As the US population continues to age, adults aged >65 years account for an increasingly large proportion of those electing for spine surgery. These patients often have complex medical conditions, and some studies have suggested that >70% of them have multiple comorbidities [1,2]. Many comorbid conditions such as diabetes mellitus (DM) have well-researched guidelines for managing medications perioperatively; however, no similar “best practices” exist for patients on antithrombotic therapies, particularly for spine surgery [3]. This is a common clinical problem encountered in the United States. In 2012, the American College of Surgeons (ACS) estimated that 250,000 patients on anticoagulants face the prospect of an invasive surgical procedure annually [4].

The management of these patients is challenging because the risks of thrombo-ischemic events increase in the presence of antithrombotics; however, the risks of bleeding complications increase in their absence. Furthermore, the degree of these risks is enormously nuanced as it varies not only among patients and specific agent but heavily on operative variables such as complexity of the procedure, duration of postoperative immobility, and magnitude of correction [5,6]. In addition, spine surgeries are associated with large blood losses, with amounts ranging from 650 mL to 2 L, further complicating the picture [7].

Guidance available in the literature is sparse in the field of spine surgery. Some studies have assessed elective surgical procedures in general, and they have focused on delineating optimal intervals for interruption and resumption of specific antithrombotics and on the value of bridging therapies. The most notable is the ACS Clinical Practice Guidelines from 2012; however, even this document is limited by its reliance on a “disproportionately large number of methodologically weak observational studies” [4]. Given the relative paucity of well-designed trials, we aimed to examine the clinical implications of using preoperative antithrombotic agents among patients undergoing elective spine surgery in a large, matched analysis.

Between January 1, 2010, and June 30, 2018, 754,274 individuals undergoing elective cervical or lumbar spine interventions were identified. Of these individuals, 19,940 underwent surgery for lumbar/cervical fracture repair and were thus excluded from the analysis. Moreover, 73,468 individuals were also excluded for being <18 years old at the time of surgery. Ultimately, the study included a total of 660,866 individuals who underwent elective spine surgery (Fig. 1). The descriptive characteristics of this unmatched population are summarized in Supplement 3. Models comparing the risks of bleeding complications and thromboembolic events in the unmatched population are summarized in Supplements 4 and 5.

Following a 1:1 matching procedure based on patient age and comorbidities, 56,476 patients remained, including 28,238 and 28,238 patients in the antithrombotic and non-antithrombotic groups, respectively. No statistical difference in patient age was found between the groups. Each cohort predominantly consisted of individuals aged >65 years (n=14,886 [52.7%]). The most prevalent comorbidities in the matched cohort were hypertension (83.56%), DM (49.89%), and hyperlipidemia (44.93%) (Table 1).

In this retrospective cohort study, data from 56,476 patients were matched and analyzed. Significantly increased rates of 90-day thrombo-ischemic events and bleeding complications were recorded among patients prescribed preoperative antithrombotic therapy. More specifically, the risks were uniformly higher for DVT, PE, MI, ischemic stroke, hematoma, and the need for transfusion in the antithrombotic group.

Although seemingly paradoxical, the increase in thrombo-ischemic events in the antithrombotic group likely has to do with the interruption of antithrombotics before surgery and the inherent invasiveness of spine procedures. Unfortunately, we could not determine the details of antithrombotic interruption before the surgery from our database analysis. However, literature has shown a higher incidence of thrombo-ischemic events in patients with spine disorders, such as an estimated incidence for DVT in non-spinal cord injury trauma patients at 6.0% (range, 0%–19%) [8]. Although poorly understood, unique aspects of spine surgery may likely explain this increased risk because studies following non-spine surgical procedures have reported a <1% incidence of perioperative ischemic events for those taking antithrombotic medications preoperatively [9,10].

The literature is clear on bleeding complications. Recently, Kim et al. [11] demonstrated that preoperative anticoagulation was a risk factor for postoperative hematoma in a retrospective analysis of 310 patients undergoing endoscopic spine surgery. Yi et al. [12] found similar results in their retrospective review of 3,720 spine surgeries and revealed preoperative anticoagulation therapy as a risk factor for postoperative epidural hematoma. Likewise, in patients on chronic anticoagulation therapy with warfarin, Young et al. [13] reported an increased need for transfusion following lumbar spine surgery. These findings underpin the difficulty of perioperative antithrombotic management. Although some antithrombotics require bridging therapy or earlier interruption to obtain good outcomes, the ideal timelines are unclear. Without consistent, evidence-based methodology, surgeons resort to personal experience and other empiric strategies to minimize adverse events, with visibly dubious outcomes.

Currently, the half-life of antithrombotics and operative variables such extent of surgery and involvement of highly vascular organs drive expert recommendations for antithrombotic management [14]. These factors are usually combined with the HAS-BLED score, which is a short scale based on patient factors such as comorbid renal impairments or hypertension, with moderative predictive value in bleeding risk quantification. To stratify the risk for thrombo-embolic events, the CHADS2 score was also used, although it has not been validated for perioperative risk stratification [15]. The results of these efforts highly vary and gain inconsistent expert recommendations [16,17]. For example, a recent study considered employing an optimal interval for the interruption of newer direct oral anticoagulants and concluded that the optimal interval may range from 1 to 5 days preoperatively, even in surgeries with similar bleeding risks [18].

Bridging is similarly controversial; however, several randomized clinical trials have investigated its value in antithrombotic agent management following surgery to minimize bleeding and thrombo-ischemic risk in a non-spine setting. For example, the PAUSE and BRIDGE trials focused on patients who had atrial fibrillation and were taking direct oral anticoagulants and warfarin, respectively. They revealed the limited role of heparin bridging following high-risk bleeding procedures [9,10]. However, these findings are not yet transplanted into clinical practice.

The timing of postoperative resumption of antithrombotic agents remains unclear. In addition, to the limited observational and small cohort studies, no strong data can be relied upon. The current ACS guidelines reflect this ambiguity by leaving it to surgeons to assess and balance patient risk factors—an inexact recommendation to say the least (Table 2) [19]. Baschera et al. [20] echoed this lack of clarity in their nationwide survey of orthopedic and neurosurgery attendings—they found wide variations in antithrombotic agent restart times, specifically with orthopedic surgeons recommending restarting platelet inhibitors earlier than their neurosurgeon colleagues.

This study has several shortcomings. First, the retrospective design and non-uniform procedures introduced heterogeneity in each group, which may not have been completely accounted for by the variables included in the analysis. All clinical data available in the Mariner database concerning diagnoses and adverse events were coded using ICD-9, ICD-10, and CPT codes. Misclassification errors are certainly present in the analysis; however, they typically appear at random and are less likely to bias the results, particularly given the large sample size of the matched analysis in this study. Second, the lack of granularity within the database made deriving firm conclusions regarding the duration of preoperative interruption timing and postoperative resumption interval of antithrombotic agents difficult. This information would have been useful to better stratify more nuanced treatment regimens. Similarly, future analysis should account for the medication type and type of surgery because significant heterogeneity may affect the risks of bleeding versus thrombosis. These are significant study limitations; however, our findings likely reflect the ambiguity in antithrombotic agent management for patients undergoing spine surgery. We hope our study inspires interest in further research to develop clear guidelines.

Given the rapidly aging US population, older patients with comorbidities make up an increasing proportion of those seeking elective spine surgeries. Surgeons are faced with the challenge of balancing operative variables with the potential risks of halting or continuing antithrombotic therapy. This study offers evidence of the increased rates of postoperative thrombo-ischemic events and bleeding complications within 90 days after surgery among patients prescribed preoperative antithrombotic agents. A better understanding of the medications used to decrease clotting, timing of withdrawal in the perioperative window, and definition of which patients are more likely to benefit from antithrombotic medications may be improved in the future.