Suicide is the intentional act of ending one's life, a deeply tragic event that affects not only individuals and families but also entire communities and nations, creating a significant burden on public health resources. Suicide is the tenth leading cause of death among people of all ages in the United States (Fazel et al., 2020). According to estimates from the World Health Organization (WHO) and the Global Burden of Disease (GBD) study, approximately 800,000 people die by suicide each year (Heinrich et al., 2022; GBD 2017 Causes of Death Collaborators, 2018). Cancer is a significant global health challenge and one of the leading causes of death worldwide, claiming over 10 million lives annually (Zaimy et al., 2017). In the United States, cancer ranks as the second leading cause of death across all age groups (Siegel et al., 2024). Projections estimate that by 2040, the number of new cancer cases will rise to 28.4 million, imposing a significant economic, healthcare, and financial burden on society (Sung et al., 2021). Increasing evidence indicates that cancer patients have a higher risk of suicide compared to the general population (O'Rourke, 2022; Misono et al., 2008; Choi and Park, 2020). Cancer patients have an increased risk of suicide, primarily driven by psychological distress, treatment-related side effects, a decline in quality of life, and severe, uncontrolled cancer-related pain (Heinrich et al., 2022; Zaorsky et al., 2019; Das, 2019). The combination of physical and emotional suffering can be overwhelming, further intensifying feelings of helplessness and despair, leading to the development of depression and anxiety, and ultimately increasing the risk of suicide in cancer patients (Hofmann et al., 2023; Ribeiro et al., 2018; Henson et al., 2019). Therefore, identifying which subgroups of cancer patients are at a higher risk of suicide could be beneficial.

Pancreatic cancer is one of the most aggressive malignancies, characterized by difficult early diagnosis and poor prognosis, with a 5-year survival rate of only 13%, and is projected become the third-leading cause of cancer-related death in 2024 (Siegel et al., 2024). Pancreatic cancer patients are more vulnerable to severe mental health disorders, especially depression and anxiety, compared to those with other solid tumors (Dengsø et al., 2020; Holland et al., 1986). It is noteworthy that pancreatic cancer has been reported to have the highest suicide rate within the 1st year after diagnosis compared to other cancers (Saad et al., 2019). However, there is currently limited research on suicide among pancreatic cancer patients, and reliable long-term predictive models for assessing individual suicide risk are lacking. Machine learning offers more accurate and reliable predictions and assessments of suicide risk compared to traditional methods (Lejeune et al., 2022). A robust predictive model can accurately assess the suicide risk for individual patients, thereby enabling clinicians to efficiently identify those at high risk and expedite their referral to mental health professionals for timely intervention, ultimately contributing to a reduction in suicide rates. In a nomogram, complex statistical prediction models are streamlined into a single numerical estimate, which accurately conveys the probability of an event, taking into account the unique conditions of each individual (Jeong et al., 2020; Balachandran et al., 2015). This approach is highly useful for practical application in clinical settings to predict the likelihood of suicide in pancreatic cancer patients. In this study, we analyzed the Surveillance, Epidemiology, and End Results (SEER) database to identify independent risk factors for suicide in pancreatic cancer patients. Additionally, we developed and validated a nomogram to predict the suicide risk in pancreatic cancer patients.

Pancreatic cancer is an extremely aggressive malignancy of the digestive system, with an estimated 517,500 deaths worldwide each year (Siegel et al., 2024). Additionally, the study reveals that the suicide risk in pancreatic cancer patients is more than five times higher than in the general population, with a standardized mortality ratio (SMR) of 6.43 (Anderson et al., 2018; Ma et al., 2022). Nevertheless, predictive models for assessing suicide risk in pancreatic cancer patients remain lacking. Therefore, it is necessary to develop a new suicide risk assessment model for a specific population group (Fazel et al., 2019). The value of machine learning is especially evident when attempting to analyze datasets that are too large or complex for a human to handle (Greener et al., 2022). This study leveraged pancreatic cancer patient data from the SEER database, employing LASSO regression and Cox proportional hazards regression for variable selection to prevent overfitting, and subsequently developed a suicide risk prediction model. The nomogram calculates cumulative scores based on relevant patient variables to predict suicide risk in pancreatic cancer patients. Implementing this tool in clinical practice will enable a more precise approach to patient treatment plans, allowing physicians to identify suicide risk more quickly and intervene at the optimal time. According to our knowledge, our research is the first multicenter retrospective study to develop a nomogram model for predicting suicide risk in pancreatic cancer patients.

Our study indicates that the risk of suicide peaks at the initial cancer diagnosis but decreases over time (Figure 2). This suggests that the psychological and emotional stress linked to the diagnosis gradually decreases as time progresses. Male gender is suggested to be a risk factor for suicide in pancreatic cancer patients, according to our findings, which is consistent with previous studies (Zaorsky et al., 2019; Chen et al., 2021; Ravaioli et al., 2020). Men are more prone than women to impulsive, aggressive behavior and alcohol abuse, which can unintentionally lead them to focus more on immediate negative emotions, thereby intensifying psychological distress and increasing the risk of suicide (Gauthier et al., 2019; Spoletini et al., 2011). Additionally, for some men with poor prognoses or late-stage cancer, particularly those with a “Hemingway-type” personality (marked by independence, a strong emphasis on achievement, and stubbornness), suicide may be considered a rational and dignified choice when life expectancy is limited (Hietanen and Lönnqvist, 1991). Pancreatic cancer patients aged 65 and older have a higher tendency toward suicide compared to those under 65. Previous studies have demonstrated that the suicide rate among elderly cancer patients is markedly higher than in the general population (Mendes et al., 2021; Dormer et al., 2008). The elevated suicide risk in elderly patients may be linked to physical comorbidities, sleep disturbances, pain issues, cognitive impairments, and mental health conditions such as depression (Henson et al., 2019; Du et al., 2020). Furthermore, elderly cancer patients are more prone to experiencing depression, helplessness, and low self-esteem, which can lead to a decline in life satisfaction and, consequently, an increased risk of suicidal tendencies (Wu and Bond, 2006).

Pancreatic cancer is more prevalent in older adults, with the risk increasing with age, particularly in those aged 65 and above, which may be another reason for the elevated risk of suicide in this population. Individuals who are unmarried, separated, divorced, or widowed exhibit a higher propensity for suicide in comparison to their married counterparts. Prior research indicates that marriage mitigates the risk of suicide, presumably due to the emotional and social support it affords, which serves to alleviate social isolation (Osazuwa-Peters et al., 2018; Fukuchi et al., 2013). Our study shows that surgical intervention and chemotherapy serve as protective factors against suicide in individuals diagnosed with pancreatic cancer, with HRs of 0.384 (surgery, 95% CI: 0.251–0.647, P < 0.001) and 0.553 (chemotherapy, 95% CI: 0.244–0.604, P < 0.001), respectively. Similar findings have also been reported in several other studies (Ma et al., 2022; Shi et al., 2022). This observation may be attributed to continuous improvements in pancreatic cancer healthcare, particularly advancements in surgical techniques and the development of new chemotherapy agents and treatment strategies (Strobel et al., 2019; Neoptolemos et al., 2018). These advancements have collectively contributed to prolonging patient survival and mitigating suffering, thereby potentially decreasing the incidence of suicide among this patient population.

These findings underscore the importance of developing personalized support and prevention strategies based on cancer prognosis. Although palliative care for terminal cancer patients has gained attention, suicide risk has remained high over the past 25 years, indicating the need for greater focus on psychosocial support, pain management, and symptom control for cancer patients (Kumar et al., 2017). The transition to a biopsychosocial medical model underscores the critical importance of evaluating and managing suicide risk, as well as implementing suicide prevention strategies and psychological interventions for cancer patients (Fazel et al., 2020; Sullivan et al., 2018; Teno et al., 2013). Specifically, individuals diagnosed with pancreatic cancer generally encounter a notably poor prognosis. It is imperative to actively enhance treatment protocols and quality of life for these patients by offering improved medical care, comprehensive support systems, and specialized psychological counseling. Such measures are essential to mitigate the elevated risk of suicide among this population. Based on our nomogram, the treating oncologist or oncology nurse in routine practice enters five readily obtainable variables (age, sex, marital status, surgery, and chemotherapy) at the patient's first visit. The nomogram immediately produces an individual suicide-risk score that categorizes patients as high-risk or low-risk. High-risk patients are referred to psycho-oncology services after diagnosis, and the score is re-evaluated at key treatment milestones, such as after surgery or at the end of chemotherapy, allowing supportive care to be adjusted promptly. This efficient process facilitates critical early risk detection and targeted psychological support within the narrow therapeutic window of pancreatic cancer care.

However, we recognize several limitations in our study. Firstly, the data used in this study comes from a retrospective analysis of the SEER database, which could potentially introduce selection bias (Park et al., 2012). Although recent geopolitical developments have altered the access conditions for the SEER database, it is important to emphasize that our study utilized a complete dataset obtained prior to April 2025, thereby ensuring the integrity and validity of the data. Secondly, the constraints of the available data within the SEER database pose challenges in acquiring additional pertinent information, such as mental health issues (including depression and anxiety), comorbidities, and substance abuse. Lastly, it is imperative to recognize that our predictive model has been subjected solely to internal validation. External validation using independent cohorts is essential for evaluating its generalizability, as the absence of such validation may introduce inherent bias into our findings.

Suicide constitutes a significant global public health concern, with patients diagnosed with pancreatic cancer exhibiting a heightened risk of suicide relative to the general population. Nevertheless, a considerable number of surgeons may not possess the requisite expertise to evaluate suicide risk subsequent to a cancer diagnosis, thereby leading to potential lapses in suicide prevention efforts. In this study, we employed a combination LASSO and Cox proportional hazards regression to identify five risk factors. Among these factors, being aged 65 years or older, male, unmarried, not having undergone surgical intervention, and not receiving chemotherapy are identified as high-risk factors for suicide in patients with pancreatic cancer. Subsequently, we developed a nomogram to predict suicide risk among pancreatic cancer patients. The nomogram exhibited robust performance with respect to discrimination, calibration, clinical applicability and generalization. This model enables clinicians to quickly identify high-risk individuals, allowing for timely and appropriate preventive measures to reduce unnecessary deaths and alleviate the burden of pancreatic cancer.