Abstract
Background
Despite therapeutic advances, gastroesophageal cancers remain associated with high mortality. While most deaths are cancer-related, improvements in therapy and supportive care may alter mortality patterns over time. This study aimed to characterize causes of death and associated clinical factors in a large real-world European cohort.
Methods
We retrospectively included 2518 patients with histologically confirmed esophageal, gastric, or gastroesophageal junction carcinomas treated at the Medical University of Vienna between 1994 and 2024. Causes of death were obtained from Austria’s national death registry and categorized using International Classification of Diseases (ICD) codes. Fine–Gray competing risk models were applied to identify factors associated with cancer- and non-cancer-related deaths.
Results
At data cutoff, 1863 patients (74%) had died. Of these, 85% of deaths were cancer-related, 12% non-cancer-related, 2% unknown, and <1% suicide. Tumor stage was the strongest predictor of cancer-related mortality, with subdistribution hazard ratios (sHR) increasing from 2.34 (95% CI 1.91–2.86) in stage 2 to 7.42 (95% CI 6.15–8.96) in stage 4 disease compared with stage 1 (all p < 0.001). A higher comorbidity burden also independently increased cancer-related mortality. Non-cancer-related death was primarily associated with older age (≥65 years; sHR 2.21, 95% CI 1.12–4.35, p = 0.021) and stomach tumor location (sHR 1.56, 95% CI 1.08–2.23, p = 0.017), while more recent diagnosis (2020–2024) was linked to a lower risk of both cancer- and non-cancer-related mortality (cancer-related: sHR 0.60, 95% CI 0.47–0.76, p < 0.001; non-cancer-related: sHR 0.43, 95% CI 0.25–0.74, p = 0.002), likely reflecting a combination of evolving treatment strategies, improvements in supportive care, and shorter follow-up in the most recent cohort.
Conclusion
Cancer progression remains the predominant cause of death in patients with gastroesophageal cancer, with tumor stage as the key prognostic factor. These high mortality rates emphasize the need for improved antitumoral treatment and timely palliative care approaches.
Keywords: gastric cancer, esophageal cancer, tumor progression, survival, cause of death
Introduction
Gastroesophageal cancers, including gastric cancer, esophageal cancer, and cancers of the gastroesophageal junction (GEJ), make up a big part of all diagnosed cases and cancer deaths worldwide.1 In 2022, gastric cancer had the fifth highest incidence and mortality rate of all cancer entities. Fewer rates of carcinomas of the esophagus were recorded, but they were still accountable for almost half a million deaths, showing the seventh highest mortality. Most cases were described in eastern Asia followed by eastern Europe, and men have been affected approximately twice as often as women.1
The prognosis of gastroesophageal cancer remains poor.2,3 Most patients are diagnosed at advanced or metastatic stages, where curative treatment is rarely feasible. Although multimodal strategies – including perioperative chemotherapy, definitive chemoradiation, targeted agents, and more recently immune checkpoint inhibitors – have improved outcomes in selected subgroups, overall survival (OS) rates remain limited compared to many other solid tumors.4,5 Consequently, mortality rates remain high even in high-resource healthcare systems.
While tumor progression accounts for the majority of deaths in this population, patients with gastroesophageal cancer are often older and frequently present with substantial comorbidities, including cardiovascular disease, metabolic disorders, and chronic pulmonary conditions, as well as varying degrees of frailty, all of which may significantly influence non-cancer-related mortality and overall outcomes. In addition, cancer- and treatment-related toxicities may increase the risk of infection, thromboembolic events, cardiovascular complications, and other non-cancer causes of death.6–9 Psychosocial distress, particularly in patients with advanced disease and impaired quality of life, may further contribute to excess mortality from potentially preventable causes, including suicide.10,11 Although suicide represents a rare outcome, it is a potentially preventable cause of death and may provide insight into unmet psychosocial and supportive care needs in patients with gastroesophageal cancer.
Distinguishing between cancer-related and non-cancer-related death is important, as age, frailty and substantial comorbidity may influence the balance between treatment efficacy and toxicity. A better understanding of cause-specific mortality may therefore support individualized treatment intensity, optimize supportive and preventive care, and facilitate individualized decision-making regarding oncology therapy. Since patients who die from non-cancer causes are no longer at risk of cancer-related death, competing-risk analyses provide more accurate estimates of cause-specific mortality than conventional time-to-event methods. Over the past three decades, the therapeutic landscape of gastroesophageal cancers has changed substantially. Improvements in systemic therapies, ranging from cytotoxic chemotherapy to targeted therapies and immune checkpoint inhibitors, as well as advances in surgical techniques, radiotherapy, supportive care, and palliative medicine, have altered the natural history of gastroesophageal cancers. These developments are expected not only to prolong survival but also to modify the relative contribution of cancer- and non-cancer-related causes of death over time. However, real-world data on detailed cause-of-death patterns - particularly from long-term European cohorts - remain scarce. Most available evidence derives from registry-based studies with limited granularity regarding clinicopathological variables, molecular characteristics, and treatment details.12–14
A comprehensive evaluation of cause-specific mortality in a well-characterized real-world cohort spanning nearly three decades offers the opportunity to quantify the relative contribution of cancer-related and non-cancer-related deaths, to identify clinicopathological and sociodemographic factors associated with specific causes of death, and to assess temporal trends reflecting advances in oncologic therapy, supportive care (aimed at treatment tolerability across the disease trajectory), and palliative care (focused on symptom management, quality of life and end-of-life care). Such data are crucial for optimizing multidisciplinary management strategies, improving supportive and psycho-oncological interventions, and guiding resource allocation. Moreover, understanding potentially preventable non-cancer deaths may help to develop targeted strategies aimed at reducing excess mortality beyond tumor control alone.
Methods
Study Design and Population
This retrospective, single-center study was conducted at the Department of Medicine I – Division of Oncology, Vienna General Hospital, Medical University of Vienna, a high-volume tertiary referral center for upper gastrointestinal malignancies in Central Europe. We included all adult patients (≥18 years) with histologically confirmed esophageal, gastric, or gastroesophageal junction carcinoma treated between 1 January 1994 and 31 December 2024. Patients without histological confirmation or with inaccessible medical records were excluded.
All patients were treated according to the contemporary standard of care at the time of diagnosis and were routinely discussed in an interdisciplinary tumor board. Tumor staging was performed post-hoc for all patients according to the Union for International Cancer Control (UICC) 8th edition TNM classification. In resectable cases where endosonography was not available to reliably distinguish between clinical stage 2 and stage 3 disease, tumors were categorized as stage 2–3.
Data Sources and Variables
Clinical and pathological data were retrospectively extracted from the institutional hospital information system (AKIM) and transferred to the password-protected “Vienna Gastroesophageal Cancer Database” (FileMaker Pro®-based). Data were pseudonymized prior to analysis.
Collected variables included demographic characteristics, lifestyle factors (nicotine, alcohol), comorbidities, tumor characteristics (location, stage, histology, molecular markers including HER2, MSI/MMR, PD-L1), treatment modalities, and disease course (recurrence/progression, survival status).
Given the long study period, year of diagnosis was categorized into six predefined 5-year intervals (1994–1999, 2000–2004, 2005–2009, 2010–2014, 2015–2019, 2020–2024) to account for major temporal changes in diagnostic procedures and therapeutic strategies, including the introduction of perioperative chemotherapy, molecularly targeted therapies, and immune checkpoint inhibitors.
Comorbidities included all physician-documented chronic medical conditions present at the time of cancer diagnosis. For statistical analyses, the total number of documented comorbidities was calculated for each patient. No validated comorbidity index (eg, Charlson Comorbidity Index) was applied because complete information required for retrospective score calculation was not consistently available across the entire study period.
Molecular markers were assessed according to routine clinical practice at the time of diagnosis. As molecular testing was introduced gradually during the study period (HER2 since 2010, PD-L1 and MSI/MMR since 2018), these variables were available only for subsets of patients diagnosed in more recent years and were therefore not included in the multivariable competing-risk analyses.
Ascertainment of Death and Cause of Death
Vital status and cause of death were obtained from the national death registry of Statistik Austria, in which each deceased individual is assigned an underlying cause of death according to the International Classification of Diseases (ICD).
Causes of death were grouped into predefined categories (eg tumor progression, cardiovascular, infectious, suicide, unknown). If the ICD code was inconclusive or insufficiently specific, additional information was retrieved from the hospital information system (AKIM) to adjudicate the most plausible cause of death. These cases were independently reviewed by two investigators, and any discrepancies were resolved by consensus.
Survival follow-up was updated until the predefined data cut-off date of 29 December 2025.
Statistical Analysis
Categorical variables are presented as frequencies and percentages and were compared using Chi-square or Fisher’s exact tests. Continuous variables are summarized using mean (standard deviation) or median (interquartile range). OS was estimated using the Kaplan–Meier method.
Cause-specific mortality was analyzed using Fine–Gray hazard regression models. Time-to-event was calculated from the date of diagnosis to death or last follow-up. Cancer-related death was defined as the primary event of interest. Non-cancer-related death, suicide, and deaths of unknown cause were treated as competing events. In secondary analyses, non-cancer-related death was modeled as the event of interest.
Multivariable models were adjusted a priori for sex, age at diagnosis (≤45, 45–65, ≥65 years), year of diagnosis (1994–1999, 2000–2004, 2005–2009, 2010–2014, 2015–2019, 2020–2024), comorbidity burden (0, 1–2, 3–5, 6–10, 11–15, >16 comorbidities), histological subtype (adenocarcinoma vs squamous cell carcinoma), tumor location (gastroesophageal junction, stomach, esophagus), and tumor stage (1, 2, 2–3, 3, 4). No imputation of missing data was performed; regression analyses were based on complete cases for the covariates included in each model. Subdistribution hazard ratios (sHRs) with 95% confidence intervals (CIs) were reported.
Cumulative incidence functions were estimated for visualization of cause-specific mortality. Forest plots were generated to display adjusted sHRs and 95% CIs. Statistical analyses were performed using R (version 4.2.3) with the packages dplyr, ggplot2, gtsummary, cmprsk, survival.
Given the exploratory and hypothesis-generating nature of this retrospective study, analyses were considered descriptive and hypothesis-generating. Therefore, no formal sensitivity analyses or correction for multiple testing were prespecified. Statistical inference should be interpreted cautiously. Effect estimates with corresponding 95% confidence intervals were considered the primary basis for interpretation, while p-values are reported as complementary measures of statistical uncertainty.
Results
Patient and Tumor Characteristics
A total of 2518 patients treated between 1994 and 2024 were included in the analysis. At the time of data cutoff, 1863 patients (74%) were deceased.
Overall, 1761 patients (70%) were male. The median age at diagnosis was 64 years (IQR 55–72). The median number of comorbidities per patient was 4 (IQR 2–6).
Histologically, 1976 tumors (78%) were adenocarcinomas. Tumor location was distributed as follows: esophagus in 821 patients (33%), GEJ in 707 (28%), and stomach in 990 (39%). Tumor stage at diagnosis was stage 1 in 438 patients (17%), stage 2 in 421 (17%), stage 2–3 in 208 (8%), stage 3 in 610 (24%), and stage 4 in 841 (33%).
Treatment Characteristics
Surgical resection of the primary tumor was performed in 1485 patients (59%), and 189 patients (8%) received definitive radiochemotherapy.
Among patients treated with curative intent, perioperative strategies included perioperative chemotherapy in 146 (21%), neoadjuvant only in 407 (60%) and adjuvant only in 131 patients (19%).
In the palliative setting - either in patients with synchronous metastatic disease or at recurrence - systemic therapy was administered as follows: first-line therapy in 927 patients (37%), second-line therapy in 389 (15%), third-line therapy in 147 (6%), fourth-line therapy in 62 (2%), and further-line therapy in 27 patients (1%).
Comparison Between Deceased and Alive Patients
Compared with patients alive at last follow-up, deceased patients were older at diagnosis and more frequently diagnosed in earlier years. Patients who were alive had a higher body mass index at diagnosis, reported lower alcohol consumption, and had fewer comorbidities.
From a tumor perspective, adenocarcinoma histology was more common among surviving patients, and survivors were more frequently diagnosed with less advanced tumor stages.
Detailed patient and comorbidity characteristics are provided in Supplementary Table 1, and tumor and treatment characteristics in Supplementary Table 2.
Cause of Death
Among the 1863 deaths, 1592 deaths (85%) were cancer-related, 221 (12%) were non cancer-related, 43 (2%) of unknown cause, and 7 (<1%) due to suicide.
Postoperative death, defined as death within one month after tumor resection irrespective of the registered cause of death, occurred in 45 patients (4% of resected patients). Postoperative mortality differed across diagnosis periods, with the highest proportion observed in patients diagnosed between 1994 and 1999 (n = 19), and generally lower rates in more recent years (2000–2004: n = 7; 2005–2009: n = 10; 2010–2014: n = 4; 2015–2019: n = 1; 2020–2024: n = 4).
Detailed subcategories of non-cancer-related deaths are presented in Table 1.
Table 1.
Subcategories of Causes of Death Among 1863 Gastroesophageal Cancer Patients
| Characteristic | N = 1863* |
|---|---|
| Cause of death | |
| Cardiovascular disease (including myocardial infarction and stroke) | 91 (5%) |
| Endocrinological disease | 8 (0%) |
| Gastrointestinal disease | 6 (0%) |
| Hematological disease | 9 (0%) |
| Infectious disease | 30 (2%) |
| Kidney disease | 3 (0%) |
| Liver disease | 8 (0%) |
| Multiorgan failure | 3 (0%) |
| Neurological disease | 7 (0%) |
| Not known | 43 (2%) |
| Other | 26 (1%) |
| Progression of cancer (not specified) | 68 (4%) |
| Progression of gastroesophageal cancer | 1,412 (76%) |
| Progression of other cancer | 112 (6%) |
| Respiratory disease | 10 (1%) |
| Suicide | 7 (0%) |
| Trauma | 20 (1%) |
| Cause of death - category | |
| Cancer-related | 1,592 (85%) |
| Not cancer-related | 221 (12%) |
| Not known | 43 (2%) |
| Suicide | 7 (0%) |
| Postoperative death | 45 (4%) |
Notes: *n (%).
Association Between Cause of Death and Clinical Characteristics
Associations between cause of death and clinical characteristics are summarized in Tables 2 and 3.
Table 2.
Associations Between Patient Characteristics and Cause of Death
| Characteristic | Cancer-Related, N = 1,592* |
Not Cancer-Related, N = 221* |
Not Known, N = 43* |
Suicide, N = 7* |
p-Value** |
|---|---|---|---|---|---|
| Sex | 0.2 | ||||
| Female | 481 (30%) | 57 (26%) | 18 (42%) | 2 (29%) | |
| Male | 1,111 (70%) | 164 (74%) | 25 (58%) | 5 (71%) | |
| Age category | 0.001 | ||||
| <=45 | 118 (7%) | 9 (4%) | 5 (12%) | 0 (0%) | |
| ≥65 | 788 (49%) | 143 (65%) | 24 (56%) | 3 (43%) | |
| >45 <65 | 686 (43%) | 69 (31%) | 14 (33%) | 4 (57%) | |
| Year of diagnosis | <0.001 | ||||
| 1994–1999 | 188 (12%) | 53 (24%) | 2 (5%) | 2 (29%) | |
| 2000–2004 | 183 (11%) | 35 (16%) | 2 (5%) | 1 (14%) | |
| 2005–2009 | 383 (24%) | 47 (21%) | 7 (16%) | 1 (14%) | |
| 2010–2014 | 370 (23%) | 41 (19%) | 9 (21%) | 1 (14%) | |
| 2015–2019 | 307 (19%) | 29 (13%) | 9 (21%) | 2 (29%) | |
| 2020–2024 | 161 (10%) | 16 (7%) | 14 (33%) | 0 (0%) | |
| BMI | 0.4 | ||||
| Normal weight (18.5–24.9) | 616 (51%) | 80 (45%) | 14 (45%) | 4 (67%) | |
| Obesity (≥30.0) | 154 (13%) | 24 (13%) | 4 (13%) | 1 (17%) | |
| Overweight (25.0–29.9) | 345 (28%) | 66 (37%) | 9 (29%) | 1 (17%) | |
| Underweight (<18.5) | 96 (8%) | 8 (4%) | 4 (13%) | 0 (0%) | |
| (Missing) | 381 | 43 | 12 | 1 | |
| Alcohol consumption | 0.053 | ||||
| Alcohol abuse | 234 (17%) | 32 (16%) | 2 (5%) | 3 (43%) | |
| Moderate alcohol consumption | 560 (40%) | 89 (45%) | 13 (34%) | 3 (43%) | |
| No alcohol consumption | 607 (43%) | 76 (39%) | 23 (61%) | 1 (14%) | |
| (Missing) | 191 | 24 | 5 | 0 | |
| Nicotine abuse | 0.7 | ||||
| Abuse | 840 (59%) | 111 (56%) | 20 (51%) | 4 (57%) | |
| No | 578 (41%) | 87 (44%) | 19 (49%) | 3 (43%) | |
| (Missing) | 174 | 23 | 4 | 0 | |
| Number of comorbidities | 0.057 | ||||
| 0 | 125 (8%) | 13 (6%) | 6 (14%) | 0 (0%) | |
| 1–2 | 375 (24%) | 42 (19%) | 15 (35%) | 2 (29%) | |
| 3–5 | 583 (37%) | 78 (35%) | 13 (30%) | 2 (29%) | |
| 6–10 | 442 (28%) | 69 (31%) | 8 (19%) | 2 (29%) | |
| 11–15 | 55 (3%) | 13 (6%) | 1 (2%) | 1 (14%) | |
| >16 | 12 (1%) | 6 (3%) | 0 (0%) | 0 (0%) | |
| Cardiovascular | 876 (55%) | 144 (65%) | 22 (51%) | 2 (29%) | 0.014 |
| Gynecological | 148 (9%) | 21 (10%) | 3 (7%) | 2 (29%) | 0.3 |
| Kidney | 199 (13%) | 41 (19%) | 4 (9%) | 0 (0%) | 0.049 |
| Liver | 368 (23%) | 70 (32%) | 9 (21%) | 3 (43%) | 0.025 |
| Endocrine | 445 (28%) | 76 (34%) | 13 (30%) | 2 (29%) | 0.3 |
| Neurological | 187 (12%) | 21 (10%) | 7 (16%) | 3 (43%) | 0.037 |
| ENT | 355 (22%) | 45 (20%) | 2 (5%) | 1 (14%) | 0.043 |
| Gastrointestinal | 691 (43%) | 101 (46%) | 7 (16%) | 3 (43%) | 0.004 |
| Musculosceletal | 445 (28%) | 53 (24%) | 17 (40%) | 2 (29%) | 0.2 |
| Infectious | 172 (11%) | 23 (10%) | 5 (12%) | 1 (14%) | >0.9 |
| Lung | 274 (17%) | 37 (17%) | 7 (16%) | 2 (29%) | 0.9 |
| Other malignancies | 241 (15%) | 29 (13%) | 8 (19%) | 1 (14%) | 0.8 |
| Rheumathological | 85 (5%) | 12 (5%) | 0 (0%) | 1 (14%) | 0.3 |
| Urological | 162 (10%) | 29 (13%) | 5 (12%) | 1 (14%) | 0.6 |
| Ophthalmological | 172 (11%) | 27 (12%) | 6 (14%) | 2 (29%) | 0.4 |
| Other | 252 (16%) | 42 (19%) | 7 (16%) | 3 (43%) | 0.2 |
| Psychiatric | 78 (5%) | 8 (4%) | 2 (5%) | 0 (0%) | 0.8 |
| Hematological | 93 (6%) | 17 (8%) | 2 (5%) | 0 (0%) | 0.6 |
| Pancreas | 32 (2%) | 6 (3%) | 1 (2%) | 0 (0%) | 0.9 |
| Genetic disorders | 13 (1%) | 2 (1%) | 0 (0%) | 0 (0%) | >0.9 |
| Dermatological | 79 (5%) | 14 (6%) | 2 (5%) | 1 (14%) | 0.6 |
| Spleen | 11 (1%) | 5 (2%) | 0 (0%) | 0 (0%) | 0.11 |
Notes: *n (%). ** Pearson’s Chi-squared test.
Table 3.
Associations Between Tumor and Treatment Characteristics and Cause of Death
| Characteristic | Cancer-Related, N = 1,592* |
Not Cancer-Related, N = 221* |
Not Known, N = 43* |
Suicide, N = 7* |
p-Value** |
|---|---|---|---|---|---|
| Histology | 0.5 | ||||
| Adeno | 1,217 (76%) | 174 (79%) | 35 (81%) | 4 (57%) | |
| SCC | 375 (24%) | 47 (21%) | 8 (19%) | 3 (43%) | |
| Tumor location | 0.012 | ||||
| Esophagus | 538 (34%) | 69 (31%) | 14 (33%) | 3 (43%) | |
| GEJ | 457 (29%) | 46 (21%) | 17 (40%) | 0 (0%) | |
| Stomach | 597 (38%) | 106 (48%) | 12 (28%) | 4 (57%) | |
| Stage | <0.001 | ||||
| 1 | 127 (8%) | 75 (34%) | 6 (14%) | 1 (14%) | |
| 2 | 244 (15%) | 54 (24%) | 9 (21%) | 3 (43%) | |
| 2–3 | 120 (8%) | 10 (5%) | 2 (5%) | 0 (0%) | |
| 3 | 405 (25%) | 45 (20%) | 10 (23%) | 1 (14%) | |
| 4 | 696 (44%) | 37 (17%) | 16 (37%) | 2 (29%) | |
| Signet ring cells | 0.4 | ||||
| Negative | 1,005 (70%) | 140 (72%) | 29 (83%) | 4 (67%) | |
| Positive | 435 (30%) | 54 (28%) | 6 (17%) | 2 (33%) | |
| (Missing) | 152 | 27 | 8 | 1 | |
| MMR | 0.11 | ||||
| Deficient | 15 (6%) | 3 (17%) | 0 (0%) | 0 (NA%) | |
| Proficient | 229 (94%) | 15 (83%) | 18 (100%) | 0 (NA%) | |
| (Missing) | 1,348 | 203 | 25 | 7 | |
| PD-L1 | 0.4 | ||||
| Negative | 47 (21%) | 1 (6%) | 3 (18%) | 0 (NA%) | |
| Positive | 181 (79%) | 15 (94%) | 14 (82%) | 0 (NA%) | |
| (Missing) | 1,364 | 205 | 26 | 7 | |
| HER2 | 0.2 | ||||
| Negative | 473 (79%) | 41 (91%) | 19 (79%) | 2 (100%) | |
| Positive | 125 (21%) | 4 (9%) | 5 (21%) | 0 (0%) | |
| (Missing) | 994 | 176 | 19 | 5 | |
| Recurrence/Progression | <0.001 | ||||
| No recurrence/progression | 154 (10%) | 160 (72%) | 13 (30%) | 5 (71%) | |
| Recurrence/Progression | 1,438 (90%) | 61 (28%) | 30 (70%) | 2 (29%) | |
| Surgical resection | 840 (53%) | 162 (73%) | 20 (47%) | 4 (57%) | <0.001 |
| (Missing) | 2 | 0 | 0 | 0 | |
| Definitive RCHT | 128 (8%) | 19 (9%) | 4 (9%) | 1 (14%) | >0.9 |
| Perioperative treatment strategy | 0.3 | ||||
| Adjuvant | 76 (20%) | 9 (26%) | 2 (25%) | 0 (0%) | |
| Neoadjuvant | 229 (62%) | 22 (65%) | 5 (63%) | 0 (0%) | |
| Perioperative | 66 (18%) | 3 (9%) | 1 (13%) | 1 (100%) | |
| First line systemic therapy | 779 (49%) | 39 (18%) | 23 (53%) | 1 (14%) | <0.001 |
| Second line systemic therapy | 339 (21%) | 12 (5%) | 8 (19%) | 0 (0%) | <0.001 |
| Third line systemic therapy | 136 (9%) | 2 (1%) | 5 (12%) | 0 (0%) | <0.001 |
| Fourth line systemic therapy | 56 (4%) | 2 (1%) | 2 (5%) | 0 (0%) | 0.2 |
| Further line systemic therapy | 24 (2%) | 1 (0%) | 2 (5%) | 0 (0%) | 0.2 |
Notes: *n (%). **Pearson’s Chi-squared test.
Age was associated with cause of death, with elderly patients more frequently dying from non-cancer-related causes. Year of diagnosis was also associated, with a decreasing proportion of non-cancer-related deaths in more recent years.
Overall comorbidity burden was not associated with cause of death. However, specific comorbidity categories were associated with non-cancer-related mortality, including cardiovascular, kidney, liver, and gastrointestinal diseases. Neurological comorbidities appeared more frequent among patients who died by suicide, while ENT-related conditions were more commonly observed in patients with unknown cause of death.
Tumor location was associated with cause of death, with stomach cancer more frequently observed among patients with non-cancer-related death. Tumor stage showed a strong association, with advanced stages predominantly linked to cancer-related mortality. In stage 1, cancer-related death occurred in 127 patients, accounting for 61% of deceased patients and 29% of all patients in this stage. In locally advanced stages (2, 2–3, 3), 769 cancer-related deaths were observed, corresponding to 85% of deceased patients and 62% of all patients in this group. In stage 4, cancer-related death occurred in 696 patients, representing 93% of deceased patients and 83% of all patients in this stage.
Recurrence or progression was strongly associated with cancer-related death.
Surgical resection was associated with cause of death, with resected patients more likely to die from non-cancer-related causes. Type of perioperative treatment strategy was not associated with cause of death, whereas receipt of palliative systemic therapy showed an association.
Survival and Cumulative Risk Assessment
Median OS for the entire cohort was 22.2 months (95% CI 21.1–23.6) (Figure 1A). Median OS of cancer-related death was 14.2 months (95% CI 13.3–15.2), of non-cancer-related cause of death 48.3 months (95% CI 32.6–68.4), unknown cause of death 22.0 months (95% CI 14.3–34.3) and suicide 43.3 months (95% CI 12.7-NA).
Figure 1.
Survival and cumulative incidence of cause-specific mortality. (A) Kaplan-Meier curve for overall survival. (B) Cumulative incidence functions for cancer-related (85% of deaths), non-cancer-related mortality (12% of deaths), suicide, and unknown cause of death, accounting for competing risks.
The cumulative incidence functions for cancer-related and non-cancer-related death are shown in Figure 1B. In the multivariable Fine–Gray model for cancer-related death, tumor stage was the strongest independent predictor. Compared with stage 1 disease, the subdistribution hazard increased progressively with advancing stage, reaching an sHR of 7.42 (95% CI 6.15–8.96; p < 0.001) for stage 4 disease. Intermediate stages were similarly associated with elevated risks compared with stage 1 disease, stage 2 was associated with an sHR of 2.34 (95% CI 1.91–2.86; p < 0.001), stage 2–3 with an sHR of 2.86 (95% CI 2.27–3.62; p < 0.001) and stage 3 with an sHR of 3.35 (95% CI 2.77–4.05; p < 0.001).
A higher comorbidity burden was independently associated with increased cancer-related mortality, particularly in patients with 6–10 comorbidities (sHR 1.34; 95% CI 1.08–1.66; p = 0.007) and 11–15 comorbidities (sHR 1.52; 95% CI 1.09–2.11; p = 0.013), compared with patients without comorbidities. Patients diagnosed in the most recent time period (2020–2024) had a lower risk of cancer-related death compared with those diagnosed between 1994 and 1999 (sHR 0.60; 95% CI 0.47–0.76; p < 0.001). Age ≥65 years showed a borderline association with increased cancer-related mortality (sHR 1.21; 95% CI 0.99–1.48; p = 0.059), whereas sex (sHR 1.02; 95% CI 0.91–1.15; p = 0.74), histological subtype (sHR 1.03; 95% CI 0.86–1.23; p = 0.74), and tumor location (stomach: sHR 0.96; 95% CI 0.84–1.09; p = 0.50; esophagus: sHR 1.07; 95% CI 0.90–1.26; p = 0.44) were not independently associated with cancer-related death.
In contrast, the Fine–Gray model for non-cancer-related death demonstrated a different risk pattern. Age ≥65 years was strongly associated with an increased risk of non-cancer-related death (sHR 2.21; 95% CI 1.12–4.35; p = 0.021). Male sex was associated with a lower risk compared with female sex (sHR 0.65; 95% CI 0.48–0.88; p = 0.005). More recent year of diagnosis was consistently associated with reduced non-cancer-related mortality, with the lowest risk observed in patients diagnosed between 2020 and 2024 (sHR 0.43; 95% CI 0.25–0.74; p = 0.002). Tumor location in the stomach was independently associated with an increased risk of non-cancer-related death compared with gastroesophageal junction tumors (sHR 1.56; 95% CI 1.08–2.23; p = 0.017). Conversely, advancing tumor stage was associated with a lower subdistribution hazard of non-cancer-related death, with stage 4 disease showing an sHR of 0.27 (95% CI 0.18–0.40; p < 0.001) compared with stage 1. Comorbidity burden was not independently associated with non-cancer-related mortality, although a trend toward increased risk was observed in patients with more than 16 comorbidities (sHR 2.56; 95% CI 0.97–6.80; p = 0.059). Histological subtype was not associated with non-cancer-related death.
Forest plots of the adjusted subdistribution hazard ratios are presented in Figure 2 for cancer-related mortality and in Figure 3 for non-cancer-related mortality (reference categories: female sex, age <65 years, 0 comorbidities, 1994–1999 diagnosis period, squamous histology, GEJ tumor location, stage 1).
Figure 2.
Forest plot of multivariable Fine-Gray competing risk regression analysis for cancer-related mortality. Adjusted subdistribution hazard ratios (sHR) with 95% confidence intervals.
Figure 3.
Forest plot of multivariable Fine-Gray competing risk regression analysis for non-cancer-related mortality.
Overall, the competing risk analyses demonstrate that tumor stage predominantly determines cancer-related mortality, whereas age is the principal determinant of non-cancer-related mortality.
Discussion
Cancer-related death accounted for the vast majority of deaths (85%) in our cohort, underscoring the aggressive nature and persistently poor prognosis of gastroesophageal malignancies despite advances in multimodal therapy.4,5 Tumor stage emerged as the strongest determinant of cancer-related mortality in the competing risk analysis, demonstrating a pronounced stepwise increase in risk from stage 2 to stage 4 disease. Notably, 93% of deaths among patients with stage 4 disease were cancer-related, emphasizing the dominant impact of tumor burden in the metastatic setting.7 Consequently, only 11% of patients with stage 4 disease were alive at last follow-up.
Importantly, cancer-related mortality was not restricted to advanced stages. Although more than half of stage 1 patients were alive at data cutoff, 61% of deaths in this subgroup were cancer-related. These findings highlight that even early-stage gastroesophageal cancer carries a substantial risk of tumor-related mortality, which is in line with data from the US.12,14 Recurrence, which was strongly associated with death in our cohort, likely contributes significantly to this observation, indicating that initial resectability does not fully eliminate long-term oncologic risk. The progressive increase in cancer-related mortality across stages reinforces the biological aggressiveness of these tumors and supports ongoing efforts to optimize multimodal treatment strategies. In this context, timely integration of palliative care may be beneficial and should be considered as part of routine management, with the aim of improving symptom management, psychosocial support, and overall quality of life. Evidence from randomized trials and guideline recommendations supports timely palliative care involvement in patients with advanced gastrointestinal and other solid tumors, demonstrating improvements in quality of life and other patient-centered outcomes.15,16 However, the impact of palliative care on cause-specific mortality in our analysis could not be assessed, as its utilization was not systematically recorded.
In our study, a higher comorbidity burden was independently associated with increased cancer-related mortality. Although comorbidity burden was not associated with cause of death in the univariate analysis, it emerged as an independent predictor of cancer-related mortality after multivariable adjustment. This finding may, at least in part, be explained by the fact that patients with multiple comorbidities are often less likely to receive the full therapeutic approach due to contraindications, reduced physiological reserve, or limited treatment tolerance.17 In contrast, comorbidity burden was not independently associated with non-cancer-related mortality. This may reflect the stronger influence of age on competing non-cancer causes of death, whereas comorbidities may primarily affect cancer-related mortality by limiting treatment intensity, as stated above.
While tumor progression was the predominant cause of death, non-cancer-related mortality represented a clinically relevant proportion (12%) of cases in our cohort with cardiovascular disease as the biggest subgroup, which is in line with other cohorts.13 Age was the principal determinant of non-cancer-related death, reflecting increased vulnerability in older patients with competing health risks, independent of comorbidities.18 In contrast, advanced tumor stage was inversely associated with non-cancer-related mortality, a pattern consistent with the competing risk phenomenon whereby patients with high tumor burden are more likely to succumb to cancer before other causes of death become clinically relevant. This theory is supported by the fact that the median OS for non–cancer-related causes of death is three times longer than that for cancer-related causes. These findings underscore the importance of considering competing health risks when caring for patients with gastroesophageal cancer, particularly elderly patients and limited disease. Thus, competing health risks may be relevant when individualizing treatment planning and follow-up.
Furthermore, tumor location in the stomach was independently associated with a higher risk of non–cancer-related death. This likely reflects the greater feasibility of curative resection in gastric cancer, as gastrectomy is technically less demanding than esophagectomy or complex GEJ resections, particularly in the earlier years of our 30-year cohort.19 As a result, stomach cancer patients might have more easily achieved long-term tumor control and ultimately died from competing non-cancer causes rather than tumor progression. Concerning advances in surgical resection, postoperative mortality varied across diagnosis periods, with the highest proportion observed in the earliest years of the study. Although this finding may reflect improvements in perioperative management, including surgical techniques, anesthesia, and postoperative care, the present study was not designed to specifically evaluate temporal changes in perioperative outcomes. Therefore, no causal conclusions can be drawn.
Although only seven suicides were observed in this cohort, this finding still poses a serious concern. However, given the very small number of events, the following observations should be interpreted with caution and considered hypothesis-generating. Whether timely psycho-oncological or palliative interventions could have mitigated this risk cannot be determined from the present retrospective analysis. A recent analysis showed that palliative care may help reduce the excess suicide risk observed in severely ill cancer patients.20 However, suicide accounted for less than 1% of deaths in our analysis, which is less than the reported 1.5% in the Austrian population in 202421 and a significantly lower rate compared to other cohorts.22,23 Furthermore, it occurred predominantly in patients with initially resectable disease, which is also not in line with current literature.10 This low rate may be explained by suicides being misclassified as trauma or unknown causes, a bias that has been described in national cause-of-death statistics.24 Although patient charts were reviewed when the cause of death was unclear, some misclassification may still have occurred, which must be considered as a potential limitation of this retrospective analysis. Given the retrospective design and reliance on recorded data, such misclassification cannot be excluded and should be considered a relevant limitation of this analysis. Another possible explanation is the comprehensive multidisciplinary care provided at our tertiary referral center, including readily available psycho-oncological support and early integration of specialist palliative care. In addition, Austria’s universal healthcare system facilitates timely access to specialized oncologic treatment and supportive services, which may reduce psychosocial distress and provide reassurance early after diagnosis. Although our study cannot establish a causal relationship, these factors may have contributed to the comparatively low observed suicide rate.
In addition, the retrospective design of this study carries the risk of residual confounding and incomplete data capture. As expected in a retrospective study spanning three decades, some variables contained missing data. The extent of missingness for each variable is reported in the respective tables, while comprehensive baseline demographic, clinicopathological, and treatment-related data were available for the majority of patients, allowing robust analyses.
Treatment-related mortality could not be analyzed as a separate category because causes of death were based primarily on ICD coding from the national death registry, which does not distinguish treatment-related toxicity from cancer-related mortality. Although postoperative mortality was reported separately using a predefined time-based definition, attribution of other treatment-related deaths remains challenging and may have resulted in misclassification.
Further limitations consider the long study period, as diagnostic methods and treatment standards evolved considerably over time, potentially affecting temporal comparisons. Moreover, patients diagnosed earlier benefited from longer observation periods, potentially introducing bias in cause-specific mortality estimates and contributing to a higher observed incidence of deaths in these earlier cohorts. Although more recent diagnosis was associated with lower cancer- and non-cancer-related mortality, this finding should be interpreted cautiously, as it may reflect a combination of advances in systemic therapies (including perioperative chemotherapy, targeted therapies, and immune checkpoint inhibitors) together with improvements in surgical techniques, radiotherapy, supportive care, and the shorter follow-up available for more recently diagnosed patients. However, the present study was not designed to evaluate the contribution of individual treatment modalities, and no causal conclusions can be drawn.
As this is a single-center study from a high-volume tertiary referral center, referral patterns and specialized multidisciplinary care may have influenced patient selection, treatment, and outcomes, thereby limiting the generalizability of our findings to other healthcare settings. Nevertheless, the predominance of cancer-related mortality observed in our cohort is consistent with reports from registry-based studies in the United States and Asia, despite differences in patient populations, treatment availability, and healthcare systems.12,14,25
Despite these limitations, the study has several notable strengths. It represents one of the largest and most detailed single-center analyses of cause-specific mortality in gastroesophageal cancer conducted at a European high-volume referral center. The cohort captures real-world clinical practice across all disease stages and treatment approaches over three decades, enhancing its external validity. In contrast to registry-based studies, this dataset offers substantial clinical detail, including tumor characteristics, treatment types, recurrence patterns, comorbidities, and lifestyle factors, enabling more nuanced analyses.
Conclusion
In summary, cancer-related mortality was strongly associated with advanced tumor stage, emphasizing the poor prognosis and high disease burden in this patient population. Notably, the substantial proportion of cancer-related deaths among patients with resectable disease indicates that long-term oncologic risk remains substantial despite curative-intent treatment. These findings highlight the need not only for improved therapeutic and surveillance strategies, but also for timely integration of palliative care to optimize symptom management and quality of life. Furthermore, the contribution of non-cancer-related mortality, particularly among older individuals, emphasizes the importance of comprehensive, multidisciplinary care addressing comorbidities and competing health risks. Given the retrospective, single-center design of this exploratory study, prospective, multicenter studies are needed to validate these findings and further refine strategies for optimizing survivorship in gastroesophageal cancer.
Funding Statement
This analysis was conducted without external funding.
Artificial Intelligence Statement
Artificial intelligence–based tools (ChatGPT (GPT-5.2), Perplexity (GPT-5.2), and Grammarly (1.175.0.0, WebUI 2.16.22)) were used exclusively for language editing and stylistic improvement of the manuscript. These tools were not used for data analysis, interpretation of results, or generation of intellectual content. All suggestions were carefully reviewed, revised where necessary, and approved by the authors, who take full responsibility for the content of the manuscript.
Abbreviations
ICD, International Classification of Diseases; OS, overall survival; CI, confidential interval; GEJ, gastroesophageal junction; HER2, human epidermal growth factor receptor 2; PD-L1, Programmed cell death receptor ligand 1; MMR, mismatch repair; ENT, Ears-nose-throat.
Data Sharing Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request after obtaining of all relevant regulatory approvals and signature of data sharing agreements with the Medical University of Vienna.
Ethics Approval - Institutional Review Board Statement
All procedures were in accordance with the ethical standards of the responsible committee on human experimentation and with the Helsinki Declaration of 1964 and later versions. The study was approved by the ethics committee of the Medical University of Vienna (reference number: 1226/2026). The risk of publication of sensitive patient data was minimized by restrictions of data accession and pseudonymized statistical analysis.
Informed Consent Statement
Due to the retrospective design no separate informed consent was necessary in the scope of this study. This is in line with local ethical standards and was approved by the ethics committee of the Medical University of Vienna (reference number: 1226/2026).
Author Contributions
All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.
Disclosure
HCP has received travel support from Eli Lilly, MSD, Novartis, Pfizer, PharmaMar, Pierre Fabre and Roche and received lecture honoraria from Eli Lilly.
JMB received travel support from Amgen via institutional nomination and honoraria for lectures, consultation or advisory board participation from MSD.
ESB received travel support from Jazz Pharmaceuticals and Daiichi Sankyo via institutional nomination and honoraria for lectures, consultation or advisory board participation from Merck and Servier.
MP has received honoraria for lectures, consultation or advisory board participation from the following for-profit companies: Bayer, Bristol-Myers Squibb, Novartis, Gerson Lehrman Group (GLG), CMC Contrast, GlaxoSmithKline, Mundipharma, Roche, BMJ Journals, MedMedia, Astra Zeneca, AbbVie, Lilly, Medahead, Daiichi Sankyo, Sanofi, Merck Sharp & Dome, Tocagen, Adastra, Gan & Lee Pharmaceuticals, Janssen, Servier, Miltenyi, Böhringer-Ingelheim, Telix, Medscape, OncLive, Medac, Nerviano Medical Sciences, ITM Oncologics GmbH, AdAcAp, Crinetics. MP reports Grants or contracts from Daiichi Sankyo, Novartis, Servier. Meeting/travel support from Medsir, Daiichi Sankyo, Servier, BMS; Leadership or fiduciary roles from EORTC, EANO, NMN, outside the submitted work.
AI-M: Participation in advisory boards organized by MSD, Servier, Daiichi Sankyo, BMS, Amgen and Astellas, lecture honoraria from Eli Lilly, Servier, BMS, MSD, Astellas, Astra Zeneca, BeiGene, Amgen, Merck and Daiichi Sankyo, consulting for Astellas, MSD, Amgen, Astra Zeneca, BeiGene, and Roche, travel support (all via institutional nomination) from BMS, Roche, Eli Lilly, Daiichi Sankyo, Jazz and BeiGene.
FA reports Honoraria from OPG Austrian Palliative Society, outside the submitted work.
All other authors declare no conflict of interest.
References
- 1.Bray F, Laversanne M, Sung H, et al. Global cancer statistics 2022: GLOBOCAN estimates of incidence and mortality worldwide for 36 cancers in 185 countries. Ca a Cancer J Clin. 2024;74(3):229–16. doi: 10.3322/caac.21834 [DOI] [PubMed] [Google Scholar]
- 2.National Cancer Institute Bethesda MD. SEER Cancer Stat Facts: stomach Cancer. 2026, Available from: https://seer.cancer.gov/statfacts/html/stomach.html. Accessed Nov 25, 2019.
- 3.National Cancer Institute Bethesda MD. SEER cancer stat facts: esophageal cancer. 2019, Available from: https://seer.cancer.gov/statfacts/html/esoph.html. Accessed Feb 26, 2026.
- 4.Lordick FCML, Castelo-Branco L, Pentheroudakis G, Sessa C, Smyth E. ESMO gastric cancer living guideline, Available from: https://www.esmo.org/living-guidelines/esmo-gastric-cancer-living-guideline. Accessed Apr 7, 2025.
- 5.Obermannová R, Alsina M, Cervantes A, et al. Oesophageal cancer: ESMO clinical practice guideline for diagnosis, treatment and follow-up. Ann Oncol. 2022;33(10):992–1004. doi: 10.1016/j.annonc.2022.07.003 [DOI] [PubMed] [Google Scholar]
- 6.Boire A, Burke K, Cox TR, et al. Why do patients with cancer die? Nat Rev Cancer. 2024;24(8):578–589. doi: 10.1038/s41568-024-00708-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Mani K, Deng D, Lin C, Wang M, Hsu ML, Zaorsky NG. Causes of death among people living with metastatic cancer. Nat Commun. 2024;15(1):1519. doi: 10.1038/s41467-024-45307-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Sturgeon KM, Deng L, Bluethmann SM, et al. A population-based study of cardiovascular disease mortality risk in US cancer patients. Eur Heart J. 2019;40(48):3889–3897. doi: 10.1093/eurheartj/ehz766 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Zaorsky NG, Churilla TM, Egleston BL, et al. Causes of death among cancer patients. Ann Oncol. 2017;28(2):400–407. doi: 10.1093/annonc/mdw604 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Sugawara A, Kunieda E. Suicide in patients with gastric cancer: a population-based study. Jpn J Clin Oncol. 2016;46(9):850–855. doi: 10.1093/jjco/hyw075 [DOI] [PubMed] [Google Scholar]
- 11.Wang Y, Yang X, Song M, Li Y, Jiao D, Zhou X. Suicide risk and mortality among patients with cancers of the digestive system: a systematic review and meta-analysis. Front Oncol. 2026;16:55968. doi: 10.3389/fonc.2026.1655968 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Lou T, Hu X, Lu N, Zhang T. Causes of death following gastric cancer diagnosis: a population-based analysis. Med Sci Monit. 2023;29:e939848. doi: 10.12659/msm.939848 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Xie SH, Chen H, Lagergren J. Causes of death in patients diagnosed with gastric adenocarcinoma in Sweden, 1970–2014: a population-based study. Cancer Sci. 2020;111(7):2451–2459. doi: 10.1111/cas.14441 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Liao J, Xu J, Huang S, et al. Cause of death among gastric cancer survivors in the United States from 2000 to 2020. Medicine. 2024;103(8):e37219. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Bojesson A, Brun E, Eberhard J, Segerlantz M. Quality of life for patients with advanced gastrointestinal cancer randomised to early specialised home-based palliative care: the ALLAN trial. Br J Cancer. 2024;131(4):729–736. doi: 10.1038/s41416-024-02764-x [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Petrillo LA, Jones KF, El-Jawahri A, Sanders J, Greer JA, Temel JS. Why and How to Integrate Early Palliative Care Into Cutting-Edge Personalized Cancer Care. Am Soc Clin Oncol Educat Book. 2024;44(3):e100038. doi: 10.1200/EDBK_100038 [DOI] [PubMed] [Google Scholar]
- 17.Søgaard M, Thomsen RW, Bossen KS, Sørensen HT, Nørgaard M. The impact of comorbidity on cancer survival: a review. Clin Epidemiol. 2013;5(Suppl 1):3–29. doi: 10.2147/clep.S47150 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Lin Y, Li H, Wu H, et al. Age-related disparities in pan-cancer mortality and causes of death: analysis of surveillance, epidemiology, and end results (SEER) data. J Cancer. 2024;15(6):1613–1623. doi: 10.7150/jca.91758 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Petrelli F, Ghidini M, Barni S, et al. Prognostic role of primary tumor location in non-metastatic gastric cancer: a systematic review and meta-analysis of 50 studies. Ann Surg Oncol. 2017;24(9):2655–2668. doi: 10.1245/s10434-017-5832-4 [DOI] [PubMed] [Google Scholar]
- 20.Listabarth S, Sommer L, Trojer A, et al. Suicide rates among patients receiving palliative care—descriptive results of a national cohort study. J Clin Med. 2026;15(6):2149. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Statistik Austria. Todesursachen. Available from: https://www.statistik.at/statistiken/bevoelkerung-und-soziales/bevoelkerung/gestorbene/todesursachen. Accessed Mar 3, 2026.
- 22.Hu X, Ma J, Jemal A, et al. Suicide Risk Among Individuals Diagnosed With Cancer in the US, 2000–2016. JAMA Netw Open. 2023;6(1):e2251863–e2251863. doi: 10.1001/jamanetworkopen.2022.51863 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Anderson C, Park EM, Rosenstein DL, Nichols HB. Suicide rates among patients with cancers of the digestive system. Psychooncology. 2018;27(9):2274–2280. doi: 10.1002/pon.4827 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Schmeckenbecher J, Kapusta ND, Krausz RM, Emilian CA. Autopsy rates and the misclassification of suicide and accident deaths. Eur J Epidemiol. 2024;39(10):1109–1126. doi: 10.1007/s10654-024-01142-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Hong J, Oh MJ, Kim B, et al. All-Cause and Cause-Specific Mortality by SEER Stage in Gastric Cancer: a Nationwide Population-Based Cohort Study. J Clin Med. 2026;15(9):3484. doi: 10.3390/jcm15093484 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data that support the findings of this study are available from the corresponding author upon reasonable request after obtaining of all relevant regulatory approvals and signature of data sharing agreements with the Medical University of Vienna.