Suicide Rates and Risk Factors among Korean Cancer Patients, 1993-2005

For several years after a cancer diagnosis, the suicide rate is higher for patients than for the general population, and evidence that cancer increases suicide risk has been accumulating in the United States (1, 2), Norway (2, 3), Sweden (2, 4), Austria (2, 5), Finland (2), and Denmark (1-6). The number of people who live with the after-effects of cancer and its treatment—disability, fear of recurrence, and economic burden—is growing (7-20). Factors related to suicide include mainly clinical characteristics such as age at diagnosis (6, 21), prognosis (5), psychiatric status (21, 22), stage (1, 3, 6), time since diagnosis (1, 6, 21), and period of diagnosis (1, 4-6), and sociodemographic characteristics such as sex (1, 3), race (1, 21), and marital status (1, 3, 21).

Cancer is one of several illnesses that increase the risk of suicide in Asian countries (22-28). In Korea, cancer has been named as a suicide risk factor when coupled with comorbidities such as psychiatric diseases (22). Here we investigated whether the suicide rate is higher among Korean cancer patients than it is in the general Korean population or in other countries. We then investigated risk factors that could explain the differences.

We identified cancer patients from the Korea Central Cancer Registry (KCCR), a nationwide registry that covers approximately 90% of the nation's new cancer cases (29). The study subjects were registered in the KCCR from 1993 to 2002 and followed up until 2005. We excluded multiple primary cancer cases and obtained suicide mortality data for both cancer patients and the general Korean population from the National Statistical Office (NSO) “Cause of Mortality” database.

Available data in the KCCR included sex, age, and cancer diagnosis made according to International Classification of Diseases for Oncology (ICD)-10 criteria (30, 31). We identified the 12 most common cancer sites, which included the 10 most common among men and the 10 most common among women, and we grouped the rest as “other” (7). We identified as suicides those deaths coded as “Suicide and Self-inflicted Injury (ICD-10 code X60-X84)” in the NSO Cause of Mortality database. The other variables we collected from that database were age, educational status, job status, and marital status, all at the time of death.

We calculated sex- and age-standardized mortality ratios (SMR) and used them to compare the suicide rate of cancer patients with that of the general Korean population. We calculated person-years at risk for each year from 1993 through 2005. We calculated the expected number of suicides by an indirect standardization method, multiplying person-years at risk by the Korean population suicide rate in each year of follow-up. We calculated SMRs by dividing the observed number of suicides by the expected number of suicides and derived 95% confidence intervals (95% CI) using a Poisson regression model (32).

To investigate the sociodemographic characteristics associated with suicide risk, we estimated rate ratios (RR) using a log-linear Poisson regression model. We built the regression model reflecting variables that correlated significantly with risk in univariate analysis (sex, age at diagnosis, year of diagnosis, time to suicide or cancer death, site of cancer, marital status, and educational status) and known suicide risk factors, namely, sex (1, 3), age at diagnosis (6, 21), marital status (1, 3), site of cancer (1, 3-5), and residential area (22, 33). Because we obtained the sociodemographic data from a mortality database, the study was necessarily limited to deceased patients. We did all analyses with STATA 10.1 (StataCorp LP), and all statistical tests were two-sided.

Among the 838,594 cancer patients registered, we excluded those who had multiple primary cancers (n = 19,392); those whose date of death (n = 746), age at diagnosis (n = 450), or sex (n = 1) was missing; and those whose cancer diagnosis was recorded on or after the date of death (because those survival times were not reliable; n = 1,710).

Finally, 816,295 cancer patients (54.7% male) with a mean survival time of 3.5 years (2.8 for men and 4.3 for women) were included in the study, which covered 3,007,294 person-years. We observed 446,778 men for 1,311,008 person-years and 369,517 women for 1,696,286 person-years. Among them, 2,065 patients committed suicide, yielding an overall suicide rate of 68.7/100,000 person-years (Table 1).

Suicide rates were higher for men (116/100,000 person-years) than for women (32/100,000 person-years). They were twice as high for cancer patients than for the general population, and the differences were larger among men (Table 1).

The suicide rate for patients of both sexes increased with age (Figs. 1 and 2). The suicide SMR, however, was highest among 20- to 39-year-olds and then decreased with age (Table 1). The sex- and age-adjusted suicide rates were the highest in the year following the cancer diagnosis than at any later time, but they remained somewhat elevated even after 5 years (Table 1).

A notable finding was that the 1993-1997 suicide rate jumped in 1998-2002. The suicide SMR was also higher for the 1998-2002 group than for the 1993-1997 group. The differences were still observed, although lessened, after we stratified the data by length of follow-up (Table 1).

Among men, the suicide SMR was especially high for pancreas, biliary, and oropharyngeal cancer, and low for stomach, colon, and prostate cancer (Table 2). Among women, the suicide SMR was highest for lung, ovarian, and pancreatic cancer, and lowest for breast, cervical, and thyroid cancer (Table 3). For both men and women, the suicide SMR was highest for pancreas, lung, and biliary cancer the first year after diagnosis (Tables 2 and 3).

In a case-only study among those who died (n = 483,572), the risk of suicide was significantly higher for men than for women (Table 4). Patients ages 20 to 39 years were the most likely to die of suicide (Table 4). The risk of suicide was significantly higher for people who were divorced, separated, or widowed versus married people, for those who did not have a high school education versus those who did, for the unemployed versus the employed, and for those who lived in large cities versus those who lived in small cities or rural areas, with all variables pertaining to the time of death (Table 4).

In this first population-based Asian study that used a nationwide cancer registry to investigate the risk of suicide among cancer patients, we found that the suicide rate among Korean cancer patients was approximately twice that of the general Korean population. Additionally, although direct comparisons must be made with care, suicide rates in the general population were notably higher in Korea than in western countries, where they range from 39.5 to 72.1 for men and from 7.5 to 20.4 for women (1-5). Suicide SMRs among cancer patients also tended to be higher in Korea (2.05 for men, 1.87 for women; Table 1) than in western countries (1.55-2.09 for men, 1.35-1.48 for women; refs. 1, 3-5).

That suicide SMRs were highest in those 29 to 39 years old may be due to younger patients feeling more discouraged than older patients, and the finding that suicide SMRs were still somewhat elevated 5 years later raises concerns about the long-term consequences of a cancer diagnosis and its treatment.

Our finding that both the suicide rate and the suicide SMR increased during the second 5-year follow-up period was unexpected. The differences were still observed after we stratified the data for length of follow-up. The increase cannot be explained by the cancer itself because national cancer survival rates improved over the study period (34). Suicide rates also increased in the general Korean population during the same period, but it did so more rapidly in cancer patients. Because lower socioeconomic status is a risk factor for suicide in the general population (1, 3, 22, 35-37), the increment in cancer patients may be attributable to the increasing socioeconomic inequality observed in South Korea (38-40). That suicide risk was higher among those with less education and those who were unemployed at the time of death supports that speculation. In fact, the Gini coefficient, which estimates the inequality of income or wealth distribution, increased in Korea from 0.269 in 1993 to 0.304 in 2005.

Our finding that suicide risk among cancer survivors was higher in large cities than in small cities or rural areas conflicts with previous reports about the general Korean population (22, 33). Studies are needed to determine how type of residential area is related to suicide risk.

That suicide rates in Korea were especially high for patients with pancreatic, lung, biliary, and oropharyngeal cancer is concordant with reports in western countries (1, 3-5). This could be partially explained by the patients' prognosis. Suicide risk was higher among patients with cancers that confer a poor prognosis, especially within the first year of diagnosis. It may also be related to the quality of life specific to the type of cancer. For example, change of appearance, speech, and swallowing ability could lead to suicidal ideation among patients with oropharyngeal cancer.

We observed unexpected patterns of suicide among patients with stomach or liver cancer, which are common cancers in Korea. The suicide SMR for stomach cancer was not as high as it is in western countries (1, 3, 4), which might follow from the better prognosis for stomach cancer in Korea (5, 6, 41). That the suicide rate for liver cancer was lower than in Sweden (4) could follow from differing etiologies: chronic hepatitis B (from vertical infection) in South Korea and chronic hepatitis C (often related to i.v. substance abuse) in Sweden (42, 43).

Our study should be interpreted in the light of its limitations. First, we could not determine which patients completed treatment and which were midtreatment, and those groups could differ in suicide propensity. For those who completed treatment, for example, socioeconomic status might be a more important risk factor than clinical characteristics. Second, our limited data did not allow for consideration of the influence of stage at diagnosis in the analysis of suicide risk. However, the correlation between those variables is inconsistent. A study in the United States showed that suicide rates were higher among cancer patients with advanced disease at diagnosis (1), whereas an international study of breast cancer patients showed the trend of increased suicide risk with increasing stage to be nonsignificant (P for heterogeneity = 0.08; ref. 2) and a Norwegian study reported the relative suicide risk to be higher for patients with nonlocalized cancers than for patients with localized cancers (P < 0.001) only in men (3).

Third, patients who were registered in 2001-2002 were followed up for <5 years. Therefore, for patients who survived >5 years, the difference in suicide between those who were diagnosed in 1993-1997 and those who were diagnosed in 1998-2002 could be underestimated because the suicide rate increased during the latter period.

Fourth, in the analysis of risk factors, our sample was limited to deceased patients, and it therefore may have been biased. Our observations, however, are consistent with reports that suicide risk in the general population is higher when social status is lower. Furthermore, suicide risk could be underestimated in lower socioeconomic groups because a higher cancer death rate could operate as a competing mortality. There is ample evidence that lower socioeconomic status is associated with poorer cancer outcomes and survival (44-47). Nevertheless, further investigation to confirm our observation is warranted.

Fifth, KCCR records may underestimate suicide rates because the KCCR depends on the reports of bereaved families. Thus, many suicides may be misclassified as traffic accident fatalities, and the deaths of homeless or unidentified persons are not counted (48). In addition, we did not consider race as a significant variable because Korea is racially homogeneous and neither the KCCR nor the NSO database collects racial data.

In spite of those limitations, our study showed that Korean cancer patients were at increased risk of suicide, and the risk was greater for the less educated and the unemployed. Further research into the suicide risk of cancer patients should extend the range of concerns from clinical factors to socioeconomic factors, and our findings suggest the need for social support and prevention efforts.

No potential conflicts of interest were disclosed.

Grant Support: National Cancer Center Grant 0710731-2. The manuscript was professionally edited by Dr. Miriam Bloom (SciWrite Biomedical Writing and Editing Services).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.