Poor sleep quality among newly diagnosed head and neck cancer patients: prevalence and associated factors

Data of the prospective NETherlands QUality of life and BIomedical Cohort study in head and neck cancer (NET-QUBIC) [18] was used. In the NET-QUBIC study, 739 newly diagnosed HNC patients from HNC centers in 5 university medical centers and 2 of their satellite hospitals in the Netherlands were included between March 2014 and June 2018. Inclusion criteria were as follows: being diagnosed with squamous cell carcinoma of the oral cavity, oropharynx, hypopharynx, larynx, or neck lymph node metastasis of an unknown primary tumor; being 18 years or older; having curative treatment intention; and being able to write, read, and speak Dutch. Exclusion criteria were having severe psychiatric comorbidity (e.g., schizophrenia, Korsakoff’s syndrome, severe dementia), lymphoma, thyroid cancer, nasopharyngeal cancer, malignancy of skin, or malignancy of salivary glands. The NET-QUBIC study was approved by the Medical Research Ethics Committee of the coordinating center (Amsterdam UMC, location VUmc, document number: 2013.301[A2018.307]-NL45051.029.13). Detailed procedure of the NET-QUBIC study can be found elsewhere [18].

The ongoing NET-QUBIC study encompasses measurements at baseline and at 3-, 6-, 12-, 24-, 36-, 48-, and 60-month follow-up. In the present study, we used the baseline data, which were collected shortly after the diagnosis and before cancer treatment was started.

Sleep quality was measured using the PSQI, which was filled in by the patients themselves [19]. Its validity and reliability have been confirmed both in general population [20] and in cancer patients [21, 22]. PSQI consists of 19 items across seven components of sleep quality and disturbances: (1) subjective sleep quality (i.e., one item “How would you rate your sleep quality overall?”), (2) sleep onset latency (i.e., two items asking time needed to fall asleep and its frequency in a week; poor sleep onset is defined as needing ≥ 30 min to fall asleep [23]), (3) sleep duration (i.e., one item “How many hours of actual sleep do you get at night?”; ≤ 6 h is associated with worse survival among cancer patients [24]), (4) sleep efficiency (i.e., a percentage calculated by dividing time asleep by time spent in bed and multiplied by 100; < 85% indicates poor efficiency [23]), (5) sleep disturbances (i.e., ten items about specific reasons for the sleep disturbances and their frequency), (6) use of sleep medication (i.e., one item “How often have you taken medicine to help you sleep?”), and (7) daytime dysfunction (i.e., two items asking the frequency of staying awake during daytime activity and the extent of difficulty to maintain enthusiasm to get things done) [19]. A total score (also called global score) is calculated by first scaling each component score into a 0 to 3 score then summing all component scores, resulting in a score ranging from 0 to 21; higher scores indicate worse sleep quality. Our main outcome, poor sleep quality, is defined by a total PSQI score of > 5 [19, 20]. In addition, we examined questions in the PSQI (not included in the total score calculation) which assessed the frequency of the respondent’s bed partner/roommate noticing certain behaviors during the respondent’s sleep. This information is clinically relevant as an indication of specific sleep disorders such as sleep apnea [25].

We examined the following sociodemographic factors: sex and age (from medical records), living situation (living together/alone, from interview), education level (low/middle/high according to the standard classification of education level in the Netherlands [26], from interview), and having a bed partner (yes/no, from the PSQI). The interview was conducted by trained field workers during house visit measurements. This interview also included other outcomes which were out of the scope of the present study [18]. Clinical characteristics (i.e., HNC subsite and stage, comorbidity, and performance status) were retrieved from medical records. Comorbidity was scored using the adult comorbidity evaluation (ACE-27). The ACE-27 measures the number and severity of 27 medical conditions, and is summarized into four categories: no comorbidity, mild comorbidity, moderate comorbidity, or severe comorbidity [27]. The ACE-27 has been validated among HNC patients [28]. The patients’ performance status (i.e., the patient’s level of functioning seen from their daily activity, physical ability, and self-care) was measured using the one-item Eastern cooperative oncology group (ECOG) score, which ranges from 0 (fully active) to 4 (completely disabled) [29].

Coping styles were measured by the self-reported 47-item Utrecht coping list (UCL) [30]. Each item ranges from 0 (never or seldom) to 3 (very often). The UCL measures active coping (i.e., evaluating the situation from all perspectives and taking action to solve the problem; 7 items), palliative reaction (i.e., finding distractions against the problem and seeking for manners to feel better; 8 items), avoidance coping (i.e., avoiding the situation and letting it to be solved by itself; 8 items), seeking social support (i.e., seeking help and understanding from the others, expressing worries; 6 items), passive coping (i.e., taking the blame for the situation, worrying about things in the past, withdrawing into oneself; 7 items), expression of emotions (i.e., expressing anger or abreaction; 3 items), and comforting thoughts (i.e., assuring one’s self that things will get better, that things could have been worse, or that the others’ may also have similar difficulties; 5 items) [30, 31]. For each coping style, a sum score was calculated, where a higher score indicates higher extent the specific coping style.

HNC symptoms were self-reported using the European organization for research and treatment of cancer quality of life questionnaire - HNC-specific module (EORTC QLQ-H&N35) [32]. All symptom scales were included: oral pain (4 items), swallowing problems (4 items), sense problems (2 items), speech problems (3 items), trouble with social eating (4 items), trouble with social contact (5 items), and less sexual interest and enjoyment (2 items). Also, we included Likert-scale single items measuring teeth problems, problems with opening mouth, dry mouth, sticky saliva, coughing, and feeling ill, as well as dichotomous single items measuring use of painkillers, use of nutritional supplements, use of feeding tube, weight loss, and weight gain. Each symptom scale and Likert-scale single item was converted into a score ranging from 0 to 100, according to the EORTC guidelines. A higher score indicates a higher level of symptoms or problems.

The following lifestyle factors were examined: physical activity, body mass index (BMI), smoking status, and alcohol intake. Physical activity was assessed using the 13-item physical activity scale for the elderly (PASE) with total score ranging from 0 to 400; a higher score indicates higher physical activity [33]. The validity and reliability of this patient-reported outcome measure has been previously confirmed among cancer patients [34]. BMI was measured by a trained field worker using a standardized procedure during house visit. Smoking status (not smoking/smoking every day at the time of assessment) and excessive alcohol consumption (> 14 units of alcohol per week for women or > 21 units of alcohol per week for men [35]) were self-reported using study-specific items.

To examine the differences among patients with good and poor sleep, first, their sociodemographic, clinical, and lifestyle characteristics, as well as coping styles and HNC symptoms, were compared using chi-square (categorical variables) and t tests or Mann-Whitney tests (continuous variables). Then, forward multivariable logistic regression analyses were performed on each category of associated factors separately (i.e., sociodemographic, coping style, clinical factors, HNC symptoms, and lifestyle factors) using a p value for entry of < 0.05. Subsequently, we performed logistic regression analyses using a forward selection of all significant variables across all categories, resulting in the final model. In all regression models, HNC symptom and item scores which ranged from 0 to 100 were rescaled into 10-point increments. All statistical analysis was performed using the IBM SPSS Statistics for Windows, Version 25.0 (Armonk, NY: IBM Corp. 2017).

Among the 739 eligible patients, 179 (24%) did not have available PSQI scores and were excluded from our analyses. Age, gender, education level, HNC subsite, and HNC stage were similar between patients with (n = 560, 76%) and without PSQI data. However, patients providing PSQI data more often lived with others, had better ECOG performance status, and less comorbidity compared with the patients not providing PSQI data (Table 1). There were 171 (21%) missing values on BMI; therefore, this variable was not included in the analysis. For all other variables, missing values ranged from 7 (1.3%) to 43 (7.7%). Since only 5 patients (0.9%) used a feeding tube, we did not include this variable in the analysis.

Our final sample of patients with PSQI data were mostly men (75%), on average 63 years old (SD = 9), and low educated (42%). Most patients (79%) lived with others (i.e., with a partner, child, and/or housemate) and 70% had a bed partner. About one-third (32%) had no comorbidity (Table 3).

The median PSQI total score was 5 (IQR = 3–8). There were 246 HNC patients who were classified as having poor sleep (PSQI > 5; 44%). The median time needed to fall asleep (i.e., sleep latency) was 13 min (IQR = 8–30) and the mean sleep duration was 7 h (SD = 1). Among all patients, 13% slept ≤ 6 h in a night, 42% had < 85% sleep efficiency, 16% could not fall asleep within 30 min at least three nights a week, and 45% reported nighttime or early morning awakening at least three times a week. At least once in a week, 15% of the patients used medication to improve their sleep and 5% experienced difficulties staying awake during the day. Questions answered by 417 bed partners or roommates revealed that 173 (43%) patients snored loudly and 13% of the patients had long breathing pause at least once a week. Mean (SD) or median (IQR), as well as the proportions of each component score, are presented in Table 2.

Univariate analyses revealed that patients with poor sleep (n = 246) were more often women, younger, diagnosed with cancer in the oral cavity, using painkillers, or using nutritional supplements (Table 3). Furthermore, compared with good sleepers, poor sleepers used less active coping and more palliative reaction, passive coping, and expression of emotions. Also, poor sleepers had worse scores of all HNC symptoms except for speech problems, problems with opening mouth, coughing, and weight changes.

Multivariate logistic regression models on the separate domains of risk factors showed that poor sleep was significantly associated with sociodemographic factors (younger age, being female), coping style (more passive coping), clinical characteristics (HNC subsite, especially cancer in the oral cavity and oropharynx compared with larynx), and HNC symptoms (oral pain, less sexual interest and enjoyment, and feeling ill). Combining all of these significant variables in a logistic regression model, we found that younger age (odds ratio [OR] per increasing year of age = 0.98, 95% CI = 0.96−1.00, p value = 0.049), being female (OR = 2.6, 95% CI = 1.7–4.1, p value < 0.001), higher passive coping style (OR = 1.18, 95% CI = 1.09–1.28, p value < 0.001), more oral pain (OR = 1.10, 95% CI = 1.01–1.19, p value = 0.023), and less sexual interest and enjoyment (OR = 1.13, 95% CI 1.06–1.20, p value < 0.001) corresponded to greater odds of having poor sleep (Table 4).

The strength of this study was that we used a large sample of 560 HNC patients from multiple hospitals in the Netherlands to investigate the association between sleep problems and various sociodemographic, clinical, and lifestyle factors as well as coping style and HNC symptoms. Moreover, our study was the first study so far to report on the prevalence of different components of sleep quality (as measured by the PSQI) among HNC patients before starting treatment. However, we found differences between patients who were included because their PSQI data were available (n = 560) compared with those that did not (n = 179); our study population consists of HNC patients with better performance status and less severe comorbidity. Our study may, therefore, underestimate the real prevalence of poor sleep quality among newly diagnosed HNC patients. In addition, the NET-QUBIC participants were not completely representative for the overall HNC population in the Netherlands [18], which may limit generalizability of our study findings to all HNC patients.