Circadian rhythm disorders in patients with advanced cancer: a scoping review

Circadian rhythms can be demonstrated in several biomarkers and behavioural activities, with rhythmical patterns occurring roughly over a 24-h period. Circadian disorders occur in patients with cancer and may be associated with poor clinical outcomes. This scoping review aimed to identify circadian rhythm research and reporting practices, circadian rhythm patterns, circadian rhythm disorders, and relevant associations of circadian rhythm disorders in patients with advanced cancer. Studies involved adult patients with locally advanced or metastatic cancer and used objective measures of circadian rhythmicity. Two independent authors completed initial screening of title and abstracts, full text reviews, data extraction, and data checking. A total of 98 articles were highlighted in the scoping review, which utilised physical activity measures (actigraphy and polysomnography), biomarkers (cortisol and melatonin), or a combination. Several circadian rhythms are commonly disordered amongst patients with advanced cancer and have significant implications for symptom burden, quality of life, and survival. It remains unclear which patients are most at risk of a circadian rhythm disorder. Significant heterogeneity exists in research and reporting practices. Standardising this approach may address discrepancies in the current literature and allow for research to focus on the most relevant parameters and approaches to improving circadian rhythmicity.


Introduction
Circadian rhythms (CRs), repeating patterns approximately every 24 h, can be observed throughout the human body in behavioural activities, such as sleeping and feeding, and biochemical and hormonal changes, such as cortisol and melatonin secretion (1).CRs are coordinated by a central "pacemaker" or "clock" situated in the suprachiasmatic nuclei within the hypothalamus that attempts to synchronise internal body clocks with the 24-h light-dark cycle (1,2).Additionally, areas within the brain and peripherally, such as endocrine organs, contain self-sustained secondary clocks (2).
In health, two well-established endocrine biomarkers of CRs are melatonin and cortisol, with levels being measurable in several samples, including serum, saliva, and urine (2)(3)(4).Serum melatonin begins to rise from around 22:00, peaking at around 04:00, before falling towards a baseline by 10:00, which persists throughout the day.Cortisol levels peak in the early morning, around 08:00, before falling during the day to a baseline at around 00:00 (2).
Physical activity demonstrates a circadian rhythm, with peak physical activity occurring around 14:00 and the most restful period centred around 03:00, although variability does exist between individuals (5).Circadian sleep and physical activity are primarily assessed using polysomnography and actigraphy (6).Although polysomnography and actigraphy are comparable, actigraphy can be applied in various settings and allows for prolonged periods of monitoring (6).Actigraphy utilises a wrist-worn device to detect physical movement during sleep and wake periods, with analysis of data producing several measures of circadian rhythmicity (6).Actigraphy is often accompanied by patient diaries, an approach supported by the American Academy of Sleep Medicine when investigating circadian rhythm sleep disorders (7).Diaries, however, can be burdensome, inaccurately completed, and subject to bias.Adult actigraphy research has focused on sleep-wake activity, particularly sleep onset-offset and the timing of activity phases.Various measures are used within research to describe the robustness of circadian rhythmicity or the timing and relationship of events over 24-h periods (see Table 1).
Circadian rhythmicity can alter during an individual's lifespan and impact on health and disease.With advancing age, activity levels decline, peak activity occurs earlier, sleep becomes shorter and more fragmented, and daytime napping increases (33,34).Circadian rhythm disorders (CRDs), where normal rhythmicity is altered, can perpetuate cancer and metabolic, neurodegenerative, psychological, and cardiovascular disease (35).CRDs are common amongst cancer patients, affecting up to 75%, and are associated with increased symptom burden, poorer quality of life, and shorter survival (36,37).Interestingly, even misalignment between preferred and actual bedtimes is associated with cancer progression (38).

Eligibility
Studies were eligible for inclusion if the patients were ≥18 years old with a diagnosis of advanced cancer (locally advanced or metastatic)."Locally advanced" differed between cancer histology and several studies included, rather than focused solely on, patients with advanced cancer.Eligible studies also had to consider objective measures of four markers of circadian rhythm disorders (sleepwake cycles, rest-activity cycles, cortisol levels, and melatonin levels) and be fully translated into English.

Screening, data extraction, and data synthesis
Two authors (CG and JP) independently screened the title and abstract for potential full-text review.Review papers identified in the initial search were also screened for additional articles.Fulltext articles were reviewed independently by two authors (CG and JP).The reference lists of included articles were searched for additional articles.Where full-text copies were not immediately available, the leading author or associated research centre was contacted, and if no full-text made available, the article was excluded.Data were extracted by a single author (CG) and confirmed independently by a second author (JP).The data extraction tool was then coded into main themes including circadian measures, circadian rhythm patterns, and the association of circadian measures with symptoms, quality of life, survival and other relevant factors.The review is presented according to the PRISMA-ScR checklist.

Results
The scoping review highlighted 98 articles, which were mainly observational in nature.The review process can be seen in Figure 1, and the results from individual studies are detailed in Tables 2A-D.
Heterogeneity was seen in the investigational approach and the reported measures of circadian rhythm.Studies assessing melatonin used between 20 and 190 patients, sampled melatonin at 1-16-h intervals, and used between 2 and 10 different time points.Studies assessing cortisol used between 13 and 210 patients and sampled at 20-min to 12-h intervals.Sampling included fixed times, time slots, and/or were reported in relation to waking and bedtime.Melatonin and cortisol studies lasted between 24 h and 3 days for most studies.
Melatonin levels rose from baseline in all participants between 18:00 and 23:00.Patients with cancer had higher baseline melatonin levels.The rise in melatonin was higher in controls and patients with cancer with relatively less in-bed to out-ofbed physical activity (threefold low I<O group, fivefold high I<O group, and sixfold in control subjects).Patients with cancer and relatively less daytime to night-time activity had earlier DLMOs (1,948 vs. 2,144, p=0.08).Significant inter-individual variation was noted.Serum melatonin levels and 24-h rhythm (12:00, 00:00) and urine 6-sulfatoxymelatonin levels (major metabolite of melatonin) (07:00, 16:00) A 24-h rhythm of melatonin and 6-sulfatoxymelatonin were present in all subjects.Serum melatonin at 00:00 was lower in patients than in control subjects (p<0.05).Urine 6-sulfatoxymelatonin at 07:00 and 16:00 was lower in patients than in control subjects (p<0.05)Serum melatonin levels and area under the curve (AUC) (08:00, 12:00, 16:00, 20:00, 24:00, 02:00, 04:00, 08:00) Cancer patients had significantly lower melatonin levels and area under the curve (AUC) than control subjects (p<0.05)."Nocturnal" melatonin levels and the AUC were significantly lower in patients with stage 3-4 cancer compared to patients with stage 0-1 cancer (p<0.05).Serum melatonin levels, 24-h rhythm (08:00, 14:00, 18:00, 22:00, 02:00, 08:00), amplitude (difference between peak and trough levels) and acrophase (time of peak level) A 24-h rhythm was noted in all subjects.The maximal peak levels were higher for control subjects, but the minimal trough levels were similar for control subjects and patients.The mean amplitude was higher for control subjects.The acrophase occurred earlier for control subjects (04:35 vs. 08:50).There was no significant difference in the 24-h rhythm and AUC between the three groups.Study 1: Melatonin levels were higher at both time points in patients than controls (p<0.0001).Stage 4 breast cancer patients had higher mean melatonin concentration than controls (p<0.0001)and higher levels at 24:00 (p<0.002) and 08:00 (p<0.0001)than stage 1-2 breast cancer patients.Advanced lung cancer patients had higher mean melatonin levels than control at both time-points (p<0.001 at 24:00, p<0.0001 at 08:00).Highest levels were in patients with SCLC.Advanced GI cancer patients had higher mean melatonin levels than control (p<0.005 at 24:00, p<0.001 at 08:00).
Increased melatonin levels at 08:00 were associated with lower performance status (r=−37, p<0.01).Study 2: Melatonin levels did not differ in breast cancer patients pre-and post-surgical removal of the primary tumour Study 3: The circadian melatonin rhythm was similar between patients and controls.24-h urinary melatonin (06:00-10:00, 10:00-14:00, 14:00-18:00, 18:00-22:00, 22:00-06:00) Cancer patients had a lower average melatonin urinary excretion and elevated levels between 06:00 and 10:00 than controls.The differences were not statistically significant.A more synchronised excretion pattern was found in controls Studies without a control group   A consistent diurnal change in cortisol levels was seen in most controls and patients, irrespective of their dichotomy index (I<O), which represents the relative difference between in-bed and out-of-bed physical activity.
Those with a high I<O (i.e., relatively less in-bed to out-of-bed activity) had a larger circadian cortisol amplitude (difference between peak and trough concentrations).
No significant difference was found in other cortisol parameters between I<O groups.There were no differences in the cortisol amplitude, MESOR (mean value) or absolute/relative timing between groups (p>0.09).There were no differences in the diurnal cortisol rhythm between groups (p>0.11).Abnormal cortisol peaks, midway through the sleep episode, were seen in a subset of patients and were associated with increased wake episodes (p=0.004),metastases to bone or organs rather than local recurrence (r=−0.37,p=0.002), use of steroids (r=0.26,p=0.03),ER negative status (r=−0.25,p=0.04) and higher a stage of initial diagnosis (r=0.31,p=0.009).
In a multivariate analysis, metastases to bone (p=0.02) and ER negative status (p=0.048)continued to be associated with the abnormal cortisol peaks.Abnormal cortisol peaks were not related to psychological traits (p>0.018).Larger abnormal peaks were associated with a shorter disease-free interval (r=−0.30,p=0.004).
The disease-free interval (DFI) and the diurnal cortisol rhythm were not associated (p>0.10).Lung cancer patients with depression had a flattened circadian cortisol pattern (less diurnal variation) compared to other groups.Lung cancer patients also had higher salivary cortisol at 00:00 compared to lung cancer patients without depression (p<0.001).
The salivary cortisol area under the curve (AUC) was significantly higher in patients with depression only than the other groups (p = 0.021).Salivary cortisol diurnal variation (VAR) was significantly lower in lung cancer patients with depression than other groups (p<0.001).
(Continued) Mean afternoon cortisol for ovarian cancer patients was 55% higher than for healthy women (p<0.0001) and similar to patients with benign disease (p=0.07).Nocturnal cortisol levels for ovarian cancer were 41.5% higher than benign disease (p=0.02) and 103% higher than healthy women (p=0.0001).
Cortisol variability of ovarian cancer patients was lower than for benign disease (p=0.023) and healthy women (p<0.0001).Adjusted for age and disease stage in the ovarian group, a higher nocturnal cortisol, and lower cortisol variability was associated with greater fatigue (p=0.005 and p = 0.01).Lower cortisol variability also associated with poorer physical wellbeing (p=0.007).Depression scores were associated with a higher nocturnal cortisol (p=0.059) and lower cortisol variability (p=0.028).A more advanced cancer stage was associated with a higher morning (r=0.23,p=0.02) and afternoon (r=0.32,p=0.002) cortisol, but not nocturnal cortisol (r=0.13,p=33).Adjusted for age and disease stage in ovarian cancer group: higher nocturnal cortisol associated with poorer physician-related PS (rated on a 0-4 scale) (p=0.043) and patient-related PS (p=0.035).Lower cortisol variability was also associated with poorer physician-rated PS (p=0.01) and poorer patient-rated PS (p=0.004).Stage 3 metastatic patients had similar 00:00 levels to 08:00 peak of controls.66 patients maintained a normal diurnal rhythm with significantly higher 08:00 levels to 00:00.Two patients lost their diurnal variation and in 8 patients the rhythm was reversed with 00:00 levels higher than 08:00.Elevated cortisol at 00:00 was associated with progressive disease but not length of survival.The diurnal cortisol slope data was split at the median point to distinguish flat and steep slopes.Flat and steep diurnal cortisol slopes had significantly different salivary cortisol levels at 12:00 (p=0.0086),17:00 (p<0.0001), and 21:00 (p<0.0001),but not at waking (p=0.4795) or waking +30 min (p=0.1364).This suggests that the differences between flat and steep cortisol slopes occur at 12:00 or later.Flatter diurnal cortisol slopes were associated with an escape from dexamethasone suppression (p=0.0042).
(Continued) Increasing age was associated with a higher evening cortisol (p=0.004)but not cortisol variability or slope."High grade" disease, and poorer physical well-being were associated with a higher night cortisol, a flattened diurnal cortisol slope, and reduced cortisol variability (all p<0.05)."Late" stage disease was also associated with higher evening cortisol (p=0.05).Shorter survival was seen with elevated night cortisol prior to surgery (HR, 1.802, p<0.001), and the diurnal cortisol slope (HR, 1.633, p=0.001).Longer survival was seen with cortisol variability (HR, 0.644, p<0.001).Estimated median survival for low evening cortisol was 7.3 years compared to 3.3 years in those with a high evening cortisol.Elevated night cortisol, a flattened diurnal cortisol slope, and reduced cortisol variability were associated with higher levels of inflammation indicated by ascitic and plasma IL-6 (all <0.05).The diurnal cortisol slope was associated to cortisol variability (r = 0.88, p<0.001).
Night cortisol was correlated with cortisol variability (r=−0.727,p<0.001) and the diurnal cortisol slope (r=0.758,p<0.001).The cortisol diurnal slope varied depending on the analytical method used.10/91 (11%) had a "positive" slope from 06:00.A flatter diurnal cortisol slope was associated with a lower morning peak and elevated evening trough.Plasma and salivary cortisol concentrations were correlated (p<0.001).
(Continued) Patients had abnormal cortisol circadian rhythms (defined as a lack of decline of 30% in the cortisol level from the morning to the afternoon).
No significant difference was reported in morning or afternoon mean cortisol levels between stable and progressive disease (significance levels not reported).Salivary cortisol (waking, +30 min, 15:00-18:00, 20:00-24:00) Salivary cortisol was increased for all patients with advanced cancer patients having approximately 3 times the healthy population normal values.
Cortisol AUC was significantly higher in advanced-stage patients compared to the low malignant potential patients (p=0.047).Diurnal cortisol levels did not significantly differ between groups or over the day (p>0.06 and p>0.73).Higher evening cortisol levels were associated with higher total depression (p=0.026) and vegetative depression (p=0.005).
(Continued)  Serum cortisol (23:00 and 08:00) Patients with liver involvement had a higher evening cortisol (p<0.0005).Nodal involvement did not impact on cortisol levels.28% had an altered circadian rhythm defined as 23:00 level >50% of 08:00 level.This was more frequent if there was nodal involvement and metastatic spread (p<0.005).
Cortisol levels and circadian rhythm were unrelated to CD4+ lymphocyte count (a prognostic marker).Patients with a "good" activity rhythm had higher cortisol ratios (between 08:00 and 16:00) compared to those with a "dampened" activity rhythm.Mean cortisol levels did not differ significantly between groups.
(Continued)          The mean r24 was 0.27, nighttime restfulness (proportion of activity in bed that falls below the median out of bed activity) was 97.21%, daytime sedentariness (proportion of activity out of bed that is below the median in bed activity) was 6.59%, mean sleep time was 386.94 min, sleep efficiency was 89%, and wake after sleep onset was 13 min.Uncoordinated rest-activity rhythm, such as poor inter-daily stability, was associated with elevated markers VEGF, TGF-beta, and MMP-9 (associated with angiogenesis, immunosuppression, epithelial-mesenchymal transition, tumour invasion, and metastasis).
(Continued) The mean r24 was 0.27.Nighttime inactivity was 97.3%, and daytime inactivity was 6.1%.Intrusive thoughts were associated with a lower r24 and with daytime sedentariness.There was no association between intrusive thoughts and nighttime inactivity.More avoidant coping was associated with a lower r24 (p<0.05) and daytime inactivity (p<0.001).A higher autocorrelation co-efficient was significantly correlated with a steeper diurnal cortisol slope (p = 0.003).The median and mean I<O were 97.5% and 95.1%.39 patients (50.6%) had altered I<O of <97.5%.There was no significant association between I<O and progression free survival.
Overall survival was associated with a more robust circadian rhythm (22.3 months vs. 14.7 months, p=0.013).This was independent of gender, treatment schedule, number of metastatic sites, rank of chemotherapy course of interest, and performance status on day 1 (p = 0.004).
Patients with an altered rhythm during chemotherapy had a higher risk of earlier death (HR, 2.12; p=0.004).There were no significant differences between I<O and response rate or overall grade 3-4 toxicity rate.
No clinical or biological parameters predicted the occurrence of a rhythm disturbance on treatment.Baseline disruption did not predict subsequent disruption, and there was no significant correlation between baseline and ontreatment I<O).
There was no significant difference in toxicity in relation to circadian parameters.
(Continued)  The mean r24 was 0.38, and the median r24 was 0.37.The mean I<O was 94.3, and the median I<O was 97.0.The I<O and r24 were associated (r = 0.74, p<0.001) I<O and r24 were also associated with meanAct (mean activity levels) (p<0.001).There were no significant associations between progression-free survival and any CircAct parameter (I<O, r24, meanAct).Patients in the lowest I<O quartile (<92.4%) had the poorest survival (12.0 months), and patients in the highest I<O quartile (≥99.2%) had better survival (23.5 months).I<O and r24 were related to survival (HR, 0.95; p<0.0001 and HR, 0.20; p=0.004).
Higher I<O, r24 and meanAct were associated with improved QoL and role functioning and less fatigue and appetite loss (all p≤0.01).
Higher I<O and r24 were associated with improved social functioning, and less pain and dyspnoea (p≤0.01).
Higher I<O and meanAct were associated with improved physical functioning (p≤0.01).
Higher I<O, r24 and mean activity was associated with less fatigue, appetite loss, and nausea/vomiting (all p≤0.05)Higher I<O was associated with less pain, constipation, and dyspnoea (all p≤0.002).
Higher I<O and r24 were associated with less depression (all p≤0.01).
Higher I<O and r24 were associated with improved global quality of life, physical functioning, social functioning, and emotional functioning (all p≤0.04).The r24 ranged from -0.06 to 0.77 with a median 0.42.The dichotomy index (I<O) ranged from 49%-100% with a median of 97%.
Patients with a higher r24 or I<O had longer survival (p<0.0001).Patients in the upper quartiles had a longer 2-year survival than those in the lower quartiles (38% vs. 8%).
Higher I<O and r24 were associated with improved global quality of life and physical functioning, and lower depression scores (p≤0.01).Higher I<O, r24, and mean activity were associated with less fatigue and less appetite loss (p<0.001).
Cortisol rhythmicity was correlated with the r24 (r=0.16,p = 0.04).Self-rated sleep disturbances were not correlated to the rest-activity rhythm or mean activity levels.A poor performance status was associated with lower r24 and I<O (p<0.0001), and lower mean activity (p = 0.04).The probability of an objective response was significantly influenced by r24 (p=0.02) and I<O (p<0.0001).

Melatonin Melatonin patterns in advanced cancer
Significant interindividual variation in melatonin circadian rhythms was noted (9,17).24-h melatonin rhythms, with a peakto-trough pattern, were noted in non-small cell lung, gastrointestinal, mixed gynaecological, and mixed cancer cohorts (9,11,17,39,40,42,43,45,46).Abnormal 24-h rhythms were noted in two studies and affected 17% of the patients with metastatic colorectal cancer and 100% of patients with small cell lung cancer (49,51).Detailed abnormalities included a smaller evening melatonin rise and earlier dim-light melatonin onset (DLMO) for patients with a gastrointestinal cancer who demonstrate more inbed to out-of-bed physical activity (lower I<O) (17).Smaller evening melatonin rises were also noted in patients with nonsmall cell lung, gastrointestinal, and cervical cancers, particularly in advanced stages (40,41,43,46).Advanced breast and lung cancer patients had higher mean melatonin levels compared to early-stage disease or controls (47).

Symptoms, quality of life, and survival
No statistically significant associations were reported between the melatonin circadian rhythm parameters and symptoms, quality of life, or survival in any of the included studies.
The change from peak to trough (cortisol slope) was unrelated to education level, marital status, age, time since recurrence, PR status, and metastatic sites in a breast cancer cohort (67).Patients with progesterone-receptor-positive breast cancer did, however, have a smaller cortisol awakening response (77).A flatter cortisol slope was associated with being male (69).
Salivary and serum cortisol were positively correlated, particularly when a strong circadian rhythm was present (16,80).
Furthermore, a study of patients with breast cancer highlighted abnormal cortisol peaks, rather than the diurnal rhythm, to be associated with a shorter disease-free interval (r= −0.30, p=0.004) (52).

Physical and psychological symptoms
A smaller cortisol awakening response, flatter cortisol slope, and less diurnal variability were associated with increased total symptom scores, individual scores for fatigue, and interference with general activity, work, and walking (54,64).
Within lung, ovarian, and mixed cancer cohorts, reduced diurnal cortisol variation, elevated evening cortisol, elevated morning cortisol, and higher area under the curve were associated with depression (19,53,54,73,74).Patients with steeper cortisol slopes, and therefore healthier rhythms, expressed less negative affect during psychological therapy and demonstrated more posttraumatic psychological growth following diagnosis, and those with flatter cortisol slopes were found to repress emotions (14, 67, 79).Abnormalities, including higher waking cortisol and lower cortisol awakening response, were associated with antidepressant use in patients with breast cancer (77).Some studies reported no significant correlations between cortisol levels, or cortisol slope, and psychological measures (12,70,73).

Quality of life
Flattened cortisol slopes, less cortisol variability, and elevated evening cortisol were associated with reduced physical well-being in patients with ovarian cancer (18,54).Conversely, cortisol rhythms were also not correlated with quality-of-life measures for patients with breast cancer (65).

Other
Abnormal cortisol peaks during sleep, coinciding with waking episodes, were reported in a subset of metastatic breast cancer patients (52).More frequent and longer lasting wake episodes and a progressive later waking time were also seen with flatter cortisol slopes (66,75,81).
The cortisol slope was correlated with the CAR, rather than waking level, and flatter slopes were associated with higher evening cortisol levels and an escape from cortisol suppression (63, 77).

Survival
A total of 12 studies commented on survival and all linked circadian disruption to survival.Stronger RARs evidence by improved dichotomy index (I<O), 24-h autocorrelation coefficient (r24), physical activity amplitude and MESOR, nighttime restfulness, sleep efficiency, and time awake spent immobile were associated with longer survival in patients with colorectal, breast, head and neck, non-small cell lung, and mixed cancer diagnoses (8,9,31,37,68,78,89,93,98,99,104,105).In a mixed cancer cohort, disordered I<O was not prognostic; however, r24 and sleep efficiency were prognostic (89).
Examples of prognostic relevance include colorectal and mixed cancer cohort patients with an I<O <97.5%, or below median I<O, having a reduced overall survival (OS) of between 2.1 and 9.7 months, and reduced progression-free survival (PFS) of 4.2 months (98,99,104).Similarly, patients with colorectal cancer and an I<O ≥99.2% had 11.5 months longer survival than those with an I<O <92.4% (105).The I<O was an independent prognostic factor when accounting for factors including age, gender, performance status, cancer diagnosis and stage, previous chemotherapy, and surgery (98,99,104).Similarly, sleep efficiency was an independent risk factor for patients with breast cancer whereby those with a sleep efficiency >85% had over a double survival compared to those with poor SE (68).However, I<O, r24, mean activity, and sleep activity parameters were reported also reported to not be significantly correlated with overall survival or progression-free survival (68,89,92,104,105).

Physical and psychological symptomatology
Abnormal circadian activity rhythms were associated with pain, fatigue, drowsiness, nausea, vomiting, anorexia, and weight loss (20,27,37,50,78,101,106).Higher I<O values were specifically associated with less pain, fatigue, anorexia, sleep disturbance, constipation, and dyspnoea, and improved sleep quality (95,100,105,106).Higher r24 values were specifically associated with less insomnia, daytime dysfunction, fatigue, anorexia, pain, and dyspnoea (20, 30, 100, 105).Higher sleep efficiency was associated with less pain (92).Greater time to sleep once in bed (sleep onset latency, SOL), wake after sleep onset (WASO), and time in bed (TIB) were associated with gastrointestinal symptoms in a mixed cancer cohort (92).Increase time spent napping was associated with increased pain, fatigue, and daytime sleepiness (92).No association between circadian activity parameters and pain or fatigue was found in a mixed cancer cohort (8).
Lower sleep efficiency, I<O, r24, and mean activity along with increased time spent napping or in bed were all associated with increased depression (37,50,75,92,93,106).A lower r24 and more daytime inactivity were associated with intrusive thoughts and avoidant coping in patients with breast cancer (71).Studies also reported that anxiety and depression were not associated with sleep-activity rhythms, including the I<O (8,17,23).

Quality of life
Circadian disruption was associated with interference with activity, work, relations, and enjoyment of life for patients with colorectal cancer (95).Improved r24, I<O, and meanAct were associated with improved global QoL, along with health, physical, social, and functioning subscores (20, 50,78,95,96,105,106).A lower amplitude and MESOR and a later acrophase were associated with worse global QoL in a mixed cancer cohort (8).The strongest correlation between an actigraphy parameter and quality of life measure in a mixed cancer cohort was the 24-h correlation coefficient (32).Studies also noted that circadian parameters were not associated with the fatigue, emotional, or cognitive subscales of quality-of-life measures (105,106).One study of patients with breast cancer noted that WASO and r24 were unrelated to global QoL (65).

Other
There were mixed reports regarding chemotherapy response and circadian rhythmicity in patients with colorectal cancer.One study noted that disordered rhythmicity during chemotherapy was associated with earlier death but not to objective response or toxicity, while another study noted objective response to be influenced by r24 and I<O (37,104).Patients receiving chemotherapy who also had evidence of circadian disruption were more likely to experience weight loss and fatigue (101).
I<O appeared to correlate with circadian temperature rhythms, self-reported physical activity, and chronotype (17).More robust circadian rhythms were associated with greater light exposure (8).
Subjective and objective measures differed for physical activity but were closely correlated for sleep (27, 102).Subjective sleep disruption and circadian disruption can occur together or independently (22).Although total sleep time (TST), SE, and WASO were associated with subjective sleep quality, physical activity measures were also not significantly different between those who report their sleep as good or poor (26,29,37,85,102).
Subjective scores of pain and physical function correlated with objective physical activity, and those using analgesia had more abnormal circadian activity rhythms (84,105).Daytime sleep, or inactivity, was related to sleep medication use, night-time sleep disturbance, daytime dysfunction, night-time sleep, and sleep quality (30,38,108).Sleep efficiency was reported to be correlated with chest metastases, hormone use, and radiotherapy (75).Patients with a higher r24 had less daytime dysfunction and less insomnia (20, 30).Circadian disruption was associated with tumour progression markers (15).

Polysomnography
Patients with cancer spent more time in bed were noted to have multiple nocturnal awakenings and had an average sleep efficiency of up to 77.2% (91,107,108).Increased daytime sleep was associated with less night-time sleep and more nocturnal awakenings in a mixed cancer cohort (108).Medications were also found to impact on sleep.Anticancer therapies were associated with increased sleep efficiency, whereas beta blocker use was associated with reduced sleep efficiency (108).Sleep efficiency was higher in women, white patients, and those with a higher education level (108).

Correlations between measures of circadian rhythm
Increased diurnal physical activity variability was associated with increased diurnal melatonin and cortisol variability along with an earlier DLMO (17).Salivary cortisol levels appeared unrelated to I<O, cortisol rhythmicity positively was correlated with r24, and more robust actigraphy rhythms were associated with a steeper cortisol slope (17,37,71).
Higher I<O and r24 were associated with improved sleep efficiency (100).
Polysomnography-derived values for sleep efficiency were lower, and wake after sleep onset higher, than actigraphy-derived values (91).

Discussion
This review supports, expands upon, and updates several previous reviews of circadian rhythmicity in patients with advanced cancer.It highlights that, for several patients with advanced cancer, disordered cortisol, melatonin, and physical activity circadian rhythms are associated with increased symptom burden, poorer quality of life, and shortened survival.Other important associations with CRDs include poorer performance status and raised biomarkers of tumour progression.
A review of rest-activity rhythms in advanced cancer patients found that CRDs are particularly evident amongst men, those undergoing chemotherapy, and those who were symptomatic (109).Additionally, circadian disruption may be seen across the cancer trajectory, with worse biopsychosocial outcomes reported in cancer survivors who have disordered cortisol rhythms (110).This review highlights that circadian rhythms may be maintained in some patients with cancer and that wide inter-individual variation exists.Future research aimed at identifying patients that are at risk of circadian rhythm disorders, and impacted by their associations, is important, particularly when considering interventional studies to improve circadian rhythms and patient outcomes.
Articles were predominantly observational in nature, and many studies lacked a control group.Causality is difficult to establish, particularly due to the bi-directional relationship between cancers and circadian rhythm disorders, and the influence from external factors.CRDs impact on several neuroendocrine-immune functions, including inflammatory responses and hormonal secretion, and predispose individuals to developing cancer (111).Cancer in turn generates a pro-inflammatory state, and increased circulating cytokines levels can disrupt circadian rhythms (111).Rest-activity patterns are influenced by age, sex, race, education, and voluntary behaviour (6,112).Cortisol values are influenced by sex, age, body mass index, menstrual cycle, sleep disturbances, renal disease, and acute illness, for example (4).The review highlights studies of patients prior to, during, and after anticancer therapies, and within the inpatient and community setting.Limited information on previous and current therapeutic regimes, and location of metastatic disease, limits the ability to synthesise findings.Potential modifying factors of circadian rhythmicity should be reported and taken into account when reviewing findings (113).
CRDs and their impact are not solely seen in cancer patients.Circadian disruption has been reported in patients with neurodegenerative conditions including Alzheimer's disease, Parkinson's disease, and Huntington's disease (114).Despite conflicting findings, evidence highlights altered rest-activity, body temperature, melatonin, and cortisol rhythms within this population and associations with physical and psychological wellbeing, and quality of life (114).At present, the similarities in circadian disruption between clinical conditions are not clear.
Furthermore, the review highlights heterogeneity in the investigation and reporting of circadian rhythms in cancer patients and a lack of threshold values to identify circadian parameter abnormalities.Several reporting measures, overlapping definitions, and an absence of clear definitions were found in the investigation of circadian rhythms, particularly when using actigraphy.Heterogeneity in actigraphy research is not limited to cancer populations.A review of 126 actigraphy studies of children highlighted a lack of standardisation in actigraphy practice, including the reporting of epoch length, artefact detection, and definition of variables (115).Additionally, within cohorts of bipolar disorder patients, over 30 possible actigraphy parametric and nonparametric measures were reported (116).In this review, only four actigraphy parameters (I<O, r24, mean activity, and SE) were associated with at least three of the areas of interest (physical symptoms, psychological symptoms, quality of life measures, and survival).Although a wealth of information can be obtained using actigraphy, the reporting parameters should be aligned with the overall study objectives to allow a clear message in the literature.Analysis of actigraphy data takes many forms and lacks standardisation (6,112).Similarly, variable sampling protocols, analysis, and reporting practices has been seen in cortisol and melatonin studies (3,110).When faced with such heterogeneity in approaches, it is challenging to make firm conclusions, and standardisation may improve research practice.The development of recommendations to identify, and subsequently report, optimal sampling processes, particularly the frequency and timing of samples, and the calculation of circadian parameters are required.
Actigraphy data can report the timing of events, duration of events, or relationship between events.Although studies may focus on "sleepwake" or "rest-activity" periods, there is significant overlap.Diagnostic criteria have been formulated for circadian rhythm sleep-wake disorders by American Academy of Sleep Medicine (117).The diagnosis considers the timing of sleep onset and offset, and the presence of jet lag or shift work, to categorise patients into seven different diagnoses.Many studies of advanced cancer patients reported actigraphy measures across the 24-h period rather than focusing on this timing of sleep onset-offset.The circadian activity rhythm disorders in cancer patients are likely separate to intrinsic circadian sleep-wake rhythm disorders.Recent international consensus recommendations have been developed for the assessment and diagnosis of circadian rest-activity rhythm disorders (CARDs) (118).The recommendations outline key modifiers of circadian rhythmicity, areas to consider within a clinical history, patient sleep and activity diary, and accelerometery during assessment, and criteria to diagnose a CARD.Diagnostic criteria of other forms of CRDs do not currently exist.
The scoping review was strengthened by using independent authors at multiple stages of the review process.Additional evidence was actively sought through hand searching review papers and reference lists.The scoping review is inclusive of available evidence and placed minimal limitations in the search strategy.It made no attempt to critically analyse the quality of evidence.Although the review aimed to focus on advanced cancer patients, several studies included non-advanced cancer patients.This approach may dampen associations, but it was felt to be more inclusive and to provide a broader insight of the topic.Furthermore, the review did not exclude several confounding factors in selected articles, such as medications and chemotherapy.This information was not available in several studies, and through exclusion, it would have limited the generalisability of the findings.Studies reporting on circadian rhythmicity in patients with cancer would benefit from detailed information on recent and existing modifiers of circadian rhythmicity, and the presence and location of metastatic disease.

Conclusion
Cancer patients, particularly those with advanced disease, are at risk of circadian rhythm disorders and significant associated complications.It remains unclear which subset of patients are most susceptible.Conflicting results within the review highlight the need for further studies to identify patient populations that are most impacted by circadian rhythm disorders.Current investigative approaches require a multiple sampling approach (blood, urine, and saliva) or a prolonged period of activity monitoring.In the clinical setting, and advanced cancer population, this may require an alternative approach.Current gaps in the literature are highlighted in Box 1.There needs to be an attempt to standardise research approaches and reporting practice within circadian rhythm research and to develop criteria to identify circadian rhythm disorders.Research standardisation and targeted approaches may help in future research aimed at developing management approaches to circadian rhythm disorders.
TABLE 2A Melatonin circadian rhythms and their associations in patients with advanced cancer.
h rhythm was found in all subjects with a peak concentration at night, and a trough concentration near waking.Mean values did not differ between the groups at any time points.
TABLE 2B Cortisol circadian rhythms and their associations in patients with advanced cancer.

TABLE 2B Continued
Cortisol concentrations differed between patient and controls (p<0.001).The cortisol awakening response (CAR) represents changes in cortisol levels within given time periods.CARi (the cortisol increase in the first 30 min after waking) and CARauc (the cortisol increase in the first 60 min from waking) were both significantly smaller in the patients compared to controls (p<0.01).A flatter diurnal cortisol slope was seen in the patients compared to controls (p<0.001).Patients with Eastern Cooperative Oncology Group (ECOG) Performance Status (PS) of 3-4 had a small CAR (p<0.001) and flatter diurnal cortisol slope (p<0.001) compared to than patients with an ECOG PS of 1-2.Patients with metastatic disease had a small CARauc than those without metastatic disease (p=0.003).Total MD Anderson Symptom Inventory scores, fatigue, and interference with general activity, work, and walking were all significant associated with a reduced CAR and a flatter diurnal cortisol slope (p<0.05).

TABLE 2B Continued
flatter cortisol slope was associated with longer average wake episodes at night (r=0.21,p=0.04).No significant relationship was found between mean waking cortisol or cortisol rise and other sleep measures.
Lower income status was associated with flatter cortisol slope (r=−0.28,p=0.008).Patients with progesterone receptor positive breast cancer had a lower waking cortisol rise (r=0.22,p=0.04).The cortisol slope was unrelated to demographics, disease-free interval, or treatment.Patients with progesterone receptor positivity had lower waking cortisol rise (p=0.04).
A flatter slope was associated with a higher 21:00 level (r=0.85,p<0.0001) and escape from cortisol suppression (r=0.30,p=0.005).No significant association was reported between the cortisol slope and CRF administration or social stress.Antidepressant use was associated with higher waking cortisol (r=0.21,p=0.04) and lower cortisol rise (r=−0.32,p=0.001).
rhythm was seen in the serum of 8/18 (44.4%) patients and in the saliva of 6/16 patients (37.5%).Patients with marked circadian rhythms had lower 08:00 and 16:00 levels.Marked cortisol rhythms were not associated with longer survival than those with altered rhythms and cortisol did not predict the clinical outcome.A higher performance status and per cent of liver replacement was associated with higher cortisol concentrations at 08:00 and 16:00.Salivary and serum cortisol were correlated, particularly for those with stronger circadian rhythms.
A circadian cortisol rhythm was seen in serum cortisol for 8 patients, and 6 patients had a significant salivary cortisol rhythm.Interindividual variation in markers of circadian rhythm.

TABLE 2C
Actigraphy-related circadian rhythms and their associations in patients with advanced cancer.13patients had a dichotomy index (I<O, ratio of activity in and out of bed) of ≤97.5% on chest accelerometry.Patients had worse I<O (p=0.008),levels of activity (p<0.0001), and rest probability P1-1 (probability of remaining in a rest state) than controls (p=0.005).The activity amplitude (between peak and trough activity levels), r24 (autocorrelation coefficient, a measure of physical activity consistency between days), RI (rhythm index, measure of quality, regularity, and consistency of rest state), average centre-ofrest time, and rest duration were not significantly different.Hospital Anxiety and Depression Scale (HADS) scores, performance status, and Pittsburgh Sleep Quality Index (PSQI) scores did not differ significantly between I<O groups.The circadian amplitude in rest-activity and sleep duration variability were the best predictors of a patient's I<O.
Patients had a lower median I<O than controls (97.8% vs. 99.6%)The lowest patient I<O was 75.7% compared to 97.2% in controls.Patients spent a longer time in bed (67 min longer on average) than controls but slept a similar amount of time.Patients had longer sleep onset latency, mean sleep motor activity, wake after sleep onset, number of awakenings more than 5 and lower sleep efficiency compared than controls.I<O was significantly correlated with sleep motor activity (movements during sleep within a given time frame), wake after sleep onset, number of awakenings more than 5 and sleep efficiency (all p=0.0001).Younger patients went to bed, and woke up, later than older patients, and had a delay in the midpoint of sleep.The actigraphic parameter to best discriminate cancer patients was I<O. was associated with higher quality-of-life index Health/Functioning domain scores (r=0.34,p=0.05).Higher satisfaction with health was associated with more stable circadian structures.Fatigue was associated with a diminished circadian quotient (r=−0.40,p=0.04), rhythm quotient (4=−0.41,p=0.03) and nightday sleep balance (r=−0.52,p<0.01).A more robust rhythm was associated with less fatigue.A higher rhythm quotient was associated with less pain (r=−0.39,p=0.04).Loss of appetite was negatively associated with night-day sleep balance (r=−0.47,p<0.01).Peak activity significantly correlated with all the Power and Ferrans QoL index domains (all p≤0.02).Robustness of circadian measures reflects all quality-of-life measured aspects.
Du-Quiton et al., 2010)ere abnormal for all cancer patients compared to controls.Same figures for actigraphy parameters asDu-Quiton et al., 2010).Higher daytime activity was associated with lower PSQI daytime dysfunction (r=−0.61,p=0.006),higher PSQI sleep quality (r= −0.48, p=0.014), and less use of self-reported sleep medication (r=−0.58,p<0.003).Higher daytime inactivity was associated with more daytime dysfunction (r= 0.54, p=0.017), lower PSQI global sleep quality (r=0.41,p=0.014),and higher self-reported use of sleep medication (r=0.39,p=0.05).A higher 24-h rhythm quotient was associated with less daytime dysfunction (r=−0.58,2p<0.01).Patients who slept well during the night and less in the day slept for longer regardless (p<0.03)Higherlevels of night-day sleep balance were associated with less nighttime sleep disturbance (r=−0.Night-day balance was correlated with the EORTC QLQ C30 domains of role (r=0.56,p<0.01)and cognitive function (r=0.45,p=0.02).(Continued) All actigraphy parameters were abnormal for cancer patients compared to controls (p<0.05).Patients were 20%-50% less active than controls.Patients had at least 3 times longer daytime inactivity, 4 times longer sleep/inactivity periods in the day (20.9% vs. 4.7%) than controls.Patients had more fragmented sleep, longer waking episodes and less nighttime sleep than controls.The longest patient night sleep episode was less than half of the controls.Patients had lower sleep efficiency (79.8% vs. 95.9%),shorter longest nighttime sleep duration (91.7 min vs. 255.6min), less activity (126.9 accelerations/min vs. 182.6),shorter daytime wake time (797.5 min vs. 947.1 min), shorter sleep time at night (284.0 min vs. 417.8min), and shorter % of time asleep at night (72.5% vs. 93.0%)than controls (all p<0.05).Patients took longer to fall asleep at night (20.8 min vs. 12.1 min), were awake more in the night (95.0 min vs. 31.1 min), slept for longer in the day (208.8min vs. 47.1 min), and had longer longest daytime sleep periods (43.0 min vs. 23.6 min) than controls (p<0.05).Overall daytime and nighttime sleep were not associated with anxiety or depression.There were no statistically significant associations between sleep-activity circadian rhythm and anxiety or depression amongst inpatients.
≥3-day actigraphy (Actigraph)Patients were less active in the day on weekdays and weekends than controls (p<0.01).Physical-rest activity was not different to age-matched and recently hospitalised non-cancer patients.Subjectively scored physical function and pain were predicted by objectively measured physical activity (p<0.0001).All actigraphy parameters were significantly different between the two groups.Patients had lower median r24 values (0.28 vs. 0.46, p<0.001), longer wake after sleep onset (68.34 min vs. 25.67 min, The wake after sleep onset ranged between 48.2 and 70.9 min.The longest daytime napping was 100.3 min.Patients with no perceived sleeping difficulty had a mean sleep onset latency of 10.2 min, wake after sleep onset of 48.2 min, total wake time of 68.5 min, total sleep time of 479.9 min, time in bed of 548.4 min and sleep efficiency of 87.3%.Subjective assessments differed from objective assessment, significance not reported.No significant differences for sleep-wake parameters were noted between men and women. No circadian activity rhythm parameter was correlated with pain, fatigue, depression, or maladaptive sleep behaviours.A lower amplitude (r=0.24,p≤0.01), lower MESOR (r=0.27,p≤0.05), and later acrophase (r=−0.23,p≤0.01) was associated with poorer global QoL.A higher down-MESOR was associated with poor global (r=−0.31,p≤0.05) and functioning QoL (r=−0.30p≤0.05).A more robust rest-activity rhythm (higher amplitude (r=0.33,p≤0.05),MESOR (r=0.42,p≤0.01), and r-squared (r=0.24,p≤0.05) was associated with greater 24-h exposure to light intensity >1,000 lux.There were no significant differences between any rest-activity rhythm variable between ECOG 2 vs. 3.An increased I<O was significantly associated with greater values of global quality of life (p<0.0001),physical (p<0.0001), and social (p<0.0001)functioning but not role (p=0.02)functioning.
Median values were a total sleep time of 7hr22, sleep efficiency of 90.56%, sleep latency of 8m40s, wake after sleep onset of 46m15s, I<O of 96.9%, bathyphase (lowest activity time) of 02:33, and average activity of 99 accelerations/minute. Patients with subjective sleep complaints had more circadian disruption (lower I<O) compared to those without sleep complaints (p=0.005).40.1% of patients had subjective and circadian disruption (I<O <97.5%), 25.3% had subjective disruption alone, 14.8% had circadian disruption alone, and 19.8% had neither.Lowest health-related quality of life scores (including global, physical, social, and role functioning), and highest symptom scores (including fatigue and appetite loss), were seen in those with subjective and circadian disruption.Subjective difficulty sleeping was not associated with actigraphy sleep parameters.
Baseline (mean values):The r24 values were 0.42 (I) and 0.36 (C).A poor r24 was seen in 11/56 in the intervention group and 18/55 in the standard care group (26% overall).The I<O values were 94.68 (I) and 92.65 (C).A poor I<O was seen in 11/56 in the intervention group and 23/55 in the standard care group (30% overall).The total sleep time was 380.32 (I) and 395.06 (C).The sleep efficiency was 88.94 (I) and 88.36 (C).The sleep onset latency was 27.14 (I) and 31.85 (C).The wake after sleep onset was 45.86 (I) and 50.56 (C).Females had stronger and less fragmented rhythms than males.Significant gender differences were seen for interdaily stability (IS, a measure of rhythm stability) (p = 0.002), intradaily variability (IV, a measure of rhythm fragmentation) (p = 0.001), relative amplitude (RA, the difference between the mean of ten consecutive hours with the highest values and the mean of five consecutive hours with highest values divided by the combined value of both) (p = 0.001), circadian function index (CFI, combined IV/IS/RA) (p < 0.001) and I < O (p = 0.008).
a sleep latency of more than 30 min and 61% slept <5 h per night.88% of patients had a sleep efficiency of <85% and 91% were awake more than 30 min after sleep onset.Considering "good" and "poor" sleepers, actigraphy measures of sleep and measures of mood were not significantly different.Subjective and objective measures of sleep efficiency, sleep latency, sleep hours and wake after sleep onset were significantly different (all <0.05).There were no significant associations between the sleep diary and actigraphy variables of interest.
Actigraphy demonstrated a mean r24 of 0.37, median r24 of 0.41, mean I<O of 92.8 and median I<O of 94.2. 13 patients had an r24 >0.28, 10 patients had I<O >25% quartile Interindividual variation was noted in circadian rhythms of activity, hormonal, and haematological markers of circadian system function.
TABLE 2D Polysomnography-related circadian rhythms and their associations in patients with advanced cancer.

TABLE 2D Continued
Studies without a control groupLung cancer patients had higher index of awakenings lasting at least 60 seconds than breast cancer patients (3.9 vs. 2.4, p=0.002).