Share this article on:

Risk Assessment of Hemorrhagic Complications Associated with Nonsteroidal Antiinflammatory Medications in Ambulatory Pain Clinic Patients Undergoing Epidural Steroid Injection

Horlocker, Terese T. MD*,; Bajwa, Zahid H. MD‡,; Ashraf, Zubaira MD‡,; Khan, Sajid MD‡,; Wilson, Jack L. MD*,; Sami, Naveed MD‡,; Peeters-Asdourian, Christine MD‡,; Powers, Christopher A. MD*,; Schroeder, Darrell R. MS†,; Decker, Paul A. MS†,; Warfield, Carol A. MD

doi: 10.1097/00000539-200212000-00041

We prospectively studied 1035 individuals undergoing 1214 epidural steroid injections to determine the risk of hemorrhagic complications. A history of bruising or bleeding was present in 176 (15%) patients. A platelet count was assessed in 77 patients before the epidural steroid injection; none was less than 100 × 109/L. Nonsteroidal antiinflammatory drugs (NSAIDs) were reported by 383 (32%) patients, including 34 patients on multiple medications. Aspirin was the most common NSAID and was noted by 158 patients, including 104 patients on 325 mg or less per day. There were no spinal hematomas (major hemorrhagic complications). Blood was noted during needle or catheter placement in 63 (5.2%) patients (minor hemorrhagic complications). NSAIDs did not increase the frequency of minor hemorrhagic complications. However, increased age, needle gauge, needle approach, needle insertion at multiple interspaces, number of needle passes, volume of injectant, and accidental dural puncture were all significant risk factors for minor hemorrhagic complications. There were 42 patients with new neurologic symptoms or worsening of preexisting complaints that persisted more than 24 h after injection; median duration of the symptoms was 3 days (range, 1–20 days). Our results confirm those of previous studies performed in obstetric and surgical populations that document the safety of neuraxial techniques in patients receiving NSAIDs. We conclude that epidural steroid injection is safe in patients receiving aspirin-like antiplatelet medications. Minor worsening of neurologic function may occur after epidural steroid injection and must be differentiated from etiologies requiring intervention.

Departments of *Anesthesiology and †Health Sciences Research, Mayo Clinic, Rochester, Minnesota, and ‡Department of Anesthesiology, Beth Israel Deaconess Medical Center, Boston, Massachusetts

July 30, 2002.

Address correspondence and reprint requests to Terese T. Horlocker, MD, Department of Anesthesiology, Mayo Clinic, Rochester, MN 55905. Address e-mail to

Epidural steroid injection is an accepted treatment for acute and chronic back pain (1,2). However, patients undergoing epidural steroid injection frequently receive nonsteroidal antiinflammatory drugs (NSAIDs) for the treatment of their preexisting pain or underlying inflammatory condition. NSAIDs, such as aspirin, ibuprofen, naproxen, and ketorolac are often considered a contraindication to epidural steroid injection because of the associated prolongation of the bleeding time and theoretical risk of spinal hematoma formation (3). Although the actual incidence of spinal hematoma is unknown, the incidence cited in the literature is estimated to be <1 in 150,000 epidural and <1 in 220,000 spinal anesthetics (4). In a review of the literature between 1906 and 1994, Vandermeulen et al. (3) reported only 61 cases of spinal hematoma associated with spinal or epidural block, including three patients who had received medications affecting platelet function (aspirin, indomethacin, and ticlopidine) immediately before or after the neuraxial technique (5–7). Two subsequent cases have been published (8,9). In two of the five cases of spinal hematoma attributed to neuraxial block in the presence of platelet active drugs, the patients were undergoing epidural steroid injection (7,8).

Although previous investigations involving surgical or obstetrical patients have demonstrated the safety of neuraxial block in patients receiving NSAIDs (10–12), the results have not routinely been applied to the outpatient pain clinic population. As a result, antiinflammatory medications are discontinued, causing increased patient discomfort, or costly platelet function tests are performed before the epidural steroid injection. This study prospectively evaluated the risk of spinal hematoma in patients undergoing epidural steroid injection. In addition, because the presence of blood during needle or catheter placement has been implicated as a risk factor for the development of spinal hematoma (3), other patient and anesthetic variables associated with traumatic (bloody) needle placement were identified.

Back to Top | Article Outline


After approval by our IRBs, 1035 individuals undergoing a total of 1214 epidural steroid injections in ambulatory pain centers were prospectively studied. Demographic data including age, sex, weight, height, and indication for epidural steroid injection were recorded. Patients were questioned regarding a history of excessive bleeding or easy bruising and the use of antiplatelet medications within 1 wk before the epidural injection. The total daily dose and time of the last dose of antiplatelet medications were recorded. The presence of antiplatelet effect was defined by the duration of platelet inhibition as measured by response to adenosine diphosphate, epinephrine, and/or collagen, i.e., 1 wk for aspirin and 1 day for ibuprofen (13). However, no actual platelet function studies were performed. Regional technique (single dose or continuous epidural), level of needle placement (cervical, thoracic, lumbar, or caudal), needle gauge, needle approach (midline or paramedian), and elicitation of a paresthesia during needle/catheter placement were noted. The presence of blood or blood-tinged fluid in the needle or catheter (minor spinal bleeding) was recorded. Bleeding from obvious trauma to soft tissue was not included. The number of needle passes by Pain Center personnel was assessed. A needle pass was defined as any readjustment of needle position that resulted in an additional needle tract. Patients were instructed to notify the Pain Center of any regional technique-related complications, including bruising at the injection site, persistent paresthesia (pain, numbness, or weakness present upon resolution of the block), postdural puncture headache, exacerbation of preexisting back or radicular pain, or new neurologic symptoms. If a subsequent visit was not anticipated within 48 h of the epidural steroid injection, an investigator attempted to contact the patient by phone. Patients with complications related to the epidural steroid injection were contacted daily until resolution. The duration of any complications was recorded.

Predictors of minor spinal bleeding and transient neurologic deficits were assessed using separate logistic regression analyses. In each case, variables found to be univariately predictive were included in a multivariate analysis to identify a set of independent predictors. For the multivariate analysis, all univariately significant variables were entered in the model in the first step, and a backward elimination procedure was used to eliminate nonsignificant variables. A P value ≤ 0.05 was considered statistically significant in all cases.

Back to Top | Article Outline


There were 1214 epidural steroid injections performed in 1035 individuals. The mean age, height, and weight were 58 ± 17 yr (range, 19–95 yr), 166 ± 10 cm, and 79 ± 18 kg, respectively. There were 739 (61%) epidural steroid injections performed on women and 475 (39%) on men. The diagnosis or indication for the epidural steroid injection was acute radiculopathy in 592 (49%), spinal stenosis in 384 (32%), chronic back pain in 118 (10%), foraminal stenosis in 29 (2%), failed back syndrome in 12 (1%), and postherpetic neuralgia in 8 (0.7%) patients. A history of bruising or bleeding was present in 176 (15%) patients. A platelet count was assessed in 77 patients before the epidural steroid injection; none was less than 100 × 109/L. (All platelet counts had been obtained by non-Pain Clinic personnel.) NSAIDs were reported by 383 (32%) patients, including 34 patients on multiple medications. Aspirin and ibuprofen were the most often administered drugs (Table 1). The use of NSAIDs was not associated with a history of bruising or bleeding.

Table 1

Table 1

More than 80% of epidural steroid injections were administered at the lumbar level. Needle placement was most often accomplished with a needle of 18-gauge or less. Needle approach was midline in 1124 (93%) and paramedian in 56 (5%) patients. In nine patients, both approaches were required to achieve successful placement. Needle insertion occurred at multiple interspaces in 65 (5%) patients. Localization of the epidural space required a single needle pass in 751 (62%) patients. However, three or more needle passes were required for 151 (12%) patients. Needle placement was attempted by the attending anesthesiologist in 151 (12%) patients and a resident or fellow in 1040 (86%) patients. In the remaining 23 (2%) patients, more than one person was required. In 343 (28%) cases, fluoroscopic guidance was used during needle placement. Confirmation of correct needle placement by contrast injection was performed in 294 (24%) of the treatments. An epidural catheter was placed in 14 patients. A paresthesia was elicited in 40 (3%) patients, whereas dural puncture occurred in 10 (0.8%) patients. Mean volume of the injectant was 8 ± 3 mL. The epidural injectant contained a local anesthetic in 738 (61%) patients, including lidocaine in 626 (52%) patients and bupivacaine in 113 (9%) patients. Triamcinolone (median dose, 80 mg) was injected in 1184 (98%) patients.

Blood was noted during needle or catheter placement in 63 (5.2%) patients, including 12 patients in who frank blood was noted. NSAID therapy and platelet count did not affect the incidence of traumatic (bloody) needle placement. There was no difference in the frequency of minor hemorrhagic complications among NSAIDs (Table 1). However, patient and regional anesthetic variables associated with increased frequency of minor hemorrhagic complications identified by logistic regression included increasing age (odds ratio, 1.3 for each 10-yr increase;P < 0.001), needle gauge larger than 18-gauge (odds ratio, 1.8;P = 0.027), needle approach (odds ratio, 2.5 for paramedian; odds ratio, 10.5 for both approaches;P = 0.005), needle placement at multiple interspaces (odds ratio, 4.3;P < 0.001), multiple needle passes (odds ratio, 2.3;P < 0.001), injectant volume <8 mL (odds ratio, 1.2;P = 0.023), accidental dural puncture (odds ratio, 8.2;P = 0.003), and multiple personnel required for successful placement (odds ratio, 4.1;P = 0.014) (Tables 2 and 3). Multivariate analysis identified increased age (odds ratio, 1.3 per 10-yr increase;P < 0.001), multiple needle passes (odds ratio, 2.4;P < 0.001), and injection volume <8 mL (odds ratio, 1.2;P = 0.016) as independent risk factors for minor hemorrhagic complications (Table 4).

Table 2

Table 2

Table 3

Table 3

Table 4

Table 4

Median time to follow-up was 3 days; 79% of patients were contacted within 7 days of their injection. Nearly 40% of patients self-administered NSAIDs in the time interval between epidural steroid injection and follow-up. Bruising at the site of needle insertion was noted in 97 (8%) patients. Postinjection bruising at the needle insertion site was associated with a history of bruising/bleeding (odds ratio, 1.7;P = 0.039). NSAID therapy did not affect the likelihood of postinjection bruising. There were also 42 patients with new neurologic symptoms (pain, numbness, and/or weakness) or worsening of preexisting complaints that persisted more than 24 h after injection. Twenty-seven patients reported a new (or exacerbated) sensory deficit, 12 patients reported weakness, and nine patients complained of new or worsened pain. The symptoms were bilateral in 18 patients. Changes in sensation and motor function did not interfere with ambulation; no additional intervention/evaluation was required. Median duration was 3 days (range, 1–20 days). The frequency of transient neurologic deficits was significantly (P = 0.043) increased for women (32 of 739 or 4.3%) compared to men (10 of 475 or 2.5%) and significantly (P = 0.039) less for those who received local anesthetic (19 of 738 or 2.6%) versus no local anesthetic (23 of 476 or 4.8%). The frequency of transient neurologic deficits was significantly associated with physician training in 1 of 151 patients (0.7%) for staff physicians, 32 of 922 patients (3.5%) for fellows, and 9 of 118 patients (7.6%) for residents. No independent variables were identified during multivariate analysis. Elicitation of a paresthesia or traumatic (bloody) needle placement during epidural steroid injection was not determined to increase the risk of postinjection neurologic symptoms.

Back to Top | Article Outline


Several investigations and case reports suggest that patients undergoing blocks for chronic pain management are at risk for neurologic complications (8,14). In a review of the American Society of Anesthesiologists Closed Claims Project, Cheney et al. (14) noted that spinal cord injuries were the leading cause of claims for nerve injury that occurred in the 1990s. Importantly, a major factor associated with spinal cord injury was block for chronic pain management, of which there were 14 claims, including 8 lumbar epidural steroid injections.

This study evaluated risk factors for hemorrhagic complications in outpatients undergoing epidural steroid injection. Nearly one third of the patients studied reported NSAID therapy before treatment. Whereas several patient and anesthetic variables for minor hemorrhagic complications were identified, NSAIDs were not a significant risk factor. NSAID therapy has been considered a relative contraindication to central neural blockade because of the associated prolongation of the bleeding time and potential for neuraxial bleeding (3). NSAIDs exert their antiplatelet effect through inhibiting the acetylation of cyclooxygenase and preventing the synthesis of thromboxane A2. Thromboxane A2 is not only a potent vasoconstrictor, but also facilitates secondary platelet aggregation and release reactions. Depending on the dose administered, aspirin (and other NSAIDs) may produce opposing effects on the hemostatic mechanism. For example, platelet cyclooxygenase is inhibited by small-dose aspirin (60–325 mg/d). However, larger doses (1.5–2 g/d) will also inhibit the production of prostacyclin (a potent vasodilator and platelet aggregation inhibitor) by vascular endothelial cells. Therefore, small-dose aspirin produces a greater antiplatelet effect than larger doses, hence the recommendation for a baby aspirin a day to prevent cerebro- or cardiovascular events. With aspirin, the effect is irreversible and present for the life of the platelet. Other NSAIDs (naproxen, piroxicam, and ibuprofen) produce a short-term defect that normalizes within one to five days (13). We did not detect a difference in the frequency of bleeding during needle placement in patients who reported aspirin doses of ≤325 mg/d compared with larger daily doses.

Several new classes of antiplatelet drugs have been recently introduced. The antiplatelet effect of the thienopyridine derivatives ticlopidine and clopidogrel results from inhibition of adenosine diphosphate-induced platelet aggregation. Because of their extended plasma half lives, clopidogrel must be discontinued five to seven days and ticlopidine 10–14 days before normal platelet function is restored. Platelet glycoprotein IIb/IIIa receptor antagonists, such as abciximab, eptifibatide, and tirofiban, not only inhibit platelet aggregation, but also interfere with platelet-fibrinogen binding and subsequent platelet-platelet interactions. Time to normal platelet aggregation after discontinuation of therapy ranges from 8 hours (eptifibatide and tirofiban) to 48 hours (abciximab). The pharmacologic differences of the thienopyridine derivatives glycoprotein IIb/IIIa inhibitors and NSAIDs make it impossible to extrapolate among the groups of drugs regarding the practice of neuraxial techniques. However, the increase in perioperative bleeding in patients undergoing cardiac and vascular surgery after receiving ticlopidine, clopidogrel, and glycoprotein IIb/IIIa antagonists (15) warrants concern regarding the risk of spinal hematoma. There have been two spinal hematomas attributed to neuraxial techniques and thienopyridine derivatives, including one patient undergoing a series of epidural steroid injections (6,8). Because no patient in our study received ticlopidine, clopidogrel, or a glycoprotein IIb/IIIa antagonist before the epidural steroid injection, our results and conclusions do not pertain to these drugs.

The risk of spinal hematoma associated with neuraxial block in a patient receiving antiplatelet medications remains controversial. Whereas Vandermeulen et al. (3) implicated antiplatelet medications in three of the 61 cases of spinal hematoma occurring after spinal or epidural techniques, several large studies have demonstrated the relative safety of neuraxial block in combination with aspirin or other NSAIDs (10–12). Indeed, the American Society of Regional Anesthesia Consensus Conference on Neuraxial Anesthesia and Anticoagulation concluded that “antiplatelet drugs, by themselves, appear to represent no added significant risk for the development of spinal hematoma in patients having epidural or spinal anesthesia (16).”

Because difficult and/or bloody needle placements are risk factors for clinically significant neuraxial bleeding (3), identification of patient, surgical, and anesthetic variables associated with traumatic needle insertion may decrease the incidence of spinal hematoma. In a prospective study of 1000 patients undergoing neuraxial block for orthopedic surgery, 39% reported preoperative NSAIDs (12). There were no spinal hematomas. Preoperative NSAID therapy was not a risk factor for bloody needle or catheter placement. In addition, there was no difference in the frequency of minor hemorrhagic complications between antiplatelet medications. However, multiple patient and anesthetic variables including female sex, increased age, a history of excessive bruising or bleeding, continuous catheter technique, large needle gauge, multiple needle passes, and difficult needle placement were significant risk factors.

Despite different patient populations, our results are strikingly similar. Factors associated with difficulty in localization of the epidural space such as needle approach, needle insertion at multiple interspaces, number of needle passes, accidental dural puncture, and the need for assistance with needle placement all affected the frequency of minor hemorrhagic complications. Increased patient age has also consistently been a suggested risk factor for both major and minor hemorrhagic events (12,17). We were unable to assess epidural catheter placement as a potential risk factor in the current study because only 14 patients underwent a continuous technique.

Although there were no spinal hematomas among the patients evaluated, it does not imply that the risk is zero for all patients. The absence of major hemorrhagic events in the 383 patients who received NSAIDs before the epidural steroid injection places the maximum risk of spinal hematoma (with 95% confidence interval) at 0.96%. In comparison, the lack of spinal hematoma among the total study population of 1214 corresponds to a maximum frequency of 0.3%. Although the rarity of spinal hematoma makes it impossible to make definitive conclusions on the safety of epidural steroid injection in combination with NSAID therapy, the lack of correlation between these drugs and minor hemorrhagic events, combined with the paucity of spinal hematomas reported among patients receiving these medications, once again suggests that NSAIDs do not significantly increase the risk of spinal hematoma in patients who undergo neuraxial techniques while receiving these medications.

We were intrigued by the transient worsening of neurologic function in patients undergoing epidural steroid injection. Although initially attributed to local anesthetic in the injectant, the occurrence of new sensory or motor findings that persisted for a number of days after the injection, as well as the presence of these symptoms in patients who received only saline as a diluent, excluded local anesthetic effect as a likely etiology. We speculate that either a pressure (ischemic) effect of the injection or an inflammatory response from one or more components of the injectant contributed to these neurologic deficits. Whereas psychosocial issues may also influence patient description of pain and function, an expanding spinal hematoma should be considered in the patient reporting significant worsening of (radicular) pain, progression of a sensory or motor deficit, or bowel/bladder dysfunction. Patients should be instructed to seek medical attention should symptoms of neurologic compromise occur. Additional studies are required to characterize and evaluate these phenomena.

In summary, our results confirm those of previous studies performed in obstetric and surgical populations that document the safety of neuraxial techniques in patients receiving NSAIDs. We conclude that epidural steroid injection is safe in patients receiving aspirin-like antiplatelet medications. However, pain clinic personnel should be aware that minor worsening of neurologic function may occur after epidural steroid injection and must be differentiated from etiologies requiring intervention.

Back to Top | Article Outline


1. Tonkovich-Quaranta LA, Winkler SR. Use of epidural corticosteroids in low back pain. Ann Pharmacother 2000; 34: 1165–72.
2. Cannon DT, Aprill CN. Lumbosacral epidural steroid injections. Arch Phys Med Rehabil 2000; 81: S87–98.
3. Vandermeulen EP, Van Aken H, Vermylen J. Anticoagulants and spinal-epidural anesthesia. Anesth Analg 1994; 79: 1165–77.
4. Tryba M. Rückmarksnahe regionalanästhesie und niedermolekulare heparine: pro. Anästhesiol Intensivmed Notfallmed Schmerzther 1993; 28: 179–81.
5. Greensite FS, Katz J. Spinal subdural hematoma associated with attempted epidural anesthesia and subsequent continuous spinal anesthesia. Anesth Analg 1980; 59: 72–3.
6. Mayumi T, Dohi S. Spinal subarachnoid hematoma after lumbar puncture in a patient receiving antiplatelet therapy. Anesth Analg 1983; 62: 777–9.
7. Williams KN, Jackowski A, Evans PJ. Epidural haematoma requiring surgical decompression following repeated cervical epidural steroid injections for chronic pain. Pain 1990; 42: 197–9.
8. Benzon HT, Wong HY, Siddiqui T, Ondra S. Caution in performing epidural injections in patients on several antiplatelet drugs. Anesthesiology 1999; 91: 1558–9.
9. Gerancher JC, Waterer R, Middleton J. Transient paraparesis after postdural puncture spinal hematoma in a patient receiving ketorolac. Anesthesiology 1997; 86: 490–4.
10. CLASP (Collaborative Low-Dose Aspirin Study in Pregnancy) Collaborative Group. CLASP: a randomized trial of low-dose aspirin for the prevention and treatment of pre-eclampsia among 9364 pregnant women. Lancet 1994; 343: 619–29.
11. Horlocker TT, Wedel DJ, Offord KP, et al. Does preoperative antiplatelet therapy increase the risk of hemorrhagic complications associated with regional anesthesia? Anesth Analg 1990; 70: 631–4.
12. Horlocker TT, Wedel DJ, Schroeder DR, et al. Preoperative antiplatelet therapy does not increase the risk of spinal hematoma associated with regional anesthesia. Anesth Analg 1995; 80: 303–9.
13. Cronberg S, Wallmark E, Soderberg I. Effect on platelet aggregation of oral administration of 10 non-steroidal analgesics to humans. Scand J Haematol 1984; 33: 155–9.
14. Cheney FW, Domino KB, Caplan RA, Posner KL. Nerve injury associated with anesthesia: a closed claims analysis. Anesthesiology 1999; 90: 1062–9.
15. Kovesi T, Royston D. Is there a bleeding problem with platelet-active drugs [editorial]? Br J Anaesth 2002; 88: 159–63.
16. Urmey WF, Rowlingson J. Do antiplatelet agents contribute to the development of perioperative spinal hematoma? Reg Anesth Pain Med 1998; 23: 146–51.
17. Horlocker TT, Wedel DJ. Neuraxial block and low-molecular-weight heparin: balancing perioperative analgesia and thromboprophylaxis. Reg Anesth Pain Med 1998; 23: 164–77.
Back to Top | Article Outline

Attention Authors!

Submit Your Papers Online

You can now have your paper processed and reviewed faster by sending it to us through our new, web-based Rapid Review System. Submitting your manuscript online will mean that the time and expense of sending papers through the mail can be eliminated. Moreover, because our reviewers will also be working online, the entire review process will be significantly faster. You can submit manuscripts electronically via There are links to this site from the Anesthesia & Analgesia website (, and the IARS website ( To find out more about Rapid Review, go to and click on “About Rapid Review.”

© 2002 International Anesthesia Research Society