Sharon R Lewis, Amanda Nicholson, Andrew F Smith, Phil Alderson

Shivering after general anaesthesia is common. It is unpleasant but can also have adverse physiological effects. Alpha-2 (α-2) adrenergic agonist receptors, which can lead to reduced sympathetic activity and central regulation of vasoconstrictor tone, are a group of drugs that have been used to try to prevent postoperative shivering. To assess the following: the effects of α-2 agonists on the prevention of shivering and subsequent complications after general anaesthesia in people undergoing surgery; the effects of α-2 agonists on the risk of inadvertent perioperative hypothermia; and whether any adverse effects are associated with these interventions. We searched the Cochrane Central Register of Controlled Trials (CENTRAL), MEDLINE, and EMBASE on 13 June 2014. Our search terms were relevant to the review question and limited to studies that assessed shivering or hypothermia. We also carried out searches of clinical trials registers, and forward and backward citation tracking. We considered all randomized controlled trials, quasi-randomized studies, and cluster-randomized studies with adult participants undergoing surgery with general anaesthesia in which an α-2 agonist was compared with another α-2 agonist or a placebo for the prevention of shivering. Two review authors independently assessed trial quality and extracted data, consulting a third review author in the case of disagreements. We used standard Cochrane methodological procedures, including an assessment of risk of bias and use of GRADEpro software to interpret findings. We included 20 studies with 1401 surgical participants comparing an α-2 agonist against a control. Thirteen studies compared clonidine with a control, whilst seven compared dexmedetomidine with a control. The doses, methods, and time of administration varied between studies: three studies gave the drug orally or as an intravenous bolus preoperatively and nine intraoperatively; one study gave the drug as an infusion starting preoperatively and seven started at varying points from anaesthetic induction to the end of surgery. Whilst all the studies were described as randomized, many provided insufficient detail on methods used. We had anticipated that attempts would be made to reduce performance bias by blinding of personnel and participants, however this was detailed in only six of the papers. Similarly, in some studies detail was lacking on methods to reduce the risk of detection bias. We therefore downgraded the quality of evidence in our 'Summary of findings' table by one level for risk of bias using the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) approach.All 20 included studies presented outcome data for postoperative shivering, and in meta-analysis α-2 agonists were shown to significantly reduce the risk of shivering (Mantel-Haenszel risk ratio 0.28, 95% confidence interval 0.18 to 0.43, P value < 0.0001). We found significant evidence of heterogeneity (I(2) = 80%) for this result that was not explained by sensitivity or subgroup analysis; we therefore downgraded the inconsistency of the evidence by one level. Although we did not feel that there were concerns with imprecision or indirectness of the data, we downgraded the quality of the evidence for the risk of publication bias following visual analysis of a funnel plot. Using GRADEpro, we rated the overall quality of the data for shivering as very low. Only one study reported the incidence of core hypothermia, whilst 12 studies measured core temperature. However, as the results for core temperature were reported in different styles, pooling the results was inappropriate. We found no studies with participant-reported outcomes such as experience of shivering or participant satisfaction. We found limited data for the outcomes of length of stay in the postanaesthetic care unit (three studies, 200 participants) and the following adverse effects: sedation (nine studies, 875 participants), bradycardia (eight studies, 716 participants), and hypotension (seven studies, 688 participants). Unpooled analysis suggested that sedation and bradycardia were significantly more common with dexmedetomidine than placebo, with all seven dexmedetomidine studies and none of the clonidine studies reporting statistically significantly higher levels of sedation as an adverse effect. There is evidence that clonidine and dexmedetomidine can reduce postoperative shivering, but patients given dexmedetomidine may be more sedated. However, our assessment of the quality of this evidence is very low.