Suicide , a feared and often preventable consequence of untreated depression, is being recognized increasingly among cancer patients1-9; yet, the magnitude of this problem in patients with pancreatic cancer is unknown. This topic is of considerable importance, because patients with adenocarcinoma of the pancreas have a 33%-76% prevalence of depression,10, 11 and pancreatic cancer is the fourth leading cause of cancer death.12 A strong association between depression and pancreatic cancer has been established since the 1930s; indeed, depression in patients with pancreatic cancer has been shown to have a stronger association than for other advanced intra-abdominal malignancies, such as gastric and colon cancer.13, 14 A recent report by the Institute of Medicine15 identified the need for greater efforts to decrease psychosocial distress among cancer patients. Patients with pancreatic cancer can be considered at increased risk of depression and suicide, and should be targeted for intervention, but data on which to base such initiatives are lacking. The aim of our study was to identify the suicide rate among patients with pancreatic cancer, using population-based data to identify patient, disease, and therapy characteristics associated with commission of suicide.
MATERIALS AND METHODS
This study was approved and exempted from Institutional Review Board review at the University of South Florida, which participates in research activities at the H. Lee Moffitt Cancer Center and Research Institute.
Patients with pancreatic adenocarcinoma were identified from the Surveillance, Epidemiology and End Results (SEER) program of the National Cancer Institute. The SEER program is a network of population-based, incident tumor registries capturing about 26% of the United States population. We used data available for public use from 1995-2005 for our analysis in an attempt to minimize temporal bias.
Study Population and Study Variables
We included all patients classified as ICD-3 code 8140/3, adenocarcinoma not otherwise specified (NOS), which refers to pancreatic adenocarcinoma and excludes other malignant neoplasms such as neuroendocrine malignancies and premalignant neoplasms. We used age at diagnosis, sex, SEER summary stage grouping (localized, regional, distant, and unstaged), survival, operative intervention or not (and, if not, if recommended or not), radiation therapy, and marital status at diagnosis as covariates. Marital status was coded as married, single, divorced, separated, widowed, or unknown. For this analysis, we combined patients into 2 groups, married or single, with the single group categorized as divorced, separated, and widowed. Patients with missing data were excluded from the results, although a sensitivity analysis was performed to check the effect on the covariates. Data on concomitant preexisting illnesses, such as depression or other psychiatric conditions, were not available.
Patients were considered to have committed suicide if the cause of death variable was coded as “suicide and self-inflicted injury (50220).” Patients with other cause of death values, including “accidents and adverse effects (50210),” “homicide and legal intervention (50230),” and “other cause of death (50300),” were not classified as suicides.
Surgical intervention is coded in the SEER database as a separate variable that indicates if an operation was performed and if it was recommended or not. The actual surgical procedure directed at the primary site is coded as a separate variable. No record of chemotherapy appears in this database.
We assigned a nonzero value (0.5 months) to patients who did not survive a full month after diagnosis, because the SEER database records survival in months. This is consistent with epidemiologic convention.16
The Student t test and the chi-square test with Fisher exact modification were used for patients who committed suicide. Standardized mortality was calculated via the ratio of obtained mortality with expected mortality from a standardized US population in 2005.17 We used the suicide rates available for ages 65-74, the age distribution of 69% of our study population. Sex-specific rates were available and therefore used for analysis. Confidence intervals were calculated with Fisher mid-P exact confidence intervals from web-based software available at http://www.sph.emory.edu/∼cdckms/exact-midP-SMR.html. Logistic regression was used to determine the association of pancreatic adenocarcinoma and patients who committed suicide. Survival characteristics were assessed using Kaplan-Meier analysis and the log-rank test. Multivariate modeling was performed using the Cox proportional hazards model. Cause-specific mortality—with suicide counted as a failure—was performed in addition to all-cause mortality. Given the small number of events, we did not use the competing risks model. Multivariate modeling was performed for both sexes, but we present data only for the males given the high univariate odds, which made multivariate models unstable in the presence of such a strong covariate. Post-model estimates were tested for validity of the model, and the Hosmer-Lemeshow test was used to assess goodness-of-fit. A P-value of .05 or less was considered statistically significant. All analyses were performed using SEER Stat version 6.4.4 (2008) and Stata/SE Version 9.0 (StataCorp LP, College Station, TX).
A total of 36,221 patients with pancreatic adenocarcinoma were identified, of whom 30 committed suicide over an observed 22,248 person-years. The incidence rate among patients with pancreatic adenocarcinoma was 135.5 per 100,000 person-years. The corresponding suicide rate in the United States in the year 2005 for people 65-74 years of age was 12.5 per 100,000 years.18 This calculation yields a standardized mortality ratio (SMR) of 10.8 (95% CI, 9.2-12.7).
Patient and Disease Characteristics
The average age at diagnosis of pancreatic adenocarcinoma was 68.5 years (interquartile range, 60-77 years), and 50.8% (N = 18,420) of the population were male. Caucasians were the predominant race in the distribution (N = 29,599 [81.9%]), whereas African Americans (N = 4068 [11.2%]) and Asian/Pacific Islanders (N = 2291[6.3%]) composed the majority of the remaining races. More patients were included in the SEER database after the year 2000 (N = 26,308 [72%]) compared with 1995-2000 (N = 9913 [27%]). The majority of the patients were married (N = 20,611 [58.8%]), whereas 3155 (8.7%) were divorced and 3812 (10.5%) were single and never married.
Characteristics of Therapy
Overall, 30,042 (83.5%) patients did not undergo any form of operative intervention, whereas 5923 (16.5%) underwent operative treatment. Operative intervention was not recommended to the majority of patients with pancreatic adenocarcinoma (N = 24,744 [68.8%]), whereas 5298 (14.7%) who were recommended operative therapy did not undergo it. Radiation therapy was given to 7317 (20.4%) of the patients. The suicide rates by patient and disease characteristics are shown in Table 1. Of the 5923 patients who underwent operative intervention, data on the type of operation at the primary site were available in 5020 patients (84.7%). Of this number, 56.4% underwent a pancreatoduodenectomy, 11.6% underwent an extended pancreatoduodenectomy/total pancreatectomy, and 9.4% underwent a localized/distal pancreatectomy. Overall, 4082 (81.3%) underwent an attempted curative resection, where 938 (18.7%) underwent a surgical procedure (but none directed at the primary site), which makes it likely they underwent a palliative/noncurative procedure.
Table 1. Suicide Rates and Standardized Mortality Ratios
|Age ≥60 y||24||15,990||150||12||10.2-14|
|Age <60 y||6||6258||96||7.7||6.2-9.3|
|Stage of disease|
|Operative intervention recommended not performed||5||2800||179||14.3||12.4-16.4|
|No operative intervention||16||11,776||136||10.8||9.2-12.6|
|Diagnosis between 1995-2000||9||6897||130||10.4||8.7-12.3|
|Diagnosis between 2001-2005||21||15,351||137||10.9||9.2-12.9|
The median overall survival for patients committing suicide was 2 months (95% CI, 0.5-3 months) compared with 4 months (95% CI, 4-5 months) for the entire group who did not commit suicide, irrespective of type of intervention (P = .04). Survival by patient characteristics is shown in Table 2.
Table 2. Median Survival by Patient/Disease Characteristics, and Operative Intervention
|Age ≥60 y, n = 27,887||4 (4-4)||<.001|
|Age <60 y, n = 8334||6 (6-6)||Referent|
|Women, n = 17,801||5 (4-5)||.02|
|Men, n = 18,420||4 (4-4)||Referent|
|Stage|| || |
| Localized, n = 2345||8 (7-8)||Referent|
| Regional, n = 10,662||8 (8-9)||.07|
| Distant, n = 20,627||3 (3-3)||<.001|
|Operative intervention, nondistant disease; n = 4601||14 (14-15)||Referent|
|Operative intervention recommended not performed, nondistant disease; n = 2308||6 (5-6)||<.001|
|No operative intervention, nondistant disease; n = 8508||6 (6-6)||<0.001|
Characteristics Associated With Patients Committing Suicide
Univariate odds for the association of suicide with male sex was 13.5 (95% CI, 3.2-56.9, P<.001). Association of suicide with other factors was investigated only in males, given the low number of events (suicides) in females (n = 2). Among males, suicide was associated with single status and surgery (Table 3). Among those who had some form of operative intervention, the median survival of those who committed suicide was 2 months (IQR 0.5-18 months), whereas the median survival among the others was 10 months (IQR 5-19 months) (P = .01).
Table 3. Multivariate Analysis of Patient and Therapy Characteristics to Determine Their Association With Suicide and Mortality in Males (n = 18,420)
|Age >60 y||2.2 (0.7-6.5)||.15||1.2 (1.1-1.2)||<.001||2.4 (0.8-7.0)||.11|
|Married||0.3 (0.1-0.6)||.002||0.8 (0.8-0.9)||<.001||0.3 (0.1-0.6)||.001|
|No operative intervention||1||—||1|| ||1||—|
|Operative intervention||2.5 (1.0-6.4))||.05||0.5 (0.4-0.5)||<.001||1.4 (0.5-3.8)||.4|
|Operative intervention recommended not performed||1.3 (0.4-4.1)||.61||0.9 (0.9-1.0)||.06||1.2 (0.4-3.7)||.7|
|Radiation administered||0.7 (0.2-2.0)||.47||0.6 (0.5-0.6)||<.001||0.4 (0.1-1.3)||.2|
Because of substantive improvement in survival of cancer patients over the last decade, the focus among both healthcare providers and patients has shifted toward improvement in quality of life and survivorship,9 especially in patients with pancreatic adenocarcinoma. The incidence of suicide among cancer patients is almost twice that of the US general population, and studies with the SEER database have demonstrated almost a fivefold increase in suicide rate among patients with gastric malignancies.9 A large study from the Danish Cancer Registry revealed a SMR for suicide of 1.7 for male cancer patients and 1.4 for female cancer patients, while Swedish and Norwegian studies in the same time period found similar population estimates for cancer patients.19-21
Within the context of the discussion of suicide and malignancy, depression is well recognized as being associated with pancreatic cancer as part of the triad of symptoms: anxiety, an impending sense of doom, and depression.22 Indeed, patients with pancreatic cancer may be even more susceptible to depression than patients with other malignancies. Previous clinical studies showed that 76% of patients with pancreatic cancer had depressive symptoms before operative therapy compared with only 20% of patients with colon cancer.23 In 1986, Holland et al found that patients with advanced pancreatic cancer reported greater depression and mood disturbances versus patients with other intra-abdominal neoplasms.24 Since then, several other studies have verified the high prevalence of depression in patients with pancreatic cancer, which has been reported to be as great as 33%-50%.10, 11
The mechanisms of depression in patients with pancreatic cancer may be multifactorial and include immunologic25, 26 (related possibly to antiserotonin antibodies), hormonal25 (increased serotonin secretion), paraneoplastic25, and/or biochemical25, 27 (acid-base, anemia, etc.) aspects. Social exposure to pancreatic cancer in the media has been nihilistic, with poor outcomes in several highly visible individuals recently compounding the negative perception of the disease and contributing to the overall feeling of doom. Severe depression has been associated with a worse survival, especially in patients with breast28 and hepatobiliary29 malignancies.30, 31 In considering the association of depression with malignancies, one must also consider the effects of confounding variables, such as cigarette smoking, alcohol use, obesity— and even the possibility of depression acting as the heralding feature of cancer.32 Although difficult to predict, depression remains a substantive issue in survivors with pancreatic cancer.
Studies have suggested that the suicide rate among depressed patients is around 2%-7%,33, 34 and that there is a high correlation between severe depression and suicide.35 In a recent study of 807,694 depressed veterans, the suicide rates were 29 per 100,000 person-years in women and 89 per 100,000 person-years in men, which are lower than those reported in our study.36 The exact reason for a higher suicide rate in patients with pancreatic adenocarcinoma remains a matter of investigation, although chronic pain37 and social nihilism may account for some of the difference.
Characteristics associated with suicide in the general population, such as male sex and unmarried status,38 were also evident in our study, but an important factor we found was the association of suicide with operative intervention. Suicides in the patients who underwent operative intervention tended to occur in the first two months postoperatively (median 2 months, IQR 0.5-18), as shown by our analysis. This observation is a novel finding and has not been reported previously. If verified prospectively, this tendency would afford clinicians a window of opportunity for effective screening and intervention in patients with pancreatic cancer, especially during their first postoperative visit. The overall SMR for suicide in patients undergoing operative intervention was less than those not undergoing operative intervention (8.4 vs 10.8, P = .46), but this SMR may culminate from the increased person-years at risk for patients undergoing operative intervention, given their improved survival (Tables 2 and 3). In addition, other factors relevant in the postoperative period, such as stress and the finding that patients who underwent operative intervention were younger (P < .001) and more likely to have locoregional disease (P < .001), may contribute to the lesser SMR.
Our study is subject to several limitations, related primarily to the nature of the data collected. Although cause of death may be subject to misclassification bias, other studies have shown that the suicide codes have been entered accurately.39 It has been suggested, however, that suicide is generally under-reported, which, if true, would only add strength to our conclusions. The roles of confounding factors and psychologic assessment are lacking in our study, although previously published studies do corroborate the high rates of distress and depression in patients with cancer. In addition, our study did not capture accurately the temporal trends in the diagnosis, the nature of operative intervention (palliative or curative intent), or any attempts at the treatment of depression and/or prevention of suicide over the study period.
In summary, suicide appears to be much more common in male patients with pancreatic cancer than in the general population, may be a preventable cause of mortality, and deserves further attention. Identification of depression and psychologic distress, with the intent of psychologic intervention, has been shown to have a beneficial impact on the quality of life of patients with cancer.15 The association between operative intervention and suicide needs to be investigated further. The reproducibility of these findings might encourage healthcare providers, and especially surgeons, to screen pancreatic cancer patients in the early postoperative period, which seems to be a high-risk period for suicide.
CONFLICT OF INTEREST DISCLOSURES
The authors made no disclosures.