Patient profiles and group comparison
A total of 45 patients (39 males and 6 females) were included in this study. The characteristics of these patients are described in Table 1. We classified CPAP adherence into 3 categories, good-adherence: used at least 4 hr/night on ≥ 70% of the nights per month (23), poor-adherence: used < 4 hr/night on > 30% of the nights per month, and withdrawal: dropped out within a year. Of the 45 patients included, 19, 16 and 10 were classified into the good-adherence (42.2%), poor-adherence (35.6%), and withdrawal (22.2%) groups, respectively. Among the three groups, there were no significant differences observed in age, body mass index (BMI), presence of hypertension, nasal symptoms, psychiatric comorbidities, regular alcohol consumption or current smoking habits; however, the differences were found in the rate of living alone (good-adherence: 15.8% vs. poor-adherence: 50% vs. withdrawal: 60%, n = 19, 16, 10, respectively, P = 0.0317). With regard to self-reported sleep performance from sleep questionnaires, similar scores were observed among the three groups for the Japanese-version of the Epworth Sleepiness Scale (ESS) (24), Pittsburgh Sleep Quality Index (PSQI) (25), and Athens Insomnia Scale (AIS) (26).
Table 1
Characteristics of patients
| All (n = 45) | Good adherence (n = 19, 42.2%) | Poor adherence (n = 16, 35.6%) | Withdrawal (n = 10, 22.2%) | P value |
Sex, male/female | 39/6 | 14/5 | 15/1 | 10/0 | − |
Age, y | 50.0 (45.5–64.5) | 58.0 (46.0–72.0) | 48.5 (44.3–52.8) | 49.0 (44.0-67.5) | 0.0598 |
BMI, kg/m2 | 25.6 (23.9–30.0) | 26.6 (23.2–30.0) | 25.5 (24.1–30.6) | 25.9 (21.7–30.0) | 0.8282 |
Living alone | 17 (37.8) | 3 (15.8) | 8 (50.0) | 6 (60.0) | 0.0317† |
Unemployed | 16 (35.6) | 10 (52.6) | 2 (12.5) | 4 (40.0) | 0.0609 |
Hypertension | 13 (28.9) | 6 (31.6) | 3 (18.8) | 3 (30.0) | 0.6843 |
Nasal symptom | 10 (22.2) | 7 (36.8) | 2 (12.5) | 1 (10.0) | 0.1769 |
Psychiatric comorbidities | 22 (48.9) | 12 (63.2) | 6 (37.5) | 4 (40.0) | 0.2601 |
Alcohol | 10 (22.2) | 4 (25.0) | 2 (12.5) | 4 (40.0) | 0.2998 |
Smoking | 15 (33.3) | 5 (26.3) | 6 (37.5) | 4 (40.0) | 0.7974 |
ESS | 10.5 (5.0–16.0) | 10.0 (5.0–15.0) | 13.0 (6.0-17.5) | 7.0 (2.3–15.8) | 0.2008 |
PSQI | 9.0 (7.0–12.0) | 8.0 (4.0-12.3) | 9.0 (8.0-12.5) | 11.5 (8.5–12.8) | 0.2972 |
AIS | 11.0 (6.0-13.5) | 8.5 (5.8–12.3) | 12.0 (6.5–13.5) | 13.5 (10.0-18.3) | 0.0812 |
Median (IQR) for Kruskal-Wallis post-hoc with Dunn‘s multiple comparisons test or number (%) for Fisher's exact test; BMI, body mass index; ESS, Epworth Sleepiness Scale; PSQI, Pittsburgh Sleep Quality Index; AIS, Athens Insomnia Scale. Characteristics of the with Good adherence, Poor adherence and Withdrawal from CPAP Use ≥ 4 hr/night at one year of CPAP therapy.
In terms of polysomnography (PSG) data, a significant difference was shown in sleep latency among the 3 groups (good-adherence: 7.8 (5.1–15.5) min vs. poor-adherence: 4.2 (1.1–9.5) min vs. withdrawal: 3.0 (0.5-7.0) min, n = 19, 16, 10, respectively, p = 0.0262), while no significant differences were found in other parameters including AHI (Table 2).
Table 2
Patient polysomnographic data
| Good adherence (n = 19) | Poor adherence (n = 16) | Withdrawal (n = 10) | P value |
TST, min | 451 (386–515) | 471 (433–533) | 389 (341–476) | 0.1955 |
Sleep latency, min | 7.8 (5.1–15.5) | 4.2 (1.1–9.5) | 3.0 (0.5-7.0) | 0.0262* |
WASO, min | 85.0 (51.7-115.9) | 52.4 (26.0-79.4) | 98.0 (38.0-145.7) | 0.1971 |
Sleep efficiency, % | 83.6 (76.7–88.8) | 89.6 (82.2–95.1) | 80.9 (64.9–92.2) | 0.1092 |
N3, %TST | 2.1 (0.1-8.0) | 2.3 (0.3–6.1) | 1.8 (0.0-5.7) | 0.7121 |
AHI, events/hr | 31.1 (21.5–53.1) | 37.6 (24.6–51.6) | 35.6 (26.4–58.3) | 0.5661 |
Median (IQR) for Kruskal-Wallis post-hoc with Dunn‘s multiple comparisons test; TST, total sleep time; WASO, wake after sleep onset; AHI, apnea-hypopnea index. |
Sleep-wake rhythm and variability
Prior to PSG, activity monitoring for at least 7 consecutive days was assessed for evaluation of sleep-wake rhythm and behavior rhythmicity. Significant differences were shown in sleep duration (good-adherence: 7.0 (6.7–8.1) hr vs. poor-adherence: 6.8 (5.5–8.2) hr vs. withdrawal: 4.7 (4.2–6.6) hr, n = 19, 16, 10, respectively, p = 0.0048) and in sleep onset (23:07 (22:06 − 0:16) vs. 0:19 (22:12 − 1:12) vs. 0:53 (23:47 − 2:49), n = 19, 15, 10, respectively, p = 0.0044) among the 3 groups, with shorter sleep duration and later sleep onset observed in the withdrawal group compared to in the good-adherence group (post-hoc Dunn's multiple comparisons test: p = 0.0033, p = 0.0043, respectively). No significant differences were shown in sleep offset (Fig. 1a).
Variability in sleep-wake rhythms was evaluated using interquartile range (IQR), with a higher IQR indicating higher variability. Significant differences were shown in sleep duration variability (good-adherence: 1.1 (0.8–1.7) hr vs. poor-adherence: 1.9 (1.6–2.6) hr vs. withdrawal: 1.7 (1.2–2.6) hr, n = 19, 16, 10, respectively, p = 0.0043) and in sleep onset variability (0.8 (0.5–1.1) hr vs. 1.4 (1.0-2.3) hr vs. 3.5 (1.2–6.2) hr, n = 19, 16, 10, respectively, p = 0.0027) were shown among the 3 groups, with higher variability in sleep onset in the poor-adherence and withdrawal groups compared to that in the good-adherence group (post-hoc Dunn's multiple comparisons test: p = 0.0449, 0.0042, respectively). No significant differences were shown in sleep offset variability (Fig. 1b).
Correlation between CPAP adherence and sleep-wake rhythm
CPAP adherence, as estimated from CPAP usage rate (% > 4 hr/night), showed a positive correlation to sleep duration (r = 0.4206, n = 45, P = 0.0040), and a negative correlation to sleep onset (r= -0.4502, n = 44, P = 0.0022), but no correlation to sleep offset (Figure S1a). Negative correlations were also shown between CPAP adherence and variability in sleep duration, and between sleep onset and sleep offset (r= -0.4536, -0.5475, -0.3484; n = 45, 45, 45, P = 0.0018, < 0.0001, 0.0190, respectively) (Figure S1b).
Receiver Operating Characteristic (ROC) curves demonstrated that a sleep duration of 4.8 hr (Figure S2a) and sleep onset at 0:26 (Fig. 2a) were the cut-off values for predicting CPAP withdrawal with the area under the ROC curve (AUC): 0.823, 0.775, n = 45, 44, P = 0.0020 and 0.0088, respectively. The sensitivity and specificity for these cut-off values were 60.0% and 94.3% for sleep duration, and 70.0% and 76.5% for sleep onset, respectively. No statistical significance was detected in the ROC curve for sleep offset. An examination of variability revealed that a sleep onset variability of over 2.5 hr can predict CPAP withdrawal (AUC: 0.754, n = 45, P = 0.0151) and the sensitivity and specificity of the cut-off value were 60.0% and 88.6%, respectively (Fig. 2b). No statistical significance was shown in the ROC curve for sleep duration variability. Furthermore, among those continuing CPAP therapy, the ROC curve demonstrated that a variability in sleep onset over 9.5 hr predicted poor CPAP adherence with AUC: 0.765, n = 35, P = 0.0077 (Figure S2b). The sensitivity and specificity for the cut-off value were 87.5% and 57.9%, respectively.
Behavior activity rhythm and CPAP adherence
Two patients showed no periodicity in their behavioral activity rhythm from χ2 periodogram analysis: one in the poor-adherence group (6.3%) and the other in the withdrawal group (12.5%). Two patients in the withdrawal group were excluded from the analysis as they only performed activity monitoring at night. Figure 3 shows representative double-plotted behavior activity profiles (a. rhythmic; c. arrhythmic) and periodograms (b. rhythmic; d. arrhythmic).
The χ2 periodogram showed significant differences among the 3 groups in the maximum Qp level, which indicates the regularity and/or amplitude of rhythms (good-adherence: 383 (343–424) vs. poor-adherence: 374 (296–488) vs. withdrawal: 288 (247–363), n = 19, 15, 7, P = 0.0370). The median of the maximum Qp level in the withdrawal group was lower than that in the good-adherence group (post-hoc Dunn's multiple comparisons test: P = 0.0425), indicating lower regularity and/or amplitude (Figure S3).
Other CPAP adherence predictors and sleep-wake rhythm
Sixteen patients living alone showed significantly lower CPAP usage rates (> 4 hr/night) compared to patients living with others (30.1 (1.3–46.7) % vs. 86.0 (26.1–98.8) %, n = 16, 29, p = 0.0024; Figure S4a). Sleep-wake rhythm variability did not differ significantly between the two groups, but patients living alone showed later sleep onset compared to those living with others (0:25 (23:31 − 1:11) vs. 23:28 (22:14 − 0:31), n = 16, 28, p = 0.0426; Figure S4b). The presence of a psychiatric comorbidity, alcohol consumption or smoking habit showed no association with CPAP adherence.