Twenty-one potentially relevant articles were identified from 115 citations. Twenty-one studies consisting of 1433 patients (Control group: 590 patients in twenty-one studies; Fentanyl group:199 patients in seven studies11–17; Sufentanil group: 156 patients in five studies18–22; Meperidine group: 488 patients in ten studies15,23–31) contained data regarding the effect of IT opioids on shivering (Flow chart: Figure 1). One included study by Han et al. 2007 investigated the effect of both, fentanyl and meperidine.15 Tables 1a and 1b summarise the systematic review of the effects of IT fentanyl, sufentanil and meperidine on shivering in women undergoing cesarean delivery. Out of twenty-one studies, ten studies investigated shivering as the primary outcome.14–16,21,24–26,28–30 Methodological validity scores determined by modified Oxford score ranged from 3 to 7.
Network meta-analysis (NMA)
The twenty-one studies included in this network meta-analysis investigated the effect of three interventions: fentanyl (seven studies11–17), sufentanil (five studies18–22), and meperidine (ten studies15,23–31), with four comparison groups. Six pairwise comparisons and four direct comparisons were conducted. Table 2 summarises the data on the effect of intrathecal opioids on the incidence of shivering. Out of the twenty-one studies comprising 1433 patients, 199 patients received fentanyl,11–17 156 patients received sufentanil,18–22 488 patients received meperidine15,23–31 and 590 patients were in the control group.
Table 2 provides the effect estimates of direct, indirect and mixed network meta-analysis with quality of evidence rating according to the GRADE approach. Figure 2 displays the network diagram comparing the various intrathecal lipophilic opioids to prevent incidence of shivering in women undergoing cesarean delivery. Supplementary file S2 and S3 show the contribution matrix and league table for comparison of all classes of drugs
Fentanyl: Data on the incidence of shivering with IT fentanyl (7 RCTs, n=199 patients) were available in all the studies.11–17 The mixed evidence from the network meta-analysis showed that the incidence of shivering was significantly lower in the IT fentanyl group compared to the control group [IT Fentanyl vs. Control: 22.11% vs. 51.94%; Pooled Odds Ratio (OR): 0.13; 95% Credible Interval (CrI): 0.04 to 0.35; P = 0.0004]. The funnel plot and influential analysis on the direct data identified Sadegh et al.2003 as the outlier and contributed the maximum heterogeneity to the end estimate. When this study was excluded and summary estimates were recalculated, the end estimate increased to 0.51(0.36 to 0.71); P<0.0001 and heterogeneity decreased to zero (not shown in the figure). The Begg’s test (P=0.089) and Egger regression test (P=0.2077) did not show any evidence of publication bias. Fail-safe N test showed 113 studies required to increase the p value to more than alpha (>0.05), indicating the absence of publication bias (not shown in the figure). Fentanyl was administered in the dose range of 7.5 to 25 microgram and there was no difference in the outcomes across this dose range (Coefficient -0.043; 95% CI: -0.0963 to 0.0103; P = 0.1139).
Sufentanil: Data on the incidence of shivering with IT sufentanil (5 RCTs, n=156 patients) were available in all the studies.18–22 The mixed evidence from the network meta-analysis showed that the incidence of the shivering was not significantly lower with IT sufentanil when compared to the control group. (IT Sufentanil vs. Control: 23.71% vs. 45.28%; OR: 0.37; 95% CrI: 0.11 to 1.22; P = 0.23). Meta-regression analysis based on the IT sufentanil dose did not change the final inference of the result (Coefficient 0.0919; 95% CI: -0.2495 to 0.4333; P = 0.5977)
Meperidine: Data on the incidence of shivering with IT meperidine were available in all the 10 studies. 15,23–31 The mixed evidence from the network meta-analysis showed that the incidence of shivering was lower in the meperidine group compared to the control group (IT Meperidine vs. Control: 15% vs. 44.2%; OR: 0.12; 95% CrI: 0.05 to 0.29; P < 0.00001). For the direct data, the Begg’s test (P = 0.7544) and Egger regression test (P=0.1628) did not show any evidence of publication bias. Fail-safe N test showed 85 studies required to increase the p value to more than alpha (>0.05), indicating the absence of publication bias. Meperidine was used in the dose range of 5-35 mg and there was no difference in the outcomes across this dose range (Coefficient -0.0215; 95% CI: -0.0649 to 0.0219; P=0.3314). Meta-regression and sensitivity analysis based on the quality of the study for the various subgroups slightly changed the end estimate, but did not change the final inference of our results (Table 3).
Side effects:
IT Fentanyl: The IT fentanyl group had a significantly lower incidence of intraoperative discomfort (IT Fentanyl vs. Control: 6.89% vs. 34%; Risk Ratio (RR): 0.19; 95% CI: 0.10-0.35; P < 0.00001), but there was no significant difference in other maternal adverse events like pruritus (IT Fentanyl vs. Control: 38.14% vs. 18.79%; RR: 2.03; 95% CI: 0.82-5.05; P = 0.13), nausea and vomiting (IT Fentanyl vs. Control: 39.10% vs. 58.20%; RR: 0.66; 95% CI: 0.42-1.05; P = 0.08) and hypotension (IT Fentanyl vs. Control: 43.57% vs. 54.47%; RR: 0.93; 95% CI: 0.78-1.12; P = 0.45).
IT Sufentanil: The IT sufentanil group had a significantly higher incidence of pruritus (IT Sufentanil vs. Control: 20.87% vs. 2.12%; RR: 6.18; 95% CI: 1.18-32.46; P = 0.03), but there was no significant difference in other maternal adverse events like hypotension (IT Sufentanil vs. Control: 40.51% vs. 55.46%; RR: 0.74; 95% CI: 0.37-1.47; P = 0.39), nausea and vomiting (IT Sufentanil vs. Control: 28.44% vs. 35.29%; RR: 0.83; 95% CI: 0.53-1.29; P = 0.40). IT sufentanil did not significantly decrease the intraoperative discomfort compared to the control group (IT Sufentanil vs. Control: 36.84% vs. 59.09%; RR: 0.62; 95% CI: 0.31-1.24; P = 0.18).
IT Meperidine: The IT Meperidine group had significantly lower incidence of intraoperative discomfort (IT Meperidine vs. Control: 2.7% vs. 13.6%; RR: 0.22; 95% CI: 0.09-0.55; P = 0.001). There was a significant increase in nausea and vomiting (IT Meperidine vs. Control: 42.7% vs. 19.4%; RR: 2.56; 95% CI: 1.14-5.75; P = 0.02), but there was no significant difference in other maternal adverse events between the two groups, like hypotension (IT Meperidine vs. Control: 46.9% vs. 41.8%; RR: 0.96; 95% CI: 0.67-1.37; P = 0.82) and pruritus (IT Meperidine vs. Control: 18.9% vs. 6%; RR: 0.63; 95% CI: 0.82-3.24; P = 0.17).
Quality of the evidence in network estimates:
Supplementary files S4 and S5 show the rankogram and various domains examined to assess the quality of evidence in the network meta-analysis. Most of the included studies in the network meta-analysis were randomized double blind controlled studies with no, or some, concerns in the study limitation. To assess the imprecision, effect estimates of the relative treatments lower than 0.95 and greater than 1.05 were considered to be clinically significant. The data were collected from different studies, across different countries, at varying time intervals and the network model showed some degree of incoherence (c2 statistics: 0.336; d(f): 2; p value: 0.846). The estimated value of between-study variance for the network meta-analysis is 0.412 indicating some heterogeneity and consistency in the network model. Overall, some of the comparisons were rated down for imprecision, heterogeneity and incoherence (inconsistency), thus the quality of the evidence for the effect estimates was low according to the GRADE approach.