Introduction

Over the last four decades and across multiple countries, the incidence rates of esophageal (EAC) and gastroesophageal junction adenocarcinoma (GEJAC) have been rising while the incidence of gastric adenocarcinoma (GAC) has been falling [1,2,3,4]. Nonetheless, GAC remains the fifth most common malignancy worldwide [5,6,7]. The incidence rates of EAC and GEJAC are 5-to-9-fold higher among males than females [1, 8, 9]. While for GAC, the incidence rate in males is approximately twice that for females in most studies [5, 6, 10]. The etiology underlying these sex disparities in incidence rates is not completely understood, with some data suggesting that sex hormones and other biological factors may play important roles [3].

The mortality rates for EAC, GEJAC and GAC continue to be some of the worst among all malignancies [1]. Despite the recognized male-female difference in incidence of EAC, GEJAC and GAC, data on potential sex disparities in overall survival (OS) of these three anatomically linked malignancies remain limited. Most studies using data that included patients with all stages have shown similar mortality rates between male and female patients or that females had better OS than males [1, 11,12,13,14,15]. Few studies have assessed the potential sex disparity in OS in patients with relapsed and metastatic EAC, GEJAC and GAC independently.

The importance of the cancer risk and survival differences between male and female patients is being increasingly recognized [16, 17]. The investigation of sex-biased survival disparity could provide critical insights in improving the outcomes of both men and women with cancer [16, 17]. The aims of this study were to examine potential male-female differences in OS among patients with relapsed and metastatic GAC, EAC and GEJAC and whether sex differences in OS vary by molecular features. Since late 2017 next-generation sequencing (NGS) was routinely performed for patients with advanced malignancy within Kaiser Permanente Northern California (KPNC). The study was conducted in the cohort of patients with advanced upper gastrointestinal adenocarcinomas whose tumors underwent NGS between 2017 and 2023. We used Cox regressing modeling for analysis adjusting for six most common genomic alterations in this cohort of patients in addition to demographics, Lauren’s classification as well as other factors.

Methods

Study population

Beginning late 2017 next-generation sequencing (NGS) was routinely performed for patients with advanced malignancy within Kaiser Permanente Northern California (KPNC). This cohort included 875 eligible KPNC patients with relapsed or de novo metastatic GAC (n = 426), GEJAC (n = 276), and EAC (n = 173). We excluded patients with esophageal squamous cell carcinoma. The molecular profiling was performed using StrataNGS (Ann Arbor, Michigan) from November 2017 to December 2023. Patient data on demographics, Charlson comorbidity index (CCI), performance status (PS), histology (Lauren’s classification) and receipt of systemic therapy were obtained from the electronic medical record (Epic) and cancer registry databases. CCI was based on the 12 month-period prior to diagnosis of relapsed or de novo metastatic EAC, GEJAC and GAC. Approximately 80.1% of female and 45.2% of male stage 4 GAC patients within KPNC had NGS performed during the study period. This study was approved by the KPNC institutional review board with waiver of consent and was conducted according to International Ethical Guidelines for Biomedical Research Involving Human Subjects (CIOMS).

NGS

During the study period, StrataNGS was a 429-gene, pan-solid tumor, NGS assay for formalin-fixed paraffin-embedded (FFPE) tumor tissue, performed on co-isolated DNA and RNA [18]. For the full study cohort, alterations in more than 110 individual genes were identified. We identified the six most common genomic alterations in this study including HER2 overexpression or amplification (HER2 + ), TP53 (mutp53), CDKN2A (mutCDKN2A), KRAS (mutKRAS), PIK3CA (mutPIK3CA) mutation and MYC amplification.

HER2+ definition

HER2 overexpression or amplification was considered positive (HER2 + ) if HER2 expression was 3+ in 10% or more tumor cells by immunohistochemistry, or positive by fluorescent in situ hybridization (FISH), or positive with six copies of ERBB2 by NGS [19]. A total of 133 patients had HER2+ tumors, of these, 99 (74.4%) received trastuzumab.

Gastroesophageal junction (GEJ) definition

Tumors were classified as GEJ based on reports by gastroenterologists who performed diagnostic upper endoscopies with a tissue biopsy and was defined as the anatomic area that represents the junction between the distal esophagus and the proximal stomach (cardia) [20].

Histology

Lauren’s classification data in patients with advanced GAC were extracted from pathology reports and included diffuse type, intestinal type, mixed type and unknown.

Helicobacter Pylori and Epstein Barr Virus Tests

Among 875 patients, 265 patients (190 GAC and 75 EAC) had Helicobacter Pylori test performed, only 13 (11 GAC and 2 EAC) had a positive test. There were only 3 patients with Epstein Barr Virus (EBV) test performed and among them one had a positive test.

Chemotherapy

Chemotherapy use (yes, no) was based on whether agents were administered after the diagnosis of relapsed or de novo metastatic disease and consisted of primarily fluorouracil-based multiagent regimens with or without a checkpoint inhibitor, or with or without Trastuzumab (for HER2 positive patients).

PD-L1 expression and therapy with immune checkpoint inhibitor

There were 314 patients with PD-L1 expression test results available, among them 195 had positive PD-L1 expression and 119 patients had negative PD-L1 expression. Approximately 50% of all patients received treatment with either an immune checkpoint inhibitor (ICI) nivolumab or pembrolizumab as single agent or in combination with chemotherapy.

Definition of gain-of-function (GOF) and non-gain-of-function (non-GOF) TP53 mutation

The mutp53 GOF in our study was defined as R175H, R248Q, R248W, R249S, R273H, R273L, and R282W, all located at the DNA binding domain (DBD) and all the other p53 alterations were grouped as non-GOF, same as in our previous study [21].

Statistical analysis

OS was measured from the date of diagnosis of relapsed or metastatic EAC, GEJAC and GAC to the date of death or end of study follow-up (March 10th, 2024), whichever came first. Patients who were still alive at the end of the study period were censored at that time. Median follow-up time was 37.6 months for females and 31.4 months for males. We used Pearson’s chi squared test to assess differences in distributions of demographics and HER2 + , mutp53, mutKRAS, mutCDKN2A, mutPIK3CA and MYC amplification. We used the one-way ANOVA test to assess differences in continuous variables. We used the Kaplan-Meier estimator (log rank test) to generate unadjusted (univariate) OS curves and estimate median OS. The number of patients at risk in the Kaplan-Meier OS curves accounted for delayed entry into the cohort at the time of receipt of NGS results (i.e., left truncation, with median study entry of 0.7 months post-diagnosis) [22]. Cox proportional regression modeling was used to estimate hazard ratios (HRs) and 95% confidence intervals (CI) for the association between anatomic subgroups or sex and OS, adjusted for covariates. Time since diagnosis of advanced adenocarcinoma was the time scale used in the regression models, with allowance for delayed entry into the cohort. Covariates included in our main regression models (and unless otherwise stated) were age (continuous), sex (male, female), ethnicity (Non-Hispanic White, Black, Asian, Hispanic, other/unknown), PS (0 to 1, 2 to 4), CCI (continuous), histology (diffuse type, intestinal type, mixed type, unknown), treatment received (yes, no). We examined the effect of specific genomic alteration in a model that included all six alterations simultaneously: (HER2+ [yes, no]), mutp53 [yes, no], mutCDKN2A [yes, no], mutKRAS [yes, no], mutPIK3CA (yes, no), and MYC amplification (yes, no, unknown).

Because patients who did not receive chemotherapy had poorer OS, we performed independent analysis of both subgroups of patients who received and who did not receive chemotherapy. The statistical analysis was performed using SAS software version 9.4, R (R Core Team, 2020).

Results

Distributions of demographic and clinical characteristics among sex and anatomic subgroups

The distributions of clinical characteristics among male (n = 224) and female (n = 202) patients with GAC, the largest anatomic group (n = 426), are presented in Table 1. Male versus female GAC patients had fewer cases with diffuse type histology by Lauren’s classification and had higher CCI but were similar with respect to all other characteristics including receipt of chemotherapy and ICI as well as the distribution of the six most common genomic alterations (Table 1). For GEJAC (n = 276), male versus female patients also appeared to be older and had more patients of Asian and White decent but similar in all other characteristics (Supplementary Table 1). For EAC (n = 173), male versus female patients appeared to be older, had higher percent of Asian and Non-Hispanic White decent, higher CCI, but similar PS, receipt of chemotherapy, and percent of genomic alterations (Supplementary Table 2).

Table 1 Demographics of male and female patients with gastric adenocarcinoma (GAC).
Full size table

Association of sex and OS among the full cohort and anatomic subgroups

Among all GAC patients (full cohort), male versus female had better OS (hazardous ratio [HR] = 0.73, [95% confidence interval [CI], 0.59–0.92]) (Fig. 1a). The median OS was 8.9 versus 7.6 months by Kaplan-Meier curves (log-rank test, P = 0.07. Fig. 1b). The number of patients still at risk (alive) at 12 months was 69 (34.1%) for females and 83 (37.0%) for males, and the number of patients still at risk (alive) at 24 months was 27 (13.4%) for females and 43 (19.2%) for males (Fig. 1b). The results for the association of sex and OS were similar when GAC analyses were restricted to patients who received chemotherapy (HR = 0.75; [95% CI, 0.60–0.94]) and to patients who did not receive chemotherapy (HR = 0.52; [95% CI, 0.31–0.89]) (Fig. 1a). The HRs of OS between male versus female among the entire GAC cohort, the sub-cohort that received chemotherapy and the sub-cohort that did not receive chemotherapy remain significant when false discovery rate (FDR) was calculated (Fig. 1a). The median OS was 2.3 versus 1.3 months for male versus female GAC patients who did not receive chemotherapy (log-rank test, P = 0.10, Fig. 1c). In contrast, male and female patients with EAC or GEJAC had similar OS (Fig. 1a and Supplementary Figs. 1 and 2). Chemotherapy was strongly associated with better OS among both male (HR = 0.24; [95% CI, 0.16–0.38]) and female (HR = 0.18; [95% CI, 0.12–0.29]) GAC patients (Fig. 1d). Using cBioPortal (https://www.cbioportal.org/study/clinicalData?id=stad_tcga_gdc) we identified patients with stage 4 gastric adenocarcinoma in The Cancer Genome Atlas (TCGA) database. There were total of 44 patients with stage 4 gastric adenocarcinoma in this portal, among them 19 patients were females and 25 were males. The HR on OS was 0.62 (0.27–1.40) favoring male patients though the confidence interval was wide due to small sample size (Table 2).

Fig. 1: Association of sex and OS among the full cohort and anatomic subgroups.
figure 1

a OS association of the full cohort. b Kaplan-Meier curves by sex in patients with gastric adenocarcinoma. c Kaplan-Meier curves by sex in patients with gastric adenocarcinoma not treated with chemotherapy. d OS and chemotherapy association by sex.

Full size image
Table 2 HR of OS between male and female patients with stage 4 gastric adenocarcinoma from TCGA dataset in cBioPortal (https://www.cbioportal.org/study/clinicalData?id=stad_tcga_gdc).
Full size table

Association of sex and OS among anatomic subgroups stratified by HER2 status

Approximately 9% of patients with GAC had HER2+ disease (Table 1). HER2+ disease was more common among patients with EAC (24.9%) or GEJAC (18.8%) (Supplemental Tables 1 and 2). Among GAC patients, males had better OS than females among both HER2+ (HR = 0.73; [95% CI, 0.58–0.92]) and HER2 negative (HR = 0.14; [95% CI, 0.02–0.93]) subgroups (Fig. 2). In contrast, there was no OS difference between male and female patients among EAC HER2 negative patients (there were only 3 EAC HER2+ female patients, sample size too small for meaningful comparison.), or among GEJAC HER2+ and HER2 negative patients (Supplementary Fig. 3).

Fig. 2: Forest plots of HRs for the association of sex with OS among GAC patients by HER2 status.
figure 2

HR hazardous ratio, OS overall survival. GAC gastric adenocarcinoma.

Full size image

Association of sex and OS among anatomic subgroups stratified by mutp53 status

Approximately 59% of male and 54% of female GAC patients had mutp53 (Table 1). Male versus female GAC patients with either wild-type p53 (wtp53, HR = 0.68; [95% CI, 0.47–0.98]) or mutp53 (HR = 0.78; [95% CI, 0.57–1.06]) had better OS (Fig. 3). Intriguingly, among GAC patients with mutp53 GOF, males appeared to have worse OS than females (HR, 1.80, [95% CI, 0.83–3.99]), while among GAC patients with mutp53 non-GOF, males versus females had better OS (HR = 0.59; [95% CI, 0.40–0.85]). In addition, this GOF versus non-GOF difference was confirmed with interaction analysis that shows significant statistical difference HR for interaction term = 2.54, [95% CI, 1.25–5.16] (Fig. 3). Nearly 90% of patients with EAC had mutp53 (Supplementary Table 1). Male versus female EAC patients with mutp53 had no OS difference (Supplementary Fig. 4). Approximately 79% of male and 68% of female patients with GEJAC had mutp53 (Supplementary Table 2), male versus female GEJAC patients with mutp53 also had no OS difference (Supplementary Fig. 4).

Fig. 3: Forest plots of HRs for the association of sex with OS among GAC patients by p53 mutation status.
figure 3

HR hazardous ratio, OS overall survival, GAC gastric adenocarcinoma.

Full size image

Association of sex and OS among anatomic subgroups stratified by mutKRAS status

Male versus female GAC patients with wild-type KRAS (wtKRAS) had substantially better OS (HR = 0.69, [95% CI, 0.54–0.88]), while male versus female GAC patients with mutKRAS had no OS difference (HR = 1.00, [95% CI, 0.40–2.22]) (Fig. 4a). Male versus female EAC and GEJAC patients with wtKRAS had no OS difference (Supplementary Fig. 5). Male versus female GEJAC patients with mutKRAS appeared to have worse OS but with wide confidence interval (CI) due to small sample size (HR = 2.04, [95% CI, 0.74–5.66]) (Supplementary Fig. 5).

Fig. 4: OS of male versus female patients by genomic alterations.
figure 4

A OS by KRAS mutation. B OS by CDKN2A mutation. C OS by PI3KCA mutation.

Full size image

Association of sex and OS among anatomic subgroups stratified by mutCDKN2A status

Male versus female GAC patients with either wild-type CDKN2A (wtCDKN2A) (HR = 0.75; [95% CI, 0.59–0.96]) or mutCDKN2A (HR = 0.67, [95% CI, 0.37–1.20]) had better OS (Fig. 4b). Male versus female EAC patients with wtCDKN2A had no OS difference (Supplementary Fig. 6). Male versus female GEJAC patients with wtCDKN2A or mutCDKN2A had no OS differences (Supplementary Fig. 6).

Association of sex and OS among anatomic subgroups stratified by mutPIK3CA status

Male versus female GAC patients with wild-type PIK3CA (wtPIK3CA) had better OS (HR = 0.69; [95% CI, 0.54–0.88]), while male versus female GAC patients with mutPIK3CA had no OS difference (Fig. 4c). Male versus female EAC or GEJAC patients with wtPIK3CA had no OS difference (Supplementary Fig. 7).

Association of sex and OS among anatomic subgroups stratified by MYC amplification status

Males had better OS than females among GAC patients with MYC amplification negative (HR = 0.72; [95% CI, 0.59–0.91]) (Fig. 5a). The sample size for the GAC patients with MYC amplification positive was too small (n = 12 for males and 11 for females) for meaningful analysis. Male versus female EAC or GEJAC patients with MYC amplification negative had no OS difference (Supplementary Fig. 8). There were too few female EAC or GEJAC patients with MYC amplification positive to allow meaningful comparison with male patients.

Fig. 5: OS of male versus female patients in patients not treated with chemotherapy.
figure 5

a OS of male versus female patients with MYC amplification negative. b OS of male versus female patients in patients not treated with chemotherapy by molecular subtype. c Summary of OS of male versus female patients by molecular subtypes.

Full size image

OS of male versus female GAC patients who did not receive chemotherapy by molecular subgroups

We performed an independent analysis of GAC patients who did not receive chemotherapy. The OS of male versus female GAC patients who did not receive chemotherapy was better across the molecular subgroups except that of mutKRAS (Fig. 5b), which is similar to the results with the full cohort.

Summary of data

Fig. 5c summarizes the data. Males had better OS than females among patients with GAC but not among patients with EAC or GEJAC. Among GAC patients, male versus female had better OS across the molecular subgroups except mutKRAS. Male versus female GAC patients with mutp53 had substantially better OS if tumor carried non-GOF but worse OS if tumor carried GOF, and this was confirmed by interaction analysis. When analysis was performed independently for patients who received chemotherapy and for patients who did not receive chemotherapy, the results were similar.

Discussion

In this study with a large dataset from a community multi-center healthcare organization, we have shown that male versus female had better OS among advanced GAC but not among advanced EAC or GEJAC patients. This modest OS difference between male versus female GAC patients appeared preserved across the molecular subgroups except mutKRAS. Intriguingly, among GAC patients with mutp53, male versus female had better OS if tumor carried non-GOF but worse OS if tumor carried GOF. Both males and females showed similar OS association with chemotherapy and HER2 positivity, therefore suggesting that chemotherapy and HER2 expression were not associated with this sex disparity in survival outcomes in patients with advanced GAC.

This sex disparity in survival outcomes among advanced GAC patients was unexpected as previous epidemiology studies including all disease stages had suggested that female patients had similar or better survival outcomes compared to male patients with GAC [1, 12, 23]. More intriguingly, OS of male versus female GAC patients with mutp53 was inversely associated with the presence of non-GOF versus GOF. To our knowledge this is the first time that mutp53 is found to be associated with a prognostic sex dichotomy in a malignancy. The mechanism of this mutp53 non-GOF versus GOF association with sex dichotomy is not clear and will be interesting to investigate further. A bioinformatic study showed that among all malignancies male cancer patients carried higher rate of mutp53 than female cancer patients and patients with mutp53 had poorer OS than patients with wtp53 [23]. Our data with EAC, GEJAC and GAC all showed similar rate of mutp53 between male and female patients. The same bioinformatic study also showed that X-chromosome encodes many proteins that interact with wtp53 and that many cancers express inhibitors of wtp53 [24, 25]. Our current finding, together with the finding in our accompanying manuscript that shows that mutp53 GOF versus non-GOF was associated with worse OS among patients with GAC but not EAC or GEJAC, adds new dimensions to the complexity and versatility of mutp53 and expands upon our previous discoveries that mutp53 non-GOF was associated with worse OS of right-sided versus left-sided CRC while mutp53 GOF versus non-GOF was associated with worse OS in left-sided CRC and pancreatic ductal adenocarcinoma [21, 26].

Sex disparities in outcomes have been previously found in other malignancies. We recently reported that male versus female patients with advanced intermediate to high-grade soft tissue sarcoma had better OS, and that male versus female patients with SMARCA4-mutated advanced malignancies had worse OS [27, 28]. In a study of stage I and II melanoma patients it was found that females had superior outcomes than males [29]. Epidemiology studies have found that in general females appear to have better outcomes than males in majority of malignancies [12, 15, 23]. However, these studies included all stages of malignancies and lacked the comparative specificities presented in our work that restricted the analysis to advanced stage only and with the most common genomic alterations integrated into the statistical modeling. The factors and mechanism underlying such a disparity remain poorly understood and could include sex hormones, tumor biology and social factors. Our study suggests that tumor biology likely plays important roles. A recent study showed that loss of Y chromosome in bladder carcinoma was associated with poor OS, and with a suppressive tumor microenvironment characterized by exhaustion of CD8+ T cells which sensitizes the tumors to immune check point inhibitors [30]. Loss of Y chromosome has been reported to be associated with poor prognosis of patients with head and neck cancer and EAC [31, 32]. Loss of Y chromosome has also been found in GAC and GAC cell lines [33, 34]. It is not clear if loss of Y chromosome plays a role in the pathology and outcomes of patients with GEJAC and GAC. Recently biological mechanisms underlying predominant female incidence in systemic lupus was found to be related to Xist, a long non-coding RNA that is only expressed in females and the primary effector of X-inactivation, using a transgenic male mouse model [35]. Interestingly, p53 directly regulates the expression of Xist [36]. Studies using TCGA and other publicly available datasets have shown sex-dependent differences in the rate of certain mutations (e.g. STK11 mutation more common in men, DMD mutation more common in women, etc.), driver mutation signatures, copy number variations, and gene expression signatures, and tumor mutation burden, etc. [37, 38]. There are also evidences that males and females have different metabolism of cancer drugs, such as rituximab and fluoropyrimidine leading to different dosing of rituximab and higher toxicities of fluoropyrimidine-based regimens in women [39, 40].

Our study has limitations. It is a retrospective study and the number of patients with some of our subgroups was relatively small. In addition, the length of follow-up for some patients remained relatively short. Also, more than half of the GAC patients did not have histology type (Lauren’s classification) data available. Our study has several strengths. Our dataset is relatively large, with 426 GAC patients and nearly 900 total patients who received comprehensive primary and specialty services from a large integrated healthcare system that consists of 21 medical centers. In addition, the diverse membership is relatively stable, and electronic records capture virtually all encounters, diagnoses, and procedures. Furthermore, our study was restricted to advanced stage only and with molecular features integrated into the statistical modeling in addition to the other important factors.

In summary, our study suggests that males versus females had better OS among patients with advanced GAC but not among patients with advanced EAC or GEJAC. This sex disparity among advanced GAC patients was preserved across all molecular subgroups except mutKRAS. In addition, the sex and OS association among GAC patients with mutp53 was inversely driven by the presence of non-GOF versus GOF. To our knowledge our data for the first-time linked a sex-dependent cancer prognosis to mutp53. Our findings provide additional evidence illustrating the importance of sex disparities in cancer outcomes which may facilitate future studies on p53 biology, especially on its mechanistic interplay with sex and spatial biology.