|Home | About | Journals | Submit | Contact Us | Français|
Although statin lipid-lowering medications likely reduce peri-operative ischemic complications, few data exist to describe statins’ effects on risk for and outcomes of atrial fibrillation following non-cardiac surgery.
To examine the association between treatment with statin medications and clinically significant post-operative atrial fibrillation (POAF) following major non-cardiac surgery.
A retrospective cohort study of patients aged 18 years or older who underwent major non-cardiac surgery between January 1, 2008 and December 31, 2008. Cases of clinically significant POAF were selected using a combination of ICD-9 codes and clinical variables. We defined statin users as those whose pharmacy data included a charge for a statin drug on the day of surgery, the day after surgery, or both.
Of 370447 patients, 10957 (3.0%) developed clinically significant POAF; overall, 79871 (21.6%) received a peri-operative statin. Patients receiving statins were generally older (68.8 vs. 61.1 years; P<0.001) and more likely to be receiving a beta-blocker (50.3% vs. 21.6%; P<0.001). Statin use was associated with a lower unadjusted rate of POAF (2.6% vs. 3.0%; P<0.001). After adjustment for patient risk factors and surgery type, odds for POAF remained significantly lower among statin-treated patients (adjusted odds ratio [AOR] 0.79; 95% confidence interval [CI] 0.71–0.87; P<0.001). Statin use was not associated with differences in cost, length of stay, or mortality among patients who developed POAF.
Treatment with statin agents appears to be associated with lower risk of clinically significant POAF following major non-cardiac surgery.
Atrial fibrillation affects approximately 2.5 million people in the United States and roughly 10% of people over 80 years of age.1 The chronic form of this disease has been recognized as an epidemic and a major contributor to rising health care costs. Recently, POAF after coronary artery bypass graft surgery (CABG) has been associated with increased morbidity and health care cost,2,3 highlighting the need for strategies to minimize its occurrence.
Anti-inflammatory medications such as corticosteroids and ketorolac have been shown to decrease the incidence of POAF,4,5 which is not unexpected, since the connection between surgery and atrial fibrillation is thought to be predominantly mediated by systemic inflammation.6–8 Unfortunately, these agents retard wound healing and increase bleeding complications. Statin medications have also been shown to decrease POAF (in the setting of cardiac surgery);9 the putative mechanism is thought to be related to the “pleiotropic” anti-inflammatory properties of statins, as well as their direct anti-arrhythmic effect on pulmonary vein tissue.10
Statins are associated with reduced mortality following non-cardiac surgery.11,12 The specific mechanisms by which statins may lower mortality in this group of patients have not been fully elucidated. A decrease in the incidence of myocardial ischemia has been implicated, but a lower incidence of post-operative arrhythmia in those receiving peri-operative statin therapy may be another mediator for this protective effect. Few studies, none large, have examined the relationship between statins and POAF following non-cardiac surgery.13 For that reason, we carried out a study to examine the association between statin use and the incidence of POAF in a large cohort of patients undergoing major non-cardiac surgery in a broad selection of hospitals in the United States.
Data collected from patients hospitalized for non-cardiac surgery at 375 United States hospitals were utilized. These data were obtained from Perspective, a database developed for quality and utilization benchmarking by Premier Incorporated, Charlotte, NC. In addition to data elements available in the standard hospital discharge file, the Perspective database contains a date-stamped log at the individual patient level of all billed items, including medications as well as laboratory, diagnostic, and therapeutic services.
Patients were included in the analysis if they were 18 years or older and underwent major non-cardiac surgery between January 1, 2008 and December 31, 2008. Surgical procedures were categorized using a health information system (APR-DRG, version 27.0, 3M Corp, Minneapolis, MN) and were considered major if the mean length of stay for patients in the diagnosis related groups was 3 days or more.14 Procedures classified as cardiac and obstetric were excluded. The included list of procedures was then organized into 9 categories of surgical type.
For purposes of sensitivity analyses, it was desirable to be able to discern new atrial fibrillation from chronic atrial fibrillation. To that end, cohort selection was focused on patients whose data had accurate and complete present-on-admission coding. These codes represent a date stamp in administrative data for secondary diagnoses; they help determine which diagnoses are present at the time of admission. 15–17 Because present-on-admission coding was implemented voluntarily in the middle of the study time period, only hospitals that had complete present-on-admission coding for all patients diagnosed with atrial fibrillation were included.
For each case the urgency of the admission (elective or emergent), hospital characteristics, and operating surgeon were recorded. Age, sex, race, and insurance status were captured. Co-morbidities were established using a combination of International Classification of Diseases, Ninth Revision, Clinical Modification (ICD -9-CM) secondary diagnoses using standard methods.18
Permission to perform the study was obtained by the institutional review board at UCSF, San Francisco, CA and informed consent was waived.
The designation of “significant” post-operative atrial fibrillation (POAF) was made using a secondary diagnosis ICD-9-CM code for atrial fibrillation (427.31) in combination with charge or procedure code data which suggested that the patient’s dysrhythmia necessitated therapy. A combination of an ICD-9-CM code of 427.31 and one of the following was required for a diagnosis of POAF: a procedure code for electrical cardioversion in the post-operative period, pharmacy charge data for intravenous beta-blockers in the post-operative period, pharmacy charge data for intravenous calcium channel blockers in the post-operative period, pharmacy charge data for intravenous amiodarone in the post-operative period, or pharmacy charge data for a new prescription of digoxin in the post-operative period.
Pharmacy records were used to identify whether a statin medication was administered during the peri-operative period. The peri-operative time window was limited to day 0, the day of surgery, and day 1, the day after surgery. The comparison groups were the set of patients who received statin medication in the peri-operative window (“statin-treated”) and the set of patients who did not receive statin medication in the peri-operative window (“non statin-treated”).
Peri-operative administration of beta-blockers, angiotensin-converting enzyme inhibitors, calcium channel blockers, anti-platelet agents, anti-arrhythmics, and digoxin were assessed using pharmacy records. A similar approach was used to assess post-operative use of medications.
In-hospital mortality, length of stay, actual hospitalization costs and 15/30-day readmission rates were obtained from the Perspective discharge file.
Summary statistics for the overall sample were constructed by using frequencies and proportions for categorical data and mean, medians, and inter-quartile ranges for continuous variables. These summary statistics were stratified by whether patients received peri-operative treatment with statin medications.
Univariate logistic regression was used to determine the unadjusted relationship (odds ratio) between peri-operative statin use and POAF. Next, a multivariate logistic regression model for POAF was constructed. A generalized estimation equations model with logit link, independence working correlation, and robust standard errors was employed. A group of likely confounders (age, for example) was specified a-priori – these were automatically included in the model for face validity. Another list of possible confounders was generated using a directed acyclic graph,19 which was constructed from general clinical knowledge and data from prior studies. Confounders from this second group were kept in the model if they were found to change the model in a statistically significant manner (p <0.05). This model provided an adjusted odds ratio (AOR) describing the relationship between peri-operative statin use and POAF. Robust standard errors were used to account for clustering by hospital, arising from unmeasured differences in practice and expertise. Finally, a model was developed to describe the association between statin use and outcomes (length of stay, costs, readmission, and mortality) for the subset of the cohort that developed POAF.
Propensity matching was employed to test whether the results from the standard model were robust. A propensity score for peri-operative treatment with statin was derived by constructing a non-parsimonious logistic model which was then used to create a score representing the likelihood of a patient receiving a statin medication. This score was then used to create pairs of patients (with each pair containing one patient who received statin and one who did not) that were matched to the 7th decimal place of likelihood of receiving statin, based on propensity score. Using this matched cohort, logistic models examining likelihood of atrial fibrillation were re-constructed. Robust standard errors were again employed to account for clustering by hospital.
Lastly, adjusted logistic models were used to estimate the marginal causal odds ratio and risk difference for use of statin therapy, both defined in terms of the average incidence of POAF that would be expected if everyone in the cohort were treated with statins, versus if no one were treated.20 This was accomplished by fitting separate models for statin treated and untreated subjects; marginal POAF probabilities were then estimated by averaging the predicted probability of POAF for each patient, calculated under both models. Bias-corrected bootstrap percentile confidence intervals were calculated for both marginal effect estimates, with re-sampling by hospital.
All analyses were performed using STATA version 10 (Stata Corporation, College Station, Texas).
A total of 370447 patients meeting eligibility criteria underwent major non-cardiac surgery during the study period and were included in the analysis. Of these, 79871 patients (21.6%) received a statin medication in the peri-operative period. Patients treated with statin medications were older and were more often white. Statin-treated patients were more likely to have been admitted electively and to list Medicare/Medicaid as their primary form of insurance. Prevalence of statin use varied among surgical groups (from 8.0% for abdominal surgeries to 35.2% in vascular surgeries).
Of 370447 total patients, 10957 (3.0%) had significant atrial fibrillation during their hospitalization. Of 79871 patients who were treated with peri-operative statin, 2114 (2.6%) developed POAF; 290576 patients were not treated with peri-operative statin, of which 8843 (3.0%) developed POAF (unadjusted odds ratio for POAF with statin use 0.87, 95% confidence interval [CI] 0.83–0.91; P<0.001). POAF occurred most frequently among patients undergoing thoracic (non-cardiac) and otolaryngological surgeries (5.8% and 6.2% respectively).
Adjustment for patient characteristics, co-morbidities, surgery type, and hospital factors in the multivariable logistic models demonstrated a more pronounced protective effect of statin use (adjusted odds ratio [AOR] 0.79; 95% CI 0.71–0.89; P<0.001). This protective association was also observed in secondary analyses within a propensity-matched subset, although with wider confidence intervals due to smaller sample sizes (AOR 0.74; 95% CI 0.57–0.95; P=0.021). The marginal effects regression model further confirmed these findings. The marginal causal odds of developing POAF decreased by 19% (95% bootstrap percentile CI −25% - −13%) with statin therapy; the corresponding marginal risk difference was −0.56% (95% bootstrap percentile CI −0.74% - −0.38%).
To examine whether statin use had a differential association with POAF, the statistical significance of interaction terms between the type of surgery, patient co-morbidities, and concomitant use of medications was tested. Although statin use overall was associated with lower odds for POAF, statin effect did not differ significantly across surgery type. Statins did have an incrementally larger protective effect against POAF in patients with diabetes (AOR 0.63; 95% CI 0.54–0.72), while the statin effect was attenuated but still significant in patients taking beta-blockers (AOR 0.89; 95% CI 0.82–0.98).
In unadjusted analyses, patients developing POAF who received peri-operative statin had lower mortality, shorter length of stay, and lower costs. After adjustment, however, any outcome benefit of statin treatment was markedly attenuated and all differences became non-significant.
In addition to the propensity-matched analyses described above, sensitivity analyses were carried out to test the robustness of our findings. In order to make sure that the observed treatment effect was not dominated by the prevention of significant POAF in patients with pre-existing atrial fibrillation, the analysis was repeated using present-on-admission codes to exclude patients coded as having pre-existing atrial fibrillation. In this subgroup, there was a persistent treatment with statins decreasing atrial fibrillation necessitating therapy (AOR 0.85; 95% CI 0.73–0.99; P=0.035). Confidence intervals were wider given the decrease in the number of outcomes.
The possibility that false positive cases may have been introduced into the outcome variable was of concern. For example, this might occur if the outcome was driven by the administration of intravenous beta-blockers in the post-operative period, when in fact the intravenous beta-blocker was administered for blood pressure control rather than for control of ventricular response rate in the setting of POAF. When the analysis was repeated with post-operative intravenous beta-blockers removed from the outcome variable, however, there was a persistent treatment effect for statins in decreasing atrial fibrillation necessitating therapy (AOR 0.85; 95% CI 0.76–0.96; P=0.007). Separate sensitivity analyses removing each of the other individual elements of the outcome variable (digoxin, amiodarone, calcium channel blockers, and cardioversion) confirmed a consistent statistically significant protective association between statin use and POAF.
Patients who had late surgery (defined as surgery on hospital day 4 or later) represent a distinct sub-population that may be more likely to have an extreme clinical presentation (i.e. critical illness or significant pre-operative complication). An analysis excluding all late surgeries showed no significant change in the statin treatment effect (AOR 0.81; 95% CI 0.72–0.92; P=0.001).
When non-steroidal anti-inflammatory agents (NSAIDs) and steroid medications were added to the model, there was no significant change in the statin treatment effect (AOR 0.79; 95% CI 0.71–0.88; P=0.001). NSAIDs (AOR 0.92; 95% CI 0.85–0.99; P=0.035) and steroids (AOR 0.88; 95% CI 0.80–0.97; P=0.012) were both associated with lower odds for POAF.
In this large cohort of non-cardiac surgical patients, administration of statin medications in the peri-operative time period was associated with markedly reduced odds of developing clinically significant POAF. Though the absolute observed effect size was small (a 0.4% reduction in POAF), the at-risk pool of patients undergoing major non-cardiac surgery is very large. This modest risk reduction could drive relevant changes in health care cost and outcomes when applied to millions of patients annually. While statins’ effectiveness varied somewhat in patients using other cardiovascular medications concomitantly, the association between reduced odds of POAF and statin use was robust across a wide range of subgroups. Of note, statin treatment had no effect on outcomes among patients who did develop POAF. Whether statins’ effectiveness in reducing POAF is due to a direct anti-arrhythmic effect or mediated through reduced risk for other complications (such as myocardial infarction or infection) cannot be definitively discerned with our data.
The routine use of peri-operative medications to reduce adverse events has been an area of intense study for the past decade. The role of beta-blockers has been rapidly evolving, with recent literature suggesting no net clinical benefit (and perhaps harm) associated with the routine peri-operative administration of beta-blockers for non-cardiac surgery.21,22 Meanwhile, peri-operative statin use has garnered increasing interest as a preventive strategy. A large study by Lindenauer et al. suggested that statin use is associated with lower mortality in a population of patients undergoing major non-cardiac surgery.11 Additionally, statins have been shown prospectively to decrease cardiovascular complications following vascular surgery.12,23 Few data exist to describe the association between statin use and incidence of POAF in non-cardiac surgery, though statins have been shown to lower POAF risk in patients undergoing CABG.9
Our findings are consistent with previous studies suggesting protective benefit of statins in surgical patients, and suggest that statins may proffer benefit to patients through the reduction of their risk for developing clinically significant POAF. There is a biologic basis to support the idea that statins may have a direct anti-arrhythmic effect,10,24,25 and our data provides supporting evidence for this mechanism. It is, however, also possible that our data are describing an indirect association between statin use, other complications of surgery, and POAF. For instance, there is current controversy as to whether statin use decreases post-operative sepsis,26–28 an effect that could cause a decrease in POAF as a downstream consequence. It is difficult to tease apart such indirect associations in these data. However, the fact that statins do not appear to be associated with any differences in subsequent outcomes among patients who develop POAF suggests that statins’ anti-arrhythmic effects are an important part of their protective profile.
It is important to point out that statin use, as defined in this paper, likely represents longitudinal use of the medication. Unlike beta-blockers, statins are infrequently started at the time of surgery or acutely during hospitalization. Therefore, the data presented likely represent longer-term use of these medications, with the attendant likelihood that other elements of the patient’s care may have been managed over the longer term as well. To address this potential bias, secondary analyses were performed using propensity score methods to account for socioeconomic or clinical factors associated with statin administration; these results were essentially identical to the findings from the standard fully adjusted regression model.
Our study has a number of limitations. We did not have the ability to track what medications patients were taking prior to hospitalization, thus preventing us from identifying which patients had statin withdrawal, longitudinal continuation of statin, or new prescription of statin. This limited our ability to assess the acute efficacy of statin therapy and the effect of statin withdrawal, which is thought to be a risk factor for cardiovascular events. 20,29 Since administrative data was used, subtle clinical information, such as the incidence of delirium, surgical site infection, or other potential contributing causes of POAF is lacking. Echocardiographic parameters which may have represented important confounding variables, such as left atrial size, were not available within the database. The timing and duration of POAF episodes were also unavailable for review -- such information could have provided further insight as to the clinical significance of each arrhythmic event. In addition, the data do not include detailed past medical history information. For that reason, the present-on-admission coding was employed to discern incident from prevalent atrial fibrillation and standard risk adjustment methodologies were used to account for patient risk factors. The use of present-on-admission codes does not, however, allow us to draw the important distinction between sub-types of pre-existing atrial fibrillation: paroxysmal, persistent, and long-standing persistent. Though the use of these codes is becoming more common, many of the pitfalls involved with their use are related to potentially variable coding practices between hospitals.30,31 To address this problem, we used data only from sites which were 100% compliant with present-on-admission coding, and limited our outcome variable to groups most likely to have had clinically significant atrial fibrillation. Moreover, sensitivity analyses showed that statin medications retained their protective effect when patients with pre-existing atrial fibrillation were excluded from the cohort. Additionally, we acknowledge that patients treated with lipid-lowering medications could have been healthier than those that were statin untreated, or that the administration of a statin may have simply been a marker for higher quality operative and peri-operative care. Reassuringly, however, our results were consistent across subgroups and in the marginal effects regression model.
Results from our large national observational study suggest that statin use is associated with significantly lower risk for atrial fibrillation following major non-cardiac surgery, and that this association is robust across a wide range of surgical procedures and patient groups. Statin use did not improve outcomes for patients in whom clinically significant POAF occurred, suggesting that statins’ effect is in primary prevention of adverse events. While our data should not be used as support for starting statins in the peri-operative period, they add to the growing body of evidence for statins’ potential protective effect in surgical patients.
FUNDING/DISCLOSURES: Dr. Auerbach and Ms. Maselli were supported by a Mid-Career Development grant K24HL098372 (NHLBI) during this research. Dr. Goldman was supported by an AHRQ K08 Mentored Clinical Scientist Development Award, Grant # 1 K08 HS018090-01 and NIH/NCRR UCSF-CTSI Grant Number UL1 RR024131. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the NIH.
Author potential conflicts of interest: None
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.