clinical articleJ neurosurg Pediatr 17:107–115, 2016

Ketorolac is a widely used NSAID that is effica-cious in controlling postoperative pain either alone or as part of a multimodal postoperative analgesic

regimen.12 Ketorolac is administered parenterally, making it an excellent choice in the perioperative setting because it does not require oral intake and can be quickly admin-istered. Though ketorolac is effective for postoperative pain management, reports of hemorrhage associated with

ketorolac use have fueled a debate as to whether it should be administered to patients undergoing procedures associ-ated with a high risk of bleeding. Because of the potential sequelae of hemorrhage in their patient population, many neurosurgeons have been reluctant to use ketorolac as part of the postoperative regimen.27

As a class, NSAIDs offer analgesic efficacy similar to that of morphine and meperidine.15 Nonselective, nonaspi-

abbreviations COX = cyclooxygenase; EVD = external ventricular drain.submitted August 4, 2014. accePted April 28, 2015.include when citing Published online October 9, 2015; DOI: 10.3171/2015.4.PEDS14411.

Routine perioperative ketorolac administration is not associated with hemorrhage in pediatric neurosurgery patientsmarlin dustin richardson, md,1 nicholas o. Palmeri,2 sarah a. williams,3 michelle r. torok, Phd,4,5 brent r. o’neill, md,1,5 michael h. handler, md,1,5 and todd c. hankinson, md, mba1,4,5

1Department of Neurosurgery and 4Adult and Child Center for Health Outcomes Research and Delivery Science, University of Colorado Anschutz Medical Campus, Aurora, Colorado; 2College of Physicians and Surgeons, Columbia University, New York, New York; 3Tulane University, New Orleans, Louisiana; and 5Pediatric Neurosurgery, Children’s Hospital Colorado, Aurora, Colorado

obJective NSAIDs are effective perioperative analgesics. Many surgeons are reluctant to use NSAIDs perioperative-ly because of a theoretical increase in the risk for bleeding events. The authors assessed the effect of routine periopera-tive ketorolac use on intracranial hemorrhage in children undergoing a wide range of neurosurgical procedures.methods A retrospective single-institution analysis of 1451 neurosurgical cases was performed. Data included de-mographics, type of surgery, and perioperative ketorolac use. Outcomes included bleeding events requiring return to the operating room, bleeding seen on postoperative imaging, and the development of renal failure or gastrointestinal tract injury.Variables associated with both the exposure and outcomes (p < 0.20) were evaluated as potential confounders for bleed-ing on postoperative imaging, and multivariable logistic regression was performed. Bivariable analysis was performed for bleeding events. Odds ratios and 95% CIs were estimated.results Of the 1451 patients, 955 received ketorolac. Multivariate regression analysis demonstrated no significant association between clinically significant bleeding events (OR 0.69; 95% CI 0.15–3.1) or radiographic hemorrhage (OR 0.81; 95% CI 0.43–1.51) and the perioperative administration of ketorolac. Treatment with a medication that creates a known bleeding risk (OR 3.11; 95% CI 1.01–9.57), surgical procedure (OR 2.35; 95% CI 1.11–4.94), and craniotomy/cra-niectomy (OR 2.43; 95% CI 1.19–4.94) were associated with a significantly elevated risk for radiographically identified hemorrhage.conclusions Short-term ketorolac therapy does not appear to be associated with a statistically significant increase in the risk of bleeding documented on postoperative imaging in pediatric neurosurgical patients and may be considered as part of a perioperative analgesic regimen. Although no association was found between ketorolac and clinically signifi-cant bleeding events, a larger study needs to be conducted to control for confounding factors, because of the rarity of these events.http://thejns.org/doi/abs/10.3171/2015.4.PEDS14411Key words ketorolac; Toradol; postoperative bleeding; postoperative hemorrhage; NSAID; neurosurgery

©AANS, 2016 J neurosurg Pediatr Volume 17 • January 2016 107

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

m. d. richardson et al.

J neurosurg Pediatr Volume 17 • January 2016108

rin NSAIDs reversibly inhibit cyclooxygenase (COX) en-zymes 1 and 2, preventing the conversion of arachidonic acid to proinflammatory prostaglandins, thus producing antipyretic and analgesic effects. COX enzymes are ubiq-uitous; therefore, NSAID administration also predisposes a number of tissues to undesired side effects. NSAID ad-ministration can result in unwanted effects in the kidneys, the gastric and intestinal mucosa, and platelets, among other tissues. NSAIDs alter platelet function by inhibit-ing the synthesis of procoagulant thomboxanes, which are stored in platelet granules. Inhibition of COX by NSAIDs has been shown to alter platelet function and prolong bleeding times in healthy volunteers,5,14,15,24 though bleed-ing times generally remain within normal limits.14 The prolongation of bleeding times and platelet function has led to numerous clinical studies that have investigated the safety of NSAID administration in postoperative patients. Some clinical safety studies have demonstrated that ad-ministration of NSAIDs leads to an increase in postop-erative bleeding events,3,9,10,19,20 whereas other studies have demonstrated the NSAIDs can be safely administered postoperatively.11, 13,14,17,18,23,25,26,28

Since 2000, we have routinely used ketorolac as a com-ponent of the perioperative analgesic regimen during the first 72 hours following pediatric neurosurgical proce-dures. We sought to assess whether this strategy was asso-ciated with an increased risk for hemorrhagic or systemic complications.

methodsstudy design and sample

Following approval by the Colorado Multiple Institu-tional Review Board, we retrospectively reviewed 2657 consecutive neurosurgery procedures that had adequate electronic data. Procedures were performed in 1593 pa-tients at Children’s Hospital Colorado between December 2006 and April 2012. Data were collected from electronic health records by 2 trained research assistants and a neu-rosurgery resident (M.R.). Ten percent of records were double entered independently to verify consistency. Dis-crepancies were resolved by the senior author.

Surgical procedures were stratified based on an estima-tion of hemorrhagic complications by the 2 senior authors (M.H.H. and T.C.H.) (Table 1). The 5 categories, in order of a priori perceived risk, were 1) craniotomy/craniectomy with dural opening (except for Chiari malformation Type I ); 2) intradural placement of a catheter or endoscope; 3) craniotomy/craniectomy without dural opening plus pro-cedures for Chiari malformation Type I); 4) spinal proce-dures; and 5) minor procedures (e.g., external ventricular drain [EVD], intracranial pressure monitor, or vagus nerve stimulator placement). Chiari malformation decompres-sion with dural opening was included in the third category because our institutional practice is to perform minimal intradural maneuvers during these procedures. Placement of EVDs was considered a minor procedure and included in Category 5 because of the shorter duration of the pro-cedure and decreased tissue manipulation (shunt passage and intraperitoneal exposure) compared with a shunt pro-cedure. For each patient, a single surgical procedure type

(Type 1–5), termed the “index procedure,” was identified (Table 1). For patients who underwent multiple operations, the procedure in the highest risk category (Table 1) was assessed, and all other procedures were excluded from analysis. In the case of multiple procedures within the highest risk category for a given patient, the first procedure chronologically was analyzed. The single exception to this was planned 2-stage craniotomy for intradural electrode monitoring followed by resection of epileptogenic tissue. In these cases, the second-stage procedure was assessed, based on the a priori belief that a resective procedure is associated with greater risk for hemorrhage than grid or electrode placement.

From the 1593 index procedures assessed, those as-sociated with an elevated risk of harboring hemorrhage preoperatively were excluded. These included 1) patients with procedures that occurred within 7 days of presenta-tion following trauma (n = 59) and 2) patients with radio-graphically documented intracranial or intraspinal hemor-rhage within 7 days prior to surgery (n = 114). Thirty-one cases had both exclusion criteria. The remaining 1451 pro-cedures composed the study sample.

covariablesIn addition to reviewing the surgical procedures, we

assembled a list of medications that are associated with increased risk for hemorrhage and termed them “pharma-cological confounders.” This list included systemic-dose anticoagulants (not low-dose prophylactic heparins), an-tiplatelet agents, valproic acid, and NSAIDS other than ketorolac. Due to ubiquitous inpatient use and undocu-mented outpatient use, ibuprofen was not considered a pharmacological confounder. While there is debate as to whether valproic acid is associated with an increased like-lihood of hemorrhage,2,6,7,29 our clinical practice has been to treat it as such. If at least 1 dose of the medication was administered during the 24 hours prior or 72 hours fol-lowing the index procedure, the patient was considered to have had a pharmacological confounder. Age, sex, surgi-cal procedure, surgeon, and craniotomy for epilepsy were also examined as covariables.

exposureExposure was defined as the administration of at least 1

dose of ketorolac during the 24 hours prior to or 72 hours following the index procedure.

outcomesOutcome measures included clinically significant post-

operative bleeding, radiographically identified postopera-

table 1. surgical procedure groups

Type 1: Intradural craniotomy/craniectomy except Chiari I malformationType 2: Intradural catheter/endoscope placementType 3: Extradural craniotomy/craniectomy + Chiari I malformationType 4: Spinal procedureType 5: Minor procedure (e.g., EVD, intracranial pressure monitor)

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

Perioperative ketorolac safety

J neurosurg Pediatr Volume 17 • January 2016 109

tive hemorrhage, and newly incident renal failure and/or gastrointestinal ulcer or hemorrhage.

Clinically significant bleeding events were defined as postoperative hemorrhages that required a return to the operating room for evacuation of a hematoma. Radio-graphically identified hemorrhage was assessed on axial imaging (CT scan or MRI) performed within 7 days of the index procedure. The presence of hemorrhage was identified from the impression section of the pediatric neuroradiologist’s final report. Postoperative hemorrhage described as “thin,” “minor,” “mild,” or “small” was not considered a positive finding. Any other description was considered positive and preliminarily labeled as bleeding on imaging. To confirm the findings, all scans noted in the radiology interpretation to harbor hemorrhage (how-ever small) were reviewed by the senior author, who was blinded to whether the subject had received ketorolac.

Renal failure was defined as a doubling of the preop-erative serum creatinine value. If all postoperative values were within the normal range in the absence of a preop-erative value, then the patient was considered to not have renal failure. If no postoperative value was available, then the outcome could not be assessed. Only 648 patients received postoperative laboratory analysis of creatinine level as a measure of renal function. The development of gastrointestinal insult was identified through analysis of clinical documentation. This included evaluation for ICD-9-CM diagnosis Codes 531–534 (ulcers in the gas-tric, duodenal, gastrojejunal and unspecified regions) and 578 (gastrointestinal hemorrhage) and correlation of these with notes from appropriate specialty services, including critical care, pediatrics, and gastroenterology. For the as-sessment of both renal failure and gastrointestinal tract injury, records were reviewed from the 30 days follow-ing the surgical procedure or the entirety of the relevant admission, whichever was longer. Records were also as-sessed to evaluate whether these conditions existed in the patients prior to the index procedure and were not consid-ered a complication if found to be preexisting.

statistical analysesDescriptive statistics were examined. Bivariable anal-

yses were performed to assess the relationship between covariables and the exposure, as well as covariables and the outcomes, to evaluate potential confounding variables. Chi-square, Fisher exact, and Wilcoxon rank-sum tests were performed. Variables that were associated with both the exposure and clinical bleeding event (p < 0.20) were evaluated as potential confounders for that relationship. Multivariable logistic regression was performed and ad-justed ORs and 95% CIs were estimated. Craniotomy for epilepsy and brain resection and the surgeon were colin-ear, so we only omitted surgeon from the multivariable model. Six surgeons performed procedures during the study period (Table 2). For purposes of uniformity of case numbers in analysis, surgeons were combined into 3 surgi-cal groups (Tables 3 and 4). Craniotomy for epilepsy was evaluated as a possible effect modifier. The multivariable model for the association between ketorolac and radio-graphic bleeding included the presence of a pharmaceu-tical confounder, the surgical procedure (dichotomized),

craniotomy for epilepsy and brain resection, and the use of ketorolac. A crude odds ratio and 95% CI were estimated to evaluate the relationship between ketorolac and clini-cally significant bleeding events, because the number of bleeding events was small (n = 7).

Statistical analyses were performed with SAS 9.3 (SAS Institute).

resultsstudy sample

A total of 1451 index procedures were identified. The median age at surgery was 4.8 years (range 0–30 years), with 453 (31.2%) procedures performed in patients less than 1 year old and 462 (31.8%) in patients 10 or more years old. Ketorolac was administered in 955 cases (65.8%). Ax-ial imaging was performed within 7 days following the index procedure in 539 cases (37.15%) (Table 2).

Two patients (0.1%) developed gastrointestinal tract ul-ceration and 2 of 648 (0.3%) suffered from documented renal failure. All of these patients received ketorolac, but the incidence of these events was too small to allow for statistical analysis. The 2 cases of documented renal fail-ure were not specifically attributable to ketorolac admin-istration, as 1 patient suffered from concurrent urinary re-tention and received another potentially nephrotoxic agent (vancomycin) and the other patient did not have a preke-

table 2. demographics of 1451 patients undergoing a pediatric neurosurgical procedure

Characteristic Frequency, No. (%)

Median age at op, yrs (range) 4.8 (0.00–30.00)Female patient 612 (42.18)Surgical procedure category Type 1 422 (29.08) Type 2 388 (26.74) Type 3 346 (23.85) Type 4 208 (14.33) Type 5 87 (6)Craniotomy for epilepsy 140 (9.65)Bleeding confounder present 45 (3.1)Surgeon 0 437 (30.12) 1 151 (10.41) 2 501 (34.53) 3 109 (7.51) 4 125 (8.61) 5 4 (0.28) 6 124 (8.55)Postop imaging w/in 7 days 539 (37.15)Experienced renal failure (n = 648) 2 (0.31)Experienced ulcer 2 (0.14)Received ketorolac 955 (65.82)Experienced bleeding event 7 (0.48)Bleeding on imaging present (n = 539) 56 (10.39)

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

m. d. richardson et al.

J neurosurg Pediatr Volume 17 • January 2016110

torolac serum creatinine level measured for comparison (though serum creatinine levels after ketorolac adminis-tration indicated a decreased glomerular filtration rate).

Seven patients experienced clinically significant bleed-ing events and are detailed as follows. Patient 1, an 18-year-old girl, developed progressive upper-extremity weakness following a complex decompression of a Chiari malforma-tion. Imaging revealed a postoperative hematoma that re-quired surgical evacuation. Patient 2 was a 15-year-old boy in whom a postoperative hematoma surrounding the pulse generator developed after vagus nerve stimulator place-ment. Patient 3, a 12-year-old boy, suffered an intraven-tricular hemorrhage following ventriculoperitoneal shunt placement. The hemorrhage, surrounding the proximal tip of the catheter, resulted in proximal shunt failure and emergency revision because of declining function on neu-rological examination. Patient 4 was a 14-year-old girl who underwent an encephaloduroarteriosynangiosis and devel-oped a postoperative hematoma at the operative site that required evacuation. Of note, she was receiving a bridging anticoagulation regimen (low-molecular-weight heparin to oral anticoagulation) at the time of the event. Patient 5, a 14-year-old boy, underwent evacuation of a subdural empyema and developed a subdural hematoma and recur-rent empyema that required surgical evacuation 4 days fol-lowing the initial surgery. Patient 6 was a 4-year-old boy who had undergone recent heart transplantation and was taking warfarin (Coumadin; Bristol-Myers Squibb Co.). He developed a spontaneous left frontal hematoma, sub-

arachnoid hemorrhage, and an acute-on-chronic subdural hematoma. Two days following the initial neurosurgery, he developed a recurrent subdural hematoma that required emergency evacuation. Patient 7, a 20-year-old man who underwent subdural grid implantation and developed aphasia following the procedure, was found to have a mod-est subdural hematoma that required evacuation.

Ketorolac administrationBivariable analysis (Table 3) demonstrated no differ-

ence in ketorolac administration based on patient age or sex. Patients who received a medication that might have increased perioperative bleeding risk were less likely to receive ketorolac (46.7% vs 66.4%; p = 0.006). Patients who underwent a minor procedure (e.g., EVD or intracra-nial pressure monitor placement) and those who under-went a procedure that involved intradural catheter place-ment (e.g., ventricular shunt or endoscopy) were less likely to receive ketorolac than patients who underwent other procedures (overall p < 0.0001). Among patients who un-derwent neurosurgical procedures, ketorolac administra-tion rates were as follows: Group 1 (craniotomy/craniec-tomy with dural opening [except for Chiari malformation Type I]), 75.6%; Group 2 (intradural placement of a cath-eter or endoscope), 32.5%; Group 3 (craniotomy/craniec-tomy without dural opening plus procedures for Chiari malformation Type I), 83.2%; Group 4, 93.8%; and Group 5 (minor procedures), 31.0%.

table 3. bivariable association between pediatric neurosurgery patients and ketorolac administration*

Variables Total (n = 1451)Ketorolac (n = 955)

No Ketorolac (n = 496) p Value

Median age at op, yrs (25th, 75th percentiles) 4.8 (0.7, 12.0) 5.0 (0.7, 12.0) 4.7 (0.5, 11.0) 0.0596 (W)Sex F 612 (42.2) 410 (42.9) 202 (40.7) 0.4196 (P) M 839 (57.8) 545 (57.1) 294 (59.3) Surgical procedure category <0.0001 (P) Type 1 422 (29.1) 319 (33.4) 103 (20.8) Type 2 388 (26.7) 126 (13.2) 262 (52.8) Type 3 346 (23.8) 288 (30.2) 58 (11.7) Type 4 208 (14.3) 195 (20.4) 13 (2.6) Type 5 87 (6.0) 27 (2.8) 60 (12.1) Medical confounder No 1406 (96.9) 934 (97.8) 472 (95.2) 0.0059 (P) Yes 45 (3.1) 21 (2.2) 24 (4.8) Craniotomy for epilepsy & brain resection† No 1311 (90.4) 848 (88.8) 463 (93.3) 0.0054 (P) Yes 140 (9.6) 107 (11.2) 33 (6.7) Collapsed surgeon group 0 588 (40.5) 357 (37.4) 231 (46.6) 0.0023 (P) 1 501 (34.5) 353 (37.0) 148 (29.8) 2 362 (24.9) 245 (25.7) 117 (23.6)

P = Pearson chi-square test; W = Wilcoxon rank-sum test.* Data given as no. (%) unless otherwise indicated. † Does not include corpus callosotomy.

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

Perioperative ketorolac safety

J neurosurg Pediatr Volume 17 • January 2016 111

bleedingA total of 7 clinically significant bleeding events were

identified (0.5%) (Table 4). There was no statistically sig-nificant difference, according to bivariable analysis, be-tween patients who did or did not receive ketorolac and the development of a significant bleeding event (p = 0.70). Among those in whom ketorolac was used (n = 955), there were 4 bleeding events (0.42%), and there were 3 events among the 496 patients who did not receive ketorolac (0.60%). Among the parameters assessed, the presence of a pharmaceutical confounder and older age were associ-ated with an elevated risk of a significant bleeding event (p = 0.018 and 0.007, respectively) (Table 4).

A total of 539 patients (37.15%) had axial imaging within 7 days following the index procedure (Table 5). Among these, there were 56 studies (10.4%) that demon-strated postoperative hemorrhage. There was no statisti-cally significant difference between patients who did or did not receive ketorolac and the presence of postoperative hemorrhage on imaging (p = 0.48). Craniotomy for epilep-sy did not act as an effect modifier for the relationship be-

tween ketorolac and postoperative hemorrhage on imag-ing (p = 0.93). Among those in whom ketorolac was used (n = 313), there were 35 positive imaging studies (11.1%), whereas there were 21 positive studies among the 226 patients who did not receive ketorolac (9.3%). The pres-ence of a pharmaceutical confounder was associated with an elevated likelihood of identifying hemorrhage on the postoperative imaging study (p = 0.038). Craniotomy for epilepsy and brain resection was associated with bleeding on imaging, as was the surgeon (Table 5). Following cra-niotomy/craniectomy, regardless of whether the dura was opened, 10.5%–13.9% of patients (Categories 1 and 3) had positive findings on postoperative imaging. Among other procedure categories, positive findings were identified in 0.0%–6% of patients. For the multivariable analysis, sur-gical procedure, therefore, was dichotomized into crani-otomy/craniectomy or other.

multivariable analysesBivariable analysis indicated that age and pharmaceuti-

cal confounder might confound the relationship between

table 4. bivariable associations between clinical and demographic pediatric neurosurgery patient characteristics and experiencing a bleeding event that required return to the operating room for evacuation of a hematoma*

Variables Total (n = 1451)Bleeding Event

p Value Yes (n = 7) No (n = 1444)

Median age at op, yrs (25th, 75th percentiles) 4.8 (0.7, 12.0) 14.0 (12.0, 18.0) 4.8 (0.7, 12.0) 0.0077 (W)Sex F 612 (42.2) 3 (42.9) 609 (42.2) 1.0000 (F) M 839 (57.8) 4 (57.1) 835 (57.8) Surgical procedure category Type 1 422 (29.1) 2 (28.6) 420 (29.1) 0.5333 (F) Type 2 388 (26.7) 3 (42.9) 385 (26.7) Type 3 346 (23.8) 1 (14.3) 345 (23.9) Type 4 208 (14.3) 0 (0.0) 208 (14.4) Type 5 87 (6.0) 1 (14.3) 86 (6.0) Medical confounder No 1406 (96.9) 5 (71.4) 1401 (97.0) 0.0179 (F) Yes 45 (3.1) 2 (28.6) 43 (3.0) Craniotomy for epilepsy & brain resection† No 1311 (90.4) 7 (100.0) 1304 (90.3) 1.0000 (F) Yes 140 (9.6) 0 (0.0) 140 (9.7) Collapsed surgeon group 0 588 (40.5) 4 (57.1) 584 (40.4) 1 501 (34.5) 3 (42.9) 498 (34.5) 0.3511 (F) 2 362 (24.9) 0 (0.0) 362 (25.1) Did patient receive ketorolac? No 496 (34.2) 3 (42.9) 493 (34.1) 0.6963 (F) Yes 955 (65.8) 4 (57.1) 951 (65.9)

F = Fisher exact test.* Data given as no. (%) unless otherwise indicated. † Does not include corpus callosotomy.

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

m. d. richardson et al.

J neurosurg Pediatr Volume 17 • January 2016112

ketorolac and clinically significant postoperative hemor-rhage. Because of the small number of events (n = 7), we were not able to assess this. However, in bivariable analy-sis, ketorolac was not associated with an elevated likeli-hood of clinically significant bleeding events (OR 0.69; 95% CI 0.15–3.10). The model fit was not optimal (c = 0.544).

In assessing the risk of a radiographically identified postoperative hemorrhage, the presence of a pharmaceu-tical confounder, the category of surgical procedure, the surgeon, and undergoing craniotomy for epilepsy reached the threshold of p < 0.20 for both the bivariable exposure and outcome relationships (Tables 2 and 4). As described, the category of surgical procedure was dichotomized and the surgeon, as a variable, was omitted due to collinearity with surgical procedure. The multivariable model, there-fore, included ketorolac administration, pharmaceutical confounder, dichotomized surgical procedure, and cra-niotomy for epilepsy and brain resection (Table 6). Ke-torolac was not associated with an elevated likelihood of radiographically identified bleeding controlling for these other factors (OR 0.81; 95% CI 0.43–1.51). In this case, the model fit was acceptable (c = 0.7).

discussionNSAIDs can be very effective components of a mul-

timodal postoperative analgesic regimen and have been shown to reduce postoperative opioid requirements.5,21 In our institution, perioperative ketorolac has been in routine use for nearly 15 years. To our knowledge, this study represents the largest assessment of perioperative bleeding events associated with ketorolac use in neuro-surgery. Bivariable analysis was performed to identify factors that could confound the relationship between ke-torolac use and hemorrhage. Multivariate analysis was performed to estimate the association between ketorolac use and radiographically identified postoperative hemor-rhage, accounting for the factors identified by the bivari-able analysis. This included pharmaceutical confounders, surgical procedure, and craniotomy for epilepsy and brain resection.

While the presence of a pharmacological confounder was associated with a higher rate of hemorrhage in bivari-able analysis, the use of ketorolac in this context did not increase the risk for hemorrhage. Because of the small number of clinically significant postoperative hemorrhage events (n = 7), we were unable to perform multivariable

table 5. bivariable associations between clinical and demographic pediatric neurosurgery patient characteristics and visualization of blood on imaging within 7 days of the index procedure*

Variables Total (n = 539)Bleeding on Imaging

p ValueYes (n = 56) No (n = 483)

Median age at op, yrs (25th, 75th percentiles) 7.0 (1.8, 13.0) 7.0 (1.9, 13.3) 7.0 (1.8, 13.0) 0. 0.6947 (W)Sex F 219 (40.6) 27 (48.2) 192 (39.8) 0.2222 (P) M 320 (59.4) 29 (51.8) 291 (60.2) Surgical procedure category Type 1 288 (53.4) 40 (71.4) 248 (51.3) 0.0610 (F) Type 2 201 (37.3) 12 (21.4) 189 (39.1) Type 3 38 (7.1) 4 (7.1) 34 (7.0) Type 4 6 (1.1) 0 (0.0) 6 (1.2) Type 5 6 (1.1) 0 (0.0) 6 (1.2) Medical confounder No 520 (96.5) 51 (91.1) 469 (97.1) 0.0377 (F) Yes 19 (3.5) 5 (8.9) 14 (2.9) Craniotomy for epilepsy & brain resection† No 477 (88.5) 41 (73.2) 436 (90.3) 0.0002 (P) Yes 62 (11.5) 15 (26.8) 47 (9.7) Collapsed surgeon group 0 192 (35.6) 18 (32.1) 174 (36.0) 1 186 (34.5) 29 (51.8) 157 (32.5) 0.0082 (P) 2 161 (29.9) 9 (16.1) 152 (31.5) Did patient receive ketorolac? No 226 (41.9) 21 (37.5) 205 (42.4) 0.4779 (P) Yes 313 (58.1) 35 (62.5) 278 (57.6)

F = Fisher exact test; P = Pearson chi-square test; W = Wilcoxon rank-sum test.* Data given as no. (%) unless otherwise indicated. † Does not include corpus callosotomy.

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

Perioperative ketorolac safety

J neurosurg Pediatr Volume 17 • January 2016 113

analysis. Although there was no statistically significant bi-variable association between ketorolac use and clinically significant postoperative hemorrhage (Table 7), this find-ing should be confirmed in larger studies in which covari-ables such as pharmaceutical confounders and age could be evaluated. The c-statistic for the bivariable model (c = 0.5) suggests that other factors we were unable to consider might help explain this relationship. Because of a decid-edly small number of events (n = 2 for each), we were un-able to assess whether ketorolac was associated with an in-creased rate of renal failure or gastrointestinal tract injury. Our findings indicate that the routine use of perioperative ketorolac in pediatric neurosurgery is not associated with an increased risk for bleeding events.

Prior investigations of postoperative bleeding risk as-sociated with NSAID use in neurosurgery have shown varying results. Bauer and colleagues4 retrospectively analyzed 51 children who underwent craniotomy for re-section of a brain tumor. Postoperatively, patients received ibuprofen alternating with acetaminophen every 4 to 6 hours. No patient suffered a symptomatic hemorrhage. Moderate hemorrhage was noted in a single patient (1.9%) and small, expected postoperative blood products were appreciated on 17.6% of postoperative imaging studies. In a population of adults undergoing elective craniotomy, Magni and colleagues23 did not identify an association between perioperative ketorolac administration and post-operative intracranial hemorrhage requiring operative evacuation. Nonoperative hemorrhage was not evaluated in the study. In contrast to these studies, Palmer and col-leagues27 found that antiplatelet agents were among the principle risk factors for the development of postoperative hematoma requiring evacuation in their principally adult neurosurgery population. While this study included more than 6600 procedures, the perioperative use of aspirin was not distinguished from other NSAIDs, substantially limiting the value of this work in the assessment of ke-torolac use.

In a randomized study evaluating the efficacy of a sin-gle dose of ketorolac versus morphine in the postopera-tive pediatric critical care setting, Lieh-Lai and colleagues reported no difference in bleeding time between the 2 groups and there were no cases of gastrointestinal hemor-rhage.22 Their study population of 102 children included 24 patients who underwent neurosurgical procedures. Cassi-nelli and colleagues8 also reported no bleeding, renal, or hepatic complications in a randomized trial of ketorolac versus placebo for adults undergoing lumbar decompres-sion. A recent meta-analysis of perioperative ketorolac use that examined 27 studies and 2314 patients demonstrated no significant increase in perioperative bleeding events and found ketorolac to be superior to opioids in control-ling postoperative pain.16

The findings of our study are consistent with these re-ports, indicating no relationship between short-term ke-torolac treatment and postoperative hemorrhage. The most notable limitations to this study are the use of retrospec-tive observational data and the small number of clinically significant bleeding events. Additionally, as the data come from a single institution, the study results may not be gen-eralizable. This limitation may specifically apply to adult populations, as there are significant differences between the volume of distribution and plasma clearance of ketoro-lac between children and adults.14

Selection bias may also affect our results. Specifically, a minority (37%) of patients received postoperative axial imaging and/or evaluation of renal function. These evalu-ations were completed primarily based on routine clinical practice but also in cases where concern regarding the pa-tient’s clinical status mandated a radiographic or labora-tory evaluation. While these patients would have a greater likelihood of abnormal findings than the overall sample, there is no bias regarding the use of ketorolac in those who were evaluated. We prioritized the inclusion of procedures that we believed to be associated with the highest risk for hemorrhage. While we believed that procedures involv-ing intradural placement of a catheter or endoscope (e.g., ventricular shunt or endoscopy) would be associated with more bleeding risk than extradural craniotomies/craniec-tomies (e.g., craniosynostosis repair or Chiari malforma-tion decompression), the proportion of radiographic hem-orrhage was higher in the latter group (10.5% vs 6%). If a substantial number of patients had a procedure in both of these categories and the intradural catheter/endoscope cases were preferentially included in the cohort, then ra-

table 6. multivariate analysis between ketorolac administration and bleeding on imaging (n = 539)

Characteristic OR 95% CI c Statistic

Ketorolac Yes 0.81 0.43–1.51 No RefBleeding confounder Yes 3.11 1.01–9.57 No RefSurgical procedure* Other procedures Ref Procedures 1 & 4 2.35 1.12–4.94Craniotomy for epilepsy & brain

resection0.7

Yes 2.43 1.19–4.94 No Ref

Ref = reference.* Procedures were dichotomized to allow for convergence of the model.

table 7. unadjusted association between ketorolac use and bleeding events and bleeding on imaging

CharacteristicKetorolac

Administration OR 95% CI

Bleeding on imaging (n = 539) Yes 1.23 0.70–2.17No Ref

Bleeding events (n = 1451)Yes 0.69 0.15–3.10No Ref

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

m. d. richardson et al.

J neurosurg Pediatr Volume 17 • January 2016114

diographic hemorrhage may be underestimated. The use of ketorolac was also not evenly distributed among the dif-ferent surgical groups. This resulted from our practice of discharging many patients who undergo shunt placement/revision or endoscopy on postoperative Day 1. In prepara-tion for this, many such patients are treated with an oral analgesic regimen including ibuprofen rather than receiv-ing parenteral ketorolac.

An additional limitation of this study is that our out-come assessment relied largely on the dictated report of the attending pediatric neuroradiologist. While all posi-tive findings were confirmed in a blinded fashion and the radiology reports were generated prospectively and in an unbiased manner, there is likely to be variability in the language chosen for the neuroradiology report, which may have influenced our detection of clinically silent radio-graphic hemorrhage. Despite these limitations, we believe this study supports the notion that ketorolac is not associ-ated with bleeding events or bleeding on imaging when used as part of a short-term perioperative pain manage-ment regimen following pediatric neurosurgery (Tables 6 and 7).

conclusionsWe retrospectively reviewed a large, single-institution

experience of perioperative (up to 72 hours) ketorolac use across the entire breadth of pediatric neurosurgical procedures. The results suggest no relationship between ketorolac use and radiographic or clinically significant bleeding events in the postoperative period. In combina-tion with previous studies regarding the efficacy of ketoro-lac in decreasing narcotic usage,1,2,5,11,14,15,20,23 as well as the rationale for minimizing the use of sedating medications following neurosurgery, this study supports ketorolac as an option in the perioperative setting among the pediatric neurosurgical population. The role of perioperative ketor-olac in pediatric neurosurgery should be further explored through larger prospective studies.

acknowledgmentsWe would like to thank Judy Gault, PhD, for her assistance with

the manuscript submission. This study was supported by a grant from NCATS/NIH (UL1 TR001082) to Dr. Hankinson.

references 1. Agrawal A, Gerson CR, Seligman I, Dsida RM:

Postoperative hemorrhage after tonsillectomy: use of ketorolac tromethamine. Otolaryngol Head Neck Surg 120:335–339, 1999

2. Anderson GD, Lin YX, Berge C, Ojemann GA: Absence of bleeding complications in patients undergoing cortical surgery while receiving valproate treatment. J Neurosurg 87:252–256, 1997

3. Bailey R, Sinha C, Burgess LP: Ketorolac tromethamine and hemorrhage in tonsillectomy: a prospective, randomized, double-blind study. Laryngoscope 107:166–169, 1997

4. Bauer DF, Waters AM, Tubbs RS, Rozzelle CJ, Wellons JC III, Blount JP, et al: Safety and utility of scheduled nonnarcotic analgesic medications in children undergoing craniotomy for brain tumor. Neurosurgery 67:353–356, 2010

5. Bauer KA, Gerson W, Wright C IV, Wang J, McNicol E,

Lanier RK, et al: Platelet function following administration of a novel formulation of intravenous diclofenac sodium versus active comparators: a randomized, single dose, crossover study in healthy male volunteers. J Clin Anesth 22:510–518, 2010

6. Cannizzaro E, Albisetti M, Wohlrab G, Schmugge M: Severe bleeding complications during antiepileptic treatment with valproic acid in children. Neuropediatrics 38:42–45, 2007

7. Carney BT, Minter CL: Is operative blood loss associated with valproic acid? Analysis of bilateral femoral osteotomy in children with total involvement cerebral palsy. J Pediatr Orthop 25:283–5, 2014

8. Cassinelli EH, Dean CL, Garcia RM, Furey CG, Bohlman HH: Ketorolac use for postoperative pain management following lumbar decompression surgery: a prospective, randomized, double-blinded, placebo-controlled trial. Spine (Phila Pa 1976) 33:1313–1317, 2008

9. Cawthorn TR, Phelan R, Davidson JS, Turner KE: Retrospective analysis of perioperative ketorolac and postoperative bleeding in reduction mammoplasty. Can J Anaesth 59:466–472, 2012

10. Chin CJ, Franklin JH, Turner B, Sowerby L, Fung K, Yoo JH: Ketorolac in thyroid surgery: quantifying the risk of hematoma. J Otolaryngol Head Neck Surg 40:196–199, 2011

11. Chin KR, Sundram H, Marcotte P: Bleeding risk with ketorolac after lumbar microdiscectomy. J Spinal Disord Tech 20:123–126, 2007

12. De Oliveira GS Jr, Agarwal D, Benzon HT: Perioperative single dose ketorolac to prevent postoperative pain: a meta-analysis of randomized trials. Anesth Analg 114:424–433, 2012

13. Forrest JB, Camu F, Greer IA, Kehlet H, Abdalla M, Bonnet F, et al: Ketorolac, diclofenac, and ketoprofen are equally safe for pain relief after major surgery. Br J Anaesth 88:227–233, 2002

14. Forrest JB, Heitlinger EL, Revell S: Ketorolac for postoperative pain management in children. Drug Saf 16:309–329, 1997

15. Gillis JC, Brogden RN: Ketorolac. A reappraisal of its pharmacodynamic and pharmacokinetic properties and therapeutic use in pain management. Drugs 53:139–188, 1997

16. Gobble RM, Hoang HLT, Kachniarz B, Orgill DP: Ketorolac does not increase perioperative bleeding: a meta-analysis of randomized controlled trials. Plast Reconstr Surg 133:741–755, 2014

17. Gupta A, Daggett C, Drant S, Rivero N, Lewis A: Prospective randomized trial of ketorolac after congenital heart surgery. J Cardiothorac Vasc Aneth 18:454–457, 2004

18. Gupta A, Daggett C, Ludwick J, Wells W, Lewis A: Ketorolac after congenital heart surgery: does it increase the risk of significant bleeding complications? Paediatr Anaesth 15:139–142, 2005

19. Judkins JH, Dray TG, Hubbell RN: Intraoperative ketorolac and posttonsillectomy bleeding. Arch Otolaryngol Head Neck Surg 122:937–940, 1996

20. Klein M, Støckel M, Rosenberg J, Gögenur I: Intraoperative ketorolac and bleeding after laparoscopic Roux-en-Y gastric by-pass surgery. Acta Chir Belg 112:369–373, 2012

21. Kokki H: Nonsteroidal anti-inflammatory drugs for postoperative pain: a focus on children. Paediatr Drugs 5:103–123, 2003

22. Lieh-Lai MW, Kauffman RE, Uy HG, Danjin M, Simpson PM: A randomized comparison of ketorolac tromethamine and morphine for postoperative analgesia in critically ill children. Crit Care Med 27:2786–2791, 1999

23. Magni G, La Rosa I, Melillo G, Abeni D, Hernandez H, Rosa G: Intracranial hemorrhage requiring surgery in

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC

Perioperative ketorolac safety

J neurosurg Pediatr Volume 17 • January 2016 115

neurosurgical patients given ketorolac: a case-control study within a cohort (2001–2010). Anesth Analg 116:443–447, 2013

24. Niemi TT, Backman JT, Syrjälä MT, Viinikka LU, Rosenberg PH: Platelet dysfunction after intravenous ketorolac or propacetamol. Acta Anaesthesiol Scand 44:69–74, 2000

25. Onal IK, Parlak E, Akdogan M, Yesil Y, Kuran SO, Kurt M, et al: Do aspirin and non-steroidal anti-inflammatory drugs increase the risk of post-sphincterotomy hemorrhage—a case-control study. Clin Res Hepatol Gastroenterol 37:171–176, 2013

26. Özkiriş M, Kapusuz Z, Yildirim YS, Saydam L: The effect of paracetamol, metamizole sodium and ibuprofen on postoperative hemorrhage following pediatric tonsillectomy. Int J Pediatr Otorhinolaryngol 76:1027–1029, 2012

27. Palmer JD, Sparrow OC, Iannotti F: Postoperative hematoma: a 5-year survey and identification of avoidable risk factors. Neurosurgery 35:1061–1065, 1994

28. Tolska HK, Takala A, Pitkäniemi J, Jero J: Post-tonsillectomy haemorrhage more common than previously described—an institutional chart review. Acta Otolaryngol 133:181–186, 2013

29. Ward MM, Barbaro NM, Laxer KD, Rampil IJ: Preoperative

valproate administration does not increase blood loss during temporal lobectomy. Epilepsia 37:98–101, 1996

disclosureThe authors report no conflict of interest concerning the materi-als or methods used in this study or the findings specified in this paper.

author contributionsConception and design: Richardson, Hankinson. Acquisition of data: Richardson, Palmeri, Williams, Hankinson. Analysis and interpretation of data: Richardson, Hankinson. Drafting the arti-cle: Richardson, Torok, Hankinson. Critically revising the article: O’Neill, Handler, Hankinson. Reviewed submitted version of manuscript: all authors. Approved the final version of the manu-script on behalf of all authors: Richardson. Statistical analysis: Richardson, Torok, Hankinson. Study supervision: Hankinson.

correspondenceMarlin Dustin Richardson, Department of Neurosurgery, Univer-sity of Colorado, 12631 E. 17th Ave., C307, Aurora, CO 80045. email: richardsondusty@hotmail.com.

Unauthenticated | Downloaded 07/13/20 09:37 AM UTC