Psychologic Distress Is Associated with Cancer-Specific Mortality among Patients with Cervical Cancer

Cervical cancer is the fourth most common cancer and cause of cancer-related death in women worldwide (1). It has a clear etiology in human papillomavirus (HPV) infection (2), while other reported risk factors include smoking (3), risky sexual behavior (4), and socioeconomic disadvantage (5). Clinical characteristics, such as stage and tumor histopathologic features, are currently the most important indicators for cervical cancer prognosis (6).

Emerging evidence from both experimental and epidemiologic studies indicates that psychologic distress might affect the progression of many cancer types (7, 8). For instance, stress accelerates cell proliferation, angiogenesis, and tumor invasion by activating the adrenergic pathway (9, 10). Moreover, chronic stress reduces cellular immunity response via adrenergic and glucocorticoid receptors on immune cells (11). These mechanisms may play a particularly strong role in solid malignancies with infection-related etiology, such as cervical cancer (12).

Few studies have addressed whether psychologic distress is associated with cervical cancer prognosis. The few studies that examined stressful life events, in terms of divorce and bereavement, and cervical cancer survival have led to inconclusive results (13–17), likely due to shortcomings in sample size and the lack of controlling for known prognostic factors. It has been postulated that divorce or widowhood may lead to delayed cancer diagnosis and less treatment, and subsequently a compromised prognosis, among patients with cervical cancer (16).

Patients with cancer are vulnerable to psychologic distress during the disease process. Our recent work has, for example, illustrated that patients receiving a cancer diagnosis were at highly increased risk of several stress-related psychiatric disorders such as depression, anxiety, stress reaction, and adjustment disorders (18). The risk started to increase already during the year before cancer diagnosis and peaked immediately after cancer diagnosis. However, to the best of our knowledge, no studies have specifically examined the association of stress-related disorders, especially the ones diagnosed immediately before and after a cancer diagnosis, with cancer-specific mortality among patients with cervical cancer.

Psychologic distress is challenging to measure. We a priori defined a range of common indicators of severe psychologic distress that are possible to ascertain in Swedish population and health registers, including clinical diagnoses of stress-related psychiatric disorders and stressful life events. We aimed to examine the potential influence of these indicators along the disease course, particularly around the time of cancer diagnosis, on the cancer-specific mortality of patients with cervical cancer, after careful adjustment for multiple sociodemographic and clinical characteristics of these patients.

As a part of the nationwide Case-Control Audit of Cervical Cancer (19), we utilized the national Swedish Cancer Register (20) to identify all patients (N = 4,533) diagnosed with an invasive cervical cancer from January 1, 2002 through December 31, 2011 in Sweden. Clinical data from all the identified patients were thoroughly reviewed by a senior gynecologist (B. Andrae), who collected and verified information through original medical charts deposited in the clinical departments where patients with cancer had been diagnosed and treated. After this review, we excluded 169 cases of cancers of noncervical origin, 36 cases of nonepithelial origin, 62 cases of noninvasive form, 12 recurrent cases, 7 cases detected on autopsy, and 2 cases with erroneous death records, leaving 4,245 patients in the final cohort. Information on clinical characteristics, including the International Federation of Gynecology and Obstetrics (FIGO) stage, histology, mode of detection, primary treatment, and adjuvant therapy, was also collected from the medical charts.

Using the unique Swedish personal identification numbers, we cross-linked the cohort to the Total Population Register, Education Register, and Longitudinal Integration Database for Health Insurance and Labor Market Studies (LISA by Swedish acronym) and obtained information on demographic characteristics of all patients, such as country of birth, marital status, and educational level. We also linked all patients to the Swedish Patient Register, which has collected nationwide information on hospital discharge records since 1987 and on >80% of hospital-based outpatient specialist visits since 2001 (21). On the basis of this, we calculated the Charlson's comorbidity index (excluding benign and malignant tumors; ref.22) at the time of cancer diagnosis, using the hospital-based care records from 1987 onward. Through record-linkage to the Migration Register and Causes of Death Register, we followed all patients from the date of cancer diagnosis until emigration out of Sweden, death, or December 31, 2012, whichever occurred first.

We specifically studied a composite of three psychiatric disorders, which are relatively common among patients with cervical cancer (23) and potentially related to psychologic distress (18, 24, 25). These included stress reaction and adjustment disorders (the 10th Swedish revision of the International Classification of Disease code [ICD-10]: F43), depression (F32-F33), and anxiety disorder (F40-F41), which are hereafter collectively referred to as stress-related disorders. In contrast to our previous work (18), substance abuse and somatoform/conversion disorders were not included in our current study, because they were very uncommon among patients with cervical cancer. As we aimed to use these disorders as a proxy for severe psychologic distress experienced in relation to the diagnosis and treatment of, or living with, cervical cancer (18), we used newly diagnosed disorders from one year before cancer diagnosis onward as the exposure; with a particular focus on disorders diagnosed within one year before, or after, cancer diagnosis. We identified the clinical diagnoses of these disorders through inpatient or outpatient hospital visits as recorded in the Patient Register, and used the first date of each hospital visit as the date of diagnosis.

We furthermore studied four types of stressful life events that are both relatively common in the population (26) and known as highly stressful (27–30), including loss of a family member due to death, severe illness of a family member, divorce, and being between jobs.

We identified the parents, spouses, children, and siblings of all patients through the Multi-Generation Register, which contains largely complete familial linkages for all Swedish residents born in 1932 and onward (31). Spouses were identified through a registered common child. We then linked all relatives to the Causes of Death Register to ascertain deaths among, and to the Patient Register to identify severe illnesses of the family members. The selection of severe illnesses was made a priori according to the Global Disease Burden 2004 (32); only illnesses that are relatively common and have a disability weight of ≥0.40 were chosen (see Supplementary Table S1 for ICD codes). Disability weight is the magnitude of health loss related to the illness and ranged from 0 to 1, with 0 implying a state that is equivalent to full health and 1 a state equivalent to death. Divorce was defined as a change of marital status from “married” to “unmarried,” according to the annual marital status recorded in LISA. Between jobs was defined as “actively seeking a job” as recorded in LISA. Among patients with a status of “between jobs,” 98% were previously employed and 91% had a previous job during the preceding two years. We did not study “unemployment” status because not being employed in itself might merely indicate a patient's own choice in relation to poor health.

We studied any first occurrence of these stressful life events from one year before cancer diagnosis onward, as recent events might better reflect a potential emotional burden among the patients (33). In the analysis of severe illness of a family member, we only considered newly diagnosed illnesses of the family members.

All deaths occurring among the patients during follow-up was ascertained from the Causes of Death Register, where we identified both the date and underlying cause of death. Our primary outcome was cancer-specific mortality, defined as having cervical cancer (ICD-10: C53) or unspecified uterine cancers (C55) listed as the underlying cause of death (34). A validation study showed that, in this Register, the agreement between death and medical records was 92% for cancer as an underlying cause of death (35). Overall mortality was studied as a secondary outcome.

We regarded stress-related disorders and stressful life events as two separate indicators of severe psychologic distress and defined them in a time-varying manner. Patients with either of the indicators during the year before cancer diagnosis were classified as exposed from the date of cancer diagnosis. Patients that were registered with an indicator after cancer diagnosis were first classified as unexposed, and then as exposed from the date of indicator occurrence. Patients without any indicator at any time during one year before cancer diagnosis or thereafter were classified as unexposed throughout.

Psychiatric disorders and stressful life events before cancer diagnosis might lead to delayed diagnosis or differential treatment modes (36, 37), which might partially explain the association between these exposures and mortality. Using binary or multinomial logistic regression models, we, therefore, compared patients with a stress-related disorder or life event during the year before cancer diagnosis to other patients regarding different clinical characteristics, including FIGO stage, histologic type, mode of detection, primary treatment, and adjuvant therapy.

We calculated the crude cancer-specific mortality rates (per 1,000 person-years) among the exposed and unexposed (reference) groups. We then derived HRs of cancer-specific mortality with 95% confidence intervals (CI) from the Cox proportional hazards model, comparing the exposed to the unexposed reference group. The proportional hazards assumption was found to hold for all variables when employing the Schoenfeld residual test. We also examined the associations for any diagnosis of stress-related disorders, any occurrence of stressful life events, as well as for individual psychiatric disorder and individual life event, in separate models. To relieve the concern of including unspecified uterine cancer in the definition of cancer-specific mortality, we performed an additional analysis by excluding deaths with an underlying cause of death coded as unspecified uterine cancers. We then repeated these analyses for overall mortality as a secondary outcome.

There is a greater risk increase of stress-related disorders immediately before and after a cancer diagnosis, compared with thereafter (18), and a varying effect of psychologic distress through the cancer disease continuum has been suggested (7). With a special interest in the one year before or after cancer diagnosis, we examined the studied associations by the timing of the occurrence of stress-related disorders and stressful life events.

To shed light on the potential impact of different characteristics of the stressful life events on the studied associations, we further examined loss of a family member due to death by cause of death (i.e., natural or unnatural cause of death) and severity of the illness of a family member by dichotomizing the disability weights (0.4–0.5 or ≥0.5).

The prevalences of stress-related disorders and stressful life events were different among patients with differing educational level achievement (34.7% among women with primary school education, 47.0% among women with high school education, and 42.9% among women with college and above education). Also, women with lower socioeconomic status may have worse cancer prognosis compared with other women (5). In our data, an increasing level of educational achievement was associated with decreasing cancer-specific mortality (105.9 per 1,000 person-years among women with primary school education, 44.5 among women with high school education, and 30.1 among women with college and above education). We, therefore, considered educational achievement as a confounder for the studied association. In Model A, we adjusted for age at cancer diagnosis, calendar year at diagnosis, highest educational level, country of birth, and number of family members at diagnosis. In Model B, we additionally adjusted for Charlson's comorbidity index, FIGO stage, histology, mode of detection, and treatment modes.

Women with lower socioeconomic status (38) or less social support (39) may furthermore be more vulnerable to psychologic distress, while clinical characteristics are impactful on prognosis. We, therefore, performed stratified analyses by the demographic and clinical characteristics of patients with cervical cancer, using Model B. An interaction term between exposure and the stratification variable was added to the model, and the P value for the interaction term was reported as a significance test for the difference between HRs.

We performed all analyses in STATA 14.2 (StataCorp LP). P < 0.05 indicated statistical significance. This study was approved by the Regional Ethical Review Board in Stockholm, which determined that, due to the population-based nature of the study, written informed consent from the study participants was not required (Dnr 2015/719-32; Dnr 2011/1026-31/4; Dnr 2011/921-32; Dnr 02-556). All investigations were conducted according to the principles in the Declaration of Helsinki.

A total of 1,797 patients (42.3%) were defined as exposed to either stress-related disorders (9.2%) or stressful life events (37.4%) from the year before cancer diagnosis onward (Table 1), of which 860 (48%) occurred within one year before or after diagnosis. The mean age at diagnosis was 53.9 years. The patients were commonly diagnosed at stage IB (39.7%), with tumors of squamous cell origin (72.6%), and were commonly symptomatic (70.8%) at the time of diagnosis. In total, 56.4% of patients underwent surgery, whereas 51.2% and 31.8% of the patients received radiotherapy and chemotherapy, respectively. During the follow-up (mean ± SD = 4.4 ± 3.1 years), 1,392 (32.8%) patients died, 1,005 (72.2%) of which died due to cervical cancer specifically.

Patients with indications of psychologic distress (either a stress-related disorder or a stressful life event) during the year before diagnosis did not differ substantially from patients without such indications, in terms of FIGO stage, histologic type, or cancer detection pattern (Table 2). Patients with a stress-related disorder during the year before cancer diagnosis were, however, more likely to receive palliative treatment compared with other patients.

Patients with indications of psychologic distress (either a stress-related disorder or a stressful life event) from one year before cancer diagnosis onward had an increased risk of cancer-specific mortality (HR 1.33; 95% CI, 1.14–1.54), compared with unexposed patients (Fig. 1). After further adjustment for clinical characteristics, the association was slightly attenuated but remained statistically significant (HR 1.26; 95% CI, 1.08–1.46). The risk was particularly elevated in those exposed to stress-related disorders (HR 1.55; 95% CI, 1.20–1.99), and was also elevated in those exposed to stressful life events (HR 1.20; 95% CI, 1.02–1.41). Such association was noted for all individual stress-related disorders and life events, although some associations were not statistically significant. Similar estimates were noted after excluding deaths with an underlying cause of death coded as unspecified uterine cancers from the definition of cause-specific death (5.5% of all cancer-specific deaths; Supplementary Fig. S1). Similar but somewhat weaker associations were found for overall mortality (Supplementary Fig. S2).

Stress-related disorders or stressful life events experienced within one year before or after diagnosis (HR 1.30; 95% CI, 1.11–1.52), but not thereafter (HR 1.12; 95% CI, 0.84 – 1.49), were associated with a higher risk of cancer-specific mortality (Fig. 2; Supplementary Table S2). Stress-related disorders were associated with an increased risk of cancer-specific mortality regardless of when the disorder was diagnosed, whereas stressful life events were associated with a higher risk only when they occurred within the year before or after cancer diagnosis. A similar pattern was noted for overall mortality (Supplementary Fig. S3 and Table S2).

A greater HR of cancer-specific mortality was suggested in those with loss of a family member due to unnatural causes, compared with that in those with loss due to natural causes (Table 3). Having been exposed to severe illness of a family member with a disability weight of over, but not below, 0.5 was clearly associated with an increased risk of cancer-specific mortality. A similar pattern was observed for overall mortality (Supplementary Table S3).

The associations of stress-related disorders or stressful life events with cancer-specific mortality were not clearly modified by demographic or clinical characteristics (Supplementary Table S4).

In this nationwide cohort study of women with cervical cancer, we found that psychologic distress, defined as either a clinical diagnosis of stress-related disorders or an occurrence of stressful life events around the time of cancer diagnosis, was associated with an increased risk of cervical cancer-specific mortality. For the first time, we showed that these associations were independent of known prognostic indicators of cervical cancer, including tumor characteristics, mode of diagnosis, and treatment.

Stress-related disorders are common among patients with cancer, both after cancer diagnosis (24, 40, 41) and during the year before cancer diagnosis (18). Except for depression, few studies have so far specifically examined these disorders in relation to cancer-specific mortality among patients with cancer (42, 43). Our findings show that clinical diagnoses of depression, anxiety, and stress reaction and adjustment disorders were not uncommon among patients with cervical cancer (9.2%). To our knowledge, no other study has reported the combined prevalence of clinically diagnosed stress-related disorders among patients with cervical cancer. A previous study from South Korea used administrative databases to ascertain the diagnosis of depression among patients with cervical cancer and reported a prevalence of 7.9% (44). The slightly lower prevalence of depression in this study (4.9%) is likely because only clinical diagnoses made by specialists or during in-patient care were included in our study. Stress-related disorders, either collectively or individually, were associated with 66%–139% excess risks of cancer-specific mortality. Most of these associations were attenuated to some extent but remained statistically significant after further adjustment for tumor characteristics, diagnostic mode, and treatment modality, lending support to the hypothesis that the influence of psychologic distress on cervical cancer survival is independent of other clinical factors. Stress-related disorders diagnosed immediately before or after cancer diagnosis might be a proxy of a severe stress response to clinical workup, the final cancer diagnosis (18), and a dramatically impaired quality of life after a diagnosis of cervical cancer (45). This subgroup of vulnerable patients could be detected during this specific time window.

Stressful life events have also been linked to increased mortality among patients with cancer (8). Previous studies have examined the association of individual events, for example, bereavement and divorce (13–15), with cervical cancer survival. Here, we found that four major life events, including bereavement, severe illness of a family member, divorce, and being between jobs, were prevalent among patients with cervical cancer (37.4%). While the collective prevalence of these life events has not been documented before, the prevalence of individual events has been reported among patients with cervical cancer. For example, divorce or separation was reported to have a prevalence of 15%–22% (14, 16), whereas the prevalence of bereavement (loss of a child or spouse due to death) was reported as 2%–14% (13, 14, 16), among patients with cervical cancer. However, these numbers cannot be compared directly with ours, because we ascertained these events around the time of cancer diagnosis and thereafter, whereas the other studies usually measured one individual event before cancer diagnosis. Collectively, these life events were associated with a 20% elevated risk of cancer-specific mortality. These results on bereavement and divorce are in line with findings of previous Nordic studies (13–15).

Experimental studies suggest that psychologic distress modulates tumor growth and progression through the dysregulation of oncogenes, angiogenesis, inflammation, and host immune responses operated by the hypothalamic–pituitary–adrenal axis and sympathetic nervous system (7). For instance, the adrenergic signaling pathway activated by stress may upregulate VEGF to increase angiogenesis around the tumor and subsequently facilitate tumor growth (46). Chronic stress may also reduce T-cell–proliferative response through the adrenergic and glucocorticoid receptors on immune cells, leading to decreased cellular immunity and immunosurveillance (11). In addition to such direct modulation, psychologic distress may also influence cancer survival through delayed diagnosis or altered treatment (36, 37). Few studies have disentangled the contributions of differential tumor characteristics, diagnostic mode, and cancer treatment from the link of psychologic distress to cancer-specific mortality (36, 47). For instance, the increased cancer-specific mortality reported among patients with gynecologic cancer with preexisting psychiatric disorders was attributed to a lower rate of surgery and radiotherapy (36). This is partly supported by our results, where patients with a stress-related disorder during the year before cancer diagnosis appeared more likely to receive palliative treatment. However, the association of stress-related disorders or life events with an increased risk of cancer-specific mortality remained robust after extensive adjustment for tumor characteristics, and mode of diagnosis/treatment.

Stressful life events occurring during the early course (i.e., the year before and the year after cancer diagnosis) were associated with a greater risk increase of cancer-specific mortality, compared with life events occurring thereafter. Together with the slightly stronger association noted for localized cancer (IB stage) compared with higher stage cancer, this finding suggests that psychologic distress plays a more significant role in cancer-specific mortality among patients with early-stage and localized tumors. This is also supported by accumulating experimental evidence indicating that psychologic distress might specifically facilitate the spread of tumor cells from the local site through surrounding blood and lymph vessels (7, 48). The temporal pattern of the associations between stress-related disorders and cancer-specific mortality was even more complicated. Although similar magnitude of associations was found for stress-related disorders diagnosed during the year before and the year after cancer diagnosis, as for stressful life events, a stronger association was noted for stress-related disorders occurring more than one year after diagnosis. The latter finding might be partly explained by the fact that stress-related disorders occurring during the later disease course of cancer might represent other aspects of disease characteristics (e.g., failure of the primary treatment and disease relapse), suggesting potential reverse causality.

One major merit of our study is the large-scale nationwide cohort design with complete follow-up, as well as prospective and independent nature of data collection, largely limiting the common sources of bias. The rich information on clinical characteristics allowed us to disentangle the direct influence of psychologic distress on cancer-specific mortality, from the influence through tumor characteristics, diagnosis, and cancer treatment.

The chosen proxies for severe psychologic distress might have only partially illustrated the psychologic distress a patient with cervical cancer is likely to experience. Because we used specialist-care–based in-patient and outpatient diagnoses to define stress-related disorders, we may have missed less symptomatic patients that were not attended by specialists, as well as individuals with psychiatric symptoms that had not sought care at all. Also, patients with other stressful life events such as financial hardship due to sickness absence were classified as unexposed. However, such misclassifications most likely would have led to an attenuated magnitude of the studied associations. Our study did not address earlier stressful exposures during the life course, which may influence cervical cancer survival. However, we chose to focus on distress around cancer diagnosis as such has the most direct implications for oncologic practice through, for example, early surveillance and intervention.

While our previous work (49) suggested a role of psychologic distress in the risk of oncogenic infection of human papillomavirus, cervical dysplasia, and invasive cervical cancer, these findings further indicate a role of psychologic distress in cervical cancer–specific mortality. If verified in independent populations, these findings may lend support to a specific and continuous role of psychologic distress in the etiology/pathogenesis of cervical cancer (49). Furthermore, given that 42% of patients with cervical cancer experienced a stress-related disorder or stressful life event along the disease course, psychologic intervention may be considered as an integral component in cervical cancer care.

In conclusion, stress-related disorders and stressful life events among patients with cervical cancer were associated with an increased risk of cancer-specific mortality. These associations were independent of tumor characteristics, diagnostic mode, and treatment modality.

No potential conflicts of interest were disclosed.

Conception and design: D. Lu, U. Valdimarsdóttir, F. Fang

Development of methodology: D. Lu, U. Valdimarsdóttir, F. Fang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Andrae, U. Valdimarsdóttir, K. Sundström, P. Sparén

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): D. Lu, U. Valdimarsdóttir, K. Sundström, K. Fall, P. Sparén, F. Fang

Writing, review, and/or revision of the manuscript: D. Lu, B. Andrae, U. Valdimarsdóttir, K. Sundström, K. Fall, P. Sparén, F. Fang

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): F. Fang

Study supervision: U. Valdimarsdóttir, P. Sparén, F. Fang

We thank Dr. Walter Ryd, University of Gothenburg, for pathologic review. We also acknowledge the contributions to data collection and management on this project of Sara Nordqvist Kleppe, Pouran Almstedt, and the late Peter Olausson. This work is supported by the Swedish Cancer Society (grant no. CAN 2017/322), Swedish Research Council for Health, Working Life, and Welfare (grant no. 2017-00531), and the Karolinska Institutet (Stockholm, Sweden; Partial Financing for New Doctoral Student, Senior Researcher Award, and Strategic Research Program in Epidemiology Young Investigator Award).

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.