Results
Study Selection and Characteristics
A total of 202 potentially relevant articles were identified initially from the electronic databases and 116 articles remained after removing duplicates. 64 articles were excluded after reading titles and abstracts because they did not meet the inclusion criteria. After reviewing the full text of the remaining 52 potentially eligible articles, 21 were excluded because they were study protocols, one was a review, 15 had no PSG data, seven were not for surgical patients, one was not an adult study, one was not an RCT, and one had no control group. Five eligible studies were identified.[13,14,18–20] The selection process is illustrated in Figure 1.
Figure 1.
The flow diagram of identifying studies through systemic search in multiple databases
The relevant baseline characteristics of each study, including descriptions of the DEX administration regimens, are summarized in Table 1. After excluding the number of dropouts reported in each study, the meta-analysis covered a sample size of 381 patients and included five RCTs. Two studies used DEX postoperatively, two used it as an adjuvant in patient controlled analgesia (PCA), and one study used DEX intraoperatively. All the PSG parameters were collected on the first night after surgery. The VAS score and Ramsay sedation score were assessed at 6 h after surgery. For PSG parameters, five studies reported SEI, AI, and the percentage of REM sleep; four studies reported the percentages of stage N1, N2, and N3 of NREM sleep, and two studies reported TST. Postoperative analgesia was provided with a patient-controlled intravenous or epidural analgesia pump to maintain the VAS ≤ 4 at rest or behavior pain score ≤ 6, except for patients in one study undergoing endoscopic sinus surgery. Supplemental analgesics were administered for patients when necessary.
Meta-analysis Results
Five studies reported the effect of DEX on SEI and AI compared to placebo groups, and the forest plots are presented in Figure 2a, b. Administration of DEX significantly improved SEI (11.25%, 95%CI = 1.91–20.59, p = 0.02) and lowered AI (-2.21, 95%CI = -3.61- -0.81, p = 0.002). Four studies reported the effect of DEX on the percentages of stage N1, N2, and N3 of NREM sleep compared to placebo groups. The comparison showed that the duration of stage N1 sleep was shortened (-11.96%, 95%CI = -22.54- -1.38, p = 0.03) and the duration of stage N2 sleep was longer (14.86%, 95%CI = 9.07–20.66, p < 0.00001) in the DEX group than in the placebo group (Figure 2c, d). There were no significant differences in the duration of stage N3 sleep (-2.09%, 95%CI = -7.51–3.33, p = 0.45) and REM sleep (-0.20%, 95%CI = -1.17–0.77, p = 0.69), as shown in Figure 2e, f. Two studies reporting TST showed that the administration of DEX significantly prolonged TST (74.55 min, 95%CI = 29.90–119.21, p = 0.001) (Figure 2g). The trial sequential analysis was performed on the SEI and AI (Figure 3). The cumulative Z-curve crossed both the conventional meta-analysis boundary and the trial sequential monitoring boundary before the accumulated information reached the required information size, which demonstrated that our results were considered to be stable.
Figure 2.
Forest plots for effects of DEX compared with placebo on sleep quality. (a) SEI, sleep efficiency index; (b) AI, arousal index; (c) N1, stage 1 of non-rapid eye movement sleep; (d) N2, stage 2 of non-rapid eye movement sleep; (e) N3, stage 3 of non-rapid eye movement sleep; (f) REM, rapid eye movement sleep; (g) TST, total sleep time; CI, confidence interval
Figure 3.
Trial sequential analysis for SEI and AI. (a) SEI, sleep efficiency index; (b) AI, arousal index
For secondary outcomes, four studies collected postoperative VAS scores, two reported Ramsay sedation scores and the incidence of POD, and one mentioned the incidence of PONV. Our meta-analysis demonstrated that DEX administration lowered the postoperative VAS score (-1.05, 95%CI = -1.81- -0.29, p = 0.007) and improved the Ramsay sedation score (0.40, 95%CI = 0.35–0.45, p < 0.00001) with no adverse effect on POD (odds ratio = 0.88, 95%CI = 0.32–2.40, p = 0.80), as shown in Figure 4. The meta-analysis for PONV was not conducted because of insufficient data. However, considering the high heterogeneity of most PSG parameters to evaluate sleep quality and secondary outcomes, we indicated that one or more studies may have influenced the results; thus, subgroup analysis was conducted to interpret the heterogeneity.
Figure 4.
Forest plots for effects of DEX compared with placebo on secondary outcomes. a. VAS score, Visual Analog Scale score; c. POD, postoperative delirium, CI, confidence interval
Subgroup Analysis
Table 2 presents subgroup analyses of postoperative PSG data stratified by severity of illness. Patients in two studies were admitted to the ICU, and others in three studies returned to the ward. We observed statistically different results on SEI (9.98%, 95%CI = 3.48–16.47, p = 0.003) and duration of stage N1 sleep (-8.09%, 95%CI = -14.93- -1.26, p = 0.02) in the ICU subgroup, with a dramatically decreased heterogeneity (I 2 = 0). There were statistical differences on AI (-2.51, 95%CI = -4.17- -0.86, p = 0.003) and duration of stage N2 sleep (17.82%, 95%CI = 13.53–22.12, p < 0.00001) in the non-ICU subgroup, but the heterogeneity remained high. The heterogeneity on AI and duration of stage N2 sleep was decreased in the ICU subgroup, however, no significant effect was observed in these two parameters. In addition, no statistical difference was found in the duration of stage N3 and REM sleep. Given that the stage N3 and REM sleep were barely achieved in the ICU patients, several parameters were not estimable or applicable, as shown in Table 2. From the results above, we cannot come up with the conclusion that the severity of illness is a source of heterogeneity in our meta-analysis. However, the overall effect of DEX on postoperative sleep quality seems to be beneficial.
Sensitivity Analysis
Sensitivity analysis for the PSG parameter was also conducted. We did not find any single study that could make a significant impact on the primary outcome, indicating that our results were reliable and statistically stable, as shown in Figure 5.
Figure 5.
Sensitivity analysis for SEI and AI. a. SEI, sleep efficiency index; b, AI, arousal index; CI, confidence interval
Risk of Bias
The risk of bias is summarized in Figure 6, which has been described for individual studies, and a summary, respectively. All trials were randomized, and most (4 out of 5, 80%) reported the methods of randomization. Four trials reported allocation concealment. Participants and personnel were blinded in four trials. Three trials were blinded to the outcome assessment. The remaining trials contained several domains that lacked clarity. The risk of bias for selective reporting was rated high in one study because of missing data in the control group. Other bias was rated high in one study because several patients were excluded for premature drug interruption, death within 30 days, and refusion to the follow-up test. Egger tests were performed and showed no publication bias based on SEI (p = 0.052) and AI (p = 0.246). Begg tests also demonstrated no publication bias on SEI (p = 0.221) and AI (p = 0.806).
Figure 6.
The risk of bias in the included studies. a: risk of bias for each study; b: risk of bias summary
BMC Anesthesiol. 2023;23(88) © 2023 BioMed Central, Ltd.
