Comorbidities are associated with poorer quality of life and functioning and worse symptoms in the 5 years following colorectal cancer surgery: Results from the ColoREctal Well‐being (CREW) cohort study

Abstract Objective More people are living with the consequences of cancer and comorbidity. We describe frequencies of comorbidities in a colorectal cancer cohort and associations with health and well‐being outcomes up to 5 years following surgery. Methods Prospective cohort study of 872 colorectal cancer patients recruited 2010 to 2012 from 29 UK centres, awaiting curative intent surgery. Questionnaires administered at baseline (pre‐surgery), 3, 9, 15, 24 months, and annually up to 5 years. Comorbidities (and whether they limit activities) were self‐reported by participants from 3 months. The EORTC QLQ‐C30 and QLQ‐CR29 assessed global health/quality of life (QoL), symptoms, and functioning. Longitudinal analyses investigated associations between comorbidities and health and well‐being outcomes. Results At baseline, the mean age of participants was 68 years, with 60% male and 65% colon cancer. Thirty‐two per cent had 1 and 40% had ≥2 comorbidities. The most common comorbidities were high blood pressure (43%), arthritis/rheumatism (32%), and anxiety/depression (18%). Of those with comorbidities, 37% reported at least 1 that limited their daily activities. Reporting any limiting comorbidities was associated with poorer global health/QoL, worse symptoms, and poorer functioning on all domains over 5‐year follow‐up. Controlling for the most common individual comorbidities, depression/anxiety had the greatest deleterious effect on outcomes. Conclusions Clinical assessment should prioritise patient‐reported comorbidities and whether these comorbidities limit daily activities, as important determinants of recovery of QoL, symptoms, and functioning following colorectal cancer. Targeted interventions and support services, including multiprofessional management and tailored assessment and follow‐up, may aid recovery of health and well‐being in these individuals.


| BACKGROUND
Colorectal cancer (CRC) is one of the most common cancers worldwide, with an estimated incidence of over 1.3 million, and this is increasing. 1 Five-year survival rates in the UK stand at 57% and 65% in the United States. 2,3 Colorectal cancer is more likely in older adults, with 60% of survivors aged over 65 years. 4 Comorbidity is defined as the presence of distinct medical condition(s) in addition to the particular index disease, in this case CRC. 5 Multiple comorbidity is progressively more common with age; thus, older CRC survivors generally present with high levels of comorbidity. 6 Colorectal cancer survivors also have higher rates of comorbid disease compared with the general population, 7 with around 40% to 50% of CRC patients reported to have ≥2 comorbidities. 8,9 Living with comorbidity after CRC diagnosis is now the norm rather than the exception. Therefore, investigation into how comorbidities affect CRC survivors' health and well-being has become increasingly important. Cancer survivors often report poorer health and well-being compared to healthy populations, and independently, long-term chronic conditions negatively influence quality of life (QoL). 10,11 Whilst there is a growing body of literature exploring the effect of comorbidities in people recovering from CRC, there is significant variability in study sample sizes, 12,13 participant characteristics, 14,15 and time points of assessment, 9,12 and it is not always possible to identify CRC-specific data in cohort studies that include multiple tumour groups. 16 In addition, investigations of the impact of comorbidities on QoL, symptoms, and functioning following a CRC diagnosis are limited by cross-sectional design, 9,17 a narrow range of outcomes, 17,18 and methods used to determine comorbidity status. 17,19 Most studies focus on the number of comorbidities, 9,20 or comorbidity severity using weighted scales, where severity is based on the predefined mortality risk of individual conditions, such as the Charlson Comorbidity Index 13,19,21 or similar indices. 22 Few studies describe patient-reported severity, such as limitations on activities caused by comorbidities. Those that do are either cross-sectional, limited to self-reported depression, do not exclusively examine the impact of comorbidity limitation on well-being, or present data from mixed tumour groups. 12,19,23 Few studies have described associations between comorbidities, and health and well-being over time. Associations with pain, fatigue, and mental well-being up to 1 year following a CRC diagnosis, and fatigue and QoL over time in longer term survivors, have been described, yet only in relation to the number of comorbidities. 18,24 The role of individual comorbid conditions is largely overlooked in studies.
Only one longitudinal study has mapped comorbidity prevalence up to 1 year; however, this study was nonpopulation based and limited to CRC survivors >65 years. 14 Similarly, no studies describe the demographic, clinical, and treatment characteristics of CRC survivors with comorbidities. Using results from the ColoREctal Well-being (CREW) study, 25 a longitudinal cohort study investigating recovery of health and well-being in the 5 years following colorectal cancer, this paper aims to determine the following: 1) The frequency of comorbidities, their limiting effects on daily activities, and the frequency of individual comorbid conditions among CRC survivors.
2) The association between comorbidities, and recovery of QoL, symptom, and functioning outcomes.
3) The demographic and clinical factors that characterise comorbid CRC survivors.

| Design
The ColoREctal Well-being (CREW) study is a prospective longitudinal cohort study of patients with nonmetastatic CRC undergoing curative intent surgery. Further details are described elsewhere. 25

| Participants
Eligible patients had a diagnosis of Dukes' stage A-C colorectal cancer, were being treated with curative intent surgery, aged ≥18 years, and able to complete questionnaires. Having a prior cancer diagnosis was an exclusion criterion.

| Procedure
Participants were recruited from 29 UK hospitals between November 2010 and March 2012. Self-report questionnaires were completed before surgery (baseline), and mailed questionnaires were sent at regular intervals: 3, 9, 15, 24 months

| Measures
Full details of the measures used in CREW have been published. 25 Measures that pertain to the findings presented in this paper are summarised below.
2.4.1 | Socio-demographic, clinical, and treatment data Clinical and treatment data were obtained (with consent) from medical notes: tumour site, Dukes' stage, nodal involvement, how CRC was detected, family history of CRC, presence of a stoma, and neoadjuvant and adjuvant treatment. Neighbourhood deprivation was derived from postcodes using the index of multiple deprivation. 26 Domestic and employment status were assessed by participant self-report in questionnaires.

| Comorbidity data
Patient self-reported comorbidity status remains an accurate method for health research against clinical record review. 27 Self-reported comorbidity data were collected at 3, 15, 24, 36, 48, and 60 months.
The list relating to 12 individual conditions or disease groups was a study-specific measure (not formally validated) informed by Ramsey et al, 12 with format informed by the Self-Administered Comorbidity Questionnaire. 28 The list ( Figure 1) asks whether a doctor has ever told the participant they have the condition, whether the condition limits typical daily activities, and the severity of such impact (ranked from 1 to 7 on Likert scale). At 24 months, an additional question asked whether each condition had been diagnosed before or after CRC diagnosis.

| QoL, symptoms, and functioning
Quality of life, symptoms, and functioning were assessed using the validated European Organisation for Research and Treatment of Cancer QoL (EORTC QLQ) core (C30) questionnaire 29  Global health status/QoL scale of the QLQ-C30 was used to represent overall QoL (comprises 2 items). Analyses of symptoms focussed on those most frequently reported in CRC 9,13,31 : pain, fatigue (from QLQ-C30), urinary, and bowel symptoms (from QLQ-CR29). Physical, role, cognitive, emotional, and social functioning was assessed using QLQ-C30 subscales.

| Statistical analysis
Subscale scores from the EORTC questionnaires were calculated according to published guidelines. 29 To avoid problems with multiple testing of a large number of individual symptoms, summary scores representing urinary symptoms and bowel symptoms were calculated by taking the mean of QLQ-CR29 subscales: (a) urinary frequency, FIGURE 1 Self-report comorbidity section of CREW questionnaires urinary incontinence, and dysuria for urinary symptoms and (b) blood and mucus in stool, stool frequency, abdominal pain, pain in buttocks/anal area/rectum, bloating, flatulence, and faecal incontinence for bowel symptoms.
Because of initial analyses indicating the stability in prevalence and chronicity of comorbidities over follow-up, statistical analyses used 3-month comorbidity data.
In the first part of the analyses, associations between the number of comorbidities reported at 3 months and baseline socio-demographic, clinical, and treatment factors were assessed using the chisquare test or chi-square test for trend, where appropriate. The index of multiple deprivation was categorised into quintiles. 26 Performance status was not captured.
For the second part, longitudinal analyses were conducted using generalised estimating equations, based on all available completed questionnaires up to 60 months. Analyses assessed the associations between EORTC subscale scores as dependent variables and comorbidities reported at 3 months (5 most prevalent comorbid conditions and the comorbidity status itself categorised as none, nonlimiting, or limiting) as independent variables at the 5% significance level. Separate models were fitted for each EORTC subscale of global health status/QoL, symptoms, and functioning.
The first set of multivariable regression models included independent variables separately in each model and was adjusted for time since surgery and those demographic, clinical, or treatment factors significantly associated with total numbers of comorbidities in the first analyses.
The second set of multivariable regression models focused on examining multiple effects of the significant comorbidity predictors. Independent (comorbidity) variables statistically significant in the first set of regression models were put together in the second set, again adjusting for time since surgery and demographic/clinical/treatment factors identified as significant in initial analyses.
Participants with missing questionnaires were included in analyses for time points for which they provided data; there was no imputation of missing questionnaires, or socio-demographic, clinical, treatment, or comorbidity data. Missing data on the EORTC measures were dealt with using published guidelines. 29,30 Longitudinal analyses involving individual comorbidities encompassed the 5 most prevalent individual conditions (small numbers restricted detailed analysis for less prevalent comorbidities and any associations of individual conditions that limited daily activities).
In line with published guidance, clinically meaningful differences were determined by a >10-point difference in EORTC subscale scores. 32 3 | RESULTS

| Participants
One thousand three hundred and fifty eligible individuals were identified. Of those eligible, 78% (n = 1055) agreed to participate, of whom 86% (n = 909) gave full consent to participate and 14% (n = 146) gave permission for only clinical data to be collected. Thirty-seven were found to be ineligible following surgery. Excluding 11 individuals who withdrew or died between consent and baseline, 861 participants consented to follow-up. This sample is representative of eligible patients treated during the recruitment period. Response rates were 88% at baseline and 69% at 60 months. Comorbidity data were available for 99% of those responding (n = 659) at 3 months and 87% (n = 324) at 60 months.
Mean age at baseline was 68 years (range 27 to 95 years). The majority were of white ethnic origin, and 60% were male. Most participants were retired (60%), and over 60% were married or living with a partner. Most participants had colon cancer (65%), 35% rectal tumours. Over 53% had Dukes' stage B, 20% had stage C1, and 12% to 14% had stage A or C2 (1% was undetermined), Eighteen per cent received neoadjuvant treatment and 46% adjuvant chemotherapy or radiotherapy.

| Frequency and impact of comorbidities
At 3 months, 28% reported no comorbidities, 32% reported 1, 23% 2, and 17% 3 or more. Of the 72% (n = 476) with comorbidities, the median number was 2. Of those with comorbidities, 37% reported at least 1 that limited their daily activities, with 13% reporting 2 or more limiting comorbidities ( Table 1). The proportion of limiting comorbidities remained consistent over time. Most participants (62% at 3 months) reported that their comorbidities limited daily activities "moderately", which remained fairly consistent over follow-up (Appendix 1).
Results suggest that the majority of comorbid conditions were diagnosed prior to CRC diagnosis (participants responded to this question at 24 months). The exceptions to this were stroke/brain haemorrhage and liver disease/cirrhosis, of which 50% and 80% (respectively) were diagnosed following CRC diagnosis. Of note is were diagnosed before CRC diagnosis in >78% of individuals.
Arthritis and heart failure were reported to be the most limiting conditions. Of participants reporting these conditions, >50% stated it limited their daily activities. Stroke/brain haemorrhage, myocardial infarction, and angina were reported as limiting by ≥40% of respondents with each condition, and >35% of participants with depression/anxiety and lung disease reported them as limiting. High blood pressure was the most prevalent, but least limiting condition.
Of participants with diabetes, 14% reported the condition as limiting (Appendix 2).

| Comorbidities and QoL, symptom, and functioning outcomes
Because of high correlation between age and employment status, only age at baseline was included in the multivariable regression analyses.
The first set of longitudinal multivariable regression models adjusted for age and time since surgery (from baseline to 60 months) illustrates that the presence of any limiting comorbidities was significantly associated with poorer global health status/QoL, symptom, and functioning outcomes across all domains (P < .001), including increased fatigue, pain, urinary, and bowel symptoms and reduced physical, role, emotional, cognitive, and social functioning (Appendix 4).
Findings illustrated clinically meaningful differences associated with the presence of limiting comorbidities across all outcomes (except for urinary and bowel symptoms). In contrast, the presence of nonlimiting comorbidities was only significantly associated with increased pain and worse physical functioning (P < .05).
Of the 5 most prevalent individual comorbid conditions reported at 3 months, arthritis and depression/anxiety were significantly associated with poorer global health status/QoL, symptom, and functioning outcomes across all domains (P < .001). Depression/anxiety appeared to have the greatest association with poorer outcomes, with clinically meaningful differences across all outcomes (except for urinary and bowel symptoms). Lung disease was also significantly associated with poorer outcomes, with the exception of urinary symptoms. Diabetes and high blood pressure were significantly associated with increased pain and poorer physical functioning, with diabetes also associated with worse urinary symptoms (Appendix 4).
Once adjusted for all significant comorbidity predictors, final multivariable regression models confirmed that the presence of any limiting comorbidities remained a statistically strong and significant predictor of all health and well-being outcomes (P < .001), with the exception of emotional functioning ( Table 2). The biggest and clinically significant differences were observed for pain, fatigue, physical, role, social, and cognitive functioning.
The presence of depression/anxiety remained a statistically significant and strong predictor of poorer outcomes across all domains, with the exception of urinary symptoms. Clinically meaningful differences were observed for global health status/QoL, fatigue, and emotional and social functioning. Arthritis, diabetes, and high blood pressure did not remain significantly associated with any outcomes. Lung disease remained statistically significant only in association with poorer global health status/QoL and physical functioning (P < .05).
For participants reporting both limiting comorbidities and depression/anxiety, differences in outcome scores were approximately doubled for domains including fatigue, pain, physical, role, and social functioning, with highly clinically significant differences in outcome scores of >20.

| DISCUSSION
This is the first paper to describe the stability of comorbidity prevalence, individual comorbidities, and patient-reported limitations of comorbidities, and demonstrate their significant associations with poorer QoL, symptoms, and functioning up to 5 years following CRC diagnosis. We demonstrate that it is not the presence of comorbidities per se, but the limitations on daily activities imposed by comorbidities, which has the greatest impact on health and well-being.   Higher scores for global health status/QoL and functioning subscales indicate better health/QoL and functioning.

| Frequency and prevalence of comorbidity
c Higher scores for symptom subscales indicate worse symptoms. d Urinary symptoms include urinary frequency, urinary incontinence and dysuria. e Bowel symptoms include blood and mucus in stool, stool frequency, abdominal pain, pain in buttocks/anal area/rectum, bloating, flatulence and faecal incontinence. ***P < .001. **P < .01. *P < .05.
The frequency of hypertension, arthritis, diabetes, and lung disease is comparable to other studies, 7,18 and reflect their prevalence in the general population. 35 Results demonstrating a ≤10% prevalence of angina, myocardial infarction, and heart failure in the CREW cohort are at odds with higher prevalence in other CRC studies, and in the general population. 8,34 This likely reflects differences in the criteria for assessing conditions, for example as collective "heart disease" or here, as separate conditions.

| Conclusions and clinical implications
Our findings highlight the importance of identifying patient-reported presence and limitations of comorbidities, as important health and well-being predictors both during and beyond CRC treatment. The stability in prevalence and severity of comorbidity suggests that CRC patients at risk of poorer outcomes up to 5 years following a diagnosis can be identified early, and appropriate support put in place. As such, key consideration should be given to optimising comorbidity management before CRC treatment, and clinical follow-up that incorporates comorbidity assessment, is individualised, and takes place as soon as possible following a CRC diagnosis.
The International Society of Geriatric Oncology recommends geriatrician involvement in the management of cancer patients with comorbidities, and treatment decisions that consider comorbidities. 39 We propose that targeted interventions and support services, including multiprofessional management and tailored assessment and follow-up, may aid recovery of health and well-being.
The ColoREctal Well-being (CREW) study highlights the importance of including conditions such as musculoskeletal and mood disorders, and patient-reported limitations, in future clinical and research assessments of comorbidity. The inclusion of self-reported health status in the assessment of comorbid CRC patients is a recommendation echoed by NICE multimorbidity guidance 40 and could help to identify CRC patients at risk of reduced health and well-being.

SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.