The Effect of Non-Steroidal Anti-Inflammatory Drugs on Tendon-to-Bone Healing: A Systematic Review with Subgroup Meta-Analysis

Abstract

Background:

There is some concern that non-steroidal anti-inflammatory drugs (NSAIDs) may impair the healing of certain musculoskeletal tissues. However, the effect of NSAIDs on the specialized fibrocartilaginous transition at the tendon-to-bone interface remains largely unknown. Thus, the purpose of the present study is to investigate the effect of NSAIDs on tendon-to-bone healing following acute injury or surgery.

Methods:

A systematic review was performed according to the 2009 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. The PubMed (MEDLINE), Embase, Cumulative Index to Nursing and Allied Health (CINAHL), and SportDiscus databases were searched from the time of database inception to May 14, 2018 to identify all clinical and basic science studies investigating the effect of NSAIDs on tendon-to-bone healing. Methodological quality was assessed using the Methodological Index for Non-Randomized Studies (MINORS) and SYstematic Review Center for Laboratory animal Research (SYRCLE) risk of bias assessment tools for clinical and basic science studies, respectively. A qualitative synthesis of the literature was performed with a subgroup meta-analysis of homogeneous basic science studies.

Results:

A total of 13 studies, including three clinical and 10 basic science studies, were included. The overall methodological quality of the included clinical studies was poor, while assessment of the basic science studies revealed consistent areas at high or unclear risk of bias. Of the included clinical studies, a single study reported a higher rate of rotator cuff repair failure with selective (COX-2) NSAID use compared with non-selective NSAID use, while no clinical failures were noted with NSAID use following distal biceps tendon repair. Basic science studies had heterogeneous outcome reporting. A subgroup analysis of homogeneous animal studies revealed no significant effect of all NSAIDs (Standard Mean Difference [SMD] -1.05, 95% Confidence Interval [CI] -2.39-0.30, p=0.13) or non-selective NSAIDs on load to failure (SMD -0.62, 95% CI -1.26-0.02, p=0.06).

Conclusions:

The current literature does not provide sufficient evidence for or against the use of NSAIDs following acute injury or surgical repair of the tendon-bone interface.

Level of Evidence: IV

Introduction

Acute and chronic injuries to the tendon-bone interface are a common cause of pain and dysfunction.^1-4 While a myriad of surgical procedures aimed at re-establishing the complex tendon-to-bone interface at various anatomic locations have been described for both acute and chronic conditions,^5-9 surgery fails to fully restore the unique composition and mechanical properties of the native tendon-to-bone interface.^10-13 In hopes of better recreating the native tendon-to-bone interface following surgical intervention, local biologic augmentation,^14-19 systemic pharmacotherapy,^20-22 and avoidance of pharmaceuticals that may inhibit tendon-to-bone healing²³ have been proposed as potential options to improve healing at the tendon-bone interface. However, consistent high-level evidence to support their clinical use is lacking.

Nonsteroidal anti-inflammatory drugs (NSAIDs) are frequently used for their analgesic and anti-inflammatory properties in acute and chronic musculoskeletal conditions. While there is some concern that NSAIDs may impair the healing of certain musculoskeletal tissues, including bone and tendon in isolation,^24-28 the effect of NSAIDs on the specialized fibrocartilaginous transition at the tendon-to-bone interface remains largely unknown. Definitively determining the potentially negative effect of NSAIDs on tissue healing is further complicated by the different mechanism of action of NSAIDs at the cellular level, including cyclooxygenase-1 (COX-1) and/or cyclooxygenase-2 (COX-2) inhibition,²⁹ and the temporal relationship between drug administration and healing.^26,30,31

To date, there remains no consensus within the literature as to the effect of NSAIDs on tendon-to-bone healing. While NSAIDs have been used postoperatively following orthopaedic procedures to limit inflammation,³² heterotopic ossification,^33,34 and pain, thus reducing opioid demand,^35,36 these beneficial uses must be considered in light of the potentially deleterious effect on healing. The purpose of the current review is to investigate the effect of NSAIDs specifically on tendon-to-bone healing by systematically reviewing the available basic science and clinical literature.

Methods

This study was performed according to the suggested methods within the Cochrane Handbook for Systematic Reviews of Interventions³⁷ and reported using the 2009 Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.³⁸

Study Eligibility

Inclusion criteria were established a priori and included the following: all English-language, full-text, original clinical or basic science studies investigating tendon-to-bone healing in the acute injury or surgical setting that utilized NSAIDs as part of a defined post-injury or postoperative protocol, reported a clinical, biochemical, and/or biomechanical assessment of healing, and utilized a co-culture model consisting of bone or osteoprogenitor cells and tenocytes or fibroblasts for in vitro studies.

Exclusion criteria included the following: non-English studies, case reports, review articles, editorials, letters to the editor, studies using only survey data, studies that failed to report a clinical, biochemical, and/or biomechanical assessment of healing, in vitro studies using isolated tenocyte or fibroblast cultures, studies describing the effect of NSAIDs on intra-tendinous healing after acute injury or surgery, isolated muscle healing, studies describing the effect of NSAIDs following nonoperative management of chronic or degenerative tendon conditions, studies that included surgical procedures that did not aim to repair the tendon of interest to its normal anatomic footprint, such as tendon transfer or anterior cruciate ligament reconstruction, and studies that did not provide a clear description or protocol for post-injury or postoperative NSAID utilization (type, duration, and/or dose).

Search Strategy

Independent searches of the PubMed (MEDLINE), Embase, Cumulative Index to Nursing and Allied Health (CINAHL), and SportDiscus databases were performed using optimized search strategies for each database (Appendix 1). All searches were performed from the time of inception of the database of interest through May 14, 2018. After completing the search, duplicates were removed and an abstract and title screen performed. A full-text review was then performed independently by three authors (KRD, DBL, SHP). In the event of a disagreement between the three independent reviewers, the senior author (JCR) made the final decision on study inclusion. References in the reviewed manuscripts and review articles identified during the search were scrutinized to identify any studies not captured by the initial search parameters.

Quality Assessment of Included Studies

Included randomized clinical studies were evaluated for bias using the Cochrane Collaboration’s bias assessment tool for randomized clinical trials.³⁹ All non-randomized clinical studies were evaluated for bias using the Methodological Index for Non-Randomized Studies (MINORS) scoring system.⁴⁰ The MINORS scoring system allows a score to be assigned to non-randomized studies, with a higher score indicating less bias. Using the MINORS scoring system, the highest possible score for non-comparative and comparative clinical studies is 16 and 24, respectively. MINORS scores were provided as an absolute point value as well as a percentage of the total possible score.

All included animal studies were evaluated for bias using the SYstematic Review Center for Laboratory animal Research (SYRCLE) risk of bias assessment tool,⁴¹ an adaption of the Cochrane Collaboration’s bias assessment tool for randomized clinical trials.³⁹ The SYRCLE bias assessment tool asks 10 questions which are scored as “yes”, “no”, or “unclear”, with “yes” answers indicating a low risk of bias, “no” answers indicating a high risk of bias, and “unclear” answers indicating insufficient details to adequately assess.

To date, there are no validated or consistently utilized risk of bias assessment tools for in vitro studies, including the A in vitro co-culture studies included in this systematic review.

Data Abstraction

Data from all included studies were abstracted independently by three reviewers (KRD, DBL, SHP) and organized within Excel (Microsoft Corp., Redmond, WA). Studies were classified as “clinical” or “basic science” with basic science studies further categorized as “animal” or “in vitro”. In addition to categorization by the generic pharmaceutical name, NSAIDs were further categorized as “non-selective” (COX-1 and COX-2 inhibition) or “selective” (COX-2 inhibition) based on the mechanism of action defined within the National Institute of Health’s PubChem database⁴² (Table 1). If a study included multiple treatment arms, only arms that investigated the effect of an NSAID in isolation were abstracted. Clinical studies were also categorized according to level of evidence.⁴³ Ultimately, data elements that were reported by at least three studies were included for qualitative and/or quantitative analysis, where appropriate.

Table 1.

Selectivity of Included NSAIDs

Non-Selective	Selective
Diclofenac	Celecoxib
Flubiprofen	Parecoxib
Ibuprofen	Valdecoxib
Indomethacin
Licofelone*
Meloxicam
Naproxen
Piroxicam

NSAID, Non-steroidal anti-inflammatory drug

*Combined cyclooxygenase/lipoxygenase (COX/LOX) inhibitor

Statistical Analysis

Due to the limited number, low level of evidence, and heterogeneity of clinical and in vitro studies, outcomes of interest were described with a qualitative synthesis. Similarly, a high degree of heterogeneity and inconsistency of outcomes reporting was noted in the included animal studies. However, a small subset of animal studies, specifically studies investigating the effect of a variety of NSAIDs on the load to failure of acute rotator cuff repair in an established rat model, were homogeneous in nature and meta-analysis was performed using a random effects model with inverse variance weighting. Standardized mean difference was used as the summary statistic to account for measurement and unit variability between studies. For the meta-analysis, if a study included multiple NSAID treatment arms, arms were combined to allow a single pairwise comparison to the control arm using weighted means and standard deviations, thus preventing a unit-of-analysis error.³⁷ Additionally, only treatment arms that initiated NSAID therapy immediately postoperatively (postoperative day zero or one) were included in the meta-analysis. If a study evaluated load to failure at multiple timepoints, only the latest (farthest from surgery) timepoint was included (range, 28-56 days). Separate meta-analyses were performed to include all NSAIDs (selective and non-selective) as well as only non-selective NSAIDs. Due to insufficient studies, a separate analysis was not performed for only selective NSAIDs. Descriptive statistics were prepared using IBM SPSS for Windows, version 25 (IBM Corp., Armonk, N.Y., U.S.A) and meta-analysis performed using Review Manager version 5.3 (The Cochrane Collaboration, Copenhagen, Denmark). Statistical comparisons with p-values <0.05 were considered statistically significant.

Results

Search Strategy

The initial search yielded 401 studies, with 126 studies immediately removed as duplicates, leaving 275 studies available for review (Figure 1). No studies were identified from other sources. After systematic screening, a total of 13 studies, including three clinical^22,44,45 and 10 basic science studies,^{21,23,28,31,46-51} met the inclusion and exclusion criteria.

Figure 1.

PRISMA flowchart for systematic study inclusion.

Clinical Studies

Study Characteristics

The effect of NSAIDs on rotator cuff repair was reported in a single study²² while distal biceps repair was assessed in two studies.^44,45 Both celecoxib and ibuprofen were evaluated in separate treatment arms in the rotator cuff repair study while both distal biceps repair studies used indomethacin 75 mg daily as the postoperative NSAID for heterotopic ossification prophylaxis (Table 2). All clinical studies initiated NSAIDs immediately postoperatively, continuing for 14 days postoperatively following rotator cuff repair and ranging from 10-42 days following distal biceps repair. In the rotator cuff repair study by Oh et al., tendon-to-bone healing was a secondary outcome, and only 82/180 patients (45.6%) were assessed for healing using magnetic resonance imaging (MRI) or ultrasound at 24 ± 2 months postoperatively. Further demographic data for the subset of patients undergoing healing assessment was not reported.²² The mean age of patients in the included distal biceps repair studies was 45.6⁴⁴ and 50.3⁴⁵ years, with follow-up duration described in a single study (range, 6-102 months).⁴⁵

Table 2.

Clinical Study Characteristics

Author (Year)	Study Design (LOE)	Intervention	Sample Size	Age ± SD	NSAID	Dose	Initiation	Duration (days)*	Follow-up (mths)	Repair Assessment	Failures†
Anakwenze (2011)	Case Series (IV)	Distal Biceps Repair	34	45.6	Indomethacin	75 mg QD	POD 0	42	6-102‡	Clinical	0 (0%)
Costopoulos (2017)	Retrospective Comparative (III)	Distal Biceps Repair	104	50.3 ± 20.0	Indomethacin	75 mg QD	POD 0	10-42	NA	Clinical	0 (0%)
	Randomized	Rotator	30	NA	Celecoxib	200 mg BID	POD 1	14	24 ± 2!!	MRI/US	11 (36.7%)
Oh (2018)	Controlled Trial (I)	Cuff Repair	27	NA	Ibuprofen	385 mg TID	POD 1	14		MRI/US	2 (7.4%)

LOE, Level of Evidence; SD, Standard Deviation; NSAID, Non-steroidal Anti-inflammatory Drug; mths, Months; QD, Daily; BID, Twice Daily; TID, Three Times Daily; POD, Postoperative Day; NA, Not Available; MRI, Magnetic Resonance Imaging; US, Ultrasound

*Duration of NSAID administration; †Listed as number (percent); ‡Listed as range; !!Listed as mean ± standard deviation

Quality Assessment of Clinical Studies

The rotator cuff repair study by Oh et al. was the only included randomized controlled trial (level I evidence) for the primary outcome of early postoperative pain. However, methodological quality was poor for the secondary outcome of tendon-to-bone healing given incomplete reporting of the secondary outcome and high risk of bias with selective reporting of the outcome (Figure 2). Scores for risk of bias assessment for non-randomized studies measured using the MINORS scoring system were 8 and 16 for the non-comparative⁴⁴ and comparative⁴⁵ studies investigating distal biceps repiar, representing 50.0% and 66.7% of the total possible MINORS score, respectively. The study by Costopoulos et al. was originally designed as a retrospective comparative study (level III evidence)⁴⁵ for the primary endpoint, development of heterotopic ossification, while the study by Anakwenze et al. was a case series (level IV evidence).⁴⁴

Figure 2.

Risk of bias assessment for a) randomized controlled clinical trials and b) animal basic science studies c) The proportion of animal basic science studies with various forms of bias.

Clinical Study Outcomes

The method of assessing tendon-to-bone healing in the rotator cuff repair study was a combination of MRI and ultrasound. Overall, 2/27 patients (7.4%) receiving ibuprofen postoperatively and 11/30 patients (36.7%) receiving celecoxib postoperatively experienced clinical failures at the tendon-to-bone interface. The rate of failure was significantly higher with postoperative celecoxib use compared with ibuprofen and the control arm of patients receiving tramadol. The method of assessing tendon-to-bone healing in both distal biceps repair studies was limited to a clinical assessment. In aggregate, none of the 138 distal biceps patients receiving NSAIDs postoperatively experienced a clinical failure, nor did any of the eight patients in the control arm of the study by Costopoulos et al.⁴⁵

Basic Science Studies

Study Characteristics

Of the 10 included basic science studies, nine were categorized as animal studies^{21,23,28,31,46-49, 51} and one as an in vitro study⁵⁰ (Table 3). Within the animal studies, acute rat rotator cuff repair was used as the model in six studies,^{21,23,28,31,46,51} acute rabbit rotator cuff repair in a single study⁴⁸, chronic rat rotator cuff repair in a single study,⁴⁹ and acute rat patellar tendon repair in a single study.⁴⁷ All basic science studies included a control arm, six included a single NSAID treatment arm,^{21,28,31,46,49,51} and four included multiple NSAID treatment arms.^23,47,48,50 Overall, 11 different NSAIDs were investigated, with celecoxib (4/10 studies)^21,23,47,48 and ibuprofen (3/10 studies)^28,47,48 the most frequently studied NSAIDs. All animal studies included treatment arms with immediate initiation of NSAIDs postoperatively (postoperative day zero or one), while two animal studies additionally included treatment arms that delayed NSAID treatment until postoperative day eight²⁸ and 11.³¹ The duration of NSAID therapy within the studies ranged from 7-42 days. All animal studies performed biomechanical assessment of healing, three studies performed biochemical assessment of healing,^46,47,49 and six performed a histologic assessment.^{23,28,31,47-49} Within the animal studies, the final assessment of healing was performed 14-84 days after surgery, with four studies investigating multiple timepoints.^23,28,46,48 In the in vitro experiment, a biochemical assessment of healing was reported using a variety of assays.⁵⁰

Table 3.

Basic Science Study Characteristics

Animal Studies
Author (Year)	Model	NSAID	Initiation	NSAID Duration (days)	Euthanasia (days)
Cabuk (2014)	Acute Rat RCR	Diclofenac	POD 0	Euthanasia*	7; 21; 42†
Chechik (2014)	Acute Rat RCR	Meloxicam	POD 0	10	21
			POD 11	10	21
Cohen (2006)	Acute Rat RCR	Celecoxib	POD 0	14	14; 28; 56†
		Indomethacin	POD 0	14	14; 28; 56†
Connizzo (2014)	Acute Rat RCR	Ibuprofen	POD 0	7	7; 14; 28†
			POD 8	7	14; 28†
Dolkart (2014)	Acute Rat RCR	Celecoxib	POD 1	20	21
Ferry (2007)	Acute Rat PTR	Celecoxib	POD 0	14	14
		Ibuprofen	POD 0	14	14
		Naproxen	POD 0	14	14
		Piroxicam	POD 0	14	14
		Valdecoxib	POD 0	14	14
Lu (2015)	Acute Rabbit RCR	Celecoxib	POD 0	7	21; 42; 84†
		Flurbiprofen	POD 0	7	21; 42; 84†
		Ibuprofen	POD 0	7	21; 42; 84†
Meadows (2017)	Acute Rat RCR	Indomethacin	POD 0	14	21
Oak (2011)	Chronic Rat RCR	Licofelone	POD 0	14	14
In Vitro Studies
Schwarting (2015)	Co-Culture	Ibuprofen	Immediate	NA	NA
		Parecoxib	Immediate	NA	NA

NSAID, Non-steroidal anti-inflammatory drug; RCR, Rotator cuff prepair; PTR, Patellar tendon repair;

POD, Postoperative day; NA, Not available

*NSAID administration continued until euthanasia timepoint(s); †Multiple timepoints assessed

Quality Assessment of Animal Studies

All studies had components at high risk or unclear risk of bias using the SYRCLE risk of bias assessment tool (Figure 3). Incomplete reporting of data, selective reporting of data, and other types of bias were consistently at lowest risk of bias, while random sequence allocation, blinding of intervention to investigators, and random housing assignments were most consistently at unclear or high risk of bias.

Figure 3.

Forest plot comparing postoperative administration of non-steroidal anti-inflammatory drugs (NSAIDs) with controls for homogeneous studies (acute rat rotator cuff repair) using a) any NSAID and b) non-selective NSAIDs.

Animal Study Outcomes

Biomechanical outcomes were most consistently reported, with load to failure, stiffness, cross-sectional area, and maximum stress reported by at least three studies. Comparative results of the studies, which included 18 total treatment arms, are listed in Table 4.^{21,23,28,31,46-49, 51} Of these studies, four studies (8/18 treatment arms)^23,31,47,48 reported a decrease in load to failure with NSAID administration when compared with the control arm at the latest follow-up. Delayed NSAID administration (postoperative day 11) resulted in decreased load to failure in a single study when compared with the control arm,³¹ while another study investigating delayed NSAID administration revealed no effect on load to failure with delayed administration.²⁸ Of the 18 treatment arms, 13 investigated non-selective NSAIDs,^{23,28,31,46-49, 51} with 4/13 (30.8%) demonstrating a decrease in load to failure compared with control arms.^23,31,47,48 Five treatment arms investigated selective NSAIDs,^21,23,47,48 with 4/5 (80.0%) demonstrating a decrease in load to failure.^23,47,48 In the subgroup meta-analysis,^23,28,46 load to failure was not significantly affected when comparing the use of all NSAIDs (Standard Mean Difference [SMD] -1.05, 95% Confidence Interval [CI] -2.39-0.30, p=0.13) or non-selective NSAIDs (SMD -0.62, 95% CI -1.26-0.02, p=0.06) with controls (Figure 3). Of the eight studies reporting results on stiffness,^{21,23,28,31,46,47,49,51} including 15 treatment arms, stiffness was significantly reduced in 2/15 treatment arms (13.3%), both of which included immediate administration of non-selective NSAIDs. Tendon cross-sectional area was decreased following NSAID administration in 1/10 treatment arms (10%) in the four studies reporting data,^23,28,47,49 while maximum stress was decreased following NSAID administration in 2/11 treatment arms (18.2) in the four studies reporting data.^21,28,47,49 Histologic outcomes were reported in six studies.^{23,28,31,47-49} Collagen orientation was reported in four studies, including 11 treatment arms,^23,28,31,47 and was the only histologic metric to warrant qualitative synthesis. At the latest assessed timepoint, NSAID administration resulted in significantly poorer collagen orientation in 3/11 treatment arms (27.3%) when compared with control arms. Biochemical outcomes were only reported in three studies,^46,47,49 and there were no consistently reported biochemical metrics between studies to warrant qualitative synthesis.

Table 4.

Comparison of Relevant Treatment Outcomes to Controls in Animal Studies

Author (year)	NSAID	Load to Failure	Stiffness	Cross-Sectional Area	Maximum Stress	Collagen Orientation
Cabuk (2014)	Diclofenac	↔	↓
Chechik (2014)	Meloxicam - T0	↔	↔			↔
	Meloxicam - T1	↓	↔			↓
Cohen (2006)	Celecoxib	↓	↔	↔		↓
	Indomethacin	↓	↔	↓		↓
Connizzo (2014)	Ibuprofen - T0	↔	↓	↔	↔	↔
	Ibuprofen - T1	↔	↔	↔	↔	↔
Dolkart (2014)	Celecoxib	↔	↔		↔
Ferry (2007)	Celecoxib	↓	↔	↔	↓	↔
	Ibuprofen	↔	↔	↔	↔	↔
	Naproxen	↔	↔	↔	↔	↔
	Piroxicam	↓	↔	↔	↓	↔
	Valdecoxib	↓	↔	↔	↓	↔
Lu (2015)	Celecoxib	↓
	Flurbiprofen	↓
	Ibuprofen	↔
Meadows (2017)	Indomethacin	↔	↔
Oak (2011)	Licofelone	↑	↑	↔	↑

NSAID = Non-steroidal anti-inflammatory drug; T0 = immediate administration; T1 = delayed administration

↔ = statistical equivalence to control; ↓ = statistical decrease compared with control; ↑ = statistical increase compared with control

In vitro Study Outcomes

A single in vitro study reported biochemical, cell viability, and cell migration results in a tendon-bone co-culture model using mouse MC3T3-E1 pre-osteoblasts and 3T3 fibroblast cell lines. Both ibuprofen and parecoxib were investigated. Ibuprofen led to reduced expression of bone gamma-carboxyglutamate protein (Bglap) within the osteoblasts and reduced expression of runt-related transcription factor (Runx2) within the osteoblast, fibroblast, and interface regions. Parecoxib similarly reduced Runx2 expression in all three regions while also reducing alkaline phosphatase (Alp1) and fibromodulin (Fmod) expression in all three regions. Ibuprofen administration resulted in a dose-dependent decrease in viability of osteoblasts, while parecoxib decreased viability in fibroblasts in a non-dose-dependent fashion. Both ibuprofen and parecoxib administration led to increased cell migration from the osteoblast region into the interface region compared with control cultures.

Discussion

Orthopaedic procedures to re-establish the tendon-to-bone interface are some of the most frequently performed ambulatory surgery procedures in the United States.^52,53 Thus, optimizing the outcomes of these procedures would serve to positively affect a large patient population. As highlighted by this review, the current literature on the effect of NSAIDs on tendon-to-bone healing is limited, not only when measured by the number of available publications, but also when considering the methodological quality of the available clinical and basic science studies. Furthermore, translating the more readily available basic science studies, most of which utilize an acute rotator cuff repair model, to clinical practice poses several challenges that must be considered before drawing strong conclusions as to the effect of NSAIDs on tendon-to-bone healing.

The dilemma to use or prohibit NSAIDs following surgical repair of the tendon-bone interface may best be illustrated by considering rotator cuff repair. Due to the relatively high failure rate following repair of large and massive rotator cuff tears,^54-56 there has been a great deal of focus on optimizing the biomechanical and biologic environment to allow restoration of the unique properties of the enthesis. Despite the lack of evidence, this concern may cause some providers to avoid the use of NSAIDs following procedures such as rotator cuff repair that are already at a high risk of structural failure. This approach is at odds with the potential pain-relieving benefits of NSAIDs, particularly in an era where multimodal pain management strategies are being encouraged due to the opioid epidemic.^35,36 Therefore, the need for further high-quality clinical studies is necessary. At the time of this systematic review, a single randomized controlled trial investigating the effect of NSAIDs on rotator cuff repair integrity as a secondary outcome was available. The study found that the use of a non-selective COX-2 inhibitor, celecoxib, led to significantly more failures at the tendon-to-bone interface compared with ibuprofen or the control arm receiving tramadol. While the study was appropriately randomized for the primary outcome and blinded, <50% of patients were available for the repair integrity outcome, thus raising concerns with incomplete reporting and selection bias. The other two included clinical studies, consisting of level III and IV evidence, evaluated distal biceps repair, which have an inherently low failure rate when compared with rotator cuff repair, making it difficult to delineate clinical factors that may lead to failure given the low event rate.^57-59 While several factors pose challenges when designing clinical studies, including choosing the optimal method to assess or define clinical failures,⁶⁰ future randomized controlled studies could serve to significantly impact practice given the high rate at which these procedures are performed as well as the relatively widespread use of NSAIDs for the treatment of musculoskeletal pain.

Compared with clinical studies, basic science studies often provide logistical advantages, including ease of randomization and inclusion of an adequate number of specimens. However, translation of findings from basic science models into clinical practice can create challenges. For example, in the included basic science studies for this review, all but one study used an acute tendon-bone repair model. Particularly when considering rotator cuff tears, this is rarely the clinical scenario. Additionally, the clinical applicability and relevance of testing methods and results, such as biomechanical testing, often are not feasible in the clinical setting, and the clinical significance of the reported results may be limited. While most of those limitations will remain inherent to the majority of basic science studies, this review highlights several modifiable areas for improvement. First, while methodological quality is often only considered in clinical trials, the methodological quality of basic science studies focusing on the effects of NSAIDs on tendon-to-bone healing is poor. In the future, making efforts to appropriately randomize specimens while blinding animal caregivers and outcome assessors would serve to significantly improve the methodological quality of basic science research on the subject. Second, the heterogeneity of outcomes reporting, including biomechanical, biochemical, and histologic outcomes, and the variability of postoperative NSAID initiation and duration, limits aggregate analysis of a homogeneous cohort. The findings from this study can hopefully serve to provide guidance for future basic science studies, providing details on reporting of relevant outcomes and areas where future research efforts, such as the temporal effect of NSAID administration, can be focused.

The current study does have several limitations, including the quality of studies available in the current literature. At the time of this review, only three clinical studies, consisting of level I, III, and IV evidence, were available for review.^22,44,45 Basic science studies, comprised primarily of animal studies, were of low methodological quality and at high risk of bias using the SYRCLE risk of bias assessment tool, due in part to inadequate randomization and lack of blinding of animal caretakers and investigators. Additionally, the ability to translate individual findings within basic science studies to clinical practice remains limited. Despite the number of basic science studies available for review, many of which included multiple treatment arms, the heterogeneity of outcome reporting limited a high-level aggregate analysis. While three animal studies did utilize homogenous methodology and were subsequently selected for a subgroup meta-analysis, the conclusions from the analysis are limited by the small number of included studies, the unknown clinical significance of load to failure as an outcome, and the variable timepoints at which load to failure was measured postoperatively, although standardized measures were used. Despite these limitations, it is important to understand the current state of the literature in order to improve study design moving forward.

Conclusions

The current literature does not provide sufficient evidence for or against the use of NSAIDs following acute injury or surgical repair of the tendon-bone interface. Due to the low methodological quality of currently available literature, future high-quality studies on the effect of NSAIDs on tendon-to-bone healing are warranted while addressing the limitations identified in the current review in order to guide both clinicians and patients moving forward.

APPENDIX 1: Search Strategy Report

Date: Database Inception – May 14, 2018

Database: PubMed (MEDLINE)

Set #		Results
1	“Anti-Inflammatory Agents, Non-Steroidal” [Pharmacological Action] OR “Cyclooxygenase Inhibitors” [Pharmacological Action] OR “Nonsteroidal anti-inflammatory drugs”[tiab] OR “Nonsteroidal anti-inflammatory drug”[tiab] OR “non-steroidal anti-inflammatory drugs”[tiab] OR “non-steroidal anti-inflammatory drug”[tiab] OR NSAIDs[tiab] OR NSAID[tiab] OR “Anti-Inflammatory Agents, Non-Steroidal”[mesh] OR “cyclooxygenase inhibitors”[tiab] OR “cyclooxygenase inhibitor”[tiab] OR “Cyclooxygenase Inhibitors/adverse effects”[mesh] OR “cox inhibitors”[tiab] OR “cox inhibitor”[tiab] OR cyclooxygenase[tiab] OR “cox-1”[tiab] OR “cox-2”[tiab] OR “cox1”[tiab] OR “cox2”[tiab]	259,439
2	“Wound Healing”[Mesh] OR “Wound Healing/drug effects”[Mesh] OR healing[tiab] OR heal[tiab] OR repair[tiab] OR repairing[tiab] OR remodeling[tiab] OR remodel[tiab]	553,302
3	“Tendons”[Mesh] OR Tendon[tiab] OR tendons[tiab] OR “Tendon Injuries”[Mesh] OR “patellar ligament”[MeSH Terms] OR “ligament”[tiab] OR “ligaments”[tiab] OR “patellar ligament”[tiab] OR “rotator cuff”[MeSH Terms] OR “rotator cuff”[tiab] OR “osteotendinous junction”[tiab] OR “bone-tendon”[tiab] OR “tendon-to-bone”[tiab]	130,443
4	#1 AND #2 AND #3	273
5	#4 NOT (Editorial[ptyp] OR Letter[ptyp] OR Case Reports[ptyp] OR Comment[ptyp])	258
6	#5 AND English[lang]	238
7	#6 AND 2017/10/02:2018/05/14[edat]	16

Database: SportDiscus

Set #		Results
1	DE “NONSTEROIDAL anti-inflammatory agents” OR TI (“Nonsteroidal anti-inflammatory drugs” OR “Nonsteroidal anti-inflammatory drug” OR “non-steroidal anti-inflammatory drugs” OR “non-steroidal anti-inflammatory drug” OR NSAIDs OR NSAID OR “cyclooxygenase inhibitors” OR “cyclooxygenase inhibitor” OR “cox inhibitors” OR “cox inhibitor” OR cyclooxygenase OR “cox-1” OR “cox-2” OR cox1 OR cox2) OR AB (“Nonsteroidal anti-inflammatory drugs” OR “Nonsteroidal anti-inflammatory drug” OR “non-steroidal anti-inflammatory drugs” OR “non-steroidal anti-inflammatory drug” OR NSAIDs OR NSAID OR “cyclooxygenase inhibitors” OR “cyclooxygenase inhibitor” OR “cox inhibitors” OR “cox inhibitor” OR cyclooxygenase OR “cox-1” OR “cox-2” OR cox1 OR cox2) OR KW (“Nonsteroidal anti-inflammatory drugs” OR “Nonsteroidal anti-inflammatory drug” OR “non-steroidal anti-inflammatory drugs” OR “non-steroidal anti-inflammatory drug” OR NSAIDs OR NSAID OR “cyclooxygenase inhibitors” OR “cyclooxygenase inhibitor” OR “cox inhibitors” OR “cox inhibitor” OR cyclooxygenase OR “cox-1” OR “cox-2” OR cox1 OR cox2)	1,923
2	DE “WOUND healing” OR TI (healing OR heal OR repair OR repairing OR remodeling OR remodel) OR AB (healing OR heal OR repair OR repairing OR remodeling OR remodel) OR KW (healing OR heal OR repair OR repairing OR remodeling OR remodel)	18,142
3	(((DE “TENDONS -- Wounds & injuries” OR DE “ACHILLES tendon -- Wounds & injuries”) OR (DE “TENDONS”)) OR (DE “LIGAMENTS -- Wounds & injuries” OR DE “COLLATERAL ligament -- Wounds & injuries” OR DE “CRUCIATE ligaments -- Wounds & injuries” OR DE “PATELLAR ligament -- Wounds & injuries”) OR (DE “LIGAMENTS”) OR (DE “SHOULDER joint -- Rotator cuff” OR DE “SHOULDER joint -- Rotator cuff -- Wounds & injuries”)) OR TI (tendon OR tendons OR ligament OR ligaments OR patellar ligament OR rotator cuff OR osteotendinous junction OR bone-tendon OR tendon-to-bone) OR AB (tendon OR tendons OR ligament OR ligaments OR patellar ligament OR rotator cuff OR osteotendinous junction OR bone-tendon OR tendon-to-bone) OR KW (tendon OR tendons OR ligament OR ligaments OR patellar ligament OR rotator cuff OR osteotendinous junction OR bone-tendon OR tendon-to-bone)	26,453
4	#1 AND #2 AND #3	54
5	#4 AND Peer Reviewed; Language: English	47
6	#5 AND

Database: Embase

Set #		Results
1	‘antiinflammatory agent’/exp OR ‘cyclooxygenase 2 inhibitor’/exp OR Nonsteroidal anti-inflammatory drugs:ab,ti,kw OR Nonsteroidal anti-inflammatory drug:ab,ti,kw OR non-steroidal anti-inflammatory drugs:ab,ti,kw OR non-steroidal anti-inflammatory drug:ab,ti,kw OR NSAIDs:ab,ti,kw OR NSAID:ab,ti,kw OR cyclooxygenase inhibitors:ab,ti,kw OR cyclooxygenase inhibitor:ab,ti,kw OR cox inhibitors:ab,ti,kw OR cox inhibitor:ab,ti,kw OR cyclooxygenase:ab,ti,kw OR “cox-1”:ab,ti,kw OR “cox-2”:ab,ti,kw OR cox1:ab,ti,kw OR cox2:ab,ti,kw	70,179
2	‘wound healing’/exp OR healing:ab,ti,kw OR heal:ab,ti,kw OR repair:ab,ti,kw OR repairing:ab,ti,kw OR remodeling:ab,ti,kw OR remodel:ab,ti,kw	712,235
3	‘tendon’/exp OR Tendon:ab,ti,kw OR tendons:ab,ti,kw OR ‘tendon injury’/exp OR “ligament”:ab,ti,kw OR “ligaments”:ab,ti,kw OR “patellar ligament”:ab,ti,kw OR ‘rotator cuff’/exp OR ‘rotator cuff injury’/exp OR “rotator cuff”:ab,ti,kw OR “osteotendinous junction”:ab,ti,kw OR “bone-tendon”:ab,ti,kw OR “tendon-to-bone”:ab,ti,kw	162,782
4	#1 AND #2 AND #3	87
5	#4 NOT (‘case report’/exp OR ‘case study’/exp OR ‘editorial’/exp OR ‘letter’/exp OR ‘note’/exp)	87
6	#5 AND [English]/lim	86
7	#6 AND Published Date: 20171001-	2

Database: CINAHL

Set #		Results
1	MH “Antiinflammatory Agents, Non-Steroidal+” OR TI (“Nonsteroidal anti-inflammatory drugs” OR “Nonsteroidal anti-inflammatory drug” OR “non-steroidal anti-inflammatory drugs” OR “non-steroidal anti-inflammatory drug” OR NSAIDs OR NSAID OR “cyclooxygenase inhibitors” OR “cyclooxygenase inhibitor” OR “cox inhibitors” OR “cox inhibitor” OR cyclooxygenase OR “cox-1” OR “cox-2” OR “cox1” OR “cox2”) OR AB (“Nonsteroidal anti-inflammatory drugs” OR “Nonsteroidal anti-inflammatory drug” OR “non-steroidal anti-inflammatory drugs” OR “non-steroidal anti-inflammatory drug” OR NSAIDs OR NSAID OR “cyclooxygenase inhibitors” OR “cyclooxygenase inhibitor” OR “cox inhibitors” OR “cox inhibitor” OR cyclooxygenase OR “cox-1” OR “cox-2” OR “cox1” OR “cox2”)	31,039
2	MH “Wound Healing+” OR TI (healing OR heal OR repair OR repairing OR remodeling OR remodel) OR AB (healing OR heal OR repair OR repairing OR remodeling OR remodel)	89,109
3	MH “Tendons+” OR Tendon OR tendons OR MH “Tendon Injuries+” OR MH “Ligaments+” OR TI (“ligament” OR “ligaments” OR “patellar ligament” OR “rotator cuff” OR “osteotendinous junction” OR “bone-tendon” OR “tendon-to-bone”) OR AB (“ligament” OR “ligaments” OR “patellar ligament” OR “rotator cuff” OR “osteotendinous junction” OR “bone-tendon” OR “tendon-to-bone”)	36,096
4	#1 AND #2 AND #3	70
5	#1 AND #2 AND #3 AND Limiters - Research Article; English Language	34
6	#5 AND Published Date: 20171001-	1