The impact of chemotherapy on cognitive performance post‐surgery in patients with colorectal cancer: A prospective cohort study

Subjective reports of cognitive impairment following chemotherapy are frequent in cancer patients. Objective cognitive impairment has been observed in cancer patients regardless of treatment regimen suggesting the relationship between cognitive impairment and chemotherapy is not clear cut. Little research has explored the effects of chemotherapy on cognition following surgery in colorectal cancer (CRC). The present study explored the effects of chemotherapy on cognitive performance in a sample of CRC patients.


| INTRODUCTION
Colorectal cancer (CRC) average 5-year relative survival rates have increased from 22% to 57% in the United Kingdom over the last 40 years. 1 This reduction in mortality has led to an increased focus on quality of life. Cancer and its treatments can lead to a range of side effects, which negatively impact on patient quality of life. Impairment of cognitive function is one such side effect. 2 The negative effects of surgery and anaesthesia on cognition have been demonstrated in CRC patients. 2,3 Adjuvant chemotherapy is routinely offered to CRC patients with high-risk stage II and stage III colorectal cancers to reduce the risk of local and systemic recurrence post-surgery in the United Kingdom. 4 Chemotherapy is ideally initiated 4-8 weeks after surgery, and its effects are likely to commence whilst some patients are still recovering from the cognitive effects of surgery and anaesthesia.
Several meta-analyses have confirmed the presence of objective cognitive deficits in patients undergoing chemotherapy across a range of cancers. 5,6 Whilst initially solely attributed to treatment with chemotherapy, hence the term 'chemobrain', it is becoming clear that a range of factors including the cancer, surgery, anaesthesia, fatigue and mood may play a role in the aetiology of cognitive impairment in cancer patients with some research indicating cognitive impairment present prior to systemic treatment. 5,[7][8][9][10][11] Research in CRC is limited with few longitudinal studies available to give accurate indications of the prevalence and incidence of cognitive impairment in CRC. 6 Existing studies suggests that CRC patients display greater cognitive impairment than matched healthy controls (43% vs 15%). 12 CRC patients exhibited impairment in attention/working memory, verbal learning/memory, and complex processing speed. 12 The relationship between chemotherapy and cognitive impairment in CRC remains unclear with research to-date presenting conflicting results, with some studies finding chemotherapy to be associated with poorer cognitive performance, but other studies finding no association. 5,12,13 Despite the publication of international guidelines for researching cognitive impairment in cancer survivors, 14 a recent systematic review in colorectal cancer highlighted the need for future research to use standardised criteria and measures to define and assess cognitive impairment. 5 Furthermore, the review called for more research into the relationships between emotional distress and cognitive impairment to attempt to clarify these associations. 5 The primary objective of the present research was to explore the impact of adjuvant chemotherapy on cognitive recovery post-surgery in CRC patients, adhering to International Cognition and Cancer Task Force (ICCTF) guidelines. 14 Specifically, comparing the frequency of cognitive deficits and changes in cognition over time between groups.
It was hypothesised that recovery from post-surgery cognitive performance would be greater in patients who did not undergo adjuvant chemotherapy. The secondary objective was to explore the relationships between emotional distress, fatigue and cognition.

| Design
This study is reported in accordance with the guidelines for Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. 15 The data reported in this paper was taken from a prospective cohort study measuring cognition in two groups: (i) patients diagnosed with CRC who had undergone surgery 'surgery only group', (ii) patients diagnosed with CRC who had surgery followed by chemotherapy 'chemotherapy group'. A third group with no cancer history, 'learning control group', were recruited to control for learning effects on the neuropsychological tests only. Data was collected at three time-points:'T1' (approx. 4 weeks post-surgery, prior to chemotherapy), 'T2' (12 weeks after first scheduled chemotherapy [approx. 4 months post-surgery]), 'T3' (3 months after last scheduled chemotherapy for patients receiving 6 months of treatment or 6 months after last scheduled chemotherapy for patients receiving 3 months of treatment (approx. 10 months post-surgery)).
Data was collected from the surgery group at equivalent time-points.

| Procedure
A consecutive series of outpatients attending oncology clinics from five London Hospitals were invited to participate in the study between April 2014 and July 2018. Inclusion criteria: aged 18 years and over; diagnosed with resectable CRC to be followed by adjuvant chemotherapy treatment (chemotherapy group) or no further systemic cancer treatment (surgery group); fluent in spoken and written English sufficient to complete the assessments (based on clinician and/or researcher observations). Exclusion criteria: prior exposure to chemotherapy; significant comorbidities which could affect ability to participate; history of stroke or other brain trauma. All exclusion criteria were identified through medical records or self-report. Eligible patients were identified and approached by a member of the research team at each Hospital. Those interested in participating were contacted by a research assistant by telephone. Participants were contacted by telephone to arrange all subsequent follow-up appointments.

| Assessments
Demographic and clinical information were collected at T1. A battery of neuropsychological assessments (NP) was used to assess cognitive function at all three time-points conducted by a trained research assistant. Initial assessment took place in clinic with follow up assessments conducted at the participants home. The battery included the three core measures recommended by the ICCTF 14  and The Benton Visual Retention Test, a test of immediate recall visual memory (number correct and number of errors recorded). 22 Depression and anxiety were assessed using the Hospital Anxiety and Depression Scale 23 and fatigue using the Functional Assessment of Chronic Illness Therapy-Fatigue (FACIT-F v4). 24 Further details of the methodology can be found in the study protocol. 25 A number of steps were taken to minimise potential bias in the study. Both the methods and analyses strictly adhered to the ICCTF guidelines. Standardised neuropsychological tests were administered by trained researchers and alternate forms were used when possible.
Tests scores were normed for age and gender and the effects of age, cancer stage and education were controlled in the analyses. The effects of learning were controlled for using data from a learning control group where appropriate. Multiple methods of analysis were used, drop-outs were adjusted for through the use of multilevel modelling, and multiple imputation was used to account for missing data at each time point. A total of 136 participants were recruited into the study: 78 in the chemotherapy group, 58 in the surgery only group. A breakdown of recruitment is given in the flow chart ( Figure 1).

| STATISTICAL ANALYSES
Analysis of missing data among all variables of the study was undertaken. Initial analyses determined participants lost to follow-up as a percentage, per group. Thereafter missing data levels per timepoint were ascertained. If missing value levels were >5% per variable at any timepoint, missing values within a timepoint were multiply imputed Composite NP scores for each participant were calculated by taking the mean z-score across 14 tests (the COWA semantic version was not a standardised score therefore excluded), at each timepoint.
The mean score across tests rather than a sum was used to retain the original metrics of the standardised scores.
In line with the ICCTF recommendations on reporting analyses based on individual tests scores 14 multiple methods were used to assess change in cognitive performance over time.

| Calculating the number of scores in deficit/ number of ppt considered in deficit
Two deficit scores (0 = not in deficit; 1 = in deficit) were calculated for each neuropsychological test score; 1.5 SD and 2 SD below the group mean score. The total number of tests in deficit (at a timepoint) for each was calculated for each participant to determine if they were in cognitive impairment as follows. Ingraham and Aiken 35 provide useful data on determining criteria for impairment in multiple test batteries.
The present study yielded 15 scores from seven neuropsychological tests. On the basis of 15 tests, when employing the criteria of 2SD below the mean, according to the Ingraham and Aiken 35 criteria 30% of the population would be expected to score at least 2SD below the mean on one test out of 15% and 1% of the population would be expected to score 2SD below the mean on at least 3 tests.

| Learning adjusted difference scores (Adj∆) between timepoints (Adj∆T1T2 and Adj∆T1T3)
For each participant the difference in each test score at follow-up, from the baseline scores was calculated by subtracting baseline zscores from follow-up z-scores. To generate learning adjusted difference scores (Adj∆T1T2 and Adj∆T1T3), the mean change in the learning control group (representative of natural learning over multiple testing) was subtracted from the individual difference scores of the chemotherapy and surgery participant scores.

| Reliable change Index (RCI)
The RCI for each NP test was computed utilising Hsu 36 method. As with the RCS, this was calculated per treatment group using treatment group-based means, standard deviation and test-retest correlations; and were done for T2 and T3 separately.
Standardised scores (transformed for each NP test), number of tests in deficit, and change scores between timepoints (i.e. learning adjusted difference scores, RCS and RCI between T1 and T2, and T1 and T3) were utilised to perform analyses between groups and over time.

| Frequency of cognitive deficits by group
The graph below (Figure 2.) shows the percentage of participants in each cancer group scoring at least two SD below the mean score on at least one test and at least three tests at each timepoint compared to the proportions expected in a normal population due to chance alone. 35 Perceptuo-motor (42% scored 2SD below the mean), memory (19% at 2SD) and executive function (16% at 2SD) were the cognitive domains most frequently affected.
At all timepoints both groups showed a greater percentage of participants with deficits on both one and three tests than would be expected by chance in the normal population. A similar pattern was observed when using the 1.5SD criterion.  showed that the groups did not significantly differ within each timepoint (p > 0.05); however, while the chemotherapy group did not show significant differences between T1, T2 or T3 in pairwise comparisons (p > 0.05), the surgery only group showed significant increases in scores between T1 and T3 (p = 0.017) and T2 and T3

| DISCUSSION
The present study examined the impact of chemotherapy on cognitive recovery post-surgery in CRC patients by comparing those who received chemotherapy after surgery and those who did not.
The proportion of cancer patients in both groups demonstrated much higher deficits in cognitive function than would be expected in the normal population. The proportion with impaired scores (one test at 2SD) was similar to those found in previous CRC research (47%-67%). 12 Although both groups demonstrated deficits in cognitive function there were no differences in cognitive performance between groups at any timepoints consistent with previous research in breast cancer. 10 However, the MLM analysis found that the surgery only group, despite being older, showed a greater improvement in overall cognitive performance over time than the chemotherapy group. This finding suggests that while the observed cognitive deficits in both groups are likely attributable to the effects of surgery, and/or the cancer itself 11 the addition of chemotherapy may slow post-surgical cognitive recovery.
This dual impact of surgery and chemotherapy may explain the inconsistent research findings regarding the effect of chemotherapy on cognition to date. 5 Research has shown no effect of chemotherapy on cognition 12 and poorer cognition in CRC patients having chemotherapy. 13 A larger sample assessed over a longer time frame is required to tease out the effects of chemotherapy and surgery and the potential role of some other element of the cancer process.
Consistent with prior research 12 there was little association between mood, fatigue and cognition supporting the assertion that the observed deficits were attributable to surgery. The chemotherapy group showed greater depression and fatigue than the surgery only group during chemotherapy, but these differences reduced post chemotherapy.
The results indicate that interventions aimed at supporting patients and reducing the impact of cognitive deficits should not be limited to those undergoing chemotherapy. 37

| Study limitations
While the use of a consecutive series of patients from multiple sites allows these findings to be generalised to the wider CRC population, the study is not without limitations. These include the small sample size and expected loss to follow-up. Recruitment was challenging, patients were approached whilst recovering from surgery and a large proportion declined to participate (59%). This may have caused sampling biases and limited the statistical power to detect differences between the patient groups. Baseline assessment took place in clinic with subsequent assessments at the participants home. The potential impact of context specific learning should be considered. 38 It was not possible to analyse different treatment regimens on cognitive performance, as these were too varied and any analyses would have limited statistical power. A substantial loss of follow-up was found (up to 40%) with many participants reporting wanting to 1064 -DWEK ET AL.
move forward with life, deteriorating health, complications, multiple healthcare appointments, and return to work and usual routine acting as barriers to participation. However, the use of covariates and MLM in the statistical analysis increased the power and partially mitigated the reduced sample size.
Despite these limitations this paper reports one of the few longitudinal studies of cognitive impairment in patients with CRC whilst controlling for learning effects and including a pre-chemotherapy cognitive function baseline. The study closely adhered to the ICCTF guidelines for both assessment and analysis, increasing the robustness of the findings and allowing cross-study comparison.

| Clinical implications
CRC patients undergoing surgery, regardless of chemotherapy status, should be monitored for cognitive difficulties and support and appropriate referral made for those with impairment which impacts their daily functioning.

| CONCLUSIONS
This study confirms the presence of objective cognitive impairment up to 10 months after surgery in people with CRC. Those not undergoing chemotherapy showed greater recovery over time. Further follow-up of this cohort will enable a better understanding of the long-term trajectory of cognitive impairment following surgery for CRC. There is now adequate evidence to suggest the need for supportive cognitive interventions for those undergoing treatment for CRC.