Improved Postoperative Outcomes Associated with Preoperative Statin Therapy

In this systematic review, we identified and analyzed randomized prospective clinical trials and retrospective observational studies published from January 1977 (when the use of statins was first described) to November 2005 that reported the effects of preoperative statin therapy in adults undergoing surgical interventions. We included data published either as full-text journal publications or scientific abstracts (published after January 2004) and excluded in vitro  and animal studies. The literature was screened for reports of preoperative statin therapy using any of the following commercially available statins: cerivastatin, fluvastatin, pravastatin, atorvastatin, simvastatin, lovastatin, and rosuvastatin. The study did not control for the effect of different types or doses of statins, because the focus of this investigation was on the clinical effects of the drug class collectively.

Two authors (K.H. and S.F.) independently performed the systematic literature search, reviewed each included study for quality, and extracted relevant data, using a standardized data extraction form. A third reviewer (B.R.) resolved any disagreements. All identified publications were assigned to one of three groups according to the type of surgical intervention: cardiac, vascular, or noncardiovascular surgery. The effect of preoperative statin therapy (compared with either placebo or no treatment) on predefined endpoints, including postoperative adverse events (specifically MI, cardiac arrhythmia, and stroke) and short-term mortality was then determined. Short-term mortality was defined as death from any cause within 30 days after surgery. Cardiac arrhythmia was defined as any occurrence of postoperative atrial fibrillation or ventricular tachycardia or fibrillation. Stroke was diagnosed if the relevant study described clinical radiologic (computed tomography or magnetic resonance imaging) evidence of a focal or global cerebral defect. Only two articles assessed the effect of statin therapy on renal function, and this endpoint was therefore not analyzed.

Relevant key words were used to build an effective literature search strategy for published data ( appendix). There was no language restriction for trial inclusion; however, no appropriate studies reported in non–English-language journals were identified. When more than one publication of data from a patient cohort existed, we included only the publication with the most complete data set. The electronic databases searched were MEDLINE, EMBASE, the Cochrane Central Register of Controlled Trials, the Cochrane Database of Systematic Reviews, the American College of Physicians Journal Club, and the Database of Abstracts of Reviews of Effects. In addition, abstracts from conferences and scientific meetings of the American Heart Association, the American Society of Anesthesiologists, the Society of Cardiovascular Anesthesiologists, the International Anesthesia Research Society, and the Society of Critical Care Medicine over the past 2 yr were searched. Further reports were identified from the bibliographies of all relevant articles, and missing information was obtained by contacting investigators. The quality of each study was assessed according to the Quality of Reporting of Meta-analyses (QUOROM) guidelines.38Each study was rated on the following factors: (1) Were participants randomized? (2) Were randomization procedures described? (3) Did the authors report numbers and reasons for dropouts? (4) Did the study include a control group? (5 Did the authors report monitoring treatment fidelity?

All statistical analyses were performed using RevMan software (Version 4.2 for Windows; The Nordic Cochrane Centre, Kobehavn, Denmark). Univariate (chi-square) analysis was performed to test for heterogeneity between studies. Both random and fixed effects models were used according to the presence or absence of significant heterogeneity among the studies. Univariate regression analysis assessed whether preoperative statin therapy reduced major postoperative morbidity and mortality. Pooled dichotomous outcomes were expressed as the odds ratio (OR) of the point estimate with the corresponding 95% confidence interval (CI). All metrics were converted so that an OR of less than 1 favored the experimental (statin) treatment over the control. To assess whether the studies retrieved for our meta-analysis are affected by publication bias, a funnel plot was constructed. The funnel plot exploits the difference between the effects in large and small studies. The effects, expressed as the logarithm of the OR, found in the studies were plotted against a measure of precision or weight of the studies (expressed as inverse variances of the logarithm of the OR).

More than 1,100 abstracts were retrieved from the screened databases as shown in the flow diagram (fig. 1). After critical appraisal, 37 records were selected for full text evaluation; of those, 22 published studies investigating the association between preoperative statin therapy and postoperative outcomes were analyzed. Fifteen publications were included in the meta-analysis (7 articles reporting on the effects in cardiac surgery, 7 articles in vascular surgery, and 1 article in noncardiovascular surgery; table 1). Seven studies, out of the selected 22 publications, were excluded. These studies and the reasons for exclusion are summarized in table 2. Of the 15 studies included in the meta-analysis, 1 study scored 5/5, 2 studies scored 3/5, 3 studies scored 2/5, and 9 studies scored 1/5. Study quality ratings did not correlate with the average study effect size.

The search strategy identified 10 studies11,12,17,18,24,27,31,32,34,36(6 published articles and 4 scientific abstracts) investigating the association between preoperative statin therapy and outcomes after cardiac surgery. Seven of these studies (including 1 prospective randomized controlled clinical trial)11were included in the meta-analysis.11,12,17,18,24,27,32Three studies were excluded.31,34,36These included the study by Mathew et al. ,34which measured cognitive dysfunction after cardiac surgery rather than our predefined endpoints; the study by Ali and Buth,31which included patients already considered in a previous publication by this author32; and the study by Riedel et al. ,36which was published as a scientific abstract only and exceeded our inclusion limit of within the past 2 yr for abstracts. The following outcomes were analyzed: short-term mortality (7 studies; n = 12,752; fig. 2A), MI (5 studies; n = 7,615; fig. 2B), cardiac arrhythmia (3 studies; n = 3,294), and stroke (3 studies; n = 4,872; fig. 3A). No significant (chi-square P > 0.10) heterogeneity was observed between these studies for any of the evaluated outcomes.

Postoperative mortality was significantly lower (1.9% vs.  3.1%; P < 0.0001; fig. 2A) in patients undergoing cardiac surgery who received preoperative statin therapy than in those who did not (table 3). No statistically significant differences were observed between the two groups with regard to postoperative cardiac arrhythmia (22.3% vs.  23.0%; P = 0.99) or stroke (2.7% vs.  3.2%; P = 0.26; fig. 3A). The incidence of MI was increased in those patients who received preoperative statin therapy (4.6% vs.  3.6%; P = 0.02; fig. 2B).

The search strategy revealed 10 studies15,19–23,28–30,35(all published articles) that investigated the association between preoperative statin therapy and outcomes after vascular surgery. Seven articles15,19,20,28–30,35were included in the meta-analysis (including 1 prospective cohort study29and 1 prospective, randomized, placebo-controlled, double-blinded clinical trial20): 2 studies by Kertai et al.  21,22were excluded because these reported postoperative outcomes of patients previously mentioned in the study by Poldermans et al.  15; in addition, we were unable to isolate the incidence of statin use in the subgroup of vascular patients reported by Lindenauer et al. ,23and this study was therefore also excluded. The following outcomes were analyzed and are summarized in table 3: stroke (4 studies; n = 2,749; fig. 3B), short-term mortality (7 studies; n = 5,373; fig. 4A), MI (5 studies; n = 2,862; fig. 4B), cardiac arrhythmia (2 studies; n = 329). No heterogeneity was observed among these studies.

Preoperative statin therapy significantly reduced postoperative mortality (1.7% vs.  6.1%; P < 0.0001; fig. 4A) in patients undergoing vascular surgery. The high mortality rate among non–statin users was mostly attributable to the study reported by Poldermans et al.  15In that study, the overall observed mortality rate was 5.8%, an incidence similar to that observed by the other included vascular studies. However, because the incidence of statin use was not available for all patients (personal communication), the authors performed a case-controlled (1:2 matching) study, and the mortality within the case-controlled group was 38%. Preoperative statin therapy was also associated with a significant reduction in the incidence of postoperative MI (2.9% vs.  6.2%; P = 0.001; fig. 4B) and stroke (2.0% vs.  3.3%; P = 0.049; fig. 3B) after vascular surgery. No statistically significant differences were observed with regard to cardiac arrhythmia (11.4% vs.  11.1%; P = 1.0).

Data from all patients undergoing either cardiac or vascular surgery were combined to investigate the overall influence of preoperative statin therapy on MI, cardiac arrhythmia, stroke, and short-term mortality after cardiovascular surgery (table 3).

Preoperative statin therapy was associated with a 2.3% absolute reduction (1.8% vs.  4.1%; P < 0.0001) and a 46% reduction in the odds (OR, 0.54; 95% CI, 0.44–0.66; fig. 5A) of early postoperative mortality in patients having either cardiac or vascular surgery. Despite this observed benefit, the incidence of MI was similar between the two groups (4.3% vs.  4.1%; P = 0.66; fig. 5Band table 3) and balanced by the significantly higher incidence of MI in patients undergoing cardiac surgery (fig. 2B) and lower incidence of MI for patients undergoing vascular surgery (fig. 4B).

In a separate analysis, the effect of statin therapy on postoperative mortality was investigated independent of the type of surgical procedure. The data from the patient populations reported in the eligible cardiac and vascular studies were combined with those reported by Lindenauer et al.  23The studies by Amar et al.  33and Riedel et al. ,26which evaluated patients undergoing thoracic surgery, were excluded because of differing study endpoints. Our meta-analysis revealed a 1.0% absolute reduction (2.2% vs.  3.2%; P < 0.0001; fig. 6) and a 44% reduction in the odds (OR, 0.56; 95% CI, 0.43–0.71) of early postoperative mortality in patients on preoperative statin therapy, irrespective of surgical procedure.

It is also important to note that the funnel plot is not symmetrical around the mean (fig. 7). Smaller studies (with variance log odds ranging between 0.0 and 0.8) tend toward larger effects (log odds 0.2–3) in the funnel plot, because small studies with smaller or negative effects are missing. Therefore, this funnel plot may be interpreted as an indication of publication bias.

This meta-analysis suggests that preoperative statin therapy significantly reduces postoperative mortality after cardiac, vascular, and noncardiovascular surgery. Specifically, the analysis of approximately 13,000 cardiac and 5,500 vascular surgery patients demonstrated a 1.2% and 4.4% absolute reduction and a 38% and 59% reduction in the risk of early postoperative mortality in patients receiving preoperative statin therapy, respectively. Only one11of the seven11,12,17,18,24,27,32cardiac surgery studies and only two20,29of the seven15,19,20,28–30,35vascular surgery studies included in this meta-analysis were prospective, randomized trials. The large number of patients included from retrospective studies12,15,17–19,24,27–32,34(including a study of 70 sites in 17 countries that revealed a threefold reduction in the incidence of early death after cardiac surgery among statin users18), however, increases the statistical power of our analysis and allows meaningful clinical interpretation of the data. Cautious interpretation is required because of the variability in study design among these studies. For example, one study considered the mortality rate for the first 3 postoperative days only.18Further, in the two studies reported by Ali and Buth,31,32the beneficial effect of statins on in-hospital mortality and postoperative morbidity was no longer evident after adjusting for other risk factors known to impact cardiac surgical outcome.

When the data were analyzed for the effect of statins irrespective of type of surgery, a 1.0% absolute reduction and 44% reduction in the risk of early postoperative mortality in patients receiving preoperative statin therapy was observed. The outcome data from the large study by Lindenauer et al.  23contributed 24% of the overall weighting in this analysis; nevertheless, both the point estimate and 95% confidence boundaries are in agreement with the combined result.

It is important to highlight several features of the article by Lindenauer et al.  23because it accounts for such a large percentage of the final analysis. First, this was a retrospective cohort study (780,591 patients) based on administrative data arising from physician documentation accessed from 329 hospitals throughout the United States. Diagnostic billing fields were then used to develop a propensity score to match patients’ comorbidities such that each quintile of propensity had a similar chance of receiving a statin. Therefore, these data are intrinsically less reliable than prospectively collected outcome data. Second, whenever the sample size is very large, statistical significance often does not equate to clinical significance, and unless different thresholds for significance are adopted, what is actually an unimportant difference can seem to be highly significant. Third, cardiovascular complications such as heart failure and MI (which may be both acute and chronic) are not easily detected from databases with no intrinsic date–time stamp, and are thus missing from this report. Despite these issues, the article by Lindenauer et al.  23does provide extremely useful information. Large sample sizes can overcome increased noise around a signal by their sheer power to find a difference, and this should not be underestimated. Last, and most important, the point estimate and confidence limits of this trial are consistent with the rest of the data, suggesting that both qualitatively and quantitatively, our pooled data may provide evidence that preoperative statin therapy is associated with improved postoperative outcomes.

It is feasible that the survival benefit observed in this meta-analysis may in fact be greater, especially when one considers that none of the retrospective studies clearly defined the reintroduction of postoperative statin therapy. Data supporting the importance of early reintroduction of statin therapy are evident from studies showing that statin withdrawal may increase the risk of adverse outcomes.8,39,40The importance of continuation of statin therapy during hospitalization is also supported by Kruger et al. ,41who report a reduction in-hospital mortality (OR, 0.39; 95% CI, 0.17–0.91; P = 0.03) in patients presenting with bacteremia if they were using statin therapy before hospitalization. This concept is further supported by the increased survival benefit reported at 1-yr postoperative follow-up in cardiac surgical patients receiving statin therapy.14 

The observed effect of statin therapy on postoperative morbidity was less clearly defined, with conflicting findings observed among studies, and we were unable to relate postoperative survival benefit by statin therapy to a reduction in cardiac morbidity, especially in the cardiac surgery patient population—where patients using preoperative statin therapy show a higher incidence of postoperative MI. This differential effect in incidence of MI may be delineated if one compares the prospective studies with the retrospective studies. In this regard, a lower incidence of postoperative MI is reported in the statin-treated patients within the prospective cardiac11and vascular20,29trials, where reintroduction of statin therapy in the postoperative period was well controlled. This benefit is offset by the apparent increase in the incidence of MI among statin users in the retrospective studies, where postoperative reintroduction of statins was most likely not controlled and only reinstituted at the time of admission to the intensive care unit or, more likely, at hospital discharge. This poor standardization of postoperative reintroduction of statin therapy may explain the conflicting findings observed in postoperative morbidity.

Although the pleiotropic effects (e.g. , antiinflammatory and antithrombotic effects) of statins are thought to play an important role in reducing the incidence of stroke,42–44our meta-analysis did not demonstrate any potential benefit of statins on reduced incidence of stroke after cardiac surgery. Mathew et al.  34also did not observe any difference between statin users and nonusers with regard to the incidence of neurocognitive dysfunction after cardiac surgery. In fact, statin therapy was associated with reduced improvement in cognitive performance 6 weeks after surgery. In that study, inflammatory markers (C-reactive protein and cytokine levels) measured after cardiopulmonary bypass did not differ between statin users and nonusers.34An interesting finding in one study, though, is the observation that statin therapy is associated with a dramatically lower incidence (3% vs.  81%; P < 0.001) of postoperative thrombocytosis.11Whether this relates to the observed decrease in stroke incidence in the vascular patients, however, remains unknown.

Meta-analysis of the studies15,19,20,28–30,35investigating the effect of statin therapy in vascular surgery patients, including two prospective, randomized trials,20,29showed a significant statin-associated benefit in terms of in-hospital mortality. Within the included studies, the studies by O’Neil-Callahan et al.  35and Ward et al.  30did not show any significant statin-associated reduction in the incidence of short-term mortality. However, a statistically significant benefit associated with statin therapy was observed by O’Neil-Callahan et al.  35for the combined endpoint of death, MI, and myocardial ischemia (OR, 0.56; 95% CI, 0.31–0.99; P = 0.046)—a benefit mostly driven by the reduced incidence of myocardial ischemia among statin users. In addition, Ward et al.  30reported that statin therapy was independently associated with improved long-term (mean follow-up period of 5.5 yr) survival (OR, 0.52; 95% CI, 0.32–0.84; P < 0.004) after adjusting for significant baseline characteristics. Further evidence for a beneficial effect of statin therapy beyond the immediate postoperative period in patients undergoing vascular surgery includes the observed reduction in all-cause (OR, 0.4; 95% CI, 0.3–0.6) and cardiovascular (OR, 0.3; 95% CI, 0.2–0.6) mortality at long-term follow-up (median follow-up 4.7 yr) after successful abdominal aortic surgery.45Further, the incidence of saphenous vein graft patency was also reported to be three times higher among statin users 2 yr after primary infrainguinal arterial reconstruction.19 

When analyzing for postoperative morbidity after vascular surgery, our meta-analysis revealed a significant reduction in the incidence of stroke (2.0% vs.  3.3%; P = 0.049) in patients receiving preoperative statin therapy. This finding was largely weighted (90.19%) by the study from Kennedy et al. ,28who observed a beneficial effect of statin pretreatment in symptomatic patients scheduled for carotid endarterectomy, with no effect observed in those patients who were asymptomatic. Durazzo et al.  20confirmed this effect not only in patients undergoing carotid endartectomy surgery, but also in patients undergoing aortic or femoropopliteal procedures. Furthermore, our analysis revealed that statin therapy was associated with a significant reduction in postoperative MI (2.9% vs.  6.2%; P = 0.001). Both prospective trials20,29reported a significant lower incidence of postoperative nonfatal acute MI. This observation may provide further support for the argument that reintroduction of statin therapy in the early postoperative period is imperative to reduce postoperative mortality and cardiovascular complications. Our analysis showed no effect of preoperative statin therapy on cardiac arrhythmias. This was probably because only two small retrospective studies19,35included data on cardiac arrhythmias, with one study reporting a small statin-associated benefit in terms of the incidence of cardiac arrhythmia,19and the other reporting a higher incidence of ventricular tachyarrhythmia among statin users.35 

Biccard et al. ,46who analyzed the two prospective trials20,29of statin use in patients undergoing vascular surgery, reported a number needed to treat of 15 to prevent the combined outcome of cardiovascular complications or death after vascular surgery. This treatment benefit was argued to be more cost-effective than statin therapy for primary and secondary prevention of coronary events. This is supported by the data derived from the article by Lindenauer et al. ,23including the observation that statin-derived benefit was evident in all but the lowest quintile of propensity score for statin therapy and the observation of an inverse correlation between cardiac risk factors and the calculated number needed to treat, with that number ranging from 186 in patients with no cardiac risk factors to 30 in patients with four or more cardiac risk factors. These findings emphasize that high-risk patients undergoing noncardiac surgery may benefit the most from preoperative statin therapy. Further, in this study, patients whose statin therapy was initiated late (2 or more days after surgery) were included in the nontreatment group, again highlighting the importance of early postoperative statin therapy.

Only two small studies26,33investigated the influence of preoperative statin therapy on postoperative morbidity in patients undergoing major noncardiovascular surgery. Both studies reported on the effect of statin therapy on postoperative outcome after thoracic surgery. Because the study by Lindenauer et al.  23included vascular patients, this study was excluded from analysis within this subgroup. Of the two small retrospective studies, one reported that in contrast to other preoperative variables, including β blockers, statin therapy was associated with a reduced risk of postoperative atrial fibrillation.33The other, a propensity score–matched study, did not demonstrate any statin-associated protective effect in the thoracic surgical population.26In fact, the authors observed a trend toward increased cardiovascular complications in those patients receiving statin therapy, suggesting that statins were a surrogate marker for underlying cardiovascular disease.

Although our meta-analysis showed an apparent survival benefit for preoperative statin therapy in patients undergoing surgical interventions, these data should be interpreted with caution because many of the studies included have important limitations. Variation in study design and the decision to include observational data and retrospective studies may be seen as a potential limitation. Meta-analyses prefer prospective, randomized, placebo-controlled trial data in favor of prospective observational or retrospective work to minimize the inherent biases of the latter studies. Immeasurable factors, such as physician bias regarding patient selection, choice of statin, and dosage may account for some of the heterogeneity and conflicting effects observed when observational or retrospective studies are included. Retrospective as well as observational data, however, can still be useful in answering meta-analysis questions (such as the one in the current study), if bias is identified and accounted for.

In the current study, there was a lack of sufficient data from prospective randomized trials to adequately evaluate the effect of preoperative statin therapy on postoperative outcomes. This may be largely attributed to the fact that until recently, statin therapy was not thought of as having acute beneficial effects. Although statins have been randomized in multiple trials in ambulatory patients, few studies exist in surgical patients. Proper methods for selection and combining studies were used to address this issue. According to the guidelines for meta-analysis of observational studies,47inclusion and exclusion criteria, study quality (indicated as quality score of each study), heterogeneity between the studies, and confounding bias of each study were carefully assessed. Fixed and random effects models were used in the absence or presence of heterogeneity between the included studies. Nonetheless, a need for prospective, randomized studies investigating the influence of statin therapy on adverse surgical outcomes still exists.

The current meta-analysis also evaluated for the presence of “publication or reporting bias.” Reporting bias tends to occur when statistically significant or “positive” studies are more likely to be accepted for publication (publication bias), published in English (language bias), published rapidly (time-lag bias), or cited more often (citation bias). Therefore, if a meta-analysis summarizes only published studies prone to these biases, the overall summary effect might be spuriously exaggerated. In the current study, the search strategy included published as well as unpublished studies without any language restriction. However, our attempt to seek all relevant research did not reveal any unpublished data. Further, small positive studies are more likely to be reported than small negative trials. To check for the presence of publication bias, a funnel plot (fig. 7) and statistical tests (e.g. , Egger test) are used.

A further limitation is that no data exist on the minimum duration of preoperative statin therapy that is required to improve postoperative outcome. These data could be further compounded by pharmacokinetic and pharmacogenomic factors. In this regard, the half-lives of the various statins range from 1.5 to 20 h. Furthermore, some studies implicate an interaction between the response to therapy or associated side effects and genetic factors,48–50suggesting that unstudied pharmacogenomic factors may influence these observed beneficial effects. A further limitation is that the majority of these studies did not control for postoperative continuation or return to statin therapy after surgery. Acute discontinuation of statin therapy has been shown to result in a rebound effect with reduced endothelial function that may increase postoperative risk in patients. Heeschen et al.  51reported that patients with acute coronary syndrome who were using statin therapy had a reduced incidence of 30-day all-cause mortality and nonfatal MI; however, in those whom the stain therapy was withdrawn, the cardiac event rate was significantly greater. Reasons for lack of prompt reinitiation of statin therapy in the postoperative period may include conservative reintroduction of postoperative oral intake, excessive nausea and vomiting, transient renal or hepatic dysfunction, and inattention to or a lack of understanding of the importance of the potential protective pleiotropic effects of these drugs by the postoperative surgical team. Other limitations include the fact that preoperative statin therapy may theoretically imply a higher overall standard of care or improved access to health care, which itself may be responsible for improved short- and long-term postoperative outcomes. Further, the observed trend toward increased cardiovascular complications in those patients receiving statin therapy may suggest that statin use may be a surrogate marker for underlying cardiovascular disease and therefore may increase the risk of postoperative complications such as MI.

Given that the studies included in this meta-analysis lack sufficient data on the side effects of statins in the perioperative period, we were unable to do a risk–benefit analysis for the use of statins in the perioperative period. Nonetheless, the observed benefit associated with preoperative statin therapy in the 223,010 patients included in our meta-analysis cannot be ignored. These studies also highlight the potential safety of statin therapy in the perioperative period, which to date has been limited. In these reported studies, Schouten et al.  29observed similar postoperative creatine kinase levels in high-risk patients undergoing major vascular surgery, with no patient experiencing muscle symptoms or rhabdomyolysis. Furthermore, neither long-term nor high-dose statin therapy was associated with adverse outcomes. Nonetheless, Durazzo et al.  20found that statin users were more likely to have elevated levels of creatine kinase and liver transaminase levels, indicating that careful monitoring for drug-related adverse effects is recommended, especially in high-risk patients.

In summary, perioperative statin therapy seems to be associated with a survival benefit, with a variable effect on postoperative cardiovascular morbidity. Larger prospective, randomized clinical trials are needed to confirm this observation and to determine the optimal timing and duration of statin therapy in the surgical setting. Until such studies are completed, it may be prudent to recommend that patients are returned to their statin therapy as soon as possible in the immediate postoperative period.

1. hydroxymethylglutaryl-coa reductase inhibitors/

2. (hydroxymethylglutaryl-coa$ and reductase? inhibitor?).mp. [mp = title, original title, abstract, name of substance word, subject heading word]

3. (hmg-coa$ and reductase$ inhibitor$).mp. [mp = title, original title, abstract, name of substance word, subject heading word]

4. simvastatin$.mp.

5. zocor$.mp.

6. fluvastatin$.mp.

7. lescol$.mp.

8. pravastatin$.mp.

9. pravachol$.mp.

10. lovastatin$.mp.

11. mevacor$.mp.

12. atorvastatin$.mp.

13. lipitor$.mp.

14. rosuvastatin$.mp.

15. crestor$.mp.

16. cerivastatin$.mp.

17. baycol$.mp.

18. pitavastatin$.mp.

19. livalo$.mp.

20. statin$.mp.

21. medostatin$.mp.

22. rosuvastatin$.mp.

23. torvastatin$.mp.

24. or/1-23 [drug terms]

25. exp perioperative care/

26. (intraoper$ or peri-oper$ or preoperat$ or postoper$).mp.

27. exp preoperative care/

28. exp surgical procedures, operative/

29. surg$.mp.

30. su.fs.

31. (pretreat$ or pretreat$).mp.

32. premedication/

33. (premedic$ or premedicat$).mp.

34. (intraoper$ or perioperat$ or postoperat$ or preoperat$).mp.

35. postoperative complications/

36. or/25-35 [patient population]

37. 36 and 24 [patients × drugs]

38. randomized controlled trial.pt.

39. controlled clinical trial.pt.

40. randomized controlled trials.sh.

41. random allocation.sh.

42. double blind method.sh.

43. single blind method.sh.

44. or/38-43

45. (animals not humans).sh.

46. 44 not 45 [hss cochrane phase i]

47. clinical trial.pt.

48. exp clinical trials/

49. (clin$ adj25 trial$).ti,ab.

50. ((singl$ or doubl$ or trebl$ or trip$) adj25 (blind$ or mask$)).ti,ab.

51. placebos.sh.

52. placebo$.ti,ab.

53. random$.ti,ab.

54. research design.sh.

55. or/47-54

56. 55 not 45

57. 56 not 46 [hss cochrane phase ii only]

58. comparative study/

59. exp evaluation studies/

60. follow-up studies.sh.

61. prospective studies.sh.

62. (control$ or prospectiv$ or volunteer$).ti,ab.

63. or/58-62

64. 63 not 45

65. 64 not (46 or 57) [hss cochrane phase iii only]

66. 46 or 57 or 65 [hss phase i-iii]

67. 66 and 37 [hss phases i-iii AND drugs AND patients]