This study prospectively enrolled patients with AGC who were treated with ICIs for the first time at the Fourth Hospital of Hebei Medical University. The enrollment process was initiated on November 9, 2019, concluded on January 3, 2024, and follow-up activities were completed on June 30, 2024. The ICIs employed in this study are all anti- programmed cell death protein 1 (PD-1) drugs, which include pembrolizumab, toripalimab, camrelizumab, tislelizumab, and sintilimab. The targeted drugs used include tyrosine kinase inhibitors (Lenvatinib, Regorafenib) and the human epidermal growth factor receptor 2 antibody (trastuzumab). The chemotherapy drugs used include paclitaxel, platinum and fluoropyrimidine. The dosage and method of the drug were in accordance with the instructions of the drug. This study was approved by the Ethics Committee of the Fourth Hospital of Hebei Medical University, Written informed consent was signed by all patients.

The following criteria need to be met for inclusion: (1) Age ≧ 18 years old; (2) Patients diagnosed with GC through pathological examination; (3) According to the 8th American Joint Committee on Cancer (AJCC) staging system, patients with stage III or IV and no surgical indications; (4) Patients with at least one assessable target lesion; (5) Patients receiving anti-PD-1-based combination therapy; (6) Patients who have not received ICIs therapy before; (7) Patients who had chest and abdominal computed tomography (CT) scans within 1 week of starting ICIs therapy; (8) patients with 0–1 by Eastern Cooperative Oncology Group performance status (ECOG-PS) score; (9) Be informed and consent to participate in this study.

The following criteria need to be met for exclusion: (1) Patients with gastric stromal cell tumor or gastric neuroendocrine tumor; (2) Patients with a history of malignant tumors of other organs; (3) Patients with brain metastasis; (4) Patients currently receiving treatment for depression or anxiety. (5) Patients who cannot cooperate with psychological scale assessment.

The evaluation of tumor response is conducted using CT scans after every 2 or 3 treatment cycles, following the guidelines outlined in version 1.1 of the Response Evaluation Criteria in Solid Tumors (RECIST v1.1) (32). The physicians performing imaging efficacy assessment and those responsible for administering ED questionnaires were kept mutually blinded, with no knowledge of each other’s work contents or information records.

Progression-free survival (PFS) was determined as the duration starting from the initial administration of ICIs therapy until disease progression, mortality, or conclusion of the study. OS was defined as the period commencing ICI-based systemic treatment and ending with death or study termination. ORR was determined by calculating the proportion of patients who exhibited a complete response (CR) or partial response (PR). Disease control rate (DCR) was calculated based on the percentage of patients with a CR, PR or stable disease (SD).

Depressive symptoms are evaluated using the 9-item Patient Health Questionnaire (PHQ-9), where a score of 5 or higher signifies the presence of depressive symptoms; anxiety symptoms are assessed through the 7-item Generalized Anxiety Disorder Questionnaire (GAD-7), with a score of 5 or higher indicating the existence of anxiety symptoms. When included patients completed the self-assessment questionnaire, a designated researcher was involved to assist them in minimizing confusion and ensuring accurate expression of their inner feelings. Patients exhibiting symptoms of depression or anxiety at baseline were classified as the ED group, whereas those without such symptoms were categorized as the no-ED group. The questionnaire survey was conducted within 1 week before ICIs treatment, and the anxiety and depression status scale was evaluated again after 2 cycles of treatment.

The PHQ-9 is extensively utilized for the evaluation of depressive symptoms among patients with cancer (33). This 9-item questionnaire encompasses the symptomatology of major depressive disorder as delineated by the American Psychiatric Association (34). Each item is rated on a scale from 0, indicating ‘not at all,’ to 3, representing ‘nearly every day,’ with an aggregate maximum score of 27 and a higher score indicates more severe depressive symptoms. In addition, the PHQ-9 has been translated into Chinese and validated in clinical research, and a threshold score of 5 demonstrates robust reliability and validity in the identification of depressive symptoms, exhibiting a sensitivity of 0.91 and a specificity of 0.77 (33, 35).

The GAD-7 is a validated self-assessment tool endorsed by the American Society for Clinical Oncology for the screening of anxiety symptoms among cancer patients (36). This scale consists of 7 items, each rated on a scale of 0-3, with a total score ranging from 0–21 and a higher score indicates more severe anxiety symptoms. GAD-7 was also translated into Chinese and has been extensively utilized in clinical research. It has demonstrated robust validity and reliability for screening anxiety symptoms among cancer patients, evidenced by a Cronbach’s α of 0.91 (36, 37).

Blood samples for adrenocorticotropic hormone (ACTH) and cortisol testing are collected by nursing staff between 6 am and 10 am, and it was determined by electrochemiluminescence method in the laboratory department of the Fourth Hospital of Hebei Medical University. The reference range for ACTH is established at 7.2-63.3 pg/mL, while the normative range for cortisol is defined as 133–537 nmol/L. Among all the patients, only one had an ACTH level above the normal upper limit. To investigate the relationship between ACTH and ED, we used a receiver operating characteristic (ROC) curve to determine the optimal cutoff value of ACTH as 25.09 pg/ml. Of the 59 patients with baseline cortisol data, only 3 had elevated cortisol levels. We used an ROC curve to determine the optimal cutoff value for cortisol at 193.95 nmol/L.

From the electronic case records, we extracted pertinent information regarding the patient, including age, gender, body Mass Index (BMI), programmed cell death-ligand 1 (PD-L1) expression combined positive score (CPS), status of human epidermal growth factor receptor 2 (HER2), tumor-node-metastasis (TNM) stage, treatment regimen, treatment lines, microsatellite instability (MSI) status, ACTH level, cortisol level, IL-6, ECOG PS, tumor response, progression status, the date of progression, survival status, the date of death. Part of the death time was obtained by telephone follow-up.

Statistical analysis was conducted using IBM SPSS 26.0, StataSE 16, R (v. 4.4.1) software. The chi-square test or Fisher’s exact test was employed to compare enumeration data. Calculate the optimal cutoff value based on the ROC curve. Survival curves were generated using the Kaplan-Meier method and compared using the log-rank test. To further reduce survival time bias, we performed a landmark analysis. False Discovery Rate (FDR) was computed using the Benjamini–Hochberg method and a 10% FDR threshold was used. Multivariate survival analysis was performed employing a Cox proportional hazards model. Utilize R (v. 4.4.1) to generate a forest plot for the Cox proportional hazards model. In order to balance the differences in baseline characteristics between the two groups, propensity score matching (PSM) analysis was conducted. Different models were compared using maximum likelihood ratio to select covariates that needed to be matched. Using logistic regression, propensity scores were calculated with the ED group as the reference. The 1:1 nearest neighbor matching method was employed with a caliper value of 0.05. After matching, PSM parallelism hypothesis test was performed to examine whether the matching results were balanced. Kaplan-Meier method was used again on the balanced data to plot survival curves and conduct Log-rank test for comparison between curves. A significance level below 0.05 was deemed to indicate statistical significance.

We initially enrolled 112 patients, excluded 4 due to lack of imaging assessment after ICIs therapy and lost 4 to follow-up, leaving 104 included in the data analysis. Flow diagram of the patient selection is presented in the STROBE flow diagram ( Figure 1 ). Among the patients included, 46 (44.2%) were in the baseline ED group, while 58 (55.8%) were in the baseline no-ED group. Regarding treatment regimen, 70 (67.3%) patients were administered ICIs in conjunction with chemotherapy; 22 (21.2%) patients were treated with a combination of ICIs and targeted therapy, while 12 (11.5%) patients underwent a regimen that included ICIs, chemotherapy, and targeted therapy. No significant differences were observed in age, gender, marital status, job status, caregiver, residence, monthly income, educational level, BMI, PD-L1 expression CPS, status of HER2, TNM stage, treatment regimen, treatment lines, MSI status, ACTH level, cortisol level, ECOG PS, between the ED group and the no-ED group. The clinical feature data are summarized in Figures 2A , B .

The median follow-up time for the overall patients was 19.667 [95% confidence interval (CI): 15.637-23.697] months with 73 (70.1%) patients died during the follow-up period. The minimum follow-up time in this study was 55 days. The median OS for the entire patient cohort was 12.267 months (95%CI: 10.061-14.473), while the median PFS was 5.467 months (95%CI: 3.778-7.156). In terms of clinical efficacy assessment, none of the patients achieved CR, but PR was observed in 5 (4.8%) cases, and SD was seen in 58 (55.8%) cases, resulting in an ORR of 4.8%(95% CI: 0.6%-9.0%) and DCR of 60.6%(95%CI: 51.0%-70.1%) respectively ( Table 1 ).

The ORR in the baseline no-ED group was 8.6% (95%CI: 1.2%-16.1%), while no patients had a tumor response of PR or CR in the baseline ED group, leading to no statistically significant difference between the two groups (P=0.138). The DCR was significantly different with 39.1%(95%CI: 24.5%-53.8%) for the ED group and 77.6%(95%CI: 66.5%-88.6%) for the no-ED group (RR: 0.504, 95%CI: 0.343-0.742, P=0.001, Table 1 ), these suggested that the no-ED group had a better response to ICIs.

In the univariate analysis of OS ( Figure 3A , Supplementary Table S1 ), the median OS for patients in the baseline ED group was 8.700 months (95% CI: 8.102-9.298), compared to 15.500 months (95% CI: 13.423-17.577) in the baseline no-ED group. This difference was statistically significant with a hazard ratio (HR) of 2.035 (95% CI: 1.272-3.254, P=0.003), and remained significant after Benjamini–Hochberg correction. This suggested that patients in the baseline ED group had approximately double the mortality risk compared to those in the baseline no-ED group. For the p-values in the subgroup analysis of OS according to baseline ED state, the FDR was 7.1% (1/14), suggesting that baseline ED remained significantly associated with the risk of death in AGC patients receiving ICIs treatment after Benjamini–Hochberg correction ( Supplementary Figure S1 , Supplementary Table S2 ). The median OS for patients exhibiting a baseline depressive state was 8.533 months (95%CI: 7.801-9.265), whereas that for those without depressive symptoms at baseline was 15.500 months (95% CI: 13.158-17.842). This difference was statistically significant with a HR of 2.231 (95% CI: 1.396-3.564, P=0.001), indicating that a depressive state was significantly associated with poorer survival outcomes in AGC patients undergoing ICIs. The median OS for patients exhibiting a baseline anxiety state was 9.433 months (95%CI: 5.742-13.124), whereas the median OS for those without anxiety was 13.800 months (95%CI: 11.639-15.961). Although the median OS in the baseline anxiety group was shorter than that of the non-anxiety group, this difference did not reach statistical significance (HR: 1.288, 95% CI:0.729-2.275, P=0.383). In addition to the aforementioned factors, the ECOG PS (HR: 1.632, 95% CI: 1.01-2.637, P=0.045), TNM stage (HR: 2.229, 95% CI: 1.191-4.172, P=0.012), and treatment lines (HR: 1.644, 95% CI: 1.016-2.66, P=0.043) were found to be significantly associated with survival outcomes in AGC patients undergoing treatment with ICIs. Following a multivariate analysis incorporating 4 factors ( Figure 3B ), the mortality risk for the ED group remained significantly elevated compared to that of the no-ED group (HR: 1.717, 95% CI: 1.055-2.795, P=0.03), suggesting that ED independently correlated with adverse survival outcomes in AGC patients undergoing ICIs treatment. The OS-related Kaplan-Meier survival curves for baseline ED state was showed in Figure 4A and depressive state was showed in Figure 4B .

In the landmark analysis for OS according to baseline ED state, the risk of death in the baseline ED group was significantly higher than that in the no-ED group in the analysis after 3 months (P=0.007), while no significant difference was observed between the two groups in the analysis within 3 months (P=0.146), as shown in Figure 4C . Due to the median OS of AGC patients receiving ICIs combination therapy exceeding 1 year, and patients who did not undergo the second imaging evaluation not being included in the statistical analysis of this study, the proportion of deaths in both groups within 3 months was small (10.9% in the ED group and 6.9% in the no-ED group), which in turn caused the P-value of the log-rank test for OS within 3 months to fail to show a significant difference.

In the univariate analysis of PFS ( Figure 5A , Supplementary Table S3 ), patients in the ED group had a median PFS of 2.933 months (95% CI: 2.528-3.338), while the median PFS for patients in the no-ED group was 7.467 months (95% CI: 5.228-9.706). This difference was statistically significant with a HR of 3.006 (95% CI: 1.922-4.100, P < 0.001), indicating the risk of progression for patients experiencing baseline ED state was approximately threefold that of their counterparts without ED. For the p-values in the subgroup analysis of PFS according to baseline ED state, the FDR was 0 (0/21), indicating that baseline ED was still significantly associated with the risk of progression in AGC patients receiving ICIs treatment after Benjamini–Hochberg correction( Supplementary Figure S2 , Supplementary Table S4 ). The median PFS for patients exhibiting a baseline depressive state was 3.100 months (95% CI: 2.571-3.629), whereas the median PFS for those without depressive symptoms was 7.033 months (95% CI: 4.911-9.155). This difference was statistically significant with a HR of 2.488 (95% CI: 1.59-3.891, P < 0.001), indicating that depressive state is associated with a higher risk of progression in AGC patients receiving ICIs therapy. Patients with baseline anxiety had a median PFS of 3.933 months (95% CI: 0.228 - 7.638), while those without anxiety had a median PFS of 5.833 months (4.157 - 7.509). The median PFS for patients with baseline anxiety state was shorter than that for non-anxious patients, but there was no statistically significant difference (HR: 1.599, 95% CI: 0.98-2.607, P=0.06). The results of multivariate analysis suggest that baseline ED is independently associated with the risk of progression in AGC patients received ICIs therapy ( Figure 5B ). The PFS-related Kaplan-Meier survival curves for baseline ED state was showed in Figure 6A and depressive state was showed in Figure 6B . In the landmark analysis for PFS according to baseline ED state, the progression risk in the baseline ED group was significantly higher than that in the no-ED group both within 3 months (P < 0.001) and after 3 months (P < 0.001), as shown in Figure 6C .

The propensity score was calculated based on five clinical characteristics, including TNM stage, HER2 status, baseline ACTH level, PD-L1 expression, and age. After matching, there were no significant differences in matched variables between the ED group and no-ED group, indicating that the balance assumption was met. The scatter plot of standardized mean differences for each covariate showed that all matched variables fell within the -0.25 to 0.25 range, which indicated that balance was achieved for all variables ( Figure 7 ). The matched data showed that there were 32 cases in the ED group and 32 cases in the no-ED group, with clinical characteristics after PSM shown in Figures 8A , B . Univariate analysis after PSM still indicated a significant association between ED and shorter PFS(HR:2.35, 95%CI:1.235-4.472, P=0.009), as well as a significant association between ED and shorter OS(HR:2.186, 95%CI:1.043-4.582, P=0.038) following ICIs treatment. Kaplan-Meier curves demonstrated significant intergroup differences in both OS(P=0.034, Figure 9A ) and PFS (P=0.007, Figure 9B ) based on ED grouping.

After 2 cycles of ICIs treatment, 80 (76.9%) patients underwent a second assessment of their depression and anxiety states. In the baseline No-ED group, 34 (42.5%) patients were in the ‘never ED’ group and 14 (17.5%) patients were in the ‘new onset ED’ group. In the baseline ED group, 28 (35.0%) patients were in the ‘persistent ED’ group and 12 (15.0%) patients were in the ‘remission from ED’ group.

In the ‘new onset ED’ group of baseline No-ED group, the median OS was 12.133 months (95%CI: 11.093-13.173), while in the ‘never ED’ group of baseline No-ED group, the median OS was 22.333 months (95%CI: 11.108-33.558). There was a significant difference between the two groups with a HR of 2.813 (95%CI:1.270-6.228, P=0.011), suggesting that ‘new onset ED’ was associated with poor survival outcomes of AGC patients undergoing ICIs treatment. In the ‘new onset ED’ group, the median PFS was 3.733 months (95%CI: 3.121-4.345), while in the ‘never ED’ group, the median PFS was 11.333 months (95%CI: 6.490-16.176). There was a significant difference between the two groups with a HR of 3.204 (95%CI: 1.526-6.730, P=0.002), suggesting that ‘new onset ED’ was associated with a higher risk of progression in AGC patients who receive ICIs treatment. The ‘ED persistently’ group from baseline ED group exhibited no significant difference in OS compared to the ‘ED remission’ group from baseline ED group (HR: 1.723, 95% CI: 0.764-3.884, P=0.190), and PFS also showed no significant difference between the two groups (HR: 1.314, 95% CI: 0.634-2.720, P=0.463). This finding indicates that the prognosis for AGC patients who experience alleviation of ED, despite no substantial improvement in overall emotional well-being following ICIs treatment, tends to be more favorable than that of patients with ongoing ED. The Kaplan–Meier curve was showed in Supplementary Figure S3A , S3B . This indicates that for GC patients with ED, timely implementation of intervention measures may help improve their prognosis during treatment with ICIs.

Among the 104 patients included in this study, 59 (56.7%) had cortisol levels measured at baseline with 48 (81.3%) in the higher cortisol group and 11 (18.6%) in the lower cortisol group. The missing data rate for cortisol level was 43.3%. In the baseline ED group, 26 (52.2%) patients exhibited higher cortisol levels, compared to 24 patients (41.4%) in the baseline no-ED group. Although a higher proportion of patients in the baseline ED group presented with higher cortisol levels, the difference between the two cohorts was not statistically significant (P = 0.137, Figure 2B ). Among patients with baseline depression, 21 (91.3%) exhibited higher cortisol levels, whereas among those without baseline depression, 27 (75.0%) demonstrated higher cortisol levels. No statistically significant difference was observed between the two groups (P = 0.174).

The median OS for the higher cortisol group was 13.900 months (95% CI: 10.238-17.562), whereas the median OS for the lower cortisol group has not yet been reached. For patients with AGC who receive ICIs treatment, higher baseline cortisol levels tend to have worse survival outcomes, but the difference is not statistically significant (HR: 2.318, 95% CI: 0.805-6.679, P=0.119, Figure 3A ).