Risk of major depressive disorder in Japanese cancer patients: A matched cohort study using employer‐based health insurance claims data

Abstract Objective Patients with cancer are at high risk of depression. However, the risk of major depressive disorder (MDD) after cancer diagnosis has not been studied in a population setting in Japan. This cohort study used a Japanese medical claims database to examine time to MDD in cancer patients and the risk of MDD (hazard ratio; HR) compared with matched cancer‐free controls. Methods Primary endpoint was time to MDD (starting 6 months before cancer diagnosis) in adult (18–74 years) cancer patients; secondary endpoint was time to MDD (6 months before to 12 months after cancer diagnosis) in a matched cohort of cancer patients and cancer‐free controls. Multivariate analyses were performed to determine HRs for all cancers and for each cancer site. Results Of 35 008 cancer patients (mean age, 53.3 years), 2201 (6.3%) were diagnosed with MDD within 66 months. Matched cancer patients (n = 30 372) had an elevated risk of MDD compared with cancer‐free controls (n = 303 720; HR [95% confidence interval] 2.96 [2.77–3.16]). MDD risk was highest in patients with multiple cancers, pancreatic cancer, and brain cancer. Compared with middle‐aged patients, risk was higher in patients <40 years old and lower in patients ≥65 years old; risk tended to be higher in women than in men. Conclusions Compared with cancer‐free individuals, Japanese patients with cancer, mostly <65 years old, had an almost threefold higher risk of developing MDD within 12 months of cancer diagnosis. Physicians should watch for MDD in cancer patients and treat when necessary.


| BACKGROUND
Being diagnosed with cancer is a stressful event that often leads to anxiety and depression. 1 The prevalence of depression in cancer patients, as reported in systematic reviews and meta-analyses, ranges from as low as 1% to as high as 77.5% and depends on the clinical setting, the type and severity of cancer, the method used to diagnose depression, and the expertise level of the diagnosing physician. 2 (95% confidence interval [CI] 12.2-17.7). 3 Importantly, depression in patients with cancer is associated with an elevated risk of cancer mortality 5 and an up to 24-fold greater risk of suicide or death by external injury, particularly in the first week after cancer diagnosis. 6,7 Depression also reduces adherence to cancer treatments, 8 is a psychological burden on the family, 9 and increases healthcare utilization, including hospitalizations. 10 Although the prevalence of depression in cancer patients has been well documented, only one large, well-controlled, matched cohort study, conducted in Sweden, has examined the incidence of newly developed mental disorders following a cancer diagnosis. In that study, an increased risk of mental disorders, including depression, began up to 10 months before cancer diagnosis, peaked immediately after diagnosis, and remained elevated above baseline for up to 10 years. 11 Further, the risk of mental disorders during the first year after cancer diagnosis was greatest among patients aged 40-64 years (ie, middle-aged patients). 11 One in two people in Japan are estimated to be diagnosed with cancer in their lifetimes. 12 Although its effect on mortality has decreased in recent years, cancer has been the leading cause of death in Japan since 1981. 13 In a Japanese retrospective study, almost half of patients with eight common types of cancer were prescribed psychotropic drugs, including antidepressants and benzodiazepines, soon after their cancer diagnoses. 14 These results suggest that Japanese patients with cancer are at high risk of developing depression. However, population-based data on the incidence of depression in Japan are limited to the World Mental Health Survey, 15,16 and no data exist on the incidence of depression in cancer patients. Furthermore, because depression is less prevalent in Japan than in other countries, 15 possibly because of social stigma associated with mental disorders, 17 it is important to determine the rate of depression among Japanese patients with cancer.
To expand our understanding of the relationship between cancer and depression, particularly in Japan, we used a large, nationwide, employer-based health insurance database to estimate the time to clinically diagnosed major depressive disorder (MDD) in more than 30 000 Japanese patients with any type of newly diagnosed cancer. In addition, we used a matched cohort of cancer patients and cancerfree controls to determine hazard ratios (HRs) for the development of MDD for all cancers and for individual cancer types. To our knowledge, this is the first study to determine HRs for depression across a broad range of cancer types.

| Study design
This cohort study used a commercially available, anonymized database of workers' medical services and prescriptions collected from multiple health insurance societies in Japan (JMDC database; JMDC Inc., Tokyo, Japan). 18 The JMDC database provides data on patient demographics, disease diagnoses based on International Statistical Classification of Diseases and Related Health Problems, 10th revision (ICD-10) codes, prescriptions, and medical procedures. In Japan, every person is covered by either a national or an employer-based comprehensive health insurance scheme. 19 The JMDC database covers employees of medium to large companies and their dependents (family members living in the same household who earn <1.3 million yen per year and who are <75 years old). In Japan, public health insurance systems exist for the elderly (≥75 years old) and some specific professions (eg, farmers, public officers). 20   before the index month (month/year of cancer diagnosis); ≥2 diagnoses (ie, ≥2 health insurance claims issued for cancer diagnosis), including the initial diagnosis, for the same cancer site within 3 months of the index month; no diagnosis of MDD between 6 and 12 months before the index month; and continuous health insurance enrollment for ≥12 months before the index month to ensure access to complete records. Cancer diagnosis and site were identified by ICD-10 codes (Table S1). The adult cancer patient cohort was followed up for as long as possible.

| Matched cohort of cancer patients and cancer-free controls
The HR for development of depression over 1 year after cancer diagnosis in cancer patients compared with cancer-free controls was investigated in the matched cohort. In this matched cohort, the cancer group included those adult cancer patients who had continuous health insurance enrollment for ≥12 months after the index month. The cancer-free group included individuals with no cancer diagnosis who were matched to cancer patients according to age, sex, and working status (working vs non-working in the health insurance scheme).
Cancer-free individuals also had to have continuous health insurance enrollment for ≥12 months before and ≥12 months after the index month of their matched cancer patient and no diagnosis of MDD between 6 and 12 months before the index month. A random sampling without replacement was performed at a ratio of cancer patients to non-cancer controls of 1:10. This matching ratio was based on the number of patients targeted for matching (approximately 30,000 cancer patients) and the availability of well-matched controls, and is in line with the ratio used in the Swedish cohort study. 11

| Statistical analysis
Analysis of time to onset of MDD in all adult cancer patients (primary endpoint) and in the matched cohort cancer and cancer-free groups (secondary endpoint) was performed using the Kaplan-Meier method.
The population at risk, cumulative number of events, cumulative number of censored patients, and percentage (95% CI) of patients without MDD (survival) are presented for every 6 months from the start of the observation period. The cumulative incidence of MDD was compared between the cancer group and the cancer-free group using a log-rank test with a two-sided significance level of 5%. In the matched cohort, multivariate analysis was performed using a Cox proportional hazards model to estimate the HR for onset of MDD in the matched cancer group versus the cancer-free group for all cancers and for each cancer site. The following variables were used for multivariate analysis: (a) cancer or cancer-free group; (b) sex/age interaction term (sex/age were included in the covariates as interaction terms as the effects of age may differ depending on sex; if the site of cancer was breast, uterine cervix, uterine corpus, ovary, or other female genitalia, a separate estimate was performed using "female/40-64 years of age" as the reference in the sex/age interaction term); and (c) working status (working vs non-working). An additional multivariate subgroup analysis was conducted in the matched cohort cancer group only, with covariates of sex, age, membership category, chemotherapy status, radiotherapy status, and surgery requiring hospitalization for ≥5 days status. If the HR or CI of the HR was <0.01 or >999.99, it is reported as "not evaluable." The data warehouse platform was Netezza    34.5% of female patients), colorectum (14.9%), stomach (9.1%), lung (6.1%), and prostate (6.0%; 11.6% of male patients); 6.5% of patients had cancers in multiple categories ( Figure 1). In the matched cohort, the incidence of MDD during the 18-month observation period was significantly greater (log-rank test P < .001) in the cancer group than in the cancer-free group (secondary endpoint; Figure 2). In the cancer group, 1199 of 30 372 (3.9%) patients were diagnosed with MDD, whereas in the cancer-free group, 4097 of 303 720 (1.3%) patients were diagnosed with MDD.

| Time to onset of MDD
Although the largest differences in MDD incidence between the cancer and cancer-free groups were seen after the index month, a small but significant difference was also seen in the 6 months before the index month (ie, before cancer diagnosis).

| All cancers
Compared with the cancer-free group, patients in the cancer group had a nearly threefold increased risk of being diagnosed with MDD between 6 months before and 12 months after a cancer diagnosis  Figure 1). Both male and T A B L E 1 Background and characteristics of patients with cancer and cancer-free individuals in Japan

| Cancer type
The risk of MDD relative to cancer-free individuals was increased in almost all cancer types (Figure 1). F I G U R E 1 Hazard ratios and 95% CIs of the risk of major depressive disorder in the matched cohort (cancer group vs cancer-free group) for each cancer type. Not shown are cancers of the small intestine (n = 103; 0.3%); bone and articular cartilage (n = 57; 0.2%); nasal cavity, paranasal sinus, and middle ear (n = 50; 0.2%); other endocrine gland (n = 38; 0.1%); other female genitalia (n = 25; 0.1%); other digestive organ (n = 10; <0.1%); and eye (n = 6; <0.1%). CI, confidence interval; CNS, central nervous system cancer types were generally consistent with the effects seen in all cancers combined (Table S2).

| Subgroup analysis (cancer patients only)
Within the matched cohort cancer group, the risk of MDD was signifi- Notably, the median age of patients in the Swedish study (69 years) was more than 10 years older than in our study (53 years for the matched groups). This difference reflects the employer-based health insurance database used in the present study. Both studies, as well as other studies of specific cancer types, suggest that younger cancer patients are at higher risk of depression than older patients. 11,[22][23][24] The risk of MDD was also somewhat higher in women than in men in our study, consistent with the Swedish study. 11 The risk of MDD was elevated across a range of cancer types, with the highest risk seen in patients with multiple cancer sites, pancreatic cancer, brain/central nervous system cancer, and oral cavity/ pharynx cancer. Understandably, the cancers with the highest risk of MDD were those with poor prognoses and/or substantial physical dysfunction. In addition, most patients with cancer at multiple sites had metastatic cancer, which is usually more advanced and associated with worse outcomes. Cancers with better prognoses (eg, thyroid cancer) generally had lower risks of MDD, although patients with breast cancer, which has a 5-year relative survival rate 25  cervical, 32 and breast cancer, 33 as well as Hodgkin lymphoma. 34 The size of our study allowed us to determine the risk of MDD for many more types of cancer than previously studied. However, the relatively low numbers of patients with some rarer types of cancer made it difficult to accurately determine HRs.
The main advantage of this study was the use of a large nationwide health insurance database, in which members could be followed across different medical facilities and irrespective of whether they visited a hospital, which allowed for a robust cohort design. The large sample size allowed analysis by cancer site, as well as multivariate analyses of the effects of age, sex, and insurance category. The high matching ratio (10:1) of cancer-free controls to cancer patients and the random selection of controls helped minimize any potential bias.
The study also included relatively long observation periods for both the overall cancer cohort (median of 32 months) and the matched cohort (18 months).

| Clinical implications
Compared with patients in Western countries, patients with depression in Japan may be more reluctant to seek medical treatment. 16 As treatment of the cancer may be the immediate priority, doctors should watch for depressive symptoms in their cancer patients, especially for certain cancer types and in those receiving chemotherapy, and, when necessary, encourage patients with cancer to seek support and treatment for their depression.

| CONCLUSION
In conclusion, this study documented that predominantly working-age Japanese patients with cancer are at high risk of developing MDD in the year following a cancer diagnosis. Physicians should watch for MDD in cancer patients for at least 1 year after cancer diagnosis, when the risk of MDD is high, and treat when necessary. Limited.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from JMDC Inc. but were used under license for the current study; therefore, restrictions apply and the data are not publicly available. For inquiries about access to the data set used in this study, please contact JMDC (https://www.jmdc.co.jp).