The Effect of Nonsteroidal Anti-inflammatory Drugs and Selective COX-2 Inhibitors on Bone Healing

Abstract

A recently published study, “Risk of Nonunion With Nonselective NSAIDs, COX-2 Inhibitors, and Opioids” by George et al (J Bone Joint Surg Am. 2020;102:1230–1238), assesses whether the use of nonselective nonsteroidal anti-inflammatory drugs (NSAIDs), selective cyclooxygenase 2 (COX-2) enzyme inhibitors, or opioids was associated with a risk of long bone fracture nonunion in Optum’s deidentified private health database. This review analyzes the study, including strengths, weaknesses, and areas for future research. The study found an association between COX-2 inhibitor and opioid use with fracture nonunion but not with nonselective NSAID use. Although the literature on this topic is varied, these results are at least partially aligned with several animal studies that show COX-2 inhibitors to be associated with fracture nonunion. The George et al study design has several important limitations, indicating that further research is needed on this topic.

Introduction

There is significant controversy surrounding nonsteroidal anti-inflammatory drugs (NSAIDs), such as ibuprofen and meloxicam, and specific cyclooxygenase 2 (COX-2) enzyme inhibitors, most notably celecoxib in the United States, and their effect on fracture healing in orthopedics. The biology of bone healing has been studied exhaustively at both the cellular and molecular levels. At its most rudimentary level, it can be divided into 3 phases: inflammatory, reparative, and remodeling. The reparative phase consists of primary or secondary bone healing, in the form of intramembranous or endochondral ossification, respectively. NSAIDs and specific COX-2 inhibitors, however, are thought to inhibit the inflammatory stage of bone healing, and studies have determined that the COX-2 enzyme plays an important role in the early stages of bone healing [6,22]. Animal studies have reliably demonstrated the deleterious effect of inflammatory inhibition on bone healing [1,9,14,16]. Specifically, inhibiting the production of prostaglandin is thought to disrupt the balance between bone resorption and absorption in fracture healing [12]. Some human clinical studies suggest that there is no effect on healing, whereas others demonstrate an increased risk of nonunion. Interestingly, recent studies have implicated opioid use as a risk factor for fracture nonunion [3,5]. The downregulation of osteoblasts, impaired callus maturation, and defective bony remodeling are some of the proposed mechanisms of fracture nonunion associated with opioid use [11]. In the midst of an opioid epidemic, these implications on bone healing highlight the importance of this discussion and the efforts to identify adequate nonopioid analgesics.

Several animal studies have demonstrated impaired or delayed fracture healing as a result of NSAID and COX-2 inhibitor exposure [1,9,14,16]. In an early study, Rø et al demonstrated a detrimental effect of indomethacin on femur fracture healing in rats, with greater formation of pseudoarthrosis at 24 days [16]. It should be noted, however, that the dose of indomethacin in this study was well beyond therapeutic levels. Similarly, Allen et al reported a dose-related delay in rat femur fracture healing when exposed to indomethacin [1]. In addition, Gerstenfeld et al showed a temporal relationship between selective COX-2 inhibitor (valdecoxib) exposure and delayed fracture healing in rats [9]. While animal studies are somewhat aligned in their consensus, human studies demonstrate predominantly mixed results [2,4,7,10,13,15,17,20]. A retrospective clinical study of 377 patients out of England showed both a delay in fracture union and a higher incidence of fracture nonunion in patients treated with any NSAIDs after their injury. The average duration of NSAID use in this study was 21.2 weeks [10]. Similarly, in a large private insurance database study, Tucker et al showed an increased nonunion rate in operatively treated tibial shaft, humeral shaft, and subtrochanteric femur fractures when any NSAIDs were used in the 90-day postoperative period [20]. Conversely, a study looking at posterior spinal fusion showed 48 hours of exposure to ketorolac postoperatively had no negative effect on fusion rates [15]. Finally, in the pediatric population, 2 studies assessing pediatric long bone fractures showed no detrimental effect of NSAID exposure on bone healing [7,13].

George et al report long bone fracture nonunion rates in a large insurance database, comparing the effects of filled nonselective NSAID prescriptions, COX-2 inhibitor prescriptions, and opioid prescriptions on the outcome of nonunion [8]. The aims of our review are to critically assess their study’s methodology and results and to suggest further avenues for addressing this important research topic.

The Article

• The Risk of Nonunion With Nonselective NSAIDS, COX-2 Inhibitors, and Opioids

• George MD, Baker JF, Leonard CE, Mehta S, Miano TA, and Hennessy S. J Bone Joint Surg Am. 2020;102:1230–1238.

This study used Optum’s deidentified private health database to assess whether the use of nonselective NSAIDs, selective COX-2 inhibitors, or opioids was associated with the risk of nonunion in long bone fractures. After exclusions, the authors identified 339,864 long bone fracture episodes in adult patients from January 1, 2000, to September 30, 2015. They ascertained medication exposure by identifying filled prescriptions for nonselective NSAIDs, selective COX-2 inhibitors, and/or opioids on the fracture date, in the 30 subsequent days, or within 90 days before the fracture. The primary outcome of nonunion was defined by the use of the International Classification of Diseases, 9th Edition (ICD-9) code for nonunion (733.82) alone in the 91 to 365 days after fracture diagnosis or a nonunion diagnosis in combination with a treating procedure within 30 days of the nonunion diagnosis.

There were 304,721 fracture episodes with no prior NSAID or COX-2 inhibitor prescriptions. Of these patients, the authors found that 22,590 (7.4%) filled an NSAID prescription and 2411 (0.8%) filled a COX-2 inhibitor prescription. In total, 387 (1.7%) patients with nonselective NSAID exposure and 69 (2.9%) with selective COX-2 inhibitor exposure had a nonunion diagnosis. Of the 269,841 patients without prior opioid use, 123,807 (45.9%) filled an opioid prescription within 30 days of the fracture, and of these, 2600 (2.1%) ended up with a nonunion diagnosis (Table 1). Nonunion diagnosis with a procedure to treat occurred in 2996 (0.9%) patients overall. The authors ultimately concluded that filling a nonselective NSAID prescription after fracture was not associated with either nonunion outcome (adjusted odds ratio [aOR] = 1.07, confidence interval [CI] = 0.93–1.23]; aOR = 1.08, CI = 0.96–1.20). However, filling a COX-2 inhibitor prescription (aOR = 1.84, CI = 1.38–2.46; aOR = 1.48, CI = 1.16–1.89) or opioid prescription (aOR = 1.69, CI = 1.53–1.86; aOR = 1.53, CI = 1.43–1.64) after fracture was associated with a greater risk of nonunion. Adjusted analyses revealed that filling a prescription for an NSAID or COX-2 inhibitor in the 90 days prior to fracture was also associated with nonunion.

Table 1.

Summary of findings in the study by George et al. [8].

	No. of fracture episodes exposed	Nonunion diagnosis	aOR (95% CI)
NSAID/COX-2 inhibitor analysis
Neither	279,720	4188 (1.5%)	Reference
Nonselective NSAID	22,590	387 (1.7%)	1.08 (0.96–1.20)
COX-2 Inhibitor	2411	69 (2.9%)	1.48 (1.16–1.89)
Total	304,721	4644 (1.5%)
Opioid analysis
No opioid	146,034	1513 (1.0%)	Reference
Opioid	123,807	2600 (2.1%)	1.53 (1.43–1.64)
Total	269,841	4113 (1.5%)

NSAID nonsteroidal anti-inflammatory drugs, COX-2 cyclooxygenase 2, aOR adjusted odds ratio, CI confidence interval..

Commentary

This study addresses the important and widely debated topic of NSAID and COX-2 inhibitor influence on fracture healing using prescription data in a large insurance patient database. Similar to other studies on the topic, the authors found COX-2 inhibitors to be associated with nonunion risk [6,8,14,21]. Nonselective NSAIDs, however, did not have an association with greater nonunion risk in their analysis. Interestingly, they also identified filled opioid prescription to be associated with nonunion risk. The strengths and weaknesses of the study design bring the findings into question and prompt opportunities for further investigation.

The strengths of this study are principally related to the size of the patient population. The sheer volume of patients provided the researchers access to numerous covariates for analysis and ample power for statistical analyses. In addition, 10 different long bone fractures or combinations of fractures were available for analysis, which allowed for several sensitivity analyses. Accordingly, they were able to study interactions between medication exposure and tibia, tibia and fibula, distal femur, and femoral neck fractures independently. Finally, using an insurance database, the authors had a uniform method for assessing filled prescriptions and could analyze the effect of multiple prescriptions as well as their timing (before or after the incident fracture).

There are significant limitations, however, inherent to the design of this study. One of the overarching deficiencies is inherent to the use of a national insurance database—all data and outcomes are based on administrative claims. This is a relatively flawed method for gathering clinically relevant information as it is prone to incorrect data entries. For example, ICD-9 codes in particular were developed for purposes of universal coding and are thus innately flawed in capturing the subtleties necessary in clinical research [18,19]. Probably the most glaring deficiency of this study, however, is tied to its method of ascertaining prescription use. As the primary exposure for analysis, use of nonselective NSAID, COX-2 inhibitor, and opioid is inferred by the filling of a prescription. Shortcomings of this method include the impossibility of assessing actual exposure to the various medications through filled prescriptions alone. In addition, patients without filled prescriptions but with exposure to over-the-counter NSAIDs or illicit opioids were missed in their analysis. Similarly, filled prescriptions may have been incompletely taken or not taken at all; such patients would be miscategorized in the analysis.

Despite animal studies demonstrating a dose-dependent relationship of nonselective NSAIDs on fracture healing, this relationship is not explored by George et al [1,8]. In fact, the effect of NSAID or COX-2 inhibitor dose on fracture healing in humans has not been directly studied and provides an opportunity for future research. In addition, Gerstenfeld et al demonstrated the temporal relationship between COX-2 inhibition and fracture healing in rats; in comparing the use of ketorolac, valdecoxib, and a control for 7 or 21 days, they found that 21 days of valdecoxib exposure had a negative but reversible effect compared with the other 2 treatments [9]. While George et al found that multiple filled prescriptions for all 3 medications were associated with greater risk of nonunion, the study design cannot separate causation from association. In fact, the pain that accompanies a nonunion can quite easily explain multiple analgesia refills. In addition, despite having access to several covariates for controlling analysis, this study lacks the granularity to control for several risk factors associated with fracture nonunion. Patient behaviors, intraoperative variables, tobacco use, and the mechanics of surgical constructs—among other known risk factors for nonunion—are impossible to measure, and thus to control for, in a massive database study. Finally, George et al found an association between opioid prescriptions and nonunion. The mechanism to describe this association is incompletely understood, but it is of significant importance given the ubiquity of opioid analgesia in the United States, and this relationship deserves further exploration.

In conclusion, this large insurance database study is one of the first to evaluate the influence of NSAIDs and COX-2 inhibitors on fracture nonunion, as well as highlighting the continued clinical relevance of this topic. Several of the major shortcomings viewed in the literature, such as dose measurement, timing of exposure, and treatment duration in humans, have yet to be studied. Although the authors do not address these areas directly, they have heightened the discussion and emphasized the need for continued research. Anecdotally, many orthopedic surgeons do practice caution in counseling patients about the theoretical risks of NSAID use. However, for the reasons mentioned above, the findings of this study do not change clinical practice in either direction. Ideally, a randomized controlled trial on this topic would provide the necessary evidence to change clinical practice. However, a well-executed prospective study could also provide adequate evidence to alter clinical practice.