The study, conducted across three academic medical centers — Nanfang Hospital of Southern Medical University, The Sixth Affiliated Hospital of Sun Yat-sen University, and Sun Yat-sen University Cancer Center — collected patients treated between September 2015 and October 2022. All had received chemotherapy or chemotherapy combined with immunotherapy before undergoing gastrectomy. The regimens reflected current standards of care: doublet or triplet combinations of fluoropyrimidine-based and platinum-based chemotherapy, including XELOX, SOX, FOLFOX, and the docetaxel-containing FLOT protocol, with some patients also receiving PD-1 immune checkpoint inhibitors every three weeks. Radiological response was assessed by enhanced computed tomography after two to three cycles using the RECIST 1.1 criteria, and the overall response rate across the cohort reached 48.4 percent.

Of the 488 patients, 80 — or 16.4 percent — achieved a pathological complete response. To ensure the finding was robust, the researchers harmonized tumor regression grading across institutions: because one center used the National Comprehensive Cancer Network classification while the other two applied the American Joint Committee on Cancer 8th edition system, three blinded gastrointestinal pathologists conducted a centralized review of all specimens, resolving discrepancies by panel consensus. The definition was strict — no residual tumor cells anywhere in the resection specimen or in any of the resected lymph nodes. Notably, seven patients who had no residual tumor in the primary lesion still harbored carcinoma in their regional lymph nodes, and all seven fell into the non-pCR group.

The survival signal was dramatic. Patients with pCR had significantly better overall survival and recurrence-free survival than those with residual disease, with neither median overall survival nor median recurrence-free survival reached in the pCR group during follow-up. For non-pCR patients, median overall survival was 62.8 months and median recurrence-free survival was 20.9 months. Recurrence rates diverged just as sharply: only 7.5 percent of pCR patients — six of eighty — relapsed, compared with 40.2 percent of non-pCR patients, or 164 of 408. Across the entire cohort, 170 patients, or 34.8 percent, experienced recurrence after neoadjuvant treatment and surgery.

But the study’s most provocative finding concerns the six pCR patients who did relapse. Every single one of them had received adjuvant therapy after surgery, while none of the seventeen pCR patients who were simply observed after surgery experienced recurrence. The authors are careful not to overinterpret this counterintuitive pattern. They suggest two possible explanations: selection bias, in which clinicians may have reserved adjuvant treatment for patients perceived to be at higher risk, or the development of chemotherapy resistance to the specific adjuvant regimen used. With only six events among eighty patients, the statistical power is limited, and the researchers explicitly caution that this observation should not be read as evidence against adjuvant therapy. Instead, they argue, it underscores the need for better risk stratification to identify the minority of pCR patients harboring residual micrometastatic disease who might genuinely benefit from additional systemic treatment.

The timing and location of relapse told another important story. There was no significant difference between the pCR and non-pCR groups in when or where recurrence occurred, echoing earlier work — including a 2011 cohort study by Fields and colleagues that found 27 percent of pCR gastric cancer patients relapsed within five years. Among all patients with recurrence in the current study, 73.5 percent developed distant metastases, 8.8 percent had local recurrence, and 6.5 percent had both. Of the six relapsed pCR patients, all were male; two developed locoregional recurrence while three developed distant disease involving the peritoneum, distant lymph nodes, or bone, and three of the six died of their recurrence during follow-up.

To predict who might achieve pCR in the first place, the team built a statistical model using logistic regression and a nomogram — a graphical scoring tool that translates patient characteristics into a personalized probability. Multivariate analysis identified radiological response under RECIST 1.1 as a powerful predictor: patients with a complete or partial radiological response had roughly eight times the odds of achieving pCR compared with those with stable or progressive disease. Even more striking, Epstein-Barr virus-encoded RNA positive status carried an odds ratio of 20.65, making EBER positivity the single strongest molecular predictor in the analysis. After excluding variables with too many missing values, the final nomogram incorporated the neoadjuvant treatment regimen and histological type as independent predictors. The model performed well, achieving a concordance index of 0.844 in the training set and 0.820 in the validation set, with calibration curves showing satisfactory agreement between predicted and observed outcomes and decision curve analysis confirming clinical utility.

The immunotherapy question yielded a nuanced answer. Adding immune checkpoint inhibitors to neoadjuvant chemotherapy significantly increased the pCR rate and cut the recurrence rate nearly threefold — 15.1 percent versus 40.3 percent for chemotherapy alone. Yet this did not translate into significantly longer recurrence-free or overall survival, with Kaplan-Meier comparisons showing no statistically significant difference between the two treatment groups. The finding mirrors results from major international trials: the phase III KEYNOTE-585 study demonstrated that pembrolizumab plus chemotherapy significantly raised pCR rates in locally advanced gastric and gastroesophageal junction cancer but failed to significantly improve event-free survival, while the DANTE trial reported promising pathological responses with atezolizumab-based perioperative therapy. The disconnect between pathological response and hard survival endpoints remains one of the field’s central puzzles, and the authors call for research integrating circulating tumor DNA, radiomics, and digital pathology to predict long-term outcomes more accurately.

Beyond pCR, the study identified clear risk factors for recurrence in the broader population: patients with five or more positive regional lymph nodes had nearly four times the recurrence risk, those with poorly differentiated tumors and neurovascular invasion were also at elevated risk, and a high lymph node ratio — above 30 percent — predicted significantly shorter overall and recurrence-free survival. Tumor regression grade itself mattered, with a three-year recurrence-free survival of 66.5 percent for patients with the best regression grades compared with 50.9 percent for those with lesser responses. Intriguingly, among patients with the best regression grades, those who skipped postoperative systemic therapy had a numerically higher three-year recurrence-free survival of 81.4 percent versus 64.4 percent, though the difference was not statistically significant.

The study has limitations the authors acknowledge candidly: its retrospective design invites selection and treatment bias, the small number of recurrence events among pCR patients limits statistical power, center heterogeneity introduced variability in practices and data collection, and the nomogram lacks external validation. Missing data on Lauren classification and PD-L1 combined positive scores further constrained the analysis. Still, as the largest sample of gastric and gastroesophageal junction cancer patients with pCR after neoadjuvant treatment and radical surgery to date, the study delivers a clear message for clinicians and patients alike: a pathological complete response is one of the most favorable outcomes in this disease, associated with dramatically lower recurrence and superior survival, but it is not a guarantee of cure. The role of adjuvant therapy for these patients remains unresolved, and the authors argue that only prospective randomized trials can settle whether additional postoperative treatment helps or harms the patients whose tumors have already vanished.

Subject of Research: Pathological complete response and survival after neoadjuvant treatment of locally advanced gastric cancer

Article Title: Characteristics and survival associated with pathological complete response following neoadjuvant treatment of locally advanced gastric cancer: a multicenter analysis in a real-world setting

Article References: Lai, M.-Y., Luo, D.-D., Chen, S., Yan, X., Tan, W.-Y., Guan, W.-L., Zhou, Z.-W., Zhao, L.-Y., & Qiu, M.-Z. (2026). Characteristics and survival associated with pathological complete response following neoadjuvant treatment of locally advanced gastric cancer: a multicenter analysis in a real-world setting. Holistic Integrative Oncology, 5(1), Article 53. https://doi.org/10.1007/s44178-026-00273-w

Image Credits: AI Generated

DOI: 10.1007/s44178-026-00273-w

Keywords: gastric cancer, pathological complete response, neoadjuvant treatment, immunotherapy, chemotherapy, recurrence, overall survival, recurrence-free survival, nomogram, tumor regression grade, gastroesophageal junction cancer, adjuvant therapy

That question took on renewed urgency at the 2024 meeting of the FDA Oncologic Drugs Advisory Committee, which re-examined the comparative efficacy of PD-1 inhibitors plus chemotherapy across PD-L1 expression levels. The committee concluded that patients whose tumors express PD-L1 at a combined positive score or tumor area positivity score of 10 or higher derive the clearest benefit from PD-1 inhibitors, while those scoring below 1 show no meaningful improvement over chemotherapy alone. The large intermediate zone between those thresholds remains contested, and even in higher-expression groups certain populations, notably patients with liver metastases, appear relatively resistant to single-target immunotherapy. Against this backdrop, a new network meta-analysis published in Cancer Reports offers the first systematic indirect comparison of a different immunotherapy architecture: cadonilimab, a tetravalent bispecific antibody that blocks both PD-1 and CTLA-4 within a single molecule.

The research team, registered in the PROSPERO repository and conducted under PRISMA guidelines, searched PubMed, Web of Science, Embase and the Cochrane Library from inception through December 31, 2024, supplementing the database results with conference abstracts from ASCO, ESMO and AACR and with FDA advisory committee documents. Their inclusion criteria were deliberately strict: only Phase III randomized controlled trials in pathologically confirmed, previously untreated, HER2-negative gastric or gastroesophageal junction cancer qualified, and every experimental arm had to combine chemotherapy with either a single-agent anti-PD-(L)1 antibody or the anti-PD-1/CTLA-4 bispecific antibody cadonilimab. From an initial pool of 4,089 publications, screening narrowed the field to eight trials that met all criteria: KEYNOTE-062, ATTRACTION-4, CheckMate-649, KEYNOTE-859, GEMSTONE-303, ORIENT-16, COMPASSION-15 and RATIONALE 305. Together these trials enrolled 7,127 patients, of whom 3,572 received chemotherapy plus an immune checkpoint inhibitor and 3,555 received chemotherapy with placebo.

Because no head-to-head trial has ever compared cadonilimab with a PD-(L)1 antibody, the investigators used a frequentist network meta-analysis, treating all platinum-fluoropyrimidine doublet backbones—XELOX, FOLFOX, SOX and fluorouracil-platinum—as a common chemotherapy comparator node, an approach supported by prior noninferiority trials and systematic reviews showing these regimens are clinically interchangeable. Overall survival, defined as time from randomization to death from any cause, served as the primary endpoint, with progression-free survival and Grade 3 to 5 treatment-related and serious adverse events as secondary measures. Risk of bias was low across the included studies when assessed with the Cochrane RoB 2 tool, and all trials were analyzed on an intention-to-treat basis. Because the network was star-shaped, with cadonilimab connected to the PD-(L)1 inhibitors only through the shared chemotherapy node, formal inconsistency testing was not possible, a structural limitation the authors acknowledge openly.

The headline finding is striking. In the intention-to-treat population, every immunotherapy regimen outperformed chemotherapy alone, but cadonilimab plus chemotherapy delivered the largest survival gain, with a hazard ratio for overall survival of 0.62 (95 percent confidence interval 0.50 to 0.77), compared with hazard ratios of 0.75 to 0.82 for the five PD-(L)1 antibodies: sugemalimab, sintilimab, pembrolizumab, tislelizumab and nivolumab. In the indirect comparison against the pooled PD-(L)1 class, cadonilimab was associated with a statistically significant 22 percent reduction in the hazard of death (hazard ratio 0.78, 95 percent confidence interval 0.62 to 0.98), with essentially no heterogeneity among trials. Treatment ranking by P-score placed cadonilimab first with a probability score of 0.9586, well ahead of the rest of the field, and progression-free survival followed the same pattern, with a hazard ratio of 0.70 (95 percent confidence interval 0.57 to 0.86) favoring the bispecific approach.

The mechanistic rationale for this advantage lies in cadonilimab’s unusual tetravalent design. Gastric adenocarcinomas commonly express the immune checkpoints PD-1 and CTLA-4, which tumors exploit to suppress T-cell responses, and preclinical work has long shown that blocking both pathways synergistically enhances antitumor immunity. Conventional combination immunotherapy with separate PD-1 and CTLA-4 antibodies, tested in the CheckMate 649 trial, was hampered by dose-limiting toxicity and failed to improve survival. Cadonilimab instead packs both binding specificities into one antibody that preferentially accumulates in tumor tissue where PD-1 and CTLA-4 are co-expressed, concentrating dual blockade at the tumor site while sparing normal tissue. That design, the authors argue, explains both the improved efficacy and the tolerability profile observed in COMPASSION-15.

Safety data appear to support that interpretation. Across the included trials, no statistically significant differences emerged between cadonilimab and PD-(L)1 inhibitors in Grade 3 or higher treatment-related adverse events (odds ratio 0.83, 95 percent confidence interval 0.54 to 1.28) or serious adverse events (odds ratio 0.89, 95 percent confidence interval 0.56 to 1.42). Egger’s tests for publication bias were non-significant across all endpoints, and leave-one-out sensitivity analyses left the direction and significance of every primary result unchanged. Still, the authors caution that the confidence intervals around the safety comparisons are wide and rest on a limited number of trials, so a clinically meaningful difference in toxicity cannot be definitively excluded.

The most consequential findings emerge from exploratory subgroup analyses. In patients with low PD-L1 expression—combined positive scores below 10—the hazard ratio for overall survival favored cadonilimab over PD-(L)1 therapy with a point estimate of 0.77 (95 percent confidence interval 0.58 to 1.01), consistent across the CPS below 1 and below 5 cutoffs, while the single-target antibodies showed little or no benefit versus chemotherapy in the same strata. Even more pronounced was the liver metastasis subgroup, where cadonilimab reduced the hazard of death by 33 percent relative to PD-(L)1 inhibitors (hazard ratio 0.67, 95 percent confidence interval 0.49 to 0.93) and halved it relative to chemotherapy alone. This matters because liver metastases deplete peripheral T cells, erode intratumoral T-cell diversity and enrich immunosuppressive regulatory T cells and macrophages, creating a microenvironment that single-agent PD-1 blockade struggles to overcome; the authors propose that simultaneous CTLA-4 inhibition, which suppresses regulatory T cells, counters this biology more effectively. Statistically significant advantages for cadonilimab also appeared in male patients and those with an ECOG performance status of 1, subgroups in which androgen-receptor signaling and heightened systemic inflammation respectively promote T-cell exhaustion that dual checkpoint blockade may more fully reverse.

The authors are careful to frame these subgroup results as hypothesis-generating. All subgroup estimates derive from study-level aggregate data rather than individual patient information, which introduces the risk of ecological bias and limits control of within-trial confounding, and the PD-L1 analyses relied on different immunohistochemistry platforms, albeit assays with demonstrated substantial to excellent concordance in gastric cancer. Ranking analyses are sensitive to network structure and cannot by themselves establish superiority. Nevertheless, the overall picture is internally consistent: for patients whose tumors express high levels of PD-L1, all immunotherapy regimens deliver clear survival benefits over chemotherapy, but for the large populations left in clinical limbo—those with low PD-L1 expression, liver metastases, or intermediate expression where regulatory debate continues—this analysis suggests that a bispecific antibody harnessing two immune escape pathways simultaneously may be the preferable first-line partner for chemotherapy. The findings stop short of Level 1 evidence, and the authors call explicitly for prospective head-to-head trials, but they mark the first quantitative indication that bispecific checkpoint engineering could redraw the treatment algorithm for advanced gastric and gastroesophageal junction cancer.

Subject of Research: Indirect comparison of cadonilimab, a PD-1/CTLA-4 bispecific antibody, versus PD-(L)1 inhibitors as first-line therapy for advanced gastric and gastroesophageal junction cancer

Article Title: Survival Outcomes With Cadonilimab Versus PD‐(L)1 Therapy in Front‐Line Gastric Cancer: A Network Meta‐Analysis

Article References: Yu, Z.-K., Yi, J.-H., Xue, J., Huang, Y.-C., Zhao, Y., Jiang, C., Yin, C.-X., Xia, B., Zhou, Y.-X., He, W.-Z., & Xia, L.-P. (2026). Survival Outcomes With Cadonilimab Versus PD ‐(L)1 Therapy in Front‐Line Gastric Cancer: A Network Meta‐Analysis. Cancer Reports, 9(9), Article e70672. https://doi.org/10.1002/cnr2.70672

Image Credits: AI Generated

DOI: 10.1002/cnr2.70672

Keywords: cadonilimab, gastric cancer, gastroesophageal junction cancer, network meta-analysis, immune checkpoint inhibitors, PD-1, CTLA-4, bispecific antibody, PD-L1 expression, liver metastasis, overall survival, COMPASSION-15