A systematic review on the prevalence of symptoms of depression, anxiety and distress in long‐term cancer survivors: Implications for primary care

Abstract Introduction Symptoms of depression, anxiety and distress are common in the first years after a cancer diagnosis, but little is known about the prevalence of these symptoms at the long term. The aim of this review was to describe the prevalence of symptoms of depression, anxiety and distress in long‐term cancer survivors, five or more years after diagnosis, and to provide implications for primary care. Methods We performed a systematic literature search in the PubMed, PsycINFO and CINAHL databases. Studies were eligible when reporting on the prevalence of symptoms of depression, anxiety and/or distress in long‐term cancer survivors (≥5 years after diagnosis), treated with curative intent. Results A total of 20 studies were included. The reported prevalence of depressive symptoms (N = 18) varied from 5.4% to 49.0% (pooled prevalence: 21.0%). For anxiety (N = 7), the prevalence ranged from 3.4% to 43.0% (pooled prevalence: 21.0%). For distress (N = 4), the prevalence ranged from 4.3% to 11.6% (pooled prevalence: 7.0%). Conclusion Prevalences of symptoms of depression, anxiety and distress among long‐term survivors of cancer do not fundamentally differ from the general population. This is reassuring for primary care physicians, as they frequently act as the primary physician for long‐term survivors whose follow‐up schedules in the hospital have been completed.

Survivors may face various physical and psychosocial sequelae entailed by their illness or treatment. These sequelae can be severe, debilitating and sometimes even permanent. Frequently reported symptoms are fatigue, cognitive impairment, sexual dysfunction, depression, anxiety and distress (Bloom, 2002;Deimling, Bowman, Sterns, Wagner, & Kahana, 2006;Yi & Syrjala, 2017). Yet, psychological problems, such as depression and anxiety, are often under-diagnosed and under-treated among patients with cancer, and need early identification (Walker, 2014).
Until now, most studies focusing on psychological problems in cancer survivors have explored their prevalence mainly within the first years after diagnosis, when patients are still in follow-up at the hospital (Stanton, 2006). For example, Krebber et al. (2014) found, in their meta-analysis, a pooled prevalence of depressive disorder among cancer patients of 14% during treatment, 9% in the first year posttreatment and 8% one year or more post-treatment (Krebber et al., 2014). Moreover, Watts, Prescott, Mason, McLeod, and Lewith (2015) reported in their systematic review pre-treatment, on-treatment and post-treatment anxiety prevalence rates in ovarian cancer patients to be 19%, 26% and 27% respectively (Watts et al., 2015). Conversely, long-term prevalence of (symptoms of) depression, anxiety and/or distress beyond the first 5 years, in cancer survivors with a range of tumour types, has received limited attention so far. Nevertheless, some specific tumour types have been studied already, for example, Maass, Roorda, Berendsen, Verhaak, and Bock (2015) reported a higher prevalence of symptoms of depression among long-term (>2 years) breast cancer survivors, compared to the general female population (Maass et al., 2015). However, none of the reviews conducted so far focused solely on survivors of cancer at least 5 years after diagnosis.
Focusing on the period beyond the first years after diagnosis is important, since most cancer follow-up schedules in hospitals discontinue after 5 years. This 5-year mark is consistent with the view commonly held by the general population, that those who have survived for 5 years have a relatively high probability of "being cured" (Deimling, 2007). In countries such as the Netherlands, in which the general practitioners (GPs) act as gatekeepers in the healthcare system, the GP typically is the primary physician for long-term cancer survivors after the follow-up period. For these GPs, it is important to know how prevalent psychological symptoms are among long-term cancer survivors, in order to be able to provide appropriate or proactive care.
Thus, the aim of this systematic review was to describe the prevalence of symptoms of depression, anxiety and/or distress in longterm cancer survivors, five or more years after diagnosis, and to provide an overview of implications, related to these psychological symptoms in long-term cancer survivors, for primary care.

| Protocol registration and report
The protocol of this review is available at PROSPERO, the international database of prospectively registered systematic reviews for health and social care (registration number CRD42018110822). The Preferred Reporting Items for Systematic Reviews (PRISMA statement) was used as a formal guideline for systematic reviews (Moher, Liberati, Tetzlaff, Altman, & PRISMA Group, 2010).

| Search strategy
A systematic search for publications in English has been conducted in the electronic databases PubMed (MEDLINE), PsycINFO (Ovid) and CINAHL (EBSCO), from 2000 to September 30, 2018. Studies were identified using a search syntax based on the PubMed strategy, which uses a combination of MeSH terms and free text terms, and included synonyms of terms related to cancer, survivor, long term, depression, anxiety, distress and prevalence. Where necessary, the syntax was adapted for use in the other databases. The PubMed search syntax can be found in the Appendix.

| Study selection
To assess whether identified studies met the selection criteria, they were independently screened on title and abstract by two authors (SFAD and MJS). Full-text articles were retrieved when there was not sufficient information to establish appropriateness for inclusion.
A manual search of reference lists of selected articles and relevant systematic reviews has been performed to identify further relevant studies. Studies were excluded for the following reasons: (a) no prevalence data; (b) no long-term cancer survivors (≥5 years after diagnosis); (c) no data on symptoms of depression, anxiety and/or distress; (d) no cancer; (e) no adult (≥18 years old) at time of diagnosis; (f) no treatment with curative intent; and/or (g) other (e.g., full text not available, design paper) ( Figure 1). In case of disagreement during the selection process, a third author (DB) decided upon the eligibility of a study.

| Data extraction and synthesis
A data extraction form was developed to record relevant study details. Data were independently extracted by two authors (OPG and MES) and included the following: (a) general study characteristics (e.g., author, year of publication, country), (b) study characteristics (e.g., recruitment period, design, setting), (c) participant characteristics (e.g., age, gender, number, tumour type, time since diagnosis, treatment) and (d) symptom characteristics (e.g., measurement, prevalence [of symptoms of depression, anxiety and/or distress]).
We defined symptoms of anxiety, depression and distress as follows: these have to be measured using a generic questionnaire or subscale of a questionnaire to assess symptoms of depression, anxiety or distress. The latter, distress, is typically defined by the presence of an adjustment disorder with or without depression/anxiety. All data were synthesised by describing the prevalence of symptoms of depression, anxiety and/or distress in adults with cancer assessed at least 5 years after diagnosis. Also, pooled point prevalences and 95% confidence intervals (CIs) were produced in STATA (version StataSE 15) and presented in forest plots. For this, we used the cut-off values of >16 for the CES-D, >8 for HADS-A and HADS-D, and values suggestive of possible depression, anxiety or distress for the other questionnaires.

| Quality assessment
The methodological quality of the included studies was scored independently by two authors (DB and CH), by applying the 14 items of the NIH's quality assessment tool for Observational Cohort and Cross-Sectional Studies (National Heart, Lung, & Blood Institute, 2014). We slightly adapted the summary score to be applicable for our review question. That is, we scored studies as "excellent" if they scored "Yes" on the following four questions: (a) Was the study population clearly specified and defined?; (b) Was the participation rate of eligible persons at least 50%?; (c) Were the exposure measures (independent variables) clearly defined, valid, reliable, and implemented consistently across all study participants?; and (d) Were the outcome measures (dependent variables) clearly defined, valid, reliable, and implemented consistently across all study participants?
Exposure was defined as a diagnosis of cancer; outcome was defined as symptoms of depression, anxiety or distress. Studies received the result "excellent" if they scored positive on all four questions, "good" if they scored positive on three out of these four questions, "fair" if they scored positive on two questions and "poor" in case of less than two positive questions. In case of disagreement, items were discussed until consensus was reached, or if necessary, a third author (SFAD) was consulted.

| Search results and characteristics of included studies
Our original search yielded 644 titles, 392 of which remained after the removal of duplicates. Of these titles, 22 met the criteria for a full-text review, of which 20 studies were subsequently included ( Figure 1). Agreement between researchers was 95.7% for title/abstract screening, and full consensus was reached after discussion. The third author was consulted in four cases during full-text selection.
Of the 18 studies detailing on the prevalence of depressive symp- N of specific measurement is given and may therefore differ from the overall number as presented in Table 1. b Prevalence for total group of cancer survivors, not specified per tumour type.

TA B L E 2 (Continued)
A total of seven studies specifically reported on symptoms of anxiety. Most of these studies used the HADS anxiety subscale (HADS-A) (N = 4). Other used measures were the BSI anxiety subscale or a self-developed questionnaire. The reported prevalence of anxiety ranged from 3.4% to 43.0%. The pooled prevalence of patients with symptoms of anxiety was estimated to be 21.0%.
In total, four studies reported on distress. Most studies used the Kessler Psychological Distress Scale (K6-scale) to measure symptoms of distress (N = 2). Other measures used were the HADS total score and the BSI total score. The prevalence of reported distress ranged from 4.3% to 11.6%. The mean reported percentage of patients with distress symptoms was estimated to be 7.0%.
Further details in regard to symptoms of depression, anxiety and distress, and the created forest plots can be found in Table 2 and Figure 2 respectively.

| Quality assessment
The summary scores of the quality assessment were "excellent" for half of the included studies (N = 9), nine of the included studies were of "good" quality and two studies were scored as having "fair" quality. The shortcomings mostly identified were (a) a participation rate of <50% (N = 6) and (b) lack of clarity in the description of the study population (N = 3). See Figure 3 for more details.

| Main findings
In this systematic review, we reported on the prevalence of symptoms of depression, anxiety and distress in cancer survivors, five or more years after diagnosis. The pooled prevalence of symptoms of depression and anxiety was 21.0%; the pooled prevalence of distress was 7.0%. Most frequently used instruments to measure these psychological symptoms were the HADS and the CES-D.

| Interpretation of the findings
Our results suggest that the prevalence of symptoms of depression, anxiety and distress among long-term survivors is comparable with, or even slightly below, the prevalence of these symptoms in the general population.
Looking more specifically at symptoms of depression, measured with the HADS-D, prevalence rates in general and elderly F I G U R E 2 Forest plots of prevalence of symptoms of depression, anxiety and distress populations between 10.0% and 23.0% have previously been reported (Djukanovic, Carlsson, & Årestedt, 2017;Hinz & Brahler, 2011). In our study, in long-term cancer survivors, the pooled prevalence of symptoms of depression, measured with this same questionnaire, was quite low, that is, only 10.0% (8.0%-13.0%). Measurement of symptoms of depression in the general population, using the CES-D, showed a prevalence of 21.0% (Smarr & Keefer, 2011), which is a bit lower compared to the 27.0% (9.0%-44.0%) in our study. The prevalence of symptoms of anxiety in the general population, according to the HADS-A, has been found to be between 10.0% and 21.0% (Djukanovic et al., 2017;Hinz & Brahler, 2011). The pooled prevalence of 20.0% (19.0%-22.0%) found in our study in cancer survivors is on the high end of this range.
These relatively low prevalence rates in long-term cancer survivors are in contrast to prevalence rates found among cancer patients during and shortly after diagnosis and treatment, when higher prevalence rates of psychological symptoms are found (Maass et al., 2015;Mitchell, Ferguson, Gill, Paul, & Symonds, 2013;Watts et al., 2015). Three aspects might explain this difference: (a) former studies did not focus specifically on the period beyond the first 5 years after cancer diagnosis, as we did. Earlier research showed that adults who have survived cancer for at least 5 years frequently identify themselves as cancer survivors and/or as expatients, rather than as victims or patients (Deimling, Bowman, & Wagner, 2007). So, the longer time since diagnosis, and potential regained trust in one's health, might positively influence psychological functioning of these survivors; (b) cancer patients who are suffering from symptoms of depression, anxiety and/or distress might be less inclined to participate or continue participation in studies regarding psychological sequelae, and herewith be missing in our prevalence data. In line with this, an earlier meta-analysis showed that higher levels of depressive symptoms predict higher mortality rates (Pinquart & Duberstein, 2010). It is therefore possible that de- reported as well (Sharpley et al., 2017). Yet, they used the Zung Selfrating Depression Scale, with a cut-off of ≥40 (range 20-80), known to lead to high numbers of false-positive classifications (Yesavage et al., 1982). On the other hand, in the study of Chambers et al. (2012), among survivors of colorectal cancer, a very low (6.0%) prevalence of symptoms of depression was reported (Chambers et al., 2012).
The prevalence rate in this study was measured among survivors in a longitudinal study, of whom about 40.0% were enrolled 60 months F I G U R E 3 Risk of bias after diagnosis. Therefore, the low prevalence could potentially be explained by healthy survivor bias.
Studies using the HADS depression and anxiety scales showed a high agreement in our review. However, there was some variability among studies using the CES-D. Interestingly, the difference among studies using the CES-D was largest for two specific studies, apparently reporting on the same patient population (Greenwald & McCorkle, 2008;McCorkle et al., 2006). We were unable to hypothesise an explanation for the difference in these studies, except that there was an age difference in the samples. The slightly older population in the study of McCorkle et al. (2006) reported a lower prevalence though, which is contradictory to most literature suggesting an increased prevalence of depressive symptoms with increasing age (Djukanovic et al., 2017). In order to enhance comparability and interpretation in future research into prevalence rates, we suggest using validated and widely used measures to enhance comparability and interpretation.

| Strengths and weaknesses
A major strength of our review is that we only included studies that presented data on the prevalence of symptoms of depression, anxiety and distress in cancer survivors, five or more years after their cancer diagnosis. This enabled us to provide valuable new insights into these symptoms, explicitly in long-term cancer survivors, most of whom are no longer being followed in routine hospital follow-up.
Focusing merely on symptoms and excluding clinical diagnoses of depression and anxiety disorders is another strength of our study. It gives a more accurate estimate of the symptoms' prevalences, which is important since clinically diagnosed patients often have specific healthcare needs and treatments. Furthermore, focusing on symptoms only is essential in case of early detection and early intervention programmes. Finally, our study presented pooled prevalences of symptoms of depression, anxiety and distress, using different questionnaires, enabling future comparisons.
Nevertheless, we also identified several weaknesses in our systematic review. First, since we included studies describing a variety of cancer types, we presented data on a heterogeneous population of cancer survivors. This hampers description of the prevalence for specific patients' groups. Due to the amount of studies found among long-term survivors, and because a fair proportion of included studies report on populations with multiple cancer types, subgroup analyses for cancer types were not possible. Second, most of the studies we included were not designed to study the prevalence of symptoms of depression, anxiety and/ or distress. Rather, these were mostly secondary outcomes in studies, designed to assess relationships between exposures and outcomes among cancer patients. Yet, due to the total number of patients included (N = 17,726), we believe our results to hold quite some value. Lastly, the quality assessment tool used in the current study was not specifically designed for prevalence studies.
However, the items used to assess overall quality were adapted to our review question.

| Implications for (primary) care and conclusion
Our results suggest that the prevalence rates of symptoms of depression, anxiety and distress among long-term cancer survivors do not fundamentally differ from the general population. Most hospitals have follow-up schedules for patients after the treatment for cancer, lasting up to 5 years after treatment. Hereafter, in healthcare systems with a gatekeeping function for GPs, these physicians function as the primary physician for long-term cancer survivors. Earlier research showed that patients with cancer frequently consult their primary care physician for psychosocial issues, starting in the first years after diagnosis (Brandenbarg et al., 2017;Roorda, Berendsen, Groenhof, van der Meer, & de Bock, 2013). Apart from providing or referring to psychological care at the short term, GPs might reassure cancer survivors that psychosocial sequelae are most prevalent in the first years and likely to decline over time. Based on our findings, there seems to be no need for primary care physician or other (primary) healthcare providers to actively screen all long-term cancer survivors for symptoms of depression, anxiety and/or distress.

ACK N OWLED G EM ENT
We would like to thank Truus van Ittersum for her help with formulating the search string and the actual literature search.

CO N FLI C T O F I NTE R E S T
No conflicts to declare.