INTRODUCTION
Insomnia is a common condition among patients with cancer, with nearly half of patients with a recent cancer diagnosis suffering from insomnia symptoms [1]. In this population, insomnia has been associated with psychiatric illness—particularly anxiety and depressive disorders [2], physical symptoms of pain or gastrointestinal symptoms [3], or anticancer modalities such as chemotherapy, radiotherapy, and hormone therapy [4]. Additionally, multiple studies have suggested that poor sleep and disruption of the circadian rhythm can play a role in tumorigenesis and cancer progression that reinforces the reciprocal causative approach [5].
Based on the two-level model of arousal by Ong et al. [6], an individual’s primary and secondary arousal influence their sleep disturbance. Primary arousal refers to “cognitions” and contents that directly interfere with sleep, which includes dysfunctional beliefs about sleep. Conversely, secondary arousals refer to “metacognition” and how individuals interpret their cognitive activity [6,7]. Dysfunctional beliefs about sleep are unreasonable, faulty, and negatively toned cognitive activities regarding reaching the expectations about sleep requirements and its impact on health and/or daytime functioning if these expectations are not met [8 ]. These beliefs are associated with poorer sleep quality, longer sleep latency, and more daytime dysfunction [9]. In turn, those beliefs and perceptions contribute to an individual’s emotional distress, aggravate their arousal, and reinforce the vicious cycle of insomnia [10]. Patients with cancer may have dysfunctional beliefs about sleep such as that their sleep problem might impair their immune function or lead to disease progression [11]. Their dysfunctional beliefs can influence their insomnia and even fear of disease progression [12]. Additionally, metacognition reportedly affects depression, anxiety [13], fear of progression [14], and even reproductive concerns [15] in patients with cancer; thus, we can speculate that metacognition on insomnia will also influence their insomnia or sleep disturbance. However, the role of metacognition on insomnia in patients with cancer is still unexplored.
The concept of metacognition refers to the process of reflecting upon one’s own mental states and cognitive processes, allowing an individual to regulate and modify mental states and cognitive processes [16]. It shifts the emphasis from thought contents to cognitive processes. Regarding insomnia, secondary arousal is characterized by the cognitive processes that occur in response to thoughts and cognition, e.g., worry and rumination directed at beliefs that have been amplified by primary arousal. Among patients with cancer, however, the role of sleep-related metacognitive process in their sleep disturbance is still not well elucidated.
Attention bias [17] can be observed among individuals with sleep disturbance. The literature suggests that patients with insomnia may have an attention bias toward sleep-related cues, which could be extended to sleep-related worry [17]. Sleep-related negative information [18] can be emphasized due to the worry regarding insomnia during the pre-sleep period. There is a possibility that attention bias contributes to the misperception of sleep-wake cycles. Patients with insomnia often go to bed early in the evening to fall asleep more easily. However, based on a two-process model of insomnia [19], early bedtime does not guarantee early sleep onset [20]. The patients also desperately want at least some sleep time and unconsciously want to sleep for a long time. The DBST index, the discrepancy between the desired time in bed (dTIB) and desired total sleep time (dTST), was developed based on this idea [21]. We previously observed the association between the DBST index and insomnia severity among the general population [21] and patients with cancer [22]. We found that patients with insomnia unconsciously have a high discrepancy between their dTIB and dTST, which is the opposite of how individuals without insomnia estimate their dTIB and dTST (i.e., “I want to have 7 hours of sleep. Um. For that, I must go to bed at 11 pm and wake-up at 6 am.”). Thus, the DBST index can be an index for the attention bias of patients with insomnia. Similarly, in the clinic, cancer patients with insomnia report a high discrepancy between dTIB and dTST, as opposed to those without insomnia who report low discrepancy.
We previously proposed the Discrepancy-Cognitive Arousal (DCA) model of insomnia among the general population [23] that showed the influence of the DBST index (attention bias) on insomnia severity through the two-level model of cognitive arousals (dysfunctional beliefs about sleep and sleep-related metacognitive process) [6]. And it is meaningful to explore the feasibility of the DCA model of insomnia among cancer patients. To measure metacognitive process in relation with insomnia can be done with the Metacognition Questionnaire-Insomnia (MCQ-I) [24]. The MCQ-I scale includes 65 items that evaluate cognitive beliefs about sleep, and owing to its large scale, applying this scale in clinical practice may be difficult, especially if regular follow-up is desired or for physically-ill patients who cannot easily undergo the lengthy interview. We previously reported the shortened forms of the MCQ-I as Metacognition Questionnaire-Insomnia-6 items (MCQI-6) and Metacognition Questionnaire-Insomnia-14 items (MCQI-14) with the help of the random forest machine learning algorithm [25]. These shortened versions have been validated among the general population. However, it has not been validated among patients with cancer. This study aimed to explore the reliability and validity of the MCQI-6 and MCQI-14 questionnaires among patients with cancer and examine the feasibility of the DCA model of insomnia among patients with cancer.
METHODS
Participants and procedure
This study was conducted as an anonymous online survey between March and June 30, 2022 at the Asan Medical Center Cancer Institute, Seoul, South Korea. Patients with cancer voluntarily responded to surveys delivered via a link on the poster at the Asan Medical Center Cancer Institute. We provided an e-gift coupon valued at approximately $5 as compensation for their participation. Sample size estimation was performed based on the rule of thumb, with at least 10 participants assigned for each scale item [26]. Accordingly, we aimed to collect responses from at least 140 participants for the 14 items, and we collected a total of 154 responses from patients with cancer. We collected participants’ demographic information on age, sex, and marital status, as well as cancer-related information on cancer type, cancer stages, and current treatment modalities. The protocol of this study was approved by the Institutional Review Board of Asan Medical Center (2022-0054), and the need to obtain informed consent was waived.
Measures
Shortened versions of MCQ-I; MCQI-6 and MCQI-14
Both the MCQI-6 and MCQI-14 were reliable and valid brief multidimensional measures of metacognitions about insomnia that are more convenient to use than the original MCQ-I. In a previous study, a multigroup confirmatory factor analysis (CFA) revealed that the MCQI-6 and MCQI-14 can both measure one’s metacognition on sleep in the same way across patients with insomnia or poor sleep. In the previous study, the Cronbach’s alpha for insomnia and poor sleep were 0.843 and 0.921, respectively [25].
Insomnia Severity Index
The Insomnia Severity Index (ISI) is a self-report scale that can assess one’s insomnia severity [27,28]. All 7 items were included, and these can be rated on a five-point Likert scale. The total score ranges from 0 to 28, and a higher total score indicates a severe degree of insomnia. Among this sample, the Cronbach’s alpha was 0.922.
Cancer-related Dysfunctional Beliefs and Attitudes about Sleep-14 items
The Cancer-related Dysfunctional Beliefs and Attitudes about Sleep-14 items (C-DBAS-14) is a self-report scale developed to assess cancer-related dysfunctional beliefs about sleep in patients with cancer [29]. It is a composite scale of Dysfunctional Beliefs and Attitude about Sleep-16 items [10] and Cancer-related Dysfunctional Beliefs about Sleep scale [11]. All 14 items can be rated using 100-mm visual analog scales, and the average total score can be calculated by adding the scores for all 14 items. In this study, we applied the scale as a Likert scale (0–10). A higher total average score indicates a higher degree of cancer-related dysfunctional beliefs about sleep. The Cronbach’s alpha among this sample was 0.951.
The DBST index
The DBST index is defined as a discrepancy between dTIB and dTST [30]. The dTIB was calculated based on the patients’ responses to the question “From what time to what time do you want to sleep?”, and the dTST was calculated based on the responses to “For how many hours do you want to sleep in a day?” The DBST was calculated as (desired hours of time in bed) - (desired hours of total sleep time).
Statistical analysis
First, the CFA with a diagonally weighted least squares estimator was conducted to examine the construct validity of the MCQI-6 and MCQI-14 among patients with cancer. The normality assumption was checked using the skewness and kurtosis for an acceptable limit of range ±2 [31]. Sampling adequacy and data suitability were checked using the Kaiser-Meyer-Olkin (KMO) value and Bartlett’s test of sphericity. In the CFA, a standardized root-mean-square residual value ≤0.05, root-mean-square-error of approximation (RMSEA) value ≤0.10, and comparative fit index (CFI) and Tucker Lewis index (TLI) values ≥0.90 were set as the satisfactory model fit [32,33]. Multi-group CFA with configural invariance testing was performed to examine whether both the MCQI-6 and MCQI-14 scales can measure the metacognition on insomnia across patients having clinical insomnia (ISI ≥15) or not. Convergent validity was explored using the Pearson’s correlation coefficient r. Reliability of internal consistency was assessed using the Cronbach’s alpha and McDonald’s Omega. We used a graded response model, and the Rasch analysis was performed to assess the psychometric properties of the MCQI-6 and MCQI-14 such as the item discrimination, difficulties, and factor structure.
Second, we examined the feasibility of the DCA model of insomnia among patients with cancer. The bootstrap method with 2,000 resamples was implemented to explore whether the dysfunctional beliefs about sleep and metacognition on insomnia may mediate the influence of the DBST index on insomnia severity. The SPSS version 21.0 (IBM Corp., Armonk, NY, USA), AMOS version 27 (IBM SPSS., Inc, Chicago, Illinois, USA), JASP version 0.14.1.0 software (JASP Team, Amsterdam, The Netherlands), and RStudio (Posit, Boston, MA, USA) were used for statistical analysis.
RESULTS
Demographic characteristics and rating scales scores among 154 patients with cancer are presented in Table 1.
The shortened 6-item of MCQ-I: the MCQI-6
The normality of each item was confirmed based on the skewness ranging -0.092 to -0.687 and kurtosis ranging -0.656 to -1.036. The KMO value was 0.873, and the Bartletts’ sphericity was significant (p<0.001). The MCQI-6 had a good model fit for the single factor model (χ2=6.965, df=9, p=0.641, CFI=1.000, TLI=0.996, RMSEA=0.000, SRMR=0.049). Factor loadings ranged between 0.666 and 0.831 (Tables 2 and 3). Multigroup CFA results showed scalar level invariance across sex and having insomnia symptoms (ISI ≥15) or not. Table 3 shows that the MCQI-6 had a good internal consistency reliability (the Cronbach’s alpha of 0.900 and McDonald’s omega of 0.902). The floor and ceiling effects for the MCQI-6 were 2.60 and 0.65, respectively. The corrected item-total correlation ranged from 0.629 to 0.775. The total score of the MCQI-6 was significantly correlated with MCQI-14 (r=0.97, p<0.01), ISI (r=0.68, p<0.01), C-DBAS-14 (r=0.78, p<0.01), and DBST index (r=0.21, p<0.05). The total MCQI-6 score was significantly higher in participants with ISI ≥8 (t[152]=9.44, p<0.001) (Table 4).
Graded response model outputs are presented in Supplementary Table 1 (in the online-only Data Supplement). Non-significant S-χ2 values (p-values adjusted for false discovery rate) and RMSEA values suggested that all items belonged to the same latent construct. Regarding slope parameters (α), all items have a very high slope. Slope parameters ranged between 1.814 and 2.865 (mean=2.463). These very high slopes indicated that all items were highly efficient in discriminating among individuals assessed by the MCQI-6. The threshold coefficients (b) in Supplementary Table 1 (in the online-only Data Supplement) and Figure 1A show that item 51 was the most difficult item. For item 51, a higher latent trait or theta had to endorse Likert-type response options—from “agree moderately” to “agree very much,” compared to other items. The scale information curve (Figure 2) suggested that the MCQI-6 would be efficient to assess insomnia among individuals with theta levels between -2.0 and 1.25. Additionally, the MCQI-6 has a good Item Response Theory (IRT) reliability (0.940).
Table 3 and Supplementary Table 2 (in the online-only Data Supplement) show the Rasch analysis outputs. Table 3 shows that the MCQI-6 had acceptable item and person reliability (0.811 and 0.856, respectively) and item and person separation index (2.068 and 2.439, respectively). Both infit and outfit mean-square instead of t-standardized fit statistics (MnSqs) (Supplementary Table 2 in the online-only Data Supplement) ranged between 0.83 and 1.22, 0.81 and 1.28, respectively. These MnSqs suggested a good model fit. Item difficulty values ranged between -0.35 and 0.64. Item 60 was the least difficult item, whereas item 23 was the most difficult item in the MCQI-6.
The shortened 14-item of MCQ-I: the MCQI-14
Normality was confirmed based on the skewness ranging -0.014 to -0.687 and kurtosis ranging -0.656 to -1.220. The KMO value was 0.948, and the Bartletts’ sphericity was significant (p<0.001). The MCQI-14 had a good model fit for the single factor model (χ2=40.616, df=77, p=1.000, CFI=1.000, TLI=0.994, RMSEA=0.000, SRMR=0.050) (Table 3). Factor loadings ranged between 0.686 and 0.839 (Table 2). Multigroup CFA results showed scalar level invariance across sex and having insomnia symptoms or not. Table 3 shows that the MCQI-14 had good internal consistency reliability (the Cronbach’s alpha of 0.946 and McDonald’s omega of 0.947). The floor and ceiling effects for the MCQI-14 were 2.60 and 0.65, respectively. The corrected item-total correlation ranged from 0.663 to 0.813. The total score of the MCQI-14 was significantly correlated with the ISI (r=0.68, p<0.01), C-DBAS-14 (r=0.81, p<0.01), and DBST index (r=0.21, p<0.05) (Table 4). The total MCQI-14 score was significantly higher in participants with ISI ≥8 (t[152]=9.31, p<0.001).
Graded response model outputs for the MCQI-14 are presented in Supplementary Table 2 (in the online-only Data Supplement). Non-significant S-χ2 values (p-values adjusted for false discovery rate) and RMSEA values suggested that all items belonged to the same latent construct. Regarding slope parameters (α), item 12 had a high slope, and the rest of the items had a very high slope. Slope parameters ranged between 1.693 and 3.098 (mean=2.261). These slopes indicated that all items were efficient in discriminating among individuals assessed by the MCQI-14. Threshold coefficients (b) in Supplementary Table 3 (in the online-only Data Supplement) and Figure 1B show that item 50 was the most difficult item. For items 31, 42, and 50, a higher latent trait or theta had to endorse Likert-type response options—from “agree moderately” to “agree very much” compared to other items. The scale information curve (Figure 2) suggested that the MCQI-14 would be efficient to assess insomnia among individuals with theta levels between -2.25 and 2.00. Additionally, the MCQI-14 had a good IRT reliability (0.940).
Table 3 shows that the MCQI-14 had acceptable item and person reliability (0.903 and 0.928, respectively) and item and person separation index (3.044 and 3.586, respectively). Both infit and outfit MnSqs (Supplementary Table 2 in the online-only Data Supplement) ranged between 0.76 and 1.39 and 0.77 and 1.31, respectively. These MnSqs suggested a good model fit. Item difficulty values ranged between -0.52 and 0.65. Item 60 was the least difficult item, and item 50 was the most difficult item in the MCQI-14.
The DCA model of insomnia among patients with cancer
In Table 4, the ISI was significantly correlated with the MCQI-6 (r=0.68, p<0.001), MCQI-14 (r=0.68, p<0.001), CDBAS-14 (r=0.73, p<0.001), and DBST index (r=0.22, p<0.001). Mediation analysis (Table 5) showed that the DBST index did not directly influence insomnia severity. However, the influence was totally mediated by cancer-related dysfunctional beliefs about sleep and metacognition on insomnia, although MCQI-14 showed a marginal significance (p=0.055).
DISCUSSION
The MCQI-6 and MCQI-14 were derived from the original MCQ-I using the RF machine learning training [25] among the general population, and we explored the reliability and validity of the two shortened versions among patients with cancer in this study. We observed that these shortened versions were reliable and valid scales that can measure the metacognition of patients with cancer on their insomnia. Additionally, we observed that the DCA model of insomnia was feasible among patients with cancer because the influence of the DBST index on insomnia severity was mediated by cancer-related dysfunctional beliefs about sleep and sleep-related metacognitive process.
The MCQI-6 and MCQI-14 were both valid and reliable among patients with cancer. The CFA revealed a good fit for model for single factor models of both the MCQI-6 and MCQI-14. Graded response model outputs showed that all items in both shortened scales were considered highly efficient in discriminating the patients. The scale information curve showed that the MCQI-14 was more informative than the MCQI-6 among patients with cancer, which is in accordance with a previous study conducted among the general population [25].
Metacognition and insomnia, dysfunctional beliefs about sleep, and the DBST index
In this study, both the MCQI-6 and MCQI-14 were significantly correlated with the ISI, C-DBAS-14, and DBST index. Metacognition is the ability to regulate mental states, beliefs, and processes, granting individuals the ability to reflect on their own mental states and cognitive processes. No study has examined the association between sleep-related metacognitive process and cancer-related dysfunctional beliefs about sleep among patients with cancer. Among the general population, metacognitive abilities were reportedly related to poor sleep quality via the mediation role of dysfunctional thought-control strategies in the pre-sleep period [34]. The observed effect suggests that inadequate metacognitive functioning, especially in understanding one’s own mind and mastery domains, is associated with poor sleep quality in healthy individuals via the mediation of dysfunctional beliefs about sleep at bedtime.
Among patients with cancer, the severity of their insomnia symptoms has been associated with cancer-related dysfunctional beliefs about sleep [11], which aligns with the results of this present study. Specifically, their unrealistic expectations or perceptual and attention bias by believing that their sleep disturbance significantly affects their immune function or disease progression have played an essential role in perpetuating insomnia. Dysfunctional beliefs about sleep in patients with cancer are associated with anxiety, depression, and sleep disorders [35].
In this study, we observed that the DBST index was significantly correlated with insomnia severity, cancer-related dysfunctional beliefs about sleep, and sleep-related metacognitive processes. We previously reported that the DBST index was significantly correlated with insomnia severity among patients with cancer [22]. The DBST index could affect insomnia severity and long sleep onset latency. Among the general population, the influence of the DBST index on insomnia severity is mediated by dysfunctional beliefs about or preoccupation with sleep [21,30]. However, no previous study has explored the relationship between the DBST index and sleep-related metacognitive process. Thus, this is the first report showing that the DBST index may be associated with metacognition on insomnia in patients with cancer.
The DCA model of insomnia among patients with cancer
We observed that the DCA model of insomnia is feasible among patients with cancer. However, this study did not show any direct effect of the DBST index on insomnia severity, unlike previous studies [21,30]; rather, sleep-related metacognition and dysfunctional beliefs about sleep mediated the relationship. The DBST index is a concept based on the idea that individuals with insomnia may desperately want to sleep for at least a short period of time (dTST) despite unconsciously wanting to have a greater amount of sleep time (dTIB). In the model proposed by Vaziri et al. [36], attentional bias of poor sleepers to sleep-related cues was proposed to influence insomnia. In another study, patients with insomnia reportedly had a prolonged reaction time when shifting their attention away from insomnia-associated images [37]. Patients with insomnia may have attention bias and difficulty in shifting attention from sleep-related stimuli. Because the DBST index was correlated with insomnia severity, we can speculate that an individual without insomnia may have no discrepancy between their dTIB and dTST. Conversely, patients with insomnia may have a high level of discrepancy (or attention bias) between their dTIB and dTST, contrasting with individuals without sleep problems who may have similar dTST and total time in bed, estimated from their dTST (i.e. “I want to have 7 hours of sleep. For that, I usually go to bed at 11 pm and wake-up at 6 am.”)
The concept of the DBST index is not exactly the same as dysfunctional beliefs about sleep in that dysfunctional beliefs may include faulty beliefs, worry, and attention bias [10]. Patients with cancer with sleep disturbance may have dysfunctional beliefs about sleep because their sleep disturbance might influence their immune function or disease progression [11,29]. In this study, these cancer-related dysfunctional beliefs mediated the influence of the DBST index on insomnia severity. However, cancer-related dysfunctional beliefs about sleep have been linked to insomnia in patients with cancer although metacognition has not been sufficiently examined as a possible cause of insomnia in such patients. This is the first report on the reliability and validity of the MCQI-6 and MCQI-14 among patients with cancer. Additionally, this study showed that metacognition on insomnia also mediated the influence of the DBST index on insomnia severity among patients with cancer. The MCQI-14 showed a marginally significant mediation effect, whereas the MCQI-6 showed a significant effect. However, we can conclude that the DCA model of insomnia is feasible among patients with cancer.
Several limitations should be acknowledged in this study. First, we conducted this survey via a professional survey platform rather than a face-to-face interview. Thus, participants might have misunderstood the exact meaning of the items in each rating scale and even the questions on the DBST index. Second, we recruited participants among patients with cancer in a single tertiary level medical center, which might limit the generalizability of the results. Third, There is no direct effect of the DBST index on insomnia severity among cancer patients, which may limit the feasibility of the DCA model. However, there was significant correlation between DBST index and ISI scores in this study. The lack of significant direct influence may come from the interaction among DBST, ISI, CDBAS-14, and MCQI-6 (or MCQI-14). Further study is needed to explore the difference of the results from that of the general population. Fourth, the feasibility of the DCA model was still unexplored among clinical sample of patients with insomnia. Further research is needed to explore the feasibility among these populations.
Conclusion
We observed that the MCQI-6 and MCQI-14 were reliable and valid to measure the metacognition of patients with cancer on their insomnia. Additionally, the DCA model of insomnia was feasible among patients with cancer. We hope this model can be applied to understand the characteristics of cognitive processes in patients with cancer who have sleep disturbance during their diagnosis and treatment trajectory.
