Article Text
PDF
Abstract
Aims Sleep Disturbance (SD) is a severe debilitating symptom in advanced cancer patients (ACP). However, routine screening of SD is uncommon. The primary aim of this study was to determine the optimal cutoff score for SD screening for Edmonton Symptom Assessment system (ESAS) sleep item using Pittsburgh Sleep Quality Index (PSQI) as a gold standard. We also determined the frequency of SD, obstructive sleep apnea symptoms (OSA) and restless leg syndrome (RLS) and factors associated with SD.
Methods We prospectively surveyed 180 consecutive ACP. Patients completed validated assessment for symptoms. We determined epidemiological performance, receiver operating characteristics, and correlations of SD.
Results SD according to PSQI was diagnosed in 112/180 (62%), and median (IQR) ESAS sleep was 5 (2-7). ESAS sleep ≥ 4 had a sensitivity of 74% and 80%, and specificity of 71% and 64% in the training and validation samples, respectively for screening of SD. The frequency of OSA was 61%; RLS was 38%. ESAS sleep was associated [r, p-value] with PSQI (0.61, <0.0001), pain (0.4, <0.0001); fatigue (0.35, <0.0001); depression (0.20, 0.006); anxiety (0.385, <0.0001); drowsiness (0.385, <0.0001), shortness of breath (0.24, <0.0014); anorexia (0.32, <0.0001), well-being (0.36, <0.0001). Multivariate analysis found well-being (OR per point 1.34, p=0.0003), pain (OR 1.21, p<0.0037), dyspnea (OR 1.16, p=0.027), and OSA (OR 0.31, P=0.003) as independent predictors of SD. There was no association between SD and survival.
Conclusions SD is frequent and ESAS SD item ≥ 4 has good sensitivity for SD screening.
- Cancer
- Symptoms and symptom management
- Supportive care
- Quality of life
- Clinical assessment
Statistics from Altmetric.com
Introduction
Sleep disturbance (SD) is one of the most common symptoms in advanced cancer patients (ACP) and is associated with a higher frequency of distress and mortality.1 ,2 The frequency of SD ranges from 24% to 95%.3 ,4 SD is experienced throughout the trajectory of illness5 ,6 and often increases during cancer treatment.7 SD contributes to and clusters with cancer-related fatigue, depression, anxiety and cognitive deficits and is associated with a significant worsening of the patient's health-related quality of life.3 ,5 ,6 ,8–10
Despite its high frequency and impact on quality of life, SD is underdiagnosed and not routinely screened in ACP receiving palliative care.11–13 Routine screening may improve early assessment and management of SD and its related symptoms and thereby the overall quality of life. Screening for the presence of sleep disorders in ACP may provide the basis for the application of treatment which will also contribute to the supportive care of these patients.1 Primary sleep disorders such as insomnia, sleep apnoea, restless leg syndrome (RLS) and circadian rhythm disturbance all have evidence-based treatments which have been shown to improve sleep quality.14 Sleep disorders have been shown to impact perception of pain, anxiety, mood disturbance and cognition, and therefore identification and management of sleep disorders may assist in the management of multiple cancer-related symptoms.15 ,16 The Pittsburgh Sleep Quality Index (PSQI) is a frequently used tool in clinical research to assess sleep quality,17 and the Edmonton Symptom Assessment System (ESAS) is frequently used as a screening tool to assess for symptoms in advanced cancer in routine cancer care.18 As ACP have a high frequency of symptoms, use of specific tools such as PSQI or other measures of SD that involve multiple questions can be cumbersome in routine clinical care. There is limited evidence of an appropriate assessment of cut-off of clinically relevant SD severity to be utilised using tools to assess symptoms in daily practice. In a preliminary study by our team, we found that a cut-off of 3/10 (0–10 scale) on the ESAS sleep item was suggestive of clinically significant SD. However, the patient population was highly selective (all patients had moderate-to-severe fatigue) and the study was not adequately powered for assessment of appropriate cut-off of clinically relevant SD severity, as it was a secondary objective. Hence, further studies are needed to evaluate screening tools for SD that can be used in routine clinical care.
There are also limited published data on the factors associated with SD in ACP such as patient characteristics, cancer-related symptoms, body mass index (BMI), daily opioid use as measured by the morphine equivalent daily dose, blood haemoglobin, albumin levels, obstructive sleep apnoea (OSA) and RLS. Prior studies in cancer and non-cancer settings have suggested their role in causation, precipitation or maintenance of SD. Therefore, the primary aim of this study was to determine the optimal cut-off score for SD screening for the ESAS sleep item using the PSQI as a gold standard in ACP. We also determined the frequency of patient-reported SD, OSA symptoms and RLS and factors associated with severity of patient-reported SD.
Methods
The institutional review board of The University of Texas MD Anderson Cancer Center (MD Anderson) approved the conduct of this prospective cross-sectional survey from October 2012 to June 2013. ACPs admitted in MD Anderson for at least 24 h were screened for eligibility and possible enrolment. Inclusion criteria included a diagnosis of advanced cancer, SD of ≥1/10 on a 0–10 scale, as measured by ESAS, normal cognition and ability to read, write and speak English. Normal cognition is based on the treating physician's assessment of cognition as per Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-V) criteria.
After obtaining signed informed consent, the eligible individuals completed a single interview with the research coordinator. During this interview, the patients completed the demographic data and the following questionnaires.
ESAS was designed by our group to assist in the assessment of 10 symptoms common in patients with cancer over the prior 24 h. The symptoms include pain, fatigue, nausea, depression, anxiety, drowsiness, shortness of breath, appetite, feelings of well-being and ‘other symptom’. In this study, we used the sleep item as the ‘other symptom.’ The severity of each symptom was rated from 0 to 10 on a numerical scale, 0 meaning that the symptom is absent and 10 meaning that it is of the worst possible severity. The instrument is both valid and reliable in the assessment of the intensity of symptoms in cancer populations.18 The ESAS Questionnaire was revised for this study. We used ‘Sleep Disturbance’ instead of ‘other symptom’
PSQI is the most widely used assessment tool to measure SD. In a recent meta-analysis of its use in clinical and non-clinical trials, a search for PSQI conducted in March 2104 returned 1512 articles.19 The next highest number for any sleep tool was 66 articles. It has been translated in over 40 languages and is widely used clinically to assess and monitor the effects of therapy. It differentiates ‘poor’ from ‘good’ sleep by measuring subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, use of sleeping medication and daytime dysfunction. Therefore, we used PSQI as a gold standard for subjective assessment of SD. Each item on a scale is graded from 0 to 3. The sum of the seven component scores is used for the global sleep score (range 0–21). A global sleep score of ≥5 was used to define SD. PSQI has internal consistency (Cronbach's α) of 0.83 overall.17
Insomnia Severity Index (ISI): The ISI is a seven item questionnaire designed to evaluate difficulties falling asleep, night-time awakening, early morning awakenings, impairment of daytime functioning due to sleep problems, noticeability of impairments, distress or worry caused by sleep difficulties, and dissatisfaction with sleep.20 Each item was rated using a five-point scale ranging from 0 (not at all) to 4 (very much), for a total score ranging from 0 to 28. A cut-off score of 8 was found to have optimal sensitivity and specificity (95% and 47%, respectively). The ISI has adequate psychometric properties and is sensitive to measure treatment response.20 The PSQI and ISI were administered with two time frames by asking patients to rate their SD for the past 24 h and past 30 days.
STOP-Bang Scoring Model: The STOP-Bang questionnaire is a validated screening tool for the detection of OSA. The eight questions have simple ‘yes’ and ‘no’ answers. A score of ≥3 predicts OSA with a sensitivity of 74.3% for moderate OSA and 79.5% for severe OSA when compared with gold standard polysomnography. The STOP-Bang test was developed initially as a screening tool for OSA in surgical patients. The questionnaire is short, easy to apply and has sensitivity for moderate-to-severe OSA between 92.9 and 100% when compared with the gold standard of polysomnography.
Screening for RLS: This single item-screening questionnaire was developed in 2007 as a rapid tool for evaluation of RLS.21 The question incorporates the clinical components used to make a diagnosis of RLS including characterisation of symptoms, timing and alleviating measures. It showed a sensitivity of 100% and a specificity of 96.8% when compared with conventional diagnostic criteria for the diagnosis of RLS.
Epworth Sleepiness Scale (ESS): The ESS is a simple questionnaire that measures the general level of daytime sleepiness. It has eight questions and patients rate on a four-point scale (0–3) their usual chances of falling asleep in eight different situations or activities.22 ESS has been validated primarily in OSA. However, it has been used in patients with and without cancer for the detection of sleep disorders such as narcolepsy, idiopathic hypersomnia and excessive daytime sleepiness.23
Hospital Anxiety Depression Scale (HADS): It is well known that SD are strongly influenced by psychosocial factors. Depression, anxiety and psychosocial stress have consistently been found to be risk factors for insomnia and other sleep disorders. The hospital anxiety and depression scale was developed in 1983 for screening of depression in medical patients. It has been studied and validated in patients with cancer, showing it to be a useful screening instrument.24
Statistical considerations
The primary aim of this study was to determine the optimal cut-off score for clinically significant SD as measured by ESAS sleep. We used PSQI as a gold standard for subjective assessment of SD. The clinically significant cut-off for the PSQI is 5.17 Patients were randomly assigned to either a training data set or a validation data set. We used PSQI from the training data set to calculate sensitivity and specificity (along with 95% score CIs) for all possible cut-off values of the ESAS sleep item, calculate area under the receiver-operating characteristic (ROC) curve and chose the best clinical cut-off where the best clinical cut-off was defined as the cut-off which maximises the sum of sensitivity and specificity subject to the constraint that sensitivity is at least 70%. A sample size of 180 patients was chosen for this study, of whom 120 were used to determine an optimal cut-off point from the ROC curves. The sample size was based on a past study by our team in ACP1 which showed that approximately 75% of the patients would have had clinically relevant SD. A sample size calculation showed that a sample of 180 patients would be able to detect SD with a moderate sensitivity of 80% (72 of the 90 patients with clinically relevant SD), with 95% score CI for sensitivity of 80%±8.3%. To address the first secondary aim, we estimated the proportion of patients with SD, OSA and RLS. To address the second secondary aim, we calculated correlation among ISI, PSQI, HADS, ESS and ESAS symptoms. In order to assess independent predictors associated with SD using the ESAS sleep item, we initially conducted a univariate logistic regression analysis. Variables that were significant at the 10% level in univariate regressions were included in a full multivariate model to which we applied stepwise logistic regression to arrive at a parsimonious model. SD and OSA were also tested to determine whether they were associated with overall survival (OS) using the Kaplan-Meier estimation method. A p value of ≤0.05 was considered statistically significant. Analyses were performed using SAS (V.9.3; SAS Institute Inc, Cary, North Carolina, USA) and IBM SPSS Statistics (V.19, IBM Armonk, New York, USA).
Results
Study participation
A total of 1057 consecutive ACP were screened; 483 were excluded. The reasons for exclusion were as follows: (1) patients <18 years (n=4); (2) no advanced cancer (n=8); (3) delirium (n=270); (4) non-English speaking (n=114); and (5) SD of 0 on a 0–10 ESAS scale (n=72). Of the 589 patients who were eligible and approached, 180 were enrolled and 409 declined to participate. Reasons for refusal were: (1) symptom distress (n=88), (2) not interested (n=277), (3) clinician refused to allow the patient to be approached (n=24) and (4) missing data (n=2).
Of the 180 patients assessed, 91/180 (51%) were female, SD according to PSQI was diagnosed in 112/180 (62%), the median (IQR) ESAS sleep item score was 5 (2–7), and PSQI for 30 days was 8 (5–11; table 1). The frequency of positive screening for OSA was 61%; and RLS was 38%. ESAS sleep item ≥4 had a sensitivity of 74% and 80%; and specificity of 71% and 64% in the training and validation samples, respectively. Table 2 shows the possible cut-off values for the ESAS sleep item range from >0 to <10 where a cut-off of >0 would be all patients with SD. The ROC curve (figure 1) shows an arae under the curve of 0.77. The best ESAS sleep cut-off score using our predefined criteria (cut-off which maximises the sum of sensitivity and specificity) was a score ≥4. The ESAS sleep score ≥4 had a sensitivity of 74% and a specificity of 71% based on our 120 patient training data set. We validated our cut-off using the 60 person validation data set, achieving a sensitivity of 80% and a specificity of 64%. ESAS sleep was associated (r, p value) with PSQI (0.61, <0.0001); ESAS pain (0.4, <0.0001); fatigue (0.35, <0.0001); depression (0.20, 0.006); anxiety (0.385, <0.0001); drowsiness (0.385, <0.0001), shortness of breath (0.24, <0.0014); anorexia (0.32, <0.0001) and a feeling of well-being (0.36, <0.0001; table 3).
Patient characteristics
Distribution of possible cut-off values of ESAS sleep item for clinically significant sleep disturbance
Correlation of ESAS—sleep item with sleep variables
Receiver-operating characteristic curve for Edmonton Symptom Assessment System sleep item using Pittsburgh Sleep Quality Index as a gold standard.
Table 4 shows univariate logistic regressions analysis predicting SD. Multivariate analysis of these data showed ESAS feeling of well-being (OR 1.34 per point, p=0.0003), ESAS pain (OR 1.21 per point, p<0.0037), ESAS shortness of breath (OR 1.16 per point, p=0.027) and OSA (OR 0.31 per point, p=0.003) as independent predictors of SD (Tjur's R2=0.47).We found no significant association between SD (p=0.34) or OSA scores (p=0.82) with OS in days (Median (IQR)=37 (13–84).
Factors associated with sleep disturbance*
Discussion
Our study findings suggest the importance of routine screening of SD using a single-item questionnaire such as ESAS sleep item, as SD is associated with distressing symptoms such as pain, shortness of breath, fatigue, depression, anxiety, anorexia and poor feeling of well-being in ACP,25 ,26 and it could also be a symptom of a more serious sleep disorder that warrants intervention. Our study confirms a high frequency of SD in ACP. This finding is important as prior literature suggests that SD is underdiagnosed.4 Our study also confirms the association of SD and severe symptoms such as pain, shortness of breath, fatigue, depression, anxiety, anorexia and poor feeling of well-being that are major contributors to significant symptom distress in ACP.25 ,26
Our data support the routine use of ESAS (with sleep item) for the screening of these patients. For screening purposes, the most important epidemiological variable is sensitivity, and our data in exploratory and validation samples suggest that a cut-off of ≥4 ESAS sleep item provides good sensitivity with reasonable specificity (table 2). This suggests that all patients who score ≥4 ESAS sleep item should undergo further assessments for diagnosis of SD. It could be argued that specificity of the ESAS sleep item score is not ideal but is clearly superior to the current standards of routine clinical assessment where no routine screening tools for SD are used.27 The relative ease in obtaining ESAS sleep item score also makes it an attractive clinical tool for busy clinicians. This tool may be used to identify patients who require further screening for causes of SD such as insomnia, OSA or RLS who may need referral to a sleep specialist.
Our findings in this study suggests that SD should be assessed in all ACP due to its high frequency especially in patients with high symptom distress are more likely to have SD, therefore, be suspected to have SD. On the basis of the results of this study, we are, however, unable to state if SD is the consequence of symptom distress or if SD was aggravating the symptom distress. At any rate, the association of SD and symptom distress provides important clinical information that suggests that both should be assessed and managed simultaneously (tables 3 and 4). The management of SD in ACP is complex due to the potential for multiple drug interactions and toxicity, and ideally management of SD should be based primarily on non-pharmacological interventions such as sleep hygiene, cognitive–behavioural therapy and elimination of drugs capable of causing SD (eg, corticosteroids, selective serotonin re-uptake inhibitors (SSRI's) antihypertensives such as ACE inhibitors, and α-blockers).9 ,16 Pharmacological interventions may be appropriate only in cases of refractory SD. Further studies are needed. The reciprocal relationship between SD and symptom distress also suggests that SD may be a target for intervention to improve the general well-being and quality of life in ACP. In addition, further research is needed to target the common underlying pathophysiologic mechanisms mediating SD and symptom distress such as circadian rhythm and immune dysregulation so as to effectively mitigate SD.28
In our study, we found a very high frequency of patients who screened positive for OSA according to the screening tool STOP-Bang questionnaire, which this was an unexpected finding for this sample. Prior studies found that OSA is associated with and sometimes confused as delirium.29 ,30 It is associated with an increased risk of cancer mortality.31 ,32 ACP have multiple factors associated with OSA including BMI changes, physical changes in anatomy due to cancer in the head and neck region, central nervous system tumours, medications such as opioids and sedatives, history of smoking, alcohol abuse and comorbidities such as hypothyroidism.33 However, the use of the STOP-Bang questionnaire has not been validated specifically in patients with cancer, but in surgical patients. There are seven of eight questions which are fairly generalisable to all subjects and not expected to be uniformly affirmative or negative in patients with cancer. Further confirmatory studies are needed to validate these findings.
On the basis of data from this study, it could be suggested that patients with significant SD should be proactively treated, and if they are refractory to treatment, a low threshold should be adopted to investigate for OSA and confirmatory tests such as polysomnography should be performed when appropriate.
In this study, a high frequency of RLS was found in ACP.34 The exact aetiology is not clear, but use of antidepressant medications, especially the, caffeine use and anaemia (iron deficiency), are common contributors to RLS and are frequently seen in patients with advanced cancer.35 Neuropathy due to chemotherapy may also contribute to RLS symptoms inACP.34
The study has several drawbacks including the inherent inability of the one-time survey to detect SD over time. Future studies are needed to better characterise SD in ACP using objective assessments such as polysomnography and/or actigraphy, proteomic and genomic correlates (eg, inflammatory cytokines such as interleukin (IL)-1β, IL-6, tumour necrosis factor-α) which would be useful to evaluate the associations with specific subjective SD with objective data and laboratory correlates.36 ,37 In addition, we have not collected information regarding all known medications associated with SD such as sedatives, antihypertensives, antipsychotics and antidepressants. The identification of these associated factors would result in the development of effective treatments for this distressing symptom in ACP targeting the causative mechanism.
Conclusion
SD is frequent and underdiagnosed in ACP. ESAS SD item ≥4 has good sensitivity for SD screening. Our study confirms the association of SD and severe symptoms such as pain, fatigue, shortness of breath, depression, anxiety, anorexia and a worse feeling of well-being. On the basis of this study, we recommend routine screening of SD using single item questionnaires such as ESAS sleep item in order to better treat ACP. Future studies are warranted to confirm the findings of this study, and this may result in the development of effective strategies alleviating SD in advanced cancer by targeting the causative mechanisms.
References
Footnotes
Funding The corresponding author SY is supported in part by the American Cancer Society (RSG-11–170–01-PCSM).
Competing interests None declared.
Patient consent Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.