Association of Atrial Fibrillation and Oral Anticoagulant Use With Perioperative Outcomes After Major Noncardiac Surgery.

BACKGROUND: We examined the association of atrial fibrillation (AF) and oral anticoagulant use with perioperative death and bleeding among patients undergoing major noncardiac surgery. METHODS AND
RESULTS: A population-based study of patients aged 66 years and older who underwent elective (n=87 257) or urgent (n=35 930) noncardiac surgery in Ontario, Canada (April 2012 to March 2015) was performed. Outcomes were compared between AF groups using inverse probability of treatment weighting using the propensity score. Of 4612 urgent surgical patients with AF, treatments before surgery included warfarin (n=1619), a direct oral anticoagulant (DOAC) (n=729), and no anticoagulation (n=2264). After urgent surgery, the death rate within 30 days was significantly higher in patients with AF compared with patients with no AF (hazard ratio [HR], 1.28; 95% confidence interval [CI], 1.12-1.45). In contrast, among 4769 elective surgical patients with AF treated with warfarin (n=1453), a DOAC (n=1165), or no anticoagulation (n=2151), prior AF was not associated with higher mortality. Comparing patients with AF who were or were not anticoagulated, there was no difference in 30-day mortality after urgent (HR, 0.95; 95% CI, 0.79-1.14) or elective (HR, 0.65; 95% CI, 0.38-1.09) surgery. There was no difference in 30-day mortality between patients with AF treated with a DOAC or warfarin after urgent (HR, 0.91; 95% CI, 0.70-1.18) or elective (HR, 1.64; 95% CI, 0.77-3.53) surgery. Bleeding and thromboembolic rates did not differ significantly among patients with AF prescribed a DOAC or warfarin.
CONCLUSIONS: Prior AF was associated with 30-day mortality among patients undergoing urgent surgery. In patients with AF, neither the preoperative use of oral anticoagulants, nor the type of agent (either a DOAC or warfarin) were associated with the rate of 30-day mortality.

Clinical Perspective

What Is New?

Among patients with atrial fibrillation undergoing major noncardiac surgery, there was no excess risk of 30‐day bleeding or death among those prescribed oral anticoagulants, and rates of these outcomes did not differ when a direct oral anticoagulant was compared with warfarin.

What Are the Clinical Implications?

Although atrial fibrillation is an independent predictor of mortality in patients undergoing urgent noncardiac surgery, use of direct oral anticoagulants or warfarin do not explain their higher perioperative risk.

Introduction

Since their approval for stroke prevention, prescription rates for direct oral anticoagulants (DOACs) have risen steadily among patients with atrial fibrillation (AF) in North America. At present, they represent over 60% of new prescriptions and over one fifth of all oral anticoagulant prescriptions for such patients.1, 2, 3, 4 Many of these patients will eventually require surgical intervention, as it is estimated that over 50 million surgical procedures are performed yearly in the United States.5 Among patients undergoing noncardiac surgery in prospective cohort registries, ≈7% have preexisting AF.6

While DOACs are being increasingly used, there is a paucity of information on the outcomes of patients with AF who are treated with anticoagulants who subsequently undergo major noncardiac surgery. Subgroup analysis of the RE‐LY (Randomized Evaluation of Long‐Term Anticoagulation Therapy) trial7 suggested that the risks of major bleeding did not differ between those prescribed dabigatran or warfarin, occurring in ≈3% of elective and 20% of urgent surgeries. More recently, an analysis of procedures performed during the ARISTOTLE (Apixaban for the Prevention of Stroke in Subjects With Atrial Fibrillation) trial8 reported a 30‐day major bleeding risk of only 1.6%, with more than one third of patients undergoing procedures without interrupting apixaban. Nonetheless, it is important to note that ≈90% of these surgeries were considered minor.

Outside of the clinical trial arena, there are limited data on perioperative outcomes of patients with AF who are treated with DOAC or warfarin. Accordingly, we examined the postoperative outcomes of patients with and without AF in Ontario, the most populous province of Canada, who subsequently underwent major noncardiac surgery of varying acuity—either elective or urgent procedures. Specifically, we examined the association between prescription of warfarin, a DOAC, or no anticoagulation in patients undergoing major surgery, with the rate of death or bleeding within 30 days. We hypothesized that outcomes among patients with AF would be influenced by the presence or absence of anticoagulation, as well as the specific type of anticoagulant prescribed. Furthermore, we hypothesized that there would be an association between the presence of preoperative AF and outcomes after urgent or elective major noncardiac surgical procedures.

Methods

Data Sources

Using each patient's unique encoded provincial health card number, we linked multiple administrative healthcare databases to create the study cohort. We used the Canadian Institute for Health Information Discharge Abstract Database (CIHI‐DAD) to identify all hospitalizations and the CIHI National Ambulatory Care Reporting System Database (NACRS) to identify all emergency department visits. Noncardiac surgical procedures were identified using the Canadian Classification of Interventions codes in the CIHI‐DAD and the CIHI Same Day Surgery database (CIHI‐SDS). We examined the Ontario Registered Persons Database to ascertain deaths. We also used the Ontario Drug Benefit prescription database to identify funded drug prescriptions filled by individuals aged 65 years and older. Finally, physician billing claims were identified using the Ontario Health Insurance Plan database. The study was approved by the research ethics board of Sunnybrook Health Sciences Centre. There was no need for informed consent since this was an analysis of a population database.

Study Cohort

We included all patients aged 66 years and older who underwent an elective or urgent noncardiac surgical procedure at an acute care hospital in Ontario, Canada, between April 1, 2012 (first date of reimbursement of DOACs by the Ontario Drug Benefit plan), and March 31, 2015. To avoid incomplete medication records, we excluded participants without at least 1 year of eligibility for prescription drug coverage through the Ontario Drug Benefit. Fourteen prespecified major noncardiac surgeries were included: abdominal aortic aneurysm repair, carotid endarterectomy, peripheral vascular surgery, femur and hip surgery, knee replacement, lung resection, gastrectomy or esophagectomy, bowel and rectal surgery, liver resection, pancreaticoduodenectomy, abdominal hysterectomy, radical prostatectomy, nephrectomy, and cystectomy.9, 10 In the event of multiple surgeries performed during the study period, only the first qualifying surgery was used as the index procedure. A complete list of procedure codes is shown in Table S1.

Patients with AF were identified using the International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Canada diagnostic code I48 recorded in any field of the CIHI‐DAD, the CIHI‐SDS, and the NACRS databases within 5 years before the index surgical procedure date. Code I48 has been previously validated and found to have a positive predictive value of 93.0% (95% confidence interval [CI], 91.6–94.2).11 Other medical comorbidities were identified by examining secondary diagnosis codes from the index admission and all diagnoses recorded on any hospital admissions within 5 years before the index surgery to enhance sensitivity for detection of comorbidities. A complete list of diagnostic codes for comorbid conditions is shown in Table S2. We excluded from the study cohort patients undergoing dialysis, those with rheumatic heart disease, and those with valve replacements, since DOAC studies have not included these patient groups. We excluded patients who were prescribed both DOACs and warfarin within 30 days and those with an insufficient medication supply to cover until the index hospitalization. We also excluded patients who were prescribed anticoagulants without a prior diagnosis of AF, as these drugs have alternate indications.

Anticoagulation Categories

Patients with AF were further classified based on their anticoagulation regimen as DOAC (dabigatran, rivaroxaban, or apixaban) users, warfarin users, or nonanticoagulated. We defined DOAC or warfarin users as those who were dispensed a new or refilled prescription for an anticoagulant with a sufficient number of days supplied such that the available days supply would cover the date of the index surgical hospitalization. Patients defined as nonanticoagulated were required to have not filled any prescription for an oral anticoagulant during the 100 days before the index surgical hospitalization. This allowed capture of all new and refilled prescriptions because the maximum duration of a single prescription in Ontario is 3 months supply. Patients who underwent noncardiac surgery with no history of AF and were not taking anticoagulants comprised a reference group.

Surgical Procedures

We categorized surgeries as urgent or elective based on the admission category variable contained within the CIHI‐DAD. Surgeries were classified in this way because discontinuation of anticoagulants can be preplanned in patients with elective surgery, but not necessarily in those undergoing urgent/emergent procedures. The date of surgery was determined from the CIHI‐DAD and CIHI‐SDS databases.

Outcomes

The primary outcome was time to death attributable to any cause within 30 days of the date of the index surgical procedure. The secondary outcome was hemorrhagic events occurring within 30 days after surgery (either during index hospital admission or subsequent emergency visits or hospitalizations), and included intracerebral, intraocular, intraarticular, gastrointestinal, or other postsurgical bleeding, as previously described.12, 13 Diagnostic codes for bleeding have been previously published, found to have 94% sensitivity and 83% specificity in validation studies, and are shown in Table S3.12 We also examined a related process measure: use of blood products, including transfusion of blood, platelets, or plasma during the index surgical admission or within 30 days after surgery, using the blood transfusion indicator in the CIHI‐DAD. Finally, we examined thromboembolic events either occurring as an in‐hospital complication or during a subsequent readmission within 30 days after surgery. Thromboembolic events were defined as myocardial infarction, stroke, transient ischemic attack, coronary thromboembolism, arterial embolism and thrombosis, intestinal ischemia, renal ischemia or infarction, vascular myelopathy, and atrial or ventricular thrombosis (see diagnostic codes in Table S4). The secondary outcomes that included hemorrhagic or thromboembolic events were treated as binary outcomes, since the timing of in‐hospital bleeding events is not available in the CIHI‐DAD.

Statistical Analysis

Categorical data were summarized as percentages, and differences between comparison groups were tested with the χ2 test. Continuous variables were summarized as medians and interquartile ranges, and were compared using the Wilcoxon rank‐sum test. Time to event was determined from the date of the surgical procedure. We used inverse probability of treatment (IPT) weighting using the propensity score to estimate the effect of: (1) anticoagulation versus no anticoagulation in patients with AF; and (2) DOAC versus warfarin among patients with AF who were anticoagulated. In the overall surgical cohort, we compared outcomes of: (1) AF versus no AF; and (2) nonanticoagulated AF versus no AF.14 Therefore, 4 sets of propensity score weighted analyses were developed. Weighted Cox proportional hazards models were used to estimate the effects of exposure on the rate of mortality in the sample weighted by the IPT weights. We used weighted logistic regression analysis for 30‐day bleeding and thromboembolic events because the in‐hospital indicator variables for these outcomes were not supplemented with information on the time of the event. For both models, a robust, sandwich‐type variance estimator was used to account for the within‐subject correlation in outcomes induced by weighting.15

All propensity‐weighted analyses accounted for the following covariates: age, sex, CHADS2 comorbidities (ie, congestive heart failure, hypertension, diabetes mellitus, stroke, or transient ischemic attack), vascular disease, prior myocardial infarction, Charlson comorbidity score (0 or 1 versus ≥2), hospital admission via the emergency department, Johns Hopkins Aggregated Diagnosis Groups categories, and additional administratively coded elements of the HAS‐BLED score including renal or liver disease, prior bleeding within the past year, and use of bleeding‐relevant drugs (ie, antiplatelet agents, nonsteroidal anti‐inflammatory drugs, and proton pump inhibitors). We also included the type of major surgery (eg, orthopedic, vascular, thoracic, abdominal, or urologic/pelvic) and hospital teaching status as covariates in the model. Weighted standardized differences were used to compare baseline covariates between exposure groups in the weighted samples.16 All statistical analyses were conducted separately in patients undergoing elective surgery and in those undergoing urgent surgery.

Statistical significance was defined by a 2‐sided P<0.05. Analyses were performed with SAS software (version 9.4; SAS Institute Inc).

Results

Study Cohorts

A total of 284 268 patients underwent a surgical procedure during the study period. After exclusions, the elective surgery cohort consisted of 87 257 patients while the urgent surgery cohort consisted of 35 930 patients (Figure 1). In the elective surgery cohort, 4769 (5.5%) patients had AF; of these, 1453 (30.5%) were prescribed warfarin, 1165 (24.4%) were prescribed a DOAC, and 2151 (45.1%) were nonanticoagulated. In the urgent surgery cohort, 4612 (12.8%) patients had AF, of whom 1619 (35.1%) were prescribed warfarin, 729 (15.8%) were prescribed a DOAC, and 2264 (49.1%) were nonanticoagulated. Within the DOAC group, dabigatran (n=998, 53%), rivaroxaban (n=687, 36%), and apixaban (n=209, 11%) were prescribed.

Figure 1

Patient flow diagram. AF indicates atrial fibrillation; DOAC, direct oral anticoagulant.

Baseline characteristics of patients within the elective cohort are presented in Table 1. Patients undergoing elective surgery with AF treated with warfarin were slightly older and exhibited higher rates of heart failure, stroke, and diabetes mellitus than patients with AF treated with a DOAC or those who were not anticoagulated. Patients treated with warfarin and nonanticoagulated AF had higher frequency of Charlson index ≥2. However, prior bleeding within 1 year was similar between the 3 AF groups. Baseline characteristics for the urgent surgical cohort are shown in Table 2. Among the urgent surgical cohort, comorbidities were of similar magnitude between anticoagulation categories. Specifically, patients taking warfarin had higher rates of heart failure; patients taking DOAC and warfarin had higher rates of prior stroke, and those who were not anticoagulated exhibited a higher prevalence of coronary disease or myocardial infarction. Overall, however, there was no significant difference in the prevalence of Charlson index ≥2 between the 3 AF groups. Prior bleeding within 1 year was highest among urgent surgical patients receiving a DOAC and those who were not anticoagulated. Patients prescribed warfarin had higher CHADS2 score, followed by those taking DOAC and those who were not anticoagulated. The most common surgical interventions were orthopedic and abdominal surgeries in both elective and urgent AF cohorts.

Table 1

Elective Surgery: Baseline Cohort Characteristics

Group	No Anticoagulationa No AF	AF With Warfarin	AF With DOAC	AF With No Anticoagulation	P Valueb
No.	82 488	1453	1165	2151
Age, median (IQR)	73 (69–78)	78 (73–82)	76 (72–81)	76 (71–81)	<0.001
Men, No. (%)	34 225 (41.5)	774 (53.3)	610 (52.4)	1151 (53.5)	0.814
Teaching hospital, No. (%)	24 885 (30.2)	509 (35.0)	360 (30.9)	724 (33.7)	0.079
Medical history, No. (%)
Coronary disease	8170 (9.9)	475 (32.7)	331 (28.4)	751 (34.9)	<0.001
Previous MI	3083 (3.7)	170 (11.7)	123 (10.6)	352 (16.4)	<0.001
Congestive heart failure	1498 (1.8)	329 (22.6)	235 (20.2)	315 (14.6)	<0.001
Cerebrovascular disease	2866 (3.5)	178 (12.3)	119 (10.2)	163 (7.6)	<0.001
PVD	3917 (4.7)	151 (10.4)	99 (8.5)	211 (9.8)	0.253
Diabetes mellitus	16 362 (19.8)	450 (31.0)	305 (26.2)	540 (25.1)	<0.001
Hypertension	31 396 (38.1)	942 (64.8)	762 (65.4)	1297 (60.3)	0.003
Hyperlipidemia	6858 (8.3)	256 (17.6)	195 (16.7)	419 (19.5)	0.114
COPD	5652 (6.9)	222 (15.3)	141 (12.1)	294 (13.7)	0.063
Chronic kidney disease	1469 (1.8)	120 (8.3)	41 (3.5)	110 (5.1)	<0.001
Malignancy	3825 (4.6)	84 (5.8)	49 (4.2)	150 (7.0)	0.005
Dementia	844 (1.0)	25 (1.7)	22 (1.9)	72 (3.3)	0.003
Charlson index ≥2	30 168 (36.6)	870 (59.9)	587 (50.4)	1225 (57.0)	<0.001
CHADS 0–2	76 848 (93.2)	1011 (69.6)	876 (75.2)	1725 (80.2)	<0.001
CHADS 3 or 4	5416 (6.6)	407 (28.0)	266 (22.8)	394 (18.3)
CHADS 5 or 6	224 (0.3)	35 (2.4)	23 (2.0)	32 (1.5)
Bleeding in past year	6002 (7.3)	214 (14.7)	162 (13.9)	314 (14.6)	0.816
Medications, No. (%)c
Antiplatelet	4791 (5.8)	41 (2.8)	14 (1.2)	223 (10.4)	<0.001
ACEI or ARB	34 242 (41.5)	836 (57.5)	633 (54.3)	998 (46.4)	<0.001
β‐Blocker	15 514 (18.8)	844 (58.1)	684 (58.7)	976 (45.4)	<0.001
Calcium channel blocker	20 000 (24.2)	531 (36.5)	422 (36.2)	676 (31.4)	0.002
Diuretic	15 888 (19.3)	599 (41.2)	401 (34.4)	537 (25.0)	<0.001
Statin	35 527 (43.1)	872 (60.0)	664 (57.0)	1075 (50.0)	<0.001
NSAID	11 749 (14.2)	66 (4.5)	75 (6.4)	210 (9.8)	<0.001
PPI	21 011 (25.5)	499 (34.3)	404 (34.7)	705 (32.8)	0.452
Surgery, No. (%)
Abdominal	12 033 (14.6)	270 (18.6)	183 (15.7)	446 (20.7)	<0.001
Orthopedic	51 015 (61.8)	878 (60.4)	724 (62.1)	1218 (56.6)
Thoracic	3481 (4.2)	58 (4.0)	63 (5.4)	138 (6.4)
Urologic or pelvic	11 733 (14.2)	124 (8.5)	101 (8.7)	198 (9.2)
Vascular	4226 (5.1)	123 (8.5)	94 (8.1)	151 (7.0)
Mini‐invasive	9174 (11.1)	185 (12.7)	149 (12.8)	292 (13.6)	0.707
Admission characteristics, No. (%)
Ambulance	556 (0.7)	28 (1.9)	14 (1.2)	49 (2.3)	0.096
Elective admission	80 604 (97.7)d	1420 (97.7)	1138 (97.7)	2102 (97.7)	0.916

ACEI indicates angiotensin converting enzyme inhibitor; ARB, angiotensin receptor blocker; COPD, chronic obstructive pulmonary disease; IQR, interquartile range; MI, myocardial infarction; NSAID, nonsteroidal anti‐inflammatory drug; PPI, proton pump inhibitor; PVD, peripheral vascular disease.

Unless otherwise indicated, all P<0.001 for comparison of atrial fibrillation (AF) with no anticoagulation vs AF.

P values for comparison of patients with AF treated with a direct oral anticoagulant (DOAC), warfarin, or no anticoagulation.

Medications prescribed within 1 year before admission.

For comparison of AF with no anticoagulation vs AF, P=not significant.

Table 2

Urgent Surgery: Baseline Cohort Characteristics

Group	No Anticoagulationa No AF	AF With Warfarin	AF With DOAC	AF With No Anticoagulation	P Valueb
No.	31 318	1619	729	2264
Age, median (IQR)	82 (75–88)	85 (80–89)	83 (78–88)	85 (79–90)	<0.001
Men, No. (%)	9692 (30.9)	594 (36.7)	264 (36.2)	875 (38.6)	0.328
Teaching hospital	8420 (26.9)c	485 (30.0)	184 (25.2)	670 (29.6)	0.047
Medical history, No. (%)
Coronary disease	4873 (15.6)	576 (35.6)	238 (32.6)	841 (37.1)	0.084
Previous MI	2627 (8.4)	288 (17.8)	115 (15.8)	482 (21.3)	<0.001
Congestive heart failure	2549 (8.1)	706 (43.6)	278 (38.1)	727 (32.1)	<0.001
Cerebrovascular disease	2386 (7.6)	330 (20.4)	150 (20.6)	350 (15.5)	<0.001
PVD	1971 (6.3)	189 (11.7)	71 (9.7)	240 (10.6)	0.331
Diabetes mellitus	6522 (20.8)	447 (27.6)	207 (28.4)	639 (28.2)	0.890
Hypertension	14 377 (45.9)	1116 (68.9)	503 (69.0)	1499 (66.2)	0.138
Hyperlipidemia	2770 (8.8)	262 (16.2)	120 (16.5)	364 (16.1)	0.971
COPD	3792 (12.1)	379 (23.4)	166 (22.8)	481 (21.2)	0.260
Chronic kidney disease	1487 (4.7)	260 (16.1)	48 (6.6)	290 (12.8)	<0.001
Malignancy	1814 (5.8)c	65 (4.0)	32 (4.4)	110 (4.9)	0.452
Dementia	5944 (19.0)	310 (19.1)	136 (18.7)	570 (25.2)	<0.001
Charlson index ≥2	12 899 (41.2)	1067 (65.9)	450 (61.7)	1501 (66.3)	0.070
CHADS 0–2	25 941 (82.8)	784 (48.4)	385 (52.8)	1316 (58.1)	<0.001
CHADS 3 or 4	4959 (15.8)	727 (44.9)	295 (40.5)	846 (37.4)
CHADS 5 or 6	418 (1.3)	108 (6.7)	49 (6.7)	102 (4.5)
Bleeding in past year	1790 (5.7)	169 (10.4)	96 (13.2)	305 (13.5)	0.014
Medications, No. (%)d
Antiplatelet	2911 (9.3)	52 (3.2)	24 (3.3)	302 (13.3)	<0.001
ACEI or ARB	11 237 (35.9)	749 (46.3)	335 (46.0)	787 (34.8)	<0.001
β‐Blocker	5974 (19.1)	869 (53.7)	406 (55.7)	880 (38.9)	<0.001
Calcium channel blocker	7520 (24.0)	590 (36.4)	230 (31.6)	582 (25.7)	<0.001
Diuretic	6388 (20.4)	795 (49.1)	331 (45.4)	748 (33.0)	<0.001
Statin	10 159 (32.4)	794 (49.0)	364 (49.9)	800 (35.3)	<0.001
NSAID	2057 (6.6)	53 (3.3)	21 (2.9)	93 (4.1)	0.198
PPI	8522 (27.2)	619 (38.2)	296 (40.6)	786 (34.7)	0.006
Surgery, No. (%)
Abdominal	5774 (18.4)c	259 (16.0)	138 (18.9)	394 (17.4)	0.136
Orthopedic	23 935 (76.4)c	1262 (77.9)	548 (75.2)	1744 (77.0)
Thoracic	153 (0.5)c	12 (0.7)	7 (1.0)	12 (0.5)
Urologic or pelvic	208 (0.7)c	6 (0.4)	6 (0.8)	SC
Vascular	1248 (4.0)c	80 (4.9)	30 (4.1)	110 (4.9)
Mini‐invasive	1012 (3.2)e	51 (3.2)	30 (4.1)	59 (2.6)	0.112
Admission characteristics, No. (%)
Ambulance	23 570 (75.3)	1309 (80.9)	563 (77.2)	1820 (80.4)	0.108
Emergency admission	27 013 (86.3)e	1402 (86.6)	630 (86.4)	1920 (84.8)	0.241

ACE indicates angiotensin‐converting enzyme inhibitor; ARB, angiotensin receptor blocker; COPD, chronic obstructive pulmonary disease; IQR, interquartile range; MI, myocardial infarction; NSAID, nonsteroidal anti‐inflammatory drug; PPI, proton pump inhibitor; PVD, peripheral vascular disease.

Unless otherwise indicated, all P<0.001 for comparison of atrial fibrillation (AF) with no anticoagulation vs AF.

P values for comparison of patients with AF treated with a direct oral anticoagulant (DOAC), warfarin, or no anticoagulation.

For comparison of AF with no anticoagulation vs AF, P<0.01.

Medications prescribed within 1 year before admission.

For comparison of AF with no anticoagulation vs AF, P=not signficiant.

The majority of the patients with elective surgery underwent surgery on the day of admission, with a median (interquartile range) time from admission to surgery of 0 (0–0) days for all 4 exposure categories: no AF, AF‐warfarin, AF‐DOAC, and AF‐no anticoagulation. The majority of patients in the urgent surgery group (>85%) were admitted via the emergency department and presented to the hospital by ambulance. In the urgent surgical cohort, the median time from admission to surgery for patients with AF was 2 (1–2) days for AF‐warfarin, 2 (1–3) days for AF‐DOAC, and 1 (1–2) day for AF‐no anticoagulation, while those with no AF underwent surgery 1 (0–2) day after admission.

Perioperative Event Rates

Unadjusted outcomes for perioperative mortality are presented in Table 3 and Figure 2. Within the elective surgery cohort, 30‐day mortality for patients with AF was 1.7% for those dispensed warfarin, 1.3% for those dispensed a DOAC, and 2.2% for those who were not anticoagulated. Mortality at 30 days was 0.6% in patients without AF who were not anticoagulated.

Table 3

Thirty‐Day Perioperative Outcomes

Elective Surgery	No Anticoagulation No AF	AF With Warfarin	AF With DOAC	AF With No Anticoagulation	P Value
No.	82 488	1453	1165	2151
Length of stay, median (IQR)	3 (3–5)	4 (3–7)	4 (3–6)	4 (3–7)	<0.001
Mortality, No. (%)	509 (0.6)	25 (1.7)	15 (1.3)	47 (2.2)	<0.001
Thromboembolic event, No. (%)	971 (1.2)	29 (2.0)	13 (1.1)	44 (2.0)	<0.001
Ischemic stroke, No. (%)	147 (0.2)	6 (0.4)	SCs	9 (0.4)	0.008
Bleeding, No. (%)	3203 (3.9)	96 (6.6)	81 (7.0)	118 (5.5)	<0.001
Blood transfusion, No. (%)	8590 (10.4)	259 (17.8)	140 (12.0)	372 (17.3)	<0.001
Platelet or plasma products, No. (%)	611 (0.7)	38 (2.6)	17 (1.5)	40 (1.9)	<0.001

AF indicates atrial fibrillation; DOAC, direct oral anticoagulant; IQR, interquartile range; SCs, small cells (unable to report because of privacy regulations).

Figure 2

Unadjusted 30‐day mortality (A) and bleeding (B) in patients undergoing elective major surgery. AF indicates atrial fibrillation; DOAC, direct oral anticoagulant.

Within the urgent surgical cohort, 30‐day mortality was 14.1% in patients with AF treated with warfarin, 11.7% in patients with AF dispensed a DOAC, and 14.3% in patients with AF who were not anticoagulated. The 30‐day mortality rate was 7.8% in patients without AF who were nonanticoagulated. Thromboembolic events are also shown in Table 3. There were significant differences between treatment groups, with the lowest rates of thromboembolic events in patients prescribed a DOAC and patients without AF. The highest rates were observed in patients with nonanticoagulated AF followed by patients treated with warfarin.

The unadjusted 30‐day bleeding outcomes are also shown in Table 3. While crude bleeding risks were highest in patients who were prescribed warfarin or a DOAC, the absolute increase in this risk was modest in comparison to patients with AF who were nonanticoagulated. For warfarin, DOAC, and nonanticoagulated AF, the respective bleeding rates were 6.6%, 7.0%, and 5.5% in the elective surgery and 10.1%, 9.1%, and 8.3% in the urgent surgery cohorts. In addition, the absolute increase in bleeding risks among patients prescribed warfarin or a DOAC were only 2% to 3% higher than the low‐risk group of nonanticoagulated patients without AF. Irrespective of anticoagulation status, the median length of stay among patients with AF was increased by ≈1 day in the elective surgery group and 2 to 3 days in the urgent surgical cohort, compared with patients without AF (Table 3).

As shown in Figure 2, the incidence of bleeding events exceeded that of mortality in those undergoing elective surgical procedures. However, the opposite effects were observed after urgent surgery, such that death was more common than bleeding (shown in Figure 3.

Figure 3

Unadjusted 30‐day mortality (A) and bleeding (B) in patients undergoing urgent major surgery. AF indicates atrial fibrillation; DOAC, direct oral anticoagulant.

Perioperative Outcomes of Patients With AF in Relation to Oral Anticoagulant Use

Comparing patients with AF who were anticoagulated versus patients with AF who were not anticoagulated, standardized differences for all covariates in the sample weighted by the IPT weights were <0.10 for both the elective and urgent surgery cohorts (Tables S5 and S6). After adjustment with IPT weighting using the propensity score, the rate of death within 30 days was similar in both patients with anticoagulated and nonanticoagulated AF undergoing elective or urgent surgical major surgery (Figure 4A). There was no significant difference in the odds of 30‐day thromboembolic events among patients undergoing elective (odds ratio, 0.88; 95% confidence interval, 0.52–1.50 [P=0.645]) or urgent (odds ratio, 0.79; 95% confidence interval, 0.59–1.07 [P=0.127]) surgery when patients with anticoagulated AF were compared to patients with nonanticoagulated AF (reference group) after IPT weighting–propensity score adjustment. The 30‐day perioperative risk of bleeding was also not significantly different among patients with AF who were or were not anticoagulated.

Figure 4

Adjusted mortality and bleeding outcomes when comparing patients with anticoagulated atrial fibrillation (AF) vs nonanticoagulated AF (A) and direct oral anticoagulant (DOAC) AF vs warfarin AF (B). HR indicates hazard ratio; OR, odds ratio.

When comparing patients with AF who were anticoagulated with either a DOAC or warfarin, standardized differences for all covariates were <0.10 in the weighted sample for the elective and urgent surgery cohorts (Tables S7 and S8). After adjustment with IPT weighting using the propensity score, the rate of death within 30 days was not significantly different in patients who were treated with a DOAC or warfarin preoperatively (Figure 4B), with no significant difference after either elective or urgent major surgical procedures. There was a trend to lower odds of 30‐day thromboembolic events after elective surgery (odds ratio, 0.57; 95% CI, 0.29–1.11 [P=0.097]) favoring DOACs over warfarin, but there was no difference between drug classes after urgent surgical procedures (odds ratio, 1.01; 95% CI, 0.63–1.64 [P=0.958]). Risk of bleeding perioperatively within 30 days was also not significantly different in the comparison between a DOAC and warfarin.

Outcomes Comparing AF Versus No AF

When patients with AF were compared with patients without AF, the comparator groups were well‐matched after weighting using the IPT weights (Tables S9 and S10). The rate of death within 30 days was increased only among patients who underwent urgent major surgery (HR, 1.28; 95% CI, 1.12–1.45) (Figure 5A). There was no significant increase in bleeding risk among patients with AF compared to those without AF (Figure 5A).

Figure 5

Adjusted mortality and bleeding outcomes when comparing patients with anticoagulated atrial fibrillation (AF) vs patients without AF (A) and nonanticoagulated AF vs patients without AF (B). HR indicates hazard ratio; OR, odds ratio.

Patients with AF who were not anticoagulated exhibited standardized differences <0.10 when compared with patients without AF after weighting with the IPT weights (Tables S11 and S12). Again, the rate of mortality was not significantly different when patients with elective surgery were compared. However, the rate of death was higher for patients with AF undergoing urgent surgery who were not anticoagulated when compared with patients without AF (1.34; 95% CI, 1.12–1.59) (Figure 5B). There was no difference in bleeding risk between patients with AF who were not anticoagulated compared with patients without AF (Figure 5B).

Discussion

AF is a prevalent condition that is treated using oral anticoagulation as a cornerstone for the prevention of stroke. In patients with nonvalvular AF, both the European Society of Cardiology and the Canadian Society of Cardiology now recommend DOACs over warfarin as the anticoagulant of choice.17, 18 Surgical procedures are commonly performed in patients who have AF, and the outcomes of these procedures can potentially be impacted by AF and the concomitant use of anticoagulants. In this study of patients undergoing major noncardiac surgery, our findings are 2‐fold. First, among patients with AF, regardless of the urgency of surgery, the use of anticoagulation, and specifically DOAC compared with warfarin, was not associated with an increased risk of death or bleeding after propensity score–weighted adjustment. Second, preexisting AF was associated with an increased rate of mortality within 30 days in patients with AF who underwent urgent but not elective surgery. Our findings suggest that AF represents an independent predictor of perioperative mortality in urgent major noncardiac surgery, which appeared not to be mediated by use of oral anticoagulation or hemorrhagic events.

Prior studies examining perioperative outcomes of patients treated with DOAC mainly included elective or minor surgical procedures with lower associated risks. In a post hoc analysis of the RE‐LY trial, involving 4591 patients, over 90% of interventions were performed electively and accurate classification of surgery into major/minor categories was only available in 28% of cases. Dabigatran was withheld for at least 24 hours and up to 5 days, depending on the perceived risk of bleeding and patients’ renal function. Based on this protocol, the 30‐day risk of major bleeding ranged from 6.1% to 7.8%, and was similar among patients treated with dabigatran 110 mg, dabigatran 150 mg, or warfarin.7 The prospective Dresden registry, which enrolled office‐ and hospital‐based patients on a nonconsecutive basis, reported on 595 DOAC‐treated patients (in whom 80% were prescribed for AF indications) who underwent 863 procedures, with ≈10% being major surgery. In this subgroup of DOAC‐treated patients who underwent elective major surgery, the all‐cause mortality and major bleeding rates at 30 days were 0.7% and 8.0%, respectively.19 Our study served to confirm and extend these findings on a population‐based level for patients undergoing elective major noncardiac surgery.

Unlike the lower‐risk surgical spectrum, there are limited data on the outcomes of patients with anticoagulated AF undergoing urgent, unplanned, major noncardiac surgery at a time when DOACs are commonly prescribed. The RE‐LY substudy reported the outcomes of 353 patients who underwent urgent surgery, of whom ≈60% were classified as major (orthopedic, abdominal, or vascular) surgery. The remainder included pacemakers, diagnostic procedures, and other same day procedures. In this cohort, the 30‐day all‐cause mortality and major bleeding rates were 3.3% and 18.9%, respectively. No differences in these outcomes were observed among patients treated with dabigatran (110 or 150 mg) or warfarin.20 In our urgent surgical AF cohort, consisting of 4769 patients, the 30‐day all‐cause mortality and major bleeding rates were 13.4% and 8.8%. When compared with the RE‐LY subset, the 30‐day mortality rate of our urgent surgical cohort was 4‐fold higher whereas bleeding rates were lower. These differences may be attributed to the population‐based approach, including consecutive patients, limiting participant bias that could impact clinical trials. Furthermore, the majority of patients (≈85%) arrived at hospital via ambulance, reflecting the urgency and unplanned nature of their clinical presentation. Finally, our urgent surgical cohort was on average 10 years older and had greater comorbidity burden than the RE‐LY cohort. All of the above factors may have contributed to the greater risk of death than competing bleeding events.

Presently, expert/consensus guidelines recommend that DOAC agents be withheld for 24 to 48 hours before performance of invasive procedures.21, 22 Notably, these recommendations are based on the largely predictable (but not readily measurable) pharmacokinetic profile of DOAC agents. To date, however, there are no published randomized trial data on the optimal timing of DOAC cessation before invasive procedures or surgeries. Although stopping DOACs for patients at 24 to 48 hours before elective surgery is logistically feasible, it is likely much more difficult to implement such recommendations when patients require urgent, unplanned, major noncardiac surgery. In our study, we observed, on average, a 24‐hour delay from hospital admission to urgent surgery between patients with AF who were treated with and without oral anticoagulation, which may have been driven by physicians’ decision to wait for the effect of anticoagulants to be attenuated between proceeding with surgery. In spite of this time delay, the 30‐day mortality and bleeding rates were similar between patients with anticoagulated and nonanticoagulated AF undergoing urgent surgery. Accordingly, our study provided pertinent data on perioperative outcomes of this high‐risk cohort during a time when reversal agents for DOAC were not clinically available.23, 24

Furthermore, our study suggested that patients with a preoperative diagnosis of AF undergoing major noncardiac surgery harbored an intrinsic risk of death, which extended beyond their existing comorbidities, irrespective of anticoagulant use. This finding was observed among patients undergoing urgent but not elective surgery. After extensive adjustment for baseline covariates, we found that patients with AF exhibited an ≈30% increased risk of death at 30 days after urgent, major noncardiac surgery when compared with patients without AF. Furthermore, our study suggested that the higher risk of death in the AF cohort did not appear to be mediated by postoperative bleeding, given the comparable rates between the 2 groups. Increasingly, AF is being recognized as a major risk factor for adverse perioperative outcomes after major noncardiac surgery. In an earlier study, van Diepen et al10 reported that patients with AF had a 69% increased risk of death when compared with patients with stable coronary artery disease. However, they did not elucidate whether use of oral anticoagulation and associated bleeding events were mechanisms underlying the increased mortality risk. McAlister et al6 reported a higher risk of cardiovascular death, heart failure, and myocardial injury among 961 patients with AF postoperatively in the VISION (Vascular Events in Noncardiac Surgery Patients Cohort Evaluation) registry. Our study adds to the growing knowledge base on the adverse prognostic impact of patients with AF who undergo noncardiac surgery. Further work, however, is needed to better delineate the mechanisms accounting for worse perioperative outcomes in the AF population.

Study Limitations

A number of limitations of this study should be noted. Because we relied on dispensing data from the provincial pharmacare database to determine anticoagulation status, we were not able to ascertain compliance and adherence of paients who filled their prescriptions, as well as the exact stop date before surgery. We also could not account for out‐of‐pocket payment of DOACs, although we anticipate that this only involved a small proportion of patients because these medications are covered under the Ontario Drug Benefit formulary. Furthermore, the consistency in bleeding rates in our study with those from clinical trial and registry data suggest that the above were not significant factors in the study. Our study did not capture the use of reversal agents or account for the use of bridging therapy since they were not available during the time of study. The data sources also did not capture medications administered during the inpatient hospitalization, specifically use of oral or parenteral anticoagulation during the early postoperative period. However, any potential differences in postoperative anticoagulation did not translate into significant differences in bleeding between treatment (or nontreatment) groups. In addition, propensity‐weighted analyses account for known confounders, and it is possible that some as‐yet unmeasured factor may have influenced the effects observed. Urgency was determined using the CIHI‐DAD, which indicates whether the hospital admission was elective or urgent and has been validated.25 However, it could not be further delineated whether urgent procedures were surgical emergencies. Finally, our study identified patients who underwent major surgery as the point of inception. Therefore, it is possible that certain high‐risk individuals who were deemed unsuitable for surgery were not represented in this study. However, our study was designed to address the outcomes of patients who underwent noncardiac surgery, and did not intend to examine the outcomes of all patients who might be eligible for major surgery.

Conclusions

Among patients who underwent major noncardiac surgery, patients with preoperative AF who were treated with oral anticoagulation had similar rates of death and bleeding when compared with nonanticoagulated patients. In addition, we found that patients who were treated with DOACs or warfarin had similar death and bleeding outcomes after major noncardiac surgery. Finally, patients with preoperative AF exhibited greater mortality risk after major noncardiac surgery performed on an urgent basis, suggesting the need for further investigations into the mechanisms underlying the adverse outcomes associated with this common arrhythmia.

Sources of Funding

The Institute for Clinical Evaluative Sciences (ICES) is supported in part by a grant from the Ontario Ministry of Health and Long‐Term Care. The opinions, results, and conclusions are those of the authors and no endorsement by the Ministry of Health and Long‐Term Care or by the ICES is intended or should be inferred. Parts of this material are based on data and information compiled and provided by CIHI. However, the analyses, conclusions, opinions, and statements expressed herein are those of the author and not necessarily those of CIHI. This research was supported by a Foundation Grant from the Canadian Institutes of Health Research (grant No. FDN 148446). Dr Abdel‐Qadir is supported by a fellowship award from the Canadian Institutes of Health Research. Dr Austin is a career investigator of the Heart and Stroke Foundation of Ontario. Dr Wijeysundera is supported by a New Investigator Award from the Canadian Institutes of Health Research and a Merit Award from the Department of Anesthesia at the University of Toronto. Dr Lee is supported by a mid‐career investigator award from the Heart and Stroke Foundation and is the Ted Rogers Chair in Heart Function Outcomes, a joint Hospital‐University Chair of the University Health Network and the University of Toronto.

Disclosures

Dr Ha has received research funding from Bayer and has received speakers’ honoraria from Bayer and Boehringer Ingelheim. The other authors have no financial conflicts of interest to disclose.

Table S1. Procedure Codes

Table S2. Comorbidities

Table S3. Bleeding and Transfusion Codes

Table S4. Thromboembolism Codes

Table S5. Anticoagulated AF vs Nonanticoagulated AF: Standardized Differences Before and After IPTW Matching in the Elective Surgery Cohort

Table S6. Anticoagulated AF vs Nonanticoagulated AF: Standardized Differences Before and After IPTW Matching in the Urgent Surgery Cohort

Table S7. DOAC vs Warfarin: Standardized Differences Before and After IPTW Matching in the Elective Surgery Cohort

Table S8. DOAC vs Warfarin: Standardized Differences Before and After IPTW Matching in the Urgent Surgery Cohort

Table S9. AF vs No AF: Standardized Differences Before and After IPTW Matching in the Elective Surgery Cohort

Table S10. AF vs No AF: Standardized Differences Before and After IPTW Matching in the Urgent Surgery Cohort

Table S11. Nonanticoagulated AF vs No AF: Standardized Differences Before and After IPTW Matching in the Elective Surgery Cohort

Table S12. Nonanticoagulated AF vs No AF: Standardized Differences Before and After IPTW Matching in the Urgent Surgery Cohort

Click here for additional data file.