Effect of Perioperative Statins on Death, Myocardial Infarction, Atrial Fibrillation, and Length of Stay:  A Systematic Review and Meta-analysis FREE

Despite advances in technology, operative techniques, and anesthetic interventions, perioperative cardiac complications remain frequent in patients with cardiac risk factors undergoing high-risk surgery.^1- 2 These events are not without consequences: a landmark study³ found that postoperative complications account for 22% of preventable deaths, 2.4 million additional hospital days, and $9.3 billion in additional charges each year.

Vascular and systemic inflammation precipitated by operative intervention are important mediators of many perioperative cardiac events.^4- 5 As the 3-hydroxy-3-methylglutaryl–coenzyme A reductase inhibitors (or “statins”) have pleiotropic properties, interest has grown in using perioperative statin treatment to mitigate the risk of unfavorable cardiac outcomes.⁶ Indeed, several extant meta-analyses have demonstrated reductions in perioperative complications in statin users. Although important, these analyses are limited in generalizability as they include only patients undergoing cardiac surgery, or mixed observational studies of chronic statin use with controlled trials of preoperative treatment.^7- 12 The question as to whether perioperative statin treatment improves surgical outcomes in patients not taking statins or in noncardiac operations thus remains unanswered.

In this meta-analysis, we evaluated the effect of acute statin treatment in statin-naive patients (ie, patients not maintained on long-term statin treatment before trial enrollment) on perioperative death, myocardial infarction (MI), atrial fibrillation, and length of hospital and intensive care unit (ICU) stay in randomized controlled studies.

A medical research librarian performed literature searches for English and non-English articles (see the [eText] for search strategy details). MEDLINE, via PubMed (1950-present), EMBASE (1946-present), Biosis (1926-present), and the Cochrane Central Register of Controlled Trials (1960-present, via Ovid) was searched using Boolean logic to incorporate a variety of terms and synonyms, including randomized controlled trial, statins, hydroxymethylglutaryl-CoA reductase inhibitors, atrial fibrillation, death, myocardial infarction, length of stay, perioperative events, inflammation, and treatment outcome. No publication date or status restrictions were placed on the searches. Controlled vocabularies were used to identify synonyms, whereas study design filters were used to limit retrieved articles to randomized controlled trials. We searched for unpublished and ongoing clinical trials via ClinicalTrials.gov and the Web portals of the International Federation of Pharmaceutical Manufacturers and the Pharmaceutical Research and Manufacturers of America. Conference abstracts were electronically searched through the Conference Proceedings Index provided by Cambridge Scientific Abstracts. Additional studies were identified by hand searches through bibliographies, direct contact with study authors, and consultation with clinical experts (J.B.F. and K.A.E.). We contacted 10 study authors with requests for additional unpublished data or clarification regarding methods. A total of 702 unique articles and conference abstracts were retrieved by this search (last updated April 11, 2011) (Figure 1). We followed the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) recommendations for this analysis.¹³

Two of us (V.C. and D.H.W.) independently determined study eligibility. Any difference in opinion regarding eligibility was resolved by consensus or by a third author (K.A.E. or S.S.). Studies were included if they met the following criteria: (1) human studies with participants 18 years or older undergoing a surgical procedure; (2) enrolled patients not maintained on long-term statin treatment (ie, statin naive); (3) prospective, randomized controlled trials that included a contemporaneous control group; and (4) available outcomes included at least 1 of the following: death, MI, atrial fibrillation, and ICU and hospital lengths of stay. We excluded studies if they (1) were observational or did not involve a surgical intervention, (2) involved percutaneous coronary interventions or cardioversion (considered nonsurgical procedures), (3) did not report clinical outcomes, or (4) included previous statin users. If a study met all the inclusion criteria but did not report outcomes of interest, we contacted the corresponding author to obtain these data. In all cases of contact, open-ended questions were used to obtain further details. If we could not obtain this information, the study was excluded from the analysis.

Data were extracted from all included studies independently and in duplicate (by V.C. and D.H.W.) on a form based on the Cochrane Collaboration data extraction template. We accepted the outcome definitions as stated by each study without independently validating or reviewing their data.

Study quality was assessed by the risk of bias in individual trials according to the method described by the Cochrane Statistical Methods Group.¹⁴ Each study was, thus, classified as being at low, unclear, or high risk for bias. Trials that reported all the quality domains using appropriate methods were considered to have a low risk of bias. For studies that did not report specific domains, we contacted the corresponding author to obtain this information. If we could not obtain these data, we considered the study to be at unclear or high risk for bias based on performance in other domains.

Treatment groups were defined as patients randomized to receive statin treatment vs those randomized to the control group before surgery. In studies reporting data from patients exposed to statins and other interventions, we excluded patients exposed to multiple interventions and included only those randomized to receive statins. In studies in which both trial arms were treated with varying doses of statins (ie, nonplacebo arm), we treated the control arm as placebo and included this study in the analysis given the relationship of statin treatment in the intervention group with the outcomes of interest. To verify the robustness of the meta-analytic results, sensitivity analyses excluding these studies were performed.

Incidence data were extracted from eligible studies. The outcomes of death, MI, and atrial fibrillation were treated as dichotomous variables, whereas hospital and ICU lengths of stay were treated as continuous variables. Length-of-stay data were recorded as mean (SD) in days. The primary measure of pooled treatment effect for death, MI, and atrial fibrillation was the relative risk (RR) of these outcomes in patients randomized to receive statins compared with control subjects. The primary measure of treatment effect for hospital and ICU length of stay was standardized mean difference in patients randomized to receive statins compared with controls. All the analyses were based on a random-effects model using the method described by DerSimonian and Laird.¹⁵ We explored heterogeneity between studies using the Cochran Q test (P < .05) and the I² statistic (I² > 50%). Publication bias was assessed by the Harbord test for dichotomous variables and the Eggers test for mean differences in continuous variables. P < .05 was indicative of statistically significant publication bias.

We conducted subgroup analyses to determine whether type of surgery or potency or duration of statin treatment–affected outcomes. Statin potency was classified according to the low-density lipoprotein–lowering effects of each statin reported in a recent meta-analysis.¹⁶ Using this definition, fluvastatin or pravastatin at any dose and simvastatin at doses of 40 mg or less were considered low potency and rosuvastatin or atorvastatin at any dose and lovastatin or simvastatin at doses of 80 mg or higher were considered high potency. Based on the distribution of treatment duration across studies, we dichotomized duration of statin therapy into 14 days or less and 15 days or more. Several sensitivity analyses were performed, including fitting fixed-effect models, excluding nonplacebo comparator arm studies, excluding cardiac surgery studies, and excluding studies of low and unclear quality.

Data management and statistical analyses were performed using a commercially available software program (STATA SE/MP, version 11; StataCorp LP). All the statistical tests were 2-tailed; P < .05 was considered statistically significant.

Of 702 unique citations identified by the present search, 15 randomized controlled studies involving 2292 patients fulfilled the eligibility criteria and were included in the systematic review (Figure 1). Eligible trials ranged from 40 to 533 patients and evaluated a variety of statin doses and regimens (Table 1). The total duration of statin treatment varied from 3 to 67 days; preoperative treatment ranged from 2 to 37 days, whereas postoperative treatment ranged from 7 to 30 days. No patients were prescribed statins before randomization in any study. Regarding surgical interventions, 2 trials involved noncardiac surgery,^17- 18 2 involved vascular surgery,^19- 20 and 11 involved cardiac surgery.^21- 31 One study used a nonplacebo control that featured treatment with a lower dose of the same statin.²⁰ The Cohen interrater κ statistic for study eligibility was 0.90 and for study abstraction was 0.86, indicative of excellent interrater agreement.

Based on the Cochrane method for assessing risk of bias, 7 of the 15 included studies were at low risk for bias,^{17,19- 21,23,26,31} 5 had unclear risk of bias,^{22,25,27- 29} and 3 were at high risk for bias^18,24,30 (Table 2).

Ten of the 15 included studies reported rates of MI (n = 2077).^{17- 21,23,25,29- 31} The unweighted incidence of perioperative MI in patients treated with perioperative statins was 4.5% (47 of 1041) vs 8.9% in controls (92 of 1036). Pooled results revealed that statin treatment was associated with a statistically significant reduction in MI (RR, 0.53; 95% CI, 0.38-0.74). The absolute risk reduction in MI attributable to statin treatment was 4.4%, conferring a number needed to treat of 23. No heterogeneity was noted across studies (I² = 0%, Cochran Q test statistic = 4.31, P = .89) (Figure 2). The Harbord test statistic suggested a low likelihood of publication bias in included studies (P = .36).

Nine of the 15 eligible studies (all involving cardiac surgery) reported the occurrence of perioperative atrial fibrillation (n = 933).^{21- 24,26- 27,29- 31} The incidence of atrial fibrillation in patients treated with perioperative statins was 19.9% (93 of 467) vs 36.3% in controls (169 of 466). Pooled results revealed that statin treatment was associated with a statistically significant reduction in perioperative atrial fibrillation (RR, 0.56; 95% CI, 0.45-0.69) (Figure 2). The estimated absolute risk reduction was 16.3%, conferring a number needed to treat of 6. No heterogeneity was noted across pooled studies (I² = 0%, Cochran Q test statistic = 6.68, P = .67). The Harbord test showed a low likelihood of publication bias (P = .89).

Although 10 of 15 studies included inpatient or 30-day mortality as an outcome of interest, death occurred in only 5 studies.^17- 20,23 Therefore, only data from these studies (n = 1436) were analyzed for this outcome. Statin treatment was associated with a 2.2% incidence of death (16 of 719) vs 3.7% in controls (26 of 711). Pooled data showed that this effect was not significant (RR, 0.62; 95% CI, 0.34-1.14; P = .13) (Figure 2). No heterogeneity was noted in the pooled studies (I² = 0%, Cochran Q test statistic = 0.83, P = .93). Results of the Harbord test suggested a low likelihood of publication bias (P = .54).

Twelve of the 15 included studies reported mean hospital length of stay in days (n = 2105).^{17- 18,20- 21,23,25- 31} Eight of these 12 studies involved only cardiac surgery. Statin treatment significantly reduced mean length of hospital stay (standardized mean difference, −0.32; 95% CI, −0.53 to −0.11) (Figure 3). However, a high degree of heterogeneity was noted across pooled studies (I² = 78.7%, Cochran Q test statistic = 51.6, P ≤ .001). The Egger test for publication bias yielded P = .07.

Nine of the 15 included studies reported ICU length of stay (n = 1503).^{17- 18,24- 30} All the measurements were converted from hours into days for analysis. Statin treatment was not associated with a significant reduction in length of ICU stay (standardized mean difference, −0.08; 95% CI, −0.25 to 0.10; P = .40) (Figure 3). Heterogeneity across included studies was noted, albeit to a lesser extent (I² = 38.3%, Cochran Q test statistic = 13.0, P = .11). The Egger test for publication bias revealed no evidence of publication bias (P = .10).

Exclusion of studies involving cardiac surgery removed 11 of the 15 eligible studies and statistically negated the reduction found in hospital length of stay. Furthermore, evaluation of the effect of statins on atrial fibrillation in noncardiac surgery was impossible as no trials remained that reported this outcome. However, statin treatment continued to demonstrate reduction in the risk of MI in noncardiac surgery. None of the other sensitivity analyses affected the conclusions (Table 3).

We performed subgroup analyses to investigate the effect of potency and duration of statin treatment on outcomes of interest. Although these analyses were limited by a lack of statistical power, the point estimates of RR reduction associated with statin treatment remained comparable, suggesting that no significant dose or duration effect existed in the included studies (Table 4).

The results of this meta-analysis of 15 randomized controlled trials suggest that perioperative statin treatment reduces the risk of atrial fibrillation and MI and decreases mean length of hospital stay in statin-naive patients undergoing cardiac and noncardiac surgery. Although we found a clinically relevant trend toward reduction in death, it did not meet the threshold for statistical significance. Together, these findings suggest that perioperative treatment with statins is an important risk-reducing intervention with clinical and economic benefits.

The quest to prevent perioperative complications has led to important advances, such as reductions in venous thromboembolism and surgical site infections.^32- 33 Despite these gains, perioperative complications, such as MI and atrial fibrillation, persist. In fact, atrial fibrillation remains the most common cardiac arrhythmia after cardiac surgery, with a prevalence as high as 60% in this setting.³⁴ Atrial fibrillation has important clinical (eg, stroke) and financial ramifications as its occurrence is associated with increased resource utilization and length of stay.³⁵ Similarly, MI remains a frequent complication of major noncardiac surgery. In the largest perioperative β-blocker trial to date, MI occurred in 5.1% of 8351 patients undergoing noncardiac surgery.³⁶ This incidence is greater in high-risk cohorts, such as that of the Coronary Artery Revascularization Prophylaxis study,³⁷ in which MI occurred in 26.5% of the 377 patients undergoing vascular surgery. In this context, the decrease in the risk of MI and atrial fibrillation attributable to perioperative statin treatment is important and highly relevant. Furthermore, the absolute risk reduction associated with perioperative statin treatment in the present analysis exceeds the benefit of statin treatment in secondary prevention settings, underscoring a unique advantage to the use of statins in the perioperative context.^38- 39

Although the precise mechanisms by which statins reduce adverse perioperative events are uncertain, numerous hypotheses abound. Major cardiac complications during surgery represent a crescendo of deleterious physiologic responses. Surgery produces a catecholamine surge from the neuroendocrine system, elevating heart rate, myocardial contractility, and oxygen demand.^4,40 In the setting of stenotic coronary artery disease, narrowed coronary lumina commonly do not accommodate this increased requirement, precipitating myocardial ischemia and cardiac events. Operative intervention also triggers an intense inflammatory response believed to be responsible for cardiac events in the absence of significant coronary artery disease.^17- 18,40 Inflammation specifically leads to changes in atheromatous plaque morphologic features, culminating in the rupture of lipid-laden “vulnerable” intimal lesions, the formation of platelet-rich thrombi, acute vessel occlusion, and ischemia.⁵ Statins modulate vascular function to counteract these natural responses. For example, statins upregulate endothelial nitric oxide synthase, leading to coronary artery vasodilation and improved flow in patients with occlusive coronary artery disease.^41- 42 Statin treatment tempers systemic and vascular inflammation, manifest by reductions in biomarkers associated with plaque instability and perioperative cardiac events.^{17- 18,39- 40} Conversely, withdrawal of chronic statin use during surgery is associated with an inflammatory “rebound” that increases the risk of adverse perioperative cardiac events.⁴³

The present findings should be considered in the context of several limitations. First, as with all meta-analyses, these results could be influenced by publication bias. Second, most of the studies eligible for this analysis included patients undergoing cardiac surgery as few randomized trials of patients undergoing noncardiac surgery exist. Therefore, the extent to which the benefits we observed for atrial fibrillation or hospital length of stay can be extrapolated to patients undergoing noncardiac surgery is unclear. Third, the inferences regarding the impact of statin treatment on hospital and ICU length of stay are limited owing to the heterogeneity of treatment effect noted in the results. Given the skewed and variable nature of this measure, caution is needed when interpreting these results.

Limitations notwithstanding, the present analysis has important implications. First, the high prevalence of atrial fibrillation and MI in cardiac surgery and the protection against these events afforded by statins in statin-naive patients suggests that perioperative statin treatment should be routinely implemented in this setting. This approach validates the notion that “acute” perioperative statin treatment specifically geared toward cardiac risk reduction in non–statin users is an effective strategy. Despite an abundance of trial-derived data, the evidence suggests that the use of statins in high-risk populations remains suboptimal (approximately 40%).⁴⁴ The perioperative consultant is, thus, presented a unique opportunity for the initiation of appropriate treatment and unparalleled risk reduction. Although we could not specifically evaluate the adverse effects of statin treatment (in part due to a lack of uniformity in definitions and universal reporting of these events), recent data suggest that the theoretical risk of these outcomes in the perioperative setting outweighs their clinical incidence.^17- 18,45

Second, owing to the significant reduction in the risk of MI, wider use of statins should be encouraged in high-risk patients undergoing noncardiac surgery. The American College of Cardiology Foundation/American Heart Association Guidelines on Perioperative Cardiovascular Evaluation in Noncardiac Surgery limit their class I recommendation on perioperative statins to continuing treatment in those already receiving this therapy.⁴⁶ The present analysis suggests that it may be appropriate for future guidelines to be broadened to endorse statin treatment in specific statin-naive populations, such as those undergoing high-risk procedures (eg, intra-abdominal, intrathoracic, or cardiac surgery) or those at high risk for cardiac events (eg, unstable coronary artery disease). The European Society of Cardiology⁴⁷ has recently endorsed the former in their perioperative guidelines, a recommendation supported by the results of the present analysis. Furthermore, in view of the potential effect on cardiac outcomes, perioperative statin treatment may merit inclusion as a process measure in the Surgical Care Improvement Project,⁴⁸ although more evidence is needed before promulgating this metric.

Third, we found that perioperative statin treatment is associated with a 30% reduction in mean length of hospital stay. Although we did not find similar benefits in length of ICU stay, this result is interesting and warrants further investigation. We do not believe that this finding is due to a “healthy user” effect, as only non–statin users and randomized controlled trials were included in these analyses.

In conclusion, we found that treatment with perioperative statins in cardiac and noncardiac surgery significantly reduced the risk of MI and atrial fibrillation and decreased the mean duration of hospitalization. These results suggest that perioperative practice and guidelines should be modified to incorporate greater use of statins in patients undergoing surgery.