Background: Nonsteroidal anti-inflammatory drugs (NSAIDs) are often avoided by orthopaedic surgeons because of their possible influence on bone-healing. This belief stems from multiple studies, in particular animal studies, that show delayed bone-healing or nonunions associated with NSAID exposure. The purpose of this review was to critically analyze the quality of published literature that evaluates the impact of NSAIDs on clinical bone-healing.
Methods: A MEDLINE and Embase search was conducted to identify all articles relating to bone and fracture-healing and the utilization of NSAIDs. All human studies, including review articles, were identified for further analysis. Non-English-language manuscripts and in vitro and animal studies were excluded. A total of twelve clinical articles and twenty-four literature reviews were selected for analysis. The quality of the clinical studies was assessed with a modified Coleman Methodology Score with emphasis on the NSAID utilization. Review articles were analyzed with regard to variability in the cited literature and final conclusions.
Results: The mean modified Coleman Methodology Score (and standard deviation) was significantly lower (p = 0.032) in clinical studies that demonstrated a negative effect of NSAIDs on bone-healing (40.0 ± 14.3 points) compared with those that concluded that NSAIDs were safe (58.8 ± 10.3 points). Review articles also demonstrated substantial variability in the number of cited clinical studies and overall conclusions. There were only two meta-analyses and twenty-two narrative reviews. The mean number (and standard deviation) of clinical studies cited was significantly greater (p = 0.008) for reviews that concluded that NSAIDs were safe (8.0 ± 4.8) compared with those that recommended avoiding them (2.1 ± 2.1). Unanimously, all reviews admitted to the need for prospective randomized controlled trials to help clarify the effects of NSAIDs on bone-healing.
Conclusions: This systematic literature review highlights the great variability in the interpretation of the literature addressing the impact of NSAIDs on bone-healing. Unfortunately, there is no consensus regarding the safety of NSAIDs following orthopaedic procedures, and future studies should aim for appropriate methodological designs to help to clarify existing discrepancies to improve the quality of care for orthopaedic patients.
Clinical Relevance: This systematic review highlights the limitations in the current understanding of the effects of NSAIDs on bone healing. Thus, withholding these medications does not have any proven scientific benefit to patients and may even cause harm by increasing narcotic requirements in cases in which they could be beneficial for pain management. This review should encourage further basic-science and clinical studies to clarify the risks and benefits of anti-inflammatory medications in the postoperative period, with the aim of improving patient outcomes.
Nonsteroidal anti-inflammatory drugs (NSAIDs) have been utilized to treat pain, fever, and inflammation since the nineteenth century and are among the most commonly utilized medications in the world today1. With few exceptions, NSAIDs have become the safest and most effective drugs to treat postoperative pain2,3. Despite these benefits, NSAIDs are thought to cause delayed bone-healing and nonunions and are often withheld following orthopaedic procedures4-11. Other studies have shown that NSAIDs are safe following fracture fixation and spinal fusions without considerable effects on bone-healing12-15. These discrepancies in the literature have been the subject of debate in numerous literature reviews16-39, instructional course lectures40,41, best-evidence topic reports42, commentaries43-48, expert opinions49, editorials50,51, and even lay-press articles52,53.
It is widely accepted that the existing literature has substantial methodological limitations; however, the actual quality of the literature and the factors that influence its divergent interpretation have not previously been characterized, to our knowledge. The goals of this systematic review are fourfold: (1) to offer a comprehensive assessment of the basic-science, pre-clinical, and clinical literature evaluating the effects of NSAIDs on bone-healing; (2) to characterize the methodological quality in the published literature addressing the effects of NSAIDs on bone-healing; (3) to identify elements in study designs that contribute to conflicting interpretations of results and recommendations in the published literature; and (4) to suggest methodological characteristics that may help future studies to better characterize the impact of NSAIDs on bone-healing.
Materials and Methods
A literature search was performed through the MEDLINE and Embase databases utilizing the following free-text and MeSH terms: nonsteroidal anti-inflammatory agents, NSAID, fracture-healing, bone healing, spinal fusion, and cyclooxygenase-2 inhibitors. All articles addressing the impact of NSAIDs on bone-healing and fracture-healing following orthopaedic procedures were identified. Specifically, we included studies that assessed either fracture-healing or spinal fusion, but we excluded those that assessed the impact of NSAIDs on bone metabolism, bone mineral density, fracture risk, heterotopic ossification, and osseous ingrowth onto metal implants. Non-English-language manuscripts, in vitro and animal studies, case reports, studies not specifically associated with bone-healing, and those performed on a pediatric population were also excluded. Literature reviews that cited pediatric studies were still included for analysis as a potential factor that may have influenced the authors’ conclusions and recommendations. All prospective and retrospective human studies and literature reviews were identified for further evaluation. References of all selected articles were reviewed to ensure that all relevant studies would be included for analysis. A total of twelve clinical articles and twenty-four literature reviews were selected (Fig. 1). The literature search was conducted on two separate occasions (January 5, 2015, and March 2, 2015) independently by two of the authors (A.M.-L. and I.D.H.) following recommendations from the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines54. Records were stored in a computer-based referenced management system (EndNote, Thomson Reuters).
Clinical studies were reviewed to specifically assess the type of NSAID utilized, dose, route of administration, length of NSAID administration, type of bone fixation, effect on bone-healing, and critical methodological limitations. A modified Coleman Methodology Score was utilized to assess the methodology of the clinical studies. The Coleman Methodology Score is based on a subsection of the Consolidated Standards of Reporting Trials (CONSORT) statement55,56, and has been validated and has been utilized by several published reports in the orthopaedic literature57-59. The Coleman Methodology Score utilizes ten criteria, giving a total score between 0 and 100 points56. A methodological score of 100 points indicates that the study is methodologically sound and largely avoids chance, various biases, and confounding factors. For the purpose of this review, the terms surgical description, postoperative rehabilitation, and diagnostic certainty were substituted for description of NSAID regimen and monitoring of postoperative NSAID utilization, and scores were adjusted accordingly to add up to 100 points. Previous studies have modified the original Coleman Methodology Score to accommodate for specific topics and study objectives57,59. The categorical rating was considered to be excellent if the score was 85 to 100 points, good if it was 70 to 84 points, fair if it was 55 to 69 points, and poor if it was ≤54 points59.
Both the clinical and review articles were analyzed with regard to country of origin, impact factor, author specialty and profession, and final interpretation and recommendation. Reviews were assessed for the quality and quantity of cited clinical studies. In addition, a twenty-seven-item checklist (www.prisma-statement.org/) was utilized to assess the quality of any systematic review or meta-analysis. Clinical studies were further analyzed with regard to the reported nonunion rates, functional outcomes (when available), and post hoc power calculations.
Statistical analysis was performed with use of SPSS version 20 (IBM). All data were checked for normal distribution (the Kolmogorov-Smirnov test, the Shapiro-Wilk test, and the Pearson skewness coefficient). The chi-square test was utilized for categorical data, and either the Student t test or the analysis of variance (ANOVA) with post hoc pairwise analysis (Bonferroni) was utilized for continuous data. A p value of <0.05 was utilized to denote significance. Pearson correlation analysis, with a 95% confidence interval (95% CI), was utilized to correlate the modified Coleman Methodology Score with the number of times that each study was cited, the journal impact factor, and the year of publication to investigate trends in methodology over time. Lastly, all post hoc power calculations were performed with use of G*Power software (University of Dusseldorf).
A complete list of the twelve clinical studies included in this analysis along with study characteristics and modified Coleman Methodology Score can be found in Table I. Based on the modified Coleman Methodology Score, none were excellent, one was good (a Coleman Methodology Score of 70 points), three were fair (a Coleman Methodology Score of 56 to 66 points), and eight were poor (a Coleman Methodology Score of 20 to 54 points). Only three of the studies were randomized controlled trials, and the other nine studies were retrospective chart reviews. Nonunion rates varied widely between studies (0% to 65.6% in the NSAID groups and 0% to 19.0% in the control groups), and only four studies showed any sort of functional outcomes. Based on the power calculations, only five of the twelve clinical studies had a power of >80%.
Overall, the research quality was poor, with a mean modified Coleman Methodology Score (and standard deviation) of 47.8 ± 14.5 points (range, 20 to 70 points). There were notably low mean scores in the type of study design (3.75 ± 6.50 points), in the postoperative NSAID monitoring (4.17 ± 4.93 points), in the procedure for assessing outcomes (2.85 ± 2.33 points), and in the description of the subject selection process (4.83 ± 2.11 points) (Table II).
Based on the study conclusions and recommendations, the mean modified Coleman Methodology Score was significantly lower (p = 0.032) in clinical studies that demonstrated a negative effect of NSAIDs on bone-healing (40.0 ± 14.3 points) compared with the studies that showed a neutral effect (58.8 ± 10.3 points) (Table III). On average, the journal impact factor, author specialty (percentage of orthopaedic surgeons), country of origin (United States compared with outside the United States), year of publication, or number of times cited by reviews did not significantly differ (p > 0.05) between studies with contrasting views on the effects of NSAIDs on bone-healing. The three studies most often cited by literature reviews (fourteen to twenty-two times) were all studies that concluded that NSAIDs had a detrimental effect on bone-healing. In contrast, none of the clinical studies that concluded that NSAID use was appropriate were cited more than ten times by the reviews.
Correlation analysis demonstrated that there was no significant relationship between the modified Coleman Methodology Score and the year of publication (r = −0.313, p = 0.321) or journal impact factor (r = −0.247, p = 0.440). Similarly, there was no correlation between the journal impact factor and the number of times that a study was cited (r = 0.137, p = 0.670).
A complete list of the twenty-four literature reviews included in this analysis along with study characteristics and the summaries of the authors’ conclusions and recommendations can be found in Table IV. There were two meta-analyses, and twenty-two narrative reviews. Unanimously, all review articles described important limitations in the available literature and discussed the need for prospective randomized controlled trials to help clarify the effects of NSAIDs on bone-healing. The four literature reviews that included studies on pediatric populations all concluded that either short-term NSAID use was safe or there was no clear contraindication to NSAID utilization during bone-healing27-29,33.
The mean number of clinical studies (and standard deviation) cited in review articles that concluded that NSAIDs were safe to use (8.0 ± 4.8) was significantly greater (p = 0.008) compared with those that recommended avoiding NSAIDs (2.1 ± 2.1) (Table V and Fig. 2). Similarly, the quality of clinical studies was higher among the studies that concluded that NSAID utilization was safe (50.9 ± 11.1) compared with those that recommended against NSAID utilization (32.7 ± 21.3) or that had no clear recommendations (39.8 ± 15.3); however, this was not significant (p = 0.156). Journal impact factor, author specialty (percentage of orthopaedic surgeons), country of origin (United States compared with outside the United States), and year of publication did not vary significantly (p > 0.05) when comparing reviews of various conclusions and recommendations. The two meta-analyses were performed in accordance with the PRISMA guidelines.
Several reviews cited identical clinical studies but showed conflicting conclusions and recommendations (data not shown). The two reviews with the highest number of clinical citations (n = 10) agreed that NSAID administration was safe as long as the duration was short (one week)29,33. Of note, both studies pointed out that NSAIDs should be avoided in patients at risk for delayed healing or if the fracture shows signs of delayed union or nonunion.
Postoperative pain control, particularly in orthopaedic patients, continues to be a challenge60,61. The use of multimodal analgesia, in which NSAIDs play an important role, has helped to make great strides in improving acute pain management and has been utilized extensively in patients undergoing total joint arthroplasty62-65. Although there is some evidence that this practice is being adopted by other orthopaedic specialties, fracture repair surgeons would often prefer not to use NSAIDs, thereby avoiding any risk of impaired healing37. However, this has led to an increase in the use of opiates, along with their many drawbacks, during the postoperative period66.
NSAIDs are known to inhibit both cyclooxygenase (COX) enzyme isoforms (COX-1 and COX-2), with varying degrees of selectivity depending on the type of NSAID. This inhibition decreases prostaglandin levels, which play a major regulatory role in all stages of the bone-healing cascade, particularly the early inflammatory phase67-69. Thus, it is reasonable to predict that NSAID administration may affect bone-healing. However, as demonstrated by this review, results from the published literature are highly variable and demonstrate methodological limitations.
Considerations from Basic-Science Studies
It is clear from the basic-science literature that studying the effects of NSAIDs on bone-healing is extremely challenging. Based on the physiological requirements of the bone, prostaglandins either can promote bone resorption by directly increasing osteoclast activity or can stimulate bone production by activating osteoblast differentiation70-73. Prostaglandins exert this range of action through a variety of G-protein-coupled receptor subtypes (EP1 to EP4). Although activation of EP2 and EP4 receptors promotes bone-healing, EP1-receptor activation has been shown to have a major inhibitory effect on bone-healing74-77. The effect of NSAIDs on the expression of these receptors has not been fully elucidated, but may help to explain some of the discrepancies between basic-science and clinical studies.
Animal studies have helped to advance our understanding of the physiological events critical to normal bone-healing78,79. This has led to the development of novel gene-therapy models (e.g., COX-2 transgene) that have shown to increase recruitment and expansion of mesenchymal stem cells, decrease cartilaginous callus formation, and increase angiogenesis-dependent cartilage remodeling within the healing fracture callus80,81. However, extrapolating the impact of NSAIDs in animal models to humans should be done with caution. The type of NSAID and its bioavailability are known to be highly variable between animal species and may not be taken into consideration in experimental designs47,82. Furthermore, there is great variability in the timing and length of exposure, which also appears to influence the impact of NSAIDs on bone-healing83,84. Although continuous and prolonged exposure to NSAIDs (greater than six weeks) may inhibit or may at least delay bone-healing in animals85-90, in a clinical setting, patients are more likely to receive short-term exposure (three to twenty-one days)4,8-12,14, and any detrimental effect to bone-healing would likely be reversible after discontinuation of the NSAID91. Furthermore, existing animal models fail to consider the major inflammatory insult associated with surgical fixation92,93.
In vitro and in vivo animal studies have helped to advance our understanding of the physiological events critical to normal bone-healing. In addition, they have provided insight into the effects of both COX-2-specific and non-specific NSAIDs on bone-healing35. However, the great variability in study design and the inherent limitations of basic-science models have led to inconsistent and variable effects of NSAIDs on bone-healing and have failed to translate into clinical practice.
Clinical Studies: Limited Quality of Evidence
From a methodological perspective, trying to understand the impact of NSAIDs on bone-healing in a clinical setting is extremely challenging. It is widely accepted that numerous confounding factors (e.g., smoking, diabetes, obesity) affect bone-healing, and controlling for these in a prospective study requires considerable time and planning94. These difficulties may help to explain the limited quality of clinical studies included in this review. Eight of the twelve studies received a modified Coleman Methodology Score of <54 points and were classified as being of poor quality. Factors that contributed to the low methodological scores were the type of study design, postoperative NSAID monitoring, procedure for assessing outcomes, and description of subject selection process. With the exception of postoperative NSAID monitoring, which was unique to this review, the areas of methodological deficiency identified in this analysis have been reported in the published literature as common weaknesses in orthopaedic trials56,58,59. Unfortunately, there was no correlation between modified Coleman Methodology Score and year of publication, which suggests that the quality of research in this area has not improved in the past decades.
The three clinical trials with the highest modified Coleman Methodology Score (62 to 70 points) were all prospective randomized controlled trials that showed no negative effect of NSAIDs on bone-healing10,12,13. Of these, only the study by Sagi et al. showed a greater risk of nonunion after six weeks of indomethacin exposure10, which supports the findings of Burd et al., whose 2003 article has been highly cited5. However, it is important to note that even in the former study, one week of indomethacin exposure was not associated with an increased risk of nonunion10. Ketorolac was also associated with delayed bone-healing in two studies after more than forty-eight hours of exposure4,8, although other studies have shown that exposure for less than forty-eight hours did not have any detrimental effects on bone-healing compared with controls14. Similar observations with regard to timing and dose have been noted elsewhere11,15. In one study, NSAID utilization was associated with a higher rate of nonunions; however, this was only true for patients who took NSAIDs on days 61 to 90 after surgery15. As others have mentioned, this was likely related to the pain associated with the nonunion, demonstrating an association, not proving causation21. Another study demonstrated a positive correlation between time to union and the amount of diclofenac used (r = 0.271; p < 0.001). Although high-dose diclofenac (>300 mg for two weeks) was associated with nonunions, normal dosing did not have a major effect11. These observations highlight the importance of monitoring the type, dose, timing, and duration of NSAID exposure to better understand how these drugs may be safely utilized in the setting of bone-healing. Unfortunately, only five of the twelve clinical studies in the current study showed some sort of long-term NSAID monitoring7,10,12,13,15.
Despite many efforts, the quality of evidence with regard to the effects of NSAIDs on bone-healing is limited, particularly in the adult literature. Although most authors have stated that a large, multicenter, randomized controlled trial is warranted, few have provided specific details of a proposed research design. Future studies should aim for a well-designed prospective trial, with clearly defined end points as suggested by Jakobsen et al.57, as well as strict NSAID-administration parameters (Table VI). At this point, to our knowledge, there have been no excellent studies in the published literature addressing the effects of NSAIDs on bone-healing; therefore, it is difficult to assess the amount of high-quality studies that would be needed to accurately determine the effects of NSAIDs on bone-healing.
Literature Reviews: Variability in Conclusions and Recommendations
The vast majority of reviews (twenty-two [91.7%] of twenty-four) in this analysis were narrative reviews. Narrative reviews offer a comprehensive, generally up-to-date summary of the available literature and provide a broad perspective on a given topic. However, in this type of review, authors have the risk of selecting research that helps to present their argument, which can result in contradictory conclusions and recommendations95. These types of discrepancies have previously been described by Geusens et al., who noted that many review articles and editorials reached different conclusions and provided conflicting recommendations despite citing the same clinical studies33. However, the present study highlights that the majority of review articles were not referencing the same literature and that the quality and quantity of clinical studies included in a literature review substantially impacted the overall message.
The only two meta-analyses in this review offered interesting perspectives on both the quality of the literature and the NSAID effects on bone-healing. In one meta-analysis, we noted an association between low-quality and moderate-quality studies and an increased reported risk of nonunions27. The lower-quality studies were long-bone fracture studies, and higher-quality studies originated from the spine literature. When data from the higher-quality spine studies were pooled into their analysis, there was no increase in the risk of nonunion with NSAID exposure. Our results agree with an association between higher-quality studies and no evidence of negative effects of NSAIDs on bone-healing; however, we were unable to find a significant association (p < 0.05) with the anatomic region. In our study, four (57.1%) of the seven clinical studies that recommended avoiding NSAID exposure were from the spine literature, compared with one (20%) of the five studies that concluded that NSAID use was acceptable; however, this difference was not significant (p = 0.293) (Table III). Furthermore, the mean modified Coleman Methodology Score was not significantly different (p = 0.705) between spine studies (50.5 ± 5.2 points) and fracture studies (46.3 ± 20.5 points). The fact that we excluded pediatric studies from our analysis might explain these differences, as two of the spine studies96,97 reviewed by Dodwell et al.27 were performed in pediatric populations. In contrast to the adult literature, in the pediatric orthopaedic literature there appears to be consensus that NSAIDs do not affect bone-healing96-100.
In the other meta-analysis28, the authors only included spinal fusion studies on adult patients and concluded that short exposure (less than fourteen days) to a normal dose of NSAIDs did not increase the risk of nonunion, which could represent a potential regimen for NSAID administration. Although the authors highlighted clear limitations in the available literature, they failed to exclude a retracted study by Reuben et al.101. This information should be taken into consideration when interpreting their results. Despite these limitations, both meta-analyses were in accordance with the PRISMA guidelines for conducting quantitative systematic reviews and, in our opinion, contributed to the quality of evidence related to NSAIDs and bone-healing. None of the reviews published after these reports (van Esch et al.31, Su and O’Connor32, Geusens et al.33, Kurmis et al.30, and Pountos et al.29) referenced the meta-analysis by Li et al.28, and only one (Geusens et al.33) referenced the meta-analysis by Dodwell et al.27.
There were several limitations to our study that must not be overlooked. Studies that did not mention the specific NSAID type, dose, and route of administration were included for the purpose of this review to help to assess elements that may affect the interpretation of the effects of NSAIDs on bone-healing. Studies missing these parameters received low modified Coleman Methodology Scores, which may have reduced the overall scores. However, these parameters were intentionally added to highlight the importance of including this information in clinical studies. In addition, two different reviewers scored each study independently, and any discrepancies were discussed with the senior author until a consensus was reached. We believe that these strict grading parameters enhanced the quality of our review and offered insight into improving future clinical studies. We did not offer insight on the fracture risk, if any, of chronic NSAID administration. However, studies that have assessed this particular point have demonstrated no increased fracture risk with either chronic or occasional use of NSAIDs102,103. We excluded studies in pediatric populations as the majority have demonstrated no detrimental effects on bone-healing96-100. It is widely accepted that pediatric patients should not be compared with adults, particularly with regard to orthopaedic problems104. However, literature reviews that cited pediatric studies were still included to highlight factors that may have affected the authors’ conclusion and recommendations. We also excluded studies examining osseous ingrowth and heterotopic ossification105-107. Although there may be overlap between these two phenomena and fracture-healing, the mechanism is not completely understood. Several studies addressing the effects of NSAIDs on spinal fusion101,108 were not included in this analysis because of fraud allegations that question the veracity of those results. In those studies, COX-2 inhibitors, particularly celecoxib, were found not to impact normal bone-healing108. It is important to note that all of the studies included in this systematic review assess non-selective NSAIDs, and, to date, to our knowledge, there have been no published reports on selective COX-2 inhibitors. Lastly, this review did not provide any pooled analysis between study populations to determine the impact of NSAIDs on bone-healing, based on the type of NSAID, nonunion rates, and functional outcomes. Although this information is critical for the clinical application of these medications, the purpose of this study was instead to offer the reader an opportunity to reflect on the quality of research that is guiding orthopaedic practice. In the age in which evidence-based medicine represents the foundation to the standard of care, the current evidence does not allow orthopaedic surgeons to accurately recommend for or against the use of NSAIDs.
Much effort has been invested in both basic-science and clinical studies to better understand the effect of NSAIDs on bone-healing. Nonetheless, the literature still does not offer a clear consensus with regard to the safety or harm of NSAID utilization following orthopaedic procedures. The great variability in the interpretation of the available evidence appears to suggest that NSAIDs may affect bone-healing, but that this effect depends on the type, dose, timing, and length of exposure. However, the purpose of this study was not to offer a definitive conclusion on the effects of NSAIDs on bone-healing, but to highlight the factors in the literature that may impact the interpretation of these results. It is important for future studies to aim for the most appropriate methodological design, but it is the responsibility of those who read the literature to be critical and, at times, skeptical of both clinical studies and literature reviews that may affect the quality of care for orthopaedic patients109.
Source of Funding: This study was planned and executed without any external funding source.
Investigation performed at the Department of Orthopaedic Surgery, Wake Forest University School of Medicine, Winston-Salem, North Carolina
Disclosure: None of the authors received payments or services, either directly or indirectly (i.e., via his or her institution), from a third party in support of any aspect of this work. One or more of the authors, or his or her institution, has had a financial relationship, in the thirty-six months prior to submission of this work, with an entity in the biomedical arena that could be perceived to influence or have the potential to influence what is written in this work. Also, one or more of the authors has had another relationship, or has engaged in another activity, that could be perceived to influence or have the potential to influence what is written in this work. The complete Disclosures of Potential Conflicts of Interest submitted by authors are always provided with the online version of the article.
- Copyright © 2016 by The Journal of Bone and Joint Surgery, Incorporated