H patients. Another limitation is the small number of patients enrolled due to the relatively lower prevalence of dMMR/MSI-H, especially for the adjuvant use of ICIs. Though promising and novel, our data need to be validated in a prospective study with larger sample.

In conclusion, this study confirmed the survival benefits of neoadjuvant ICIs with the longest follow-up ever reported and provided the first evidence of adjuvant ICIs in dMMR/MSI-H CRC and GC.