Hazard ratios were estimated from flexible parametric survival models, allowing the effect of cancer diagnosis to vary over time. A spline with 5 df (4 intermediate knots and 2 knots at each boundary, placed at quintiles of distribution of events) was used for the baseline rate, while 3 df was used for the time-varying effect. All models were adjusted for sex, age, and calendar period at the reference date (2001-2003, 2004-2006, or 2007-2009) as well as educational level (high, middle, or low/unknown).
Hazard ratios were estimated from flexible parametric survival models, allowing the effect of cancer diagnosis to vary over time. A spline with 5 df (4 intermediate knots and 2 knots at each boundary, placed at quintiles of distribution of events) was used for the baseline rate, while 3 df was used for the time-varying effect. Severe cancer includes pancreatic, hepatic, and esophageal cancers. All models were adjusted for sex, age, and calendar period at the reference date (2001-2003, 2004-2006, or 2007-2009) as well as educational level (high, middle, or low/unknown). CNS indicates central nervous system.
Hazard ratios were estimated from flexible parametric survival models, allowing the effect of cancer diagnosis to vary over time. A spline with 5 df (4 intermediate knots and 2 knots at each boundary, placed at quintiles of distribution of events) was used for the baseline rate, while 3 df was used for the time-varying effect. All models were adjusted for sex, age, and calendar period at the reference date (2001-2003, 2004-2006, or 2007-2009) as well as educational level (high, middle, or low/unknown). Alcohol and tobacco abuse or dependence was excluded from the group of substance abuse disorders.
Solid lines and scatters represent the percentages of medication use among patients with cancer, whereas dashed lines represent use among cancer-free individuals. Lines were plotted using locally weighted scatterplot smoothing. The spike plots indicate a statistically significant difference in medication use on a daily basis between patients with cancer and cancer-free individuals (P < .05 corrected for multiple testing). Severe cancer includes pancreatic, hepatic, and esophageal cancers. CNS indicates central nervous system.
eTable 1. Characteristics of Cancer Patients and the Cancer-Free Individuals at the Reference Date
eTable 2. Hazard Ratios and 95% Confidence Intervals of Depression, Anxiety, Substance Abuse, Somatoform/Conversion Disorder, Stress Reaction and Adjustment Disorder Before and After a Cancer Diagnosis, Comparing Patients With Locally Advanced or Metastatic Cancer to Patients With Localized or Non-metastatic Cancers
eFigure 1. Hazard Ratios and 95% Confidence Intervals of Depression, Anxiety, Substance Abuse, Somatoform/Conversion Disorder, Stress Reaction and Adjustment Disorder (Defined Through the Main Discharge Diagnosis of Inpatient Care Alone) Before and After a Cancer Diagnosis, a Matched Cohort Study in Sweden, 1999-2010
eFigure 2. Cumulative Incidences and 95% Confidence Intervals of Individual Mental Disorders Before and After a Cancer Diagnosis, a Matched Cohort Study in Sweden, 1999-2010
eFigure 3. Percentages of Individual Psychiatric Medication Use Before and After a Cancer Diagnosis, a Matched Cohort Study in Sweden, 2006-2011
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Lu D, Andersson TML, Fall K, et al. Clinical Diagnosis of Mental Disorders Immediately Before and After Cancer Diagnosis: A Nationwide Matched Cohort Study in Sweden. JAMA Oncol. 2016;2(9):1188–1196. doi:10.1001/jamaoncol.2016.0483
Psychiatric comorbidities are common among patients with cancer. However, whether or not there is increased risk of mental disorders during the diagnostic workup leading to a cancer diagnosis was unknown.
To examine the relative risks of depression, anxiety, substance abuse, somatoform/conversion disorder, and stress reaction/adjustment disorder during the periods before and after cancer diagnosis compared with individuals without cancer.
Design, Setting, and Participants
Nationwide matched cohort study from January 1, 2001, to December 31, 2010, in a Swedish population and health registers.
Main Outcomes and Measures
We estimated the time-varying hazard ratios (HRs) of the first clinical diagnosis of the studied mental disorders from 2 years before cancer diagnosis, through the time of diagnosis, and until 10 years after diagnosis, adjusting for age, sex, calendar period, and educational level. To assess milder mental conditions and symptoms, we further assessed the use of related psychiatric medications for patients with cancer diagnosed during 2008-2009.
The study included 304 118 patients with cancer and 3 041 174 cancer-free individuals who were randomly selected from the Swedish population and individually matched to the patients with cancer on year of birth and sex. The median age at diagnosis for the patients with cancer was 69 years, and 46.9% of the patients were female. The relative rate for all studied mental disorders started to increase from 10 months before cancer diagnosis (HR, 1.1; 95% CI, 1.1-1.2), peaked during the first week after diagnosis (HR, 6.7; 95% CI, 6.1-7.4), and decreased rapidly thereafter but remained elevated 10 years after diagnosis (HR, 1.1; 95% CI, 1.1-1.2). The rate elevation was clear for all main cancers except nonmelanoma skin cancer and was stronger for cancers of poorer prognosis. Compared with cancer-free individuals, increased use of psychiatric medications was noted from 1 month before cancer diagnosis and peaked around 3 months after diagnosis among patients with cancer.
Conclusions and Relevance
Patients diagnosed as having cancer had increased risks of several common mental disorders from the year before diagnosis. These findings support the existing guidelines of integrating psychological management into cancer care and further call for extended vigilance for multiple mental disorders starting from the time of the cancer diagnostic workup.
Patients with cancer are known to have an increased risk of psychiatric symptoms and disorders,1 cardiovascular diseases,2 and suicide.3 Historically, severe comorbidities among patients with cancer have largely been attributed to the adverse effects of cancer treatments, the complex pathophysiology of the underlying malignancy, or simply the burden of living with a progressing and potentially fatal disease.4,5 Not only does living with cancer induce severe psychological stress, but being diagnosed as having a cancer is also highly stressful. Recently, greatly increased risks of suicide and cardiovascular events shortly after cancer diagnosis were reported among patients with cancer in different countries, across various cancer types, and among both young and adult patients.6-12 However, these reports likely only represent the tip of the iceberg for the enormous psychological turmoil related to a newly received cancer diagnosis.
Psychiatric comorbidities, such as depression, anxiety, substance abuse, somatoform/conversion disorder, and stress reaction/adjustment disorder, are common among patients with cancer1,13 and may reflect the broader spectrum of psychological distress experienced by these patients. Although depression has been extensively studied,14 evidence is mainly limited to breast cancer15 and rarely addressed during the immediate period after a cancer diagnosis.13 Moreover, other mental disorders have received less attention.1,13 On the other hand, although mental disorders are not established risk factors for cancer in general,16 some conditions, including addiction to alcohol and tobacco, might lead to higher risk of alcohol-related and smoking-related cancers.17
Most previous related studies are cross-sectional in nature,1 and a longitudinal description of the mental disorder burden among patients with cancer is lacking to date. Furthermore, most clinical and research efforts have addressed the postdiagnostic period of cancer, focusing on survivorship, the end-of-life stage, and, increasingly, the immediate periods after diagnosis and primary cancer treatment.18-20 However, accumulating evidence suggests that the cancer diagnostic workup may also introduce severe psychological distress.21-24 Whether or not the diagnostic workup leading to a cancer diagnosis increases the risk of mental disorders is largely unknown. To this end, we aimed to investigate risk changes in several common and potentially stress-related mental disorders from the cancer diagnostic workup to postdiagnosis.
Question Is there increased risk of mental disorders, including depression, anxiety, substance abuse, somatoform/conversion disorder, and stress reaction/adjustment disorder, temporally associated with the diagnostic workup leading to a cancer diagnosis?
Findings In this nationwide matched cohort study, highly increased risk of the aforementioned mental disorders was noted from 10 months before cancer diagnosis, peaked during the first week after diagnosis, and decreased rapidly thereafter but remained elevated 10 years after diagnosis.
Meaning Extended vigilance for multiple mental disorders is needed in cancer care, starting from the time of the cancer diagnostic workup.
Based on the Swedish Population and Housing Census in 1990, we identified all individuals born and living in Sweden in 1990 (N = 7 792 012). Using the unique national identification numbers, we followed up these individuals from January 1, 1991, to December 31, 2010, through cross-linkages to the Swedish Cancer Register, Patient Register, Prescribed Drug Register, Cause of Death Register, and Migration Register. The Cancer Register has been available since 1958 and approaches 100% complete data. The Patient Register has collected nationwide information on hospital discharge records since 1987 and on more than 80% of hospital-based outpatient visits since 2001. The Prescribed Drug Register collects information on drugs redeemed with a prescription from all pharmacies in Sweden since July 2005, including prescription, dispensings, and the Anatomic Therapeutic Chemical codes.
Within the nationwide study base, we subsequently conducted a matched cohort study. We first identified 326 404 adult patients (≥18 years old) with a first diagnosis of malignancy from January 1, 2001, to December 31, 2009. Cancer cases detected through autopsy were excluded (n = 4221). In the present study, we defined the 2 years before the date of cancer diagnosis as the prediagnostic period, potentially representing the time window from the emerging prediagnostic cancer symptoms25,26 through the diagnostic workup.27 The period from the date of cancer diagnosis onward was correspondingly defined as the postdiagnostic period.
Because we aimed to study the effect of the cancer diagnostic workup on newly emerging mental disorders, individuals with previous psychiatric comorbidities before entry to the cohort (ie, 2 years before the date of diagnosis) were excluded, leaving 304 118 patients with cancer in the analysis. We then randomly selected 10 individuals per patient with cancer from the study base who were free of cancer at the diagnosis date of the index patient with cancer (ie, the reference date) and free of psychiatric comorbidities at entry to the cohort (ie, 2 years before the reference date) (n = 3 041 174). Patients with cancer and cancer-free individuals were individually matched on year of birth and sex. All participants were then followed up from 2 years before the reference date (earliest possible entry on January 1, 1999) to a first diagnosis of mental disorder, death, emigration, or December 31, 2010, whichever came first. Among all cancer-free individuals, 192 598 (6.3%) developed a cancer after the reference date, and follow-up of these individuals was also censored at the cancer diagnosis. The study was approved by the Central Ethical Review Board in Stockholm, Sweden, which waived the requirement of obtaining informed consent from the study participants.
According to the Seventh Swedish Revision of the International Classification of Diseases Codes, we classified cancers into 9 subgroups, including prostate, breast (female only), colorectal, lung, nonmelanoma skin, central nervous system (CNS), lymphatic/hematopoietic, severe (pancreatic, hepatic, and esophageal), and other cancers (eTable 1 in the Supplement). We obtained information on tumor stage for prostate, breast, colorectal, and lung cancers from the Cancer Register for patients diagnosed since 2004 onward. We defined T3/T4/N1-3 as locally advanced stage and M1 as distant metastasis.
Any first-ever inpatient or outpatient hospital visit with a mental disorder as one of the registered diagnoses during the follow-up was identified from the Patient Register using the Tenth Swedish Revision of the International Classification of Disease Codes (codes F10-F99). As the outcome of interest, we defined a composite group of the 5 most commonly diagnosed mental disorders among patients with cancer,13 including Tenth Swedish Revision of the International Classification of Disease Codes for stress reaction/adjustment disorder (code F43), depression (codes F32-F33), anxiety (codes F40-F41), substance abuse (codes F11-F16 and F18-F19), and somatoform/conversion disorder (codes F44-F45), which have also been suggested as potentially related to psychological stress.28-30 Because alcohol consumption and tobacco use are known risk factors for multiple cancers,31,32 we excluded alcohol and tobacco abuse or dependence from the group of substance abuse disorders.
To assess the potential importance of differential surveillance of mental disorders between patients with cancer and cancer-free individuals, in a sensitivity analysis, we used any first-ever inpatient hospital visit with a mental disorder as the main discharge diagnosis as the outcome of interest, assuming that differential reporting is less relevant for severer mental conditions.
To assess milder mental conditions and symptoms without attended specialist care (not recorded in the Patient Register), we further assessed the use of psychiatric medications related to the studied mental disorders, including Anatomic Therapeutic Chemical codes for antidepressants (code N06A), anxiolytics (code N05B), and hypnotics/sedatives (code N05C) through the Prescribed Drug Register. For this analysis, we included 68 104 patients with cancer and 681 040 cancer-free individuals with a reference date during 2008-2009. These participants were followed up from 2 years before the reference date (earliest possible entry on January 1, 2006) to death, emigration, December 31, 2011, or 2 years after the reference date, whichever came first. We defined current use of the studied medications as previously described.33 Briefly, we defined current use of antidepressants or anxiolytics as the 4 months after each prescription claim, whereas we defined current use of hypnotics/sedatives as the 10 days after each prescription claim.
In the primary analysis, we investigated the rate of mental disorders from 2 years before diagnosis (year −2), through the date of diagnosis, to 10 years after diagnosis (year 10). Given a generally violated proportional hazards assumption, we estimated hazard ratios (HRs) and 95% CIs of the mental disorders derived from flexible parametric survival models, which allow the HRs to change over time.34 Time since cohort entry was used as the underlying timescale. To capture the dramatically increased rates around the date of cancer diagnosis, HRs were modeled separately for the prediagnostic and postdiagnostic periods. Individuals with a mental disorder diagnosed during the prediagnostic period were excluded from the analysis of the postdiagnostic period (5316 patients with cancer and 37 170 cancer-free individuals). The corresponding cancer-free individuals of the 5316 patients were also excluded (n = 52 481). We adjusted for the matching variables, in addition to educational level, in all survival models as recommended.35
The analysis was further stratified by age, sex, calendar period, educational level, and region of residence. We separately analyzed 8 major cancer types and the 4 most common types, as well as locally advanced and localized cancers along with metastatic and nonmetastatic cancers. We further compared cancers of poor prognosis (ie, lung, CNS, and severe cancers) with other cancers. Finally, in addition to combining all studied mental disorders as one group, we separately assessed rate changes of the individual disorders. Given the overall lack of statistical power during the first 6 months of follow-up (mainly due to few individuals with a newly diagnosed mental disorder among patients with cancer), we focused on the period from 18 months before cancer diagnosis onward in these subgroup analyses.
In addition to the aforementioned relative measurements, we calculated the absolute risks of these mental disorders among both patients with cancer and their matched cancer-free individuals by plotting cumulative incidence curves for the individual disorders. Competing risk models were applied for the postdiagnostic period to control for higher mortality rates among patients with cancer compared with their cancer-free counterparts after diagnosis.
In the secondary analysis, we calculated the proportions for current use of any of the studied psychiatric medications as well as specific medications from 18 months before cancer diagnosis to 2 years after cancer diagnosis. After Bonferroni correction for multiple testing, the differences in current use on a daily basis between patients with cancer and cancer-free individuals were examined using the χ2 test. Analyses were performed using statistical software (SAS, version 9.4; SAS Institute and Stata, version 13.1; StataCorp LP). P < .05 indicated statistical significance.
The median age at diagnosis for the patients with cancer was 69 years, and 46.9% of the patients were female (eTable 1 in the Supplement). In total, we identified 3355 patients and 10 296 patients with a new diagnosis of the studied mental disorders during the prediagnostic period and the postdiagnostic period, respectively.
The overall relative rate of mental disorders started to increase from 10 months before cancer diagnosis (HR, 1.1; 95% CI, 1.1-1.2) and peaked during the week after diagnosis (Figure 1). Although the magnitude of the rate increase decreased rapidly thereafter, the rate remained elevated by year 10. Using only the main diagnosis of inpatient care as the definition for mental disorders (19.9% of all diagnoses) barely changed the temporal pattern, although the magnitude of the rate increase dropped slightly (eFigure 1 in the Supplement).
The rate increase immediately before and after cancer diagnosis was greater among women compared with men (Table). Younger and better-educated patients had higher HRs after diagnosis but not before diagnosis compared with older and less-educated patients.
Largely similar results were observed for all major cancer types except nonmelanoma skin cancer (Figure 2). The rate increase was greater for cancers of poor prognosis both before and after diagnosis compared with other cancers (P < .001 for both). Patients with locally advanced or metastatic cancers had no further increased rate compared with patients with localized or nonmetastatic cancers either before or after diagnosis (eTable 2 in the Supplement). However, toward the end of follow-up, patients with locally advanced cancer and those with metastatic breast and lung cancers had greater rate increases than the others.
Largely similar results were also observed for all individual mental disorders studied (Figure 3). The magnitude of the rate elevation was comparable before cancer diagnosis across all disorders, whereas the highest rate increase was noted for stress reaction/adjustment disorder immediately after cancer diagnosis.
Among patients with cancer, depression had the highest cumulative incidence during the study, followed by anxiety and stress reaction/adjustment disorder (eFigure 2 in the Supplement). Compared with cancer-free individuals, patients with cancer had higher cumulative incidences of most mental disorders except somatoform/conversion disorder.
Increased use of psychiatric medications was noted from 1 month before cancer diagnosis (12.2% vs 11.7%, corrected P = .04), peaked around 3 months after diagnosis (18.1% vs 11.9%, corrected P < .001), and decreased slowly thereafter but remained elevated 2 years after diagnosis (15.4% vs 12.7%, corrected P < .001) (Figure 4). Increased use was most pronounced for lung, CNS, and severe cancers, whereas no increased use was noted for nonmelanoma skin cancer. Findings of specific psychiatric medications were similar (eFigure 3 in the Supplement).
To the best of our knowledge, our study represents the first and largest study to date estimating the burden and its temporal pattern of mental disorders that are potentially related to psychological stress among adult patients with cancer during the prediagnostic to postdiagnostic periods. In line with previous studies,1,13 we showed that anxiety and stress reaction/adjustment disorder were also common among patients with cancer, in addition to depression. Most important, patients with cancer had dramatically increased risks of depression, anxiety, stress reaction/adjustment disorder, and substance abuse (excluding alcohol and tobacco abuse or dependence), as well as somatoform/conversion disorder, both during the year before and the year after diagnosis. The dramatic risk increases noted immediately after cancer diagnosis corroborate with previous findings on highly increased risks of cardiovascular diseases and suicide right after cancer diagnosis,6-12 whereas the clearly elevated risks during the year before diagnosis may suggest the effect of prediagnostic cancer symptoms as well as the severe stress of undergoing clinical evaluation for a suspected malignancy.
To a varying extent, there was a similar pattern of risk elevation for most cancer types, with the smallest (if any) risk increase observed for nonmelanoma skin cancer. Patients with prostate cancer also had smaller risk elevation compared with other cancers both before and after diagnosis, likely as a result of increasing knowledge about the disease and its perceived benign prognosis in the general population.36 However, no clear variation was noted between either locally advanced and localized cancers or between metastatic and nonmetastatic cancers until the end of the first year after diagnosis, when patients having the same cancers were compared with each other. This pattern lends support to the view that, during the diagnostic workup, the experience of psychological distress does not highly correlate with the severity of a malignancy, which seems to be in line with previous findings that the prevalence of depression and psychological distress did not differ greatly by tumor stage.1 However, the slightly greater risk increase among patients with locally advanced cancer beyond 1 year after diagnosis might indicate a direct effect of the disease course, treatment adverse effects, and disease coping after the primary cancer treatment.
In Sweden, the waiting times during the cancer diagnostic workup vary greatly across cancer types. In 2011, the median intervals between the first referral to a specialist for cancer evaluation and the final diagnosis were 11 to 12 days for colorectal cancer and 76 to 82 days for nonmetastatic prostate cancer.27 The process from the first symptom onset or the first health care contact to the final diagnosis is presumably much longer.27 Although data are scant, a growing body of evidence suggests that the workup for a suspected cancer, regardless of the eventual diagnosis result, clearly affects health and well-being, particularly in terms of the psychosocial spheres of function.22,24,37 Such effect is believed to be transient among individuals receiving an initial clear result (ie, no evidence of cancer), whereas it is more persistent among individuals eventually diagnosed as having a malignancy.21
The start of risk increase for the studied mental disorders around 10 months before cancer diagnosis may likely suggest a debut of cancer symptoms at a population level, in line with earlier reported increased sickness absence before cancer diagnosis.26 Besides the potential effect of psychological stress experienced during the diagnostic workup, prediagnostic cancer symptoms26 and paraneoplastic syndrome (eg, anemia25) may also contribute to the increased risk of mental disorders and increased use of psychiatric medications during the prediagnostic period. By contrast, the fact that no clear risk increase was noted within the second year before cancer diagnosis suggests that patients with cancer did not have a higher risk of mental disorders at baseline.
One limitation of our study is the use of clinical diagnoses through inpatient and outpatient specialist care as the outcome measurement, which likely represented only more severe demonstrations of the studied mental disorders. The similar temporal pattern (but larger magnitude) of increased use of psychiatric medications further implies the effect of cancer diagnosis on both severe and milder mental conditions. Therefore, taking together both clinical diagnoses and medication use likely depicts the entire extent of mental burden around the time of cancer diagnosis, highlighting the clinical significance of our findings. However, because the studied medications are not exclusively prescribed for mental disorders or symptoms, future studies should specify the underlying reasons for prescription of such medications.
Register diagnoses of depression, anxiety, substance abuse, and stress reaction/adjustment disorder have been shown to be of good quality in general.38-41 The validity of conversion disorder diagnosis might remain a concern. Because it is rare (5% of all somatoform/conversion disorders), results of somatoform/conversion disorder stayed unchanged after excluding conversion disorder.
We had no information on the exact time course of cancer diagnosis and were unable to disentangle whether the registered date of diagnosis coincided with the date when the patient was informed about the diagnosis. Therefore, further studies with detailed information on the cancer diagnostic workup are warranted. Although our findings highlight the importance of a timely psychological intervention throughout the cancer diagnosis, it remains to be explored whether or not such intervention should be specifically tailored for patients with cancer compared with individuals without such life-threatening conditions. Finally, the generalizability of the present findings in more recent periods or other populations needs to be investigated.
A major strength of our study is the large-scale, population-based cohort design and the complete follow-up. Given the prospectively and independently collected information on both cancer and mental disorders, our study has minimal selection and information biases. The use of flexible parametric models further enabled us to investigate the temporal pattern of risk increase in mental disorders around the time of cancer diagnosis.
Patients recently diagnosed as having cancer experience dramatically increased risks of several potentially stress-related mental disorders both immediately before and after cancer diagnosis. Our findings support the existing guidelines of integrating psychological management into cancer care42 and call for extended vigilance for multiple mental disorders starting from the time of the cancer diagnostic workup.
Correction: This article was corrected on May 26, 2016, to fix errors in the Figures and Table.
Accepted for Publication: February 15, 2016.
Corresponding Author: Donghao Lu, MD, Department of Medical Epidemiology and Biostatistics, Karolinska Institutet, Nobels Väg 12A, PO Box 281, 17177 Stockholm, Sweden (email@example.com).
Published Online: April 28, 2016. doi:10.1001/jamaoncol.2016.0483.
Author Contributions: Drs Lu and Fang had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.
Study concept and design: Lu, Andersson, Fall, Valdimarsdóttir, Fang.
Acquisition, analysis, or interpretation of data: Lu, Andersson, Fang.
Drafting of the manuscript: Lu, Fang.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Lu, Andersson.
Obtained funding: Hultman, Czene, Fang.
Administrative, technical, or material support: Hultman, Czene, Fang.
Study supervision: Fang.
Conflict of Interest Disclosures: None reported.
Funding/Support: This work was supported by grant CAN 2014/417 from the Swedish Cancer Society; by grant 2012-0498 from the Swedish Research Council for Health, Working Life and Welfare; by partial new doctoral student financing from the Karolinska Institutet (Dr Lu); and by the Swedish Society for Medical Research and a research associate award from the Karolinska Institutet (Dr Fang).
Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Previous Presentation: Preliminary results were presented at the 45th Annual Meeting of the International Society of Psychoneuroendocrinology; September 8-11, 2015; Edinburgh, Scotland.