Trajectories of Cognitive Symptoms in Sick-Listed Cancer Survivors

Simple Summary The effects of cognitive symptoms on the ability to work are of major concern for cancer survivors. The aim of our study was to explore trajectories of cognitive functioning in sick-listed cancer survivors with work capacity. We found that cognitive functioning improved between two and four years after first day of sick leave, although cognitive symptoms remained of clinical concern in cancer survivors who are non-durable work-disabled (partly or fully). This underlines the importance to provide partly and fully, non-durable work disabled cancer survivors with evidence-based treatment options for their self-perceived cognitive symptoms. Abstract Many non-central nervous system (CNS) cancer survivors experience cognitive symptoms, which may affect their self-perceived work ability. Little is known about trajectories of self-perceived cognitive functioning in cancer survivors in the period after work disability assessment. Therefore, we evaluated: (1) trajectories of self-reported cognitive functioning, in cancer survivors with work capacity, (2) differences in trajectories of self-reported cognitive functioning between three work disability groups, and (3) explanatory factors of trajectories of self-reported cognitive functioning. Participants (n = 206) were assessed on self-reported cognitive functioning at three time points between two and four years after first day of sick leave. A statistically significant improvement in cognitive functioning was found in the total group (β = 4.62, SE = 0.91, p < 0.001). When comparing cancer survivors in different work disability groups, similar trajectories of cognitive functioning were observed. Fatigue was the only factor found to be associated with the reported trajectory (β = −0.23, SE = 0.086, p = 0.08). Self-perceived cognitive functioning scores remained considerably lower than the mean score of the general Dutch population, indicating that cognitive symptoms are a persistent problem in sick-listed cancer survivors and that evidence-based treatment options are warranted.


Introduction
The past decades have seen rapid advances in cancer treatment, resulting in improved survival rates [1]. About 40-50% of cancer survivors are diagnosed at a working age [2]. Long term effects of cancer and cancer treatment on the ability to work are of major concern for cancer survivors and their family [3]. About 89% (range 84-94%) of cancer survivors, who were employed at the time of diagnosis, is able to (partly) return to work (RTW) within 2 years after diagnosis [4]. However, previous studies showed that cancer survivors have less working hours, higher unemployment rates, and higher work disability rates than their healthy counterparts [5]. Therefore, exploring and understanding factors associated factors of trajectories of self-reported cognitive functioning. Results of this study may help to identify sick-listed cancer survivors who may benefit from cognitive rehabilitation interventions to cope with cognitive symptoms, and therefore sustain and/or improve their work ability.

Study Sample and Procedure
We used data from a prospective cohort study. The aim of this prospective cohort study was to identify prognostic factors associated with work disability in sick-listed cancer survivors. Patients who approached the maximum term of 2 years sick leave and applied for a work disability benefit were selected from the SSA registries, and were followed for two years (i.e., up to four years after first day of sick leave). A detailed description of the initial prospective cohort study has been reported elsewhere [13,25].
For the current study, we included cancer survivors who: (1) had a confirmed diagnosis of non-CNS cancer at baseline, (2) had an employment contract at time of diagnosis, (3) were occupationally active (i.e., had a paid employment contract or were involved in therapeutic work (i.e., work that allows for a gradual and flexible build-up of workload and working hours)), and (4) were between 18 and 64 years of age when approaching two years of sick leave. Also, for the current study, a selection was made of cancer survivors with work capacity, based on their work disability assessment (i.e., cancer survivors who are: (1) able to work (i.e., less than 35% loss of former wages earned), (2) partly, nondurable work disabled (i.e., between 35 to 80% loss of former wages earned), or (3) fully, non-durable work disabled (i.e., between 80 to 100% loss of former wages earned)). Hence, those assessed as (4) fully, durable work disabled (i.e., between 80 to 100% loss of former wages earned; no change expected) were excluded. Cancer survivors were also excluded if they were still receiving chemotherapy and/or radiotherapy. Furthermore, in order to sustain homogeneity regarding work situation (e.g., employer security, return to work situation, vulnerability for work disability), cancer survivors who were self-employed, were working in a sheltered workplace (i.e., a workplace that employs workers with disabilities), or were registered as 'social security safety netters' (i.e., workers whose temporary employment contract ended during sick leave, temporary agency workers and workers who are unemployed [25]) were excluded.
At baseline (i.e., at two years after first day of sick leave (T0)), potentially eligible participants received a package, by post at their home address, that included an information leaflet, a baseline questionnaire including questions on sociodemographic characteristics, treatment-related characteristics, and patient-reported outcomes, as well as an informed consent form. Follow-up questionnaires on patient-reported outcomes were sent one year (T1), and two years (T2) after baseline.

Sociodemographic Characteristics
Sociodemographic characteristics, including age, gender and education, were obtained via questionnaire. A questionnaire was used to acquire information on general employment issues and on work accommodation, including employment sector, type of employment (fixed/temporary), years of work experience, and working hours and days per week according to employment contract.

Clinical and Treatment-Related Characteristics
Clinical information, including cancer site, month/year of diagnosis, received (and future) treatment(s), (i.e., surgery, radiation, chemotherapy, immunotherapy, targeted treatment, hormonal therapy, other medication), and recurrence(s) of the disease, were obtained via self-report.

Work Disability
The level of work disability, expressed as the first three categories based on wage loss in percentage of former wage, was retrieved from the SSA: (1) able to work (i.e., less than 35% loss of former wages earned), (2) partly, non-durable work disabled (i.e., between 35 to 80% loss of former wages earned), and (3) fully, non-durable work disabled (i.e., between 80 to 100% loss of former wages earned). The classification of work disability did not change over time.

Patient-Reported Outcomes Cognitive Functioning
Self-perceived cognitive functioning was measured using the Cognitive Functioning scale of the Dutch validated version of the EORTC QLQ C-30 [26]. This scale consists of two items assessing difficulty concentrating and memory problems. Items are scored on a 4-point scale ranging from 1 (not at all), to 4 (very much). The scores were reversed and linearly transformed to a scale ranging from 0 to 100. A higher score indicates a better cognitive functioning.

Fatigue, Depression and Coping
Fatigue was measured using the Functional Assessment of Chronic Illness therapy-Fatigue (FACIT-F) [27]. This questionnaire consists of 13 items. Items are scored on a 5-point scale, ranging from not at all to very much. A higher score indicates less fatigue (range 0-52).
Depression was measured using the 20 item Center for Epidemiologic Studies depression Scale (CES-D) [28]. Items are scored on a 4-points scale, ranging from none of the time to most of the time. A higher score indicates more depressive symptoms (range 0-60).
Coping was measured using the 47-item Utrecht Coping List (UCL) [29]. In total, the UCL consists of 7 subscales that measure different coping strategies: active tackling, seeking social support, palliative reacting, avoiding, passive reacting, reassuring thoughts and expression of emotions [30]. Items are scored on a 4-point scale, ranging from 'hardly ever' to 'very often'. A higher score on a subscale indicates an increased tendency using that specific coping strategy. Scoring ranges differ per coping strategy.

Statistical Analysis
Descriptive statistics were used to characterize the study sample. Scores on the patient-reported outcomes questionnaires were calculated according to published scoring algorithms. To evaluate the trajectory of cognitive functioning, we used a mixed modelling approach based on maximum likelihood estimation with random intercept and slope. This approach takes into account the within and between person variability, and deals adequately with missing data [31]. Little's Test was used to check the assumption for missing data completely at random [32]. We checked for the presence of a linear effect of time and added a quadratic term to determine whether an initial improvement or deterioration in cognitive functioning was followed by a deceleration of this effect. The choice between covariance structures (i.e., unstructured, compound symmetry, and autoregressive) and a model including only a linear effect of time and a model including both a linear and quadratic effect of time was based on model fit statistics: the Bayesian information criterion (BIC) and Akaike information criterion (AIC) [33,34].
Normative data for the Cognitive Functioning scale of the EORTC QLQ C-30 was used to interpret cognitive functioning scores, whereby the mean cognitive functioning score of the general Dutch population is 93 [35]. We also used a threshold for clinical importance, based on external criteria reflecting the clinical relevance of the problem (i.e., a score below 75 indicates clinically important cognitive symptoms) [36].
We determined whether trajectories of cognitive functioning scores differed between cancer survivors in the three included groups of work disability. Differences in change from baseline to follow-up between groups were accompanied by effect sizes (ES). ES were calculated based on the estimated marginal means. Effect sizes of 0.2 were considered small, 0.5 moderate, and 0.8 large. Effect sizes of ≥0.5 were considered to be clinically relevant [37].
Finally, we performed exploratory analysis in which we evaluated the following potential explanatory factors of cognitive functioning trajectories: (1) sociodemographic characteristics (i.e., educational level), (2) treatment-related characteristics (i.e., treatment modality, intensity and time since completion of treatment), and (3) patient-reported outcomes (i.e., fatigue, depression and coping style). All effects were considered statistically significant at a p-value smaller than 0.05. Statistical analyses were performed using Statistical Package for the Social Sciences (SPSS), version 26 (SPSS Inc., IBM Corp., Armonk, NY, USA).

Sample Characteristics
Recruitment and participant flow of the original cohort have been reported in detail elsewhere [25]. Based on inclusion criteria for the current study, a total of 206 cancer survivors (62 men and 144 women) with a mean age of 51.0 years (SD 6.8; range 28-63 years) were included in the analyses. Information on baseline sociodemographic and treatmentrelated characteristics of the sample are provided in Table 1. Breast cancer was the most prevalent cancer type among female cancer survivors (n = 104; 72%). Urogenital cancer was the most prevalent cancer type among male cancer survivors (n = 10; 16%). The mean time since diagnosis at baseline was 23.2 months (SD 9.3; range 15-98 months). Cancer survivors worked on average up to 31.9 h/week (S.D 7.9, range 7-48 h), and 68% of the cancer survivors (n = 141) worked full time at time of cancer diagnosis. Loss to follow-up at T1 and T2 were 13% (n = 26) and 16% (n = 33), respectively. 38 (18) Mean and standard deviations of cognitive functioning for the total group and for the three work disability groups separately can be found in Table 2.

Trajectory of Cognitive Functioning
After inspection of the fit indices (AIC and BIC), we found that the model including a linear effect of time showed the best fit for the total group. A statistically significant improvement in cognitive functioning was observed (β = 4.62, SE = 0.91, p < 0.001). The effect size for change in cognitive functioning in the total group was 0.26 at one-year followup (T0-T1), and 0.36 at two-year follow up (T0-T2). The negative correlation between the intercept and the linear growth parameter (β = −100.38, SE = 50.05, p = 0.045) indicated that cancer survivors with relatively high cognitive functioning at baseline had a slower linear increase, whereas cancer survivors with low cognitive functioning at baseline had a faster linear increase in cognitive functioning. Cognitive functioning scores of the total group ( Figure 1) were below the mean cognitive functioning score of the general Dutch population at all timepoints. Furthermore, cognitive functioning scores of the total group were below the threshold of clinical importance at T0 and T1, and equal to this threshold at T2.

Trajectory of Cognitive Functioning
After inspection of the fit indices (AIC and BIC), we found that the model including a linear effect of time showed the best fit for the total group. A statistically significant improvement in cognitive functioning was observed (β = 4.62, SE = 0.91, p < 0.001). The effect size for change in cognitive functioning in the total group was 0.26 at one-year follow-up (T0-T1), and 0.36 at two-year follow up (T0-T2). The negative correlation between the intercept and the linear growth parameter (β = −100.38, SE = 50.05, p = 0.045) indicated that cancer survivors with relatively high cognitive functioning at baseline had a slower linear increase, whereas cancer survivors with low cognitive functioning at baseline had a faster linear increase in cognitive functioning. Cognitive functioning scores of the total group ( Figure 1) were below the mean cognitive functioning score of the general Dutch population at all timepoints. Furthermore, cognitive functioning scores of the total group were below the threshold of clinical importance at T0 and T1, and equal to this threshold at T2.

Level of Work Disability and Trajectories of Cognitive Functioning
No significant difference in changes in cognitive functioning between the three groups were found (Figure 2). At baseline, however, cancer survivors who are non-durable work disabled (either party or fully) had significantly worse cognitive functioning compared to cancer survivors who are considered able to work (i.e., those below the 35%

Level of Work Disability and Trajectories of Cognitive Functioning
No significant difference in changes in cognitive functioning between the three groups were found (Figure 2). At baseline, however, cancer survivors who are non-durable work disabled (either party or fully) had significantly worse cognitive functioning compared to cancer survivors who are considered able to work (i.e., those below the 35% loss of former wages threshold) (β = −14.26, SE = 5.30, p = 0.008 and β = −19.74, SE = 6.75, p = 0.004, respectively). Cognitive functioning scores of the separate work disability groups ( Figure 1) were below the mean cognitive functioning score of the general Dutch population at all timepoints. Furthermore, cognitive functioning scores of cancer survivors who are considered able to work, were below the threshold of clinical importance at T0 and above this threshold at T1 and T2. Cognitive functioning scores of cancer survivors who are non-durable work disabled (either party or fully), were below the threshold of clinical importance at all timepoints.
( Figure 1) were below the mean cognitive functioning score of the general Dutch population at all timepoints. Furthermore, cognitive functioning scores of cancer survivors who are considered able to work, were below the threshold of clinical importance at T0 and above this threshold at T1 and T2. Cognitive functioning scores of cancer survivors who are non-durable work disabled (either party or fully), were below the threshold of clinical importance at all timepoints.

Factors Associated with Trajectories of Cognitive Symptoms
In the total group, educational level was not a significant predictor of changes in cognitive functioning. Neither were treatment-related characteristics, such as treatment modality, treatment intensity (e.g., local versus systemic treatment) and time since completion of treatment (Table 3). In addition, educational level and treatment-related characteristic were not associated with cognitive functioning at baseline. More fatigue at baseline was associated with lower cognitive functioning at baseline (β = 1.53, SE = 0.21, p < 0.001). Furthermore, changes over time in cognitive functioning were predicted by baseline fatigue; more fatigue at baseline was associated with a faster improvement of cognitive functioning (β = −0.23, SE = 0.086, p = 0.008). Depressive symptoms at baseline were associated with worse cognitive functioning at baseline (β = −1.53, SE = 0.22, p < 0.001), but were not a significant predictor of the changes over time in cognitive functioning. Higher scores on the coping subscales passive reaction and expression of emotion at baseline were associated with lower cognitive functioning at baseline (β = −2.43, SE = 0.71, p = 0.001 and β = −4.93, SE = 1.64, p = 0.003, respectively). However, subscales of coping did not significantly predict changes over time in cognitive functioning.

Factors Associated with Trajectories of Cognitive Symptoms
In the total group, educational level was not a significant predictor of changes in cognitive functioning. Neither were treatment-related characteristics, such as treatment modality, treatment intensity (e.g., local versus systemic treatment) and time since completion of treatment (Table 3). In addition, educational level and treatment-related characteristic were not associated with cognitive functioning at baseline. More fatigue at baseline was associated with lower cognitive functioning at baseline (β = 1.53, SE = 0.21, p < 0.001). Furthermore, changes over time in cognitive functioning were predicted by baseline fatigue; more fatigue at baseline was associated with a faster improvement of cognitive functioning (β = −0.23, SE = 0.086, p = 0.008). Depressive symptoms at baseline were associated with worse cognitive functioning at baseline (β = −1.53, SE = 0.22, p < 0.001), but were not a significant predictor of the changes over time in cognitive functioning. Higher scores on the coping subscales passive reaction and expression of emotion at baseline were associated with lower cognitive functioning at baseline (β = −2.43, SE = 0.71, p = 0.001 and β = −4.93, SE = 1.64, p = 0.003, respectively). However, subscales of coping did not significantly predict changes over time in cognitive functioning.

Discussion
This is the first study exploring trajectories of self-perceived cognitive functioning in sick-listed cancer survivors with work capacity. In the total group, cognitive functioning improved after work disability assessment (i.e., between two and four years after first day of sick leave), although the effect size of the change in cognitive functioning over time was small to medium. When comparing cancer survivors in the three groups of work disability, quite similar trajectories of cognitive functioning were observed. Fatigue at two years after first day of sick leave was the only factor found to be associated with the reported trajectory of cognitive functioning.
In our study, we found a small to medium change in cognitive functioning over time in a group a sick-listed cancer survivors with work capacity. Mean baseline and followup cognitive functioning scores of our sample are of clinical concern, according to the threshold for clinical importance established by Giesinger et al. [36]. When compared to the mean cognitive functioning score of the general Dutch population [35], their scores are considerably lower. Together, these observations indicate that cognitive symptoms are a serious and long-term problem in sick-listed cancer survivors with assessed work capacity.
Cancer survivors who were non-durable work disabled (party or fully) reported worse cognitive functioning compared to those assessed as being able to work, two years after first day of sick leave. Cognitive functioning improved to a clinically acceptable level in this latter group, whereas cognitive functioning continued to be of clinical concern for the non-durable work disabled cancer survivors. In practice, this means that in those survivors who are able to work, recovery takes place up to four years after the first day of sick leave, whereas in non-durable work disabled (partly or fully) survivors, expectations for recovery are lower based on our results.
A number of factors were found to be associated with cognitive functioning at two years of sick leave; cancer survivors with higher levels of fatigue, depression, passive reactive coping, and emotionally expressive coping, reported worse cognitive functioning. These findings are consistent with previous research demonstrating fatigue, mood and coping to be associated with cognitive symptoms in cancer patients [15,23,24,38], suggesting that these symptoms are likely to co-occur and cluster [39]. Fatigue was the only factor found to be (weakly) associated with subsequent change in cognitive functioning. Potentially, patients with more fatigue at baseline may have had more room for improve-ment, whereas patients with less fatigue may improve to a smaller degree. Since fatigue is associated with cognitive symptoms, patients with more fatigue at baseline might exhibited a stronger decrease of cognitive symptoms over time. This finding is in line with several (intervention) studies, in which it was demonstrated that lower baseline symptom levels are predictive of improvement of symptoms over time [40][41][42][43][44]. However, the association found in current study needs to be interpreted with caution due to the small effect.
In previous literature, it has been demonstrated that cognitive symptoms may occur in patients receiving systemic cancer treatments, such as chemo-, endocrine-and immunotherapy. Local treatment (i.e., surgery and/or radiotherapy) may also be associated with cognitive symptoms, however this association has been found to be less pronounced [18,45,46]. The current study found no association between treatment-related characteristics (i.e., treatment modality, intensity, and time since completion of treatment) and trajectories of self-perceived cognitive functioning. There are several possible explanations for this result. At first, the majority (84%) of our sample received systemic cancer treatment and it is therefore difficult to investigate a potential differential impact of therapies on changes in cognitive function. Furthermore, cancer survivors without cognitive symptoms may have already returned to work before two years of sick leave, and thus did not need to apply for a work disability grant. In that case, our sample might reflect cancer survivors with a more unfavorable prognosis as to cognitive functioning, regardless of cancer treatmentrelated characteristics. Another possible explanation for this finding may be that functional status (e.g., physical, emotional, cognitive, social functioning) of cancer patients prior to cancer treatment may predict the trajectory of self-perceived cognitive symptoms over time, rather than cancer treatment-related characteristics [47]. This indicates the importance of assessing functional status and symptom burden prior to treatment if aiming at identifying patients vulnerable for cognitive symptoms.
A strength of this study is its longitudinal design, which made it possible to identify trajectories of cognitive symptoms with follow-up up to four years after the first day of sick leave of these cancer survivors. Furthermore, as a result of identification of participants from the SSA registries, the current study included cancer survivors, assessed for work disability after two years of sick leave, from the entire Dutch working population [14].
Our study had several limitations. First, the division of the cohort into three groups in terms of level of work disability resulted in relatively small fully, non-durable work disabled group of cancer survivors (n = 38), and may have underestimated the effect of level of work disability on the trajectory of cognitive symptoms. A second limitation was the absence of neuropsychological assessment to measure cognitive functioning. However, cognitive functioning assessed using self-reports may be more sensitive to fluctuations in fatigue and mood than neuropsychological assessment [19].
The results of current study underscore the importance to provide partly and fully, non-durable work disabled cancer survivors with evidence-based treatment options for their self-perceived cognitive symptoms, especially since this group of cancer survivors is expected to (partly) RTW based on the assessment of the SSA. So far, several intervention studies for cognitive symptoms in cancer survivors have been conducted [19,48]. However, due to methodological limitations (e.g., small sample sizes, limited control groups, lack of inclusion criteria regarding cognitive functioning) and absence of work-related outcome measures, no firm conclusions can be drawn from these intervention studies and further research into the effectiveness of interventions for cognitive symptoms in cancer survivors with work capacity is needed. Furthermore, our findings suggest that if improvement of cognitive symptoms is a desired outcome, management of cognitive symptoms together with fatigue, mood and coping may be a beneficial approach. Currently, an RCT is underway that evaluates such an approach for cognitive symptoms at work [49].

Conclusions
Our results indicated that cognitive symptoms are of clinical concern in sick-listed cancer survivors with work capacity, up to four years after the first day of sick leave.
Although the cognitive symptoms improved over time, cognitive functioning continued to be of clinical concern in cancer survivors who are non-durable work-disabled (partly or fully). Therefore, this population is in need of options to manage their cognitive symptoms and thereby improve their ability to RTW and sustain their employability at the longer term.