Prior Intra-articular Corticosteroid Injection Within 3 Months May Increase the Risk of Deep Infection in Subsequent Joint Arthroplasty: A Meta-analysis

Abstract

Background

Intra-articular injections containing a corticosteroid are used frequently, and periprosthetic joint infection is a serious complication after total joint arthroplasty. There is debate regarding whether intra-articular corticosteroid injections before arthroplasty increase periprosthetic joint infection after surgery.

Questions/purposes

(1) Does a previous intra-articular corticosteroid injection increase the odds of infection after subsequent hip or knee arthroplasty? (2) Does this risk vary based on how soon before the arthroplasty (such as less than 3 months before surgery) the injection is administered?

Methods

Using the PubMed, Embase, Cochrane Library, and Web of Science databases from inception to July 2021, we searched for comparative studies in English on patients who received intra-articular corticosteroid injections before arthroplasty and that tracked the frequency of infection after arthroplasty. We extracted data on the risk of infection after subsequent joint arthroplasty. The keywords included “corticosteroid,” “steroid,” “arthroplasty,” “knee replacement,” and “hip replacement.” Eleven retrospective, comparative studies from four countries were included, of which 10 reported the specific diagnosis criteria and one did not. These articles included data on 173,465 arthroplasties in the hip or knee, as well as of 73,049 injections and 100,416 control patients. The methodologic quality of the included studies was evaluated according to the Newcastle-Ottawa Quality Assessment Scale; the articles’ scores ranged from 6 to 7 (the score itself spans 0 to 9, with higher scores representing better study quality). We found no evidence of publication bias based on the Egger test, and tests of heterogeneity generally found heterogeneity, so a random-effects model was used of our meta-analyses. A meta-analysis was performed with Review Manager 5.3 software and Stata version 12.0 software.

Results

Overall, there were no differences in the odds of periprosthetic joint infection between the injection group and the control group among patients who received any kind of injection (odds ratio 1.22 [95% CI 0.95 to 1.58]; p = 0.12). However, in a subgroup analysis, there was a higher OR for postoperative PJI in patients with an intra-articular corticosteroid injection in the knee or hip within 3 months (OR 1.39 [95% CI 1.04 to 1.87]; p = 0.03). There were no differences in the infection risk in patients who had injections between 3 and 6 months before arthroplasty (OR 1.19 [95% CI 0.95 to 1.48]; p = 0.13) or between 6 and 12 months before arthroplasty.

Conclusion

The current evidence suggests ipsilateral intra-articular corticosteroid injections within 3 months before arthroplasty were associated with an increased risk of periprosthetic joint infection during subsequent joint arthroplasty. We recommend against performing total joint arthroplasty on a patient who has received an intra-articular corticosteroid injection within 3 months. Further high-quality studies on this topic from registries, national databases, or insurance company data are still required to confirm and extend our findings.

Level of Evidence

Level III, therapeutic study.

Introduction

Osteoarthritis is a chronic and debilitating disease, and a major cause of joint pain and functional disability worldwide [4, 33]. Total joint arthroplasty is an effective treatment for selected patients with osteoarthritis [42], but before surgery, many patients try nonsurgical approaches, such as intra-articular injections of corticosteroids, hyaluronic acid, platelet-rich plasma, or stem cells [8, 38, 44]. There is limited evidence as to the efficacy of these treatments [1, 6, 17, 22, 29, 47], and the clinical practice guidelines of the American Academy of Orthopaedic Surgeons has downgraded its recommendation regarding corticosteroid injections (to evidence only of moderate strength on this point) and recommends against the use of intra-articular hyaluronic acid injections for patients with symptomatic knee osteoarthritis [2].

Periprosthetic joint infection (PJI) is a serious complication and a major challenge in total joint arthroplasty (TJA). PJI has been reported to be the most common indication for revision TKA and the third most common reason for revision THA [23, 24]. The incidence of PJI in knee and hip arthroplasty has been reported to be 0.5% to 2.83% [15, 27, 40], and its increasing cost represents a tremendous economic burden both for patients and healthcare systems [24, 31]. There is conflicting evidence as to whether injections of corticosteroids before arthroplasty increase the risk of PJI after surgery [26, 37] or do not increase this risk [13, 19, 20, 21, 26]. The results of previous systematic reviews and meta-analyses also have been inconsistent on this important point [11, 45, 48]. In addition, the difference in PJI risk as a function of time between injection and arthroplasty was not conducted in previous systematic reviews and meta-analyses. Although there has been some emerging evidence from high-volume database studies that suggest injections before arthroplasty are associated with increased PJI risk [7, 9, 16, 40], the timing evaluated in those studies has varied from 4 weeks to 6 months.

We therefore performed a systematic review and meta-analysis of comparative studies in which we asked: (1) Does a previous intra-articular corticosteroid injection increase the odds of infection after subsequent hip or knee arthroplasty? (2) Does this risk vary based on how soon before the arthroplasty (such as less than 3 months before surgery) the injection is administered?

Materials and Methods

Search Strategy and Selection Criteria

We systematically searched the PubMed, Embase, Web of Science, and Cochrane Library electronic databases to retrieve eligible articles from the inception of the databases until July 2021. We used the following terms: “corticosteroid,” “steroid,” “arthroplasty,” “knee replacement,” and “hip replacement” using the Boolean operators “AND” and “OR” by title. We hand-searched the references from included articles to ensure no relevant articles were missed. The current work was performed according to the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) [36].

The inclusion criteria were as follows: original articles, studies comparing patients who received intra-articular corticosteroid injections in the hip or knee before arthroplasty with patients who did not receive injections, full-text articles in the English language, and studies reporting clinical outcomes during follow-up that included PJI. We excluded cases reports, abstracts, letters, review articles, grey literature (unpublished studies, articles posted to preprint servers), studies evaluating a combination of corticosteroid and hyaluronic acid, those evaluating injections of other medication types, studies of intravenous injections, studies of people with chronic steroid use, studies evaluating injections given perioperatively, studies with duplicate or overlapping data, studies whose data could not be converted and merged, and animal experiments.

Data Extraction

The articles’ titles and abstracts were reviewed to verify their relevance to PJI risk after arthroplasty in patients who received intra-articular corticosteroid injections before arthroplasty compared with patients who did not receive any injections. Studies that did not meet our criteria were excluded and potentially eligible articles were independently selected by two investigators (QL, KC). For the chosen articles, the full texts were read carefully and subsequently analyzed.

Next, general data were extracted from the included articles, including the first author, year of publication, injection types, number of hips or knees, the time from the injection to arthroplasty, follow-up duration, postoperative PJI, and diagnostic criteria of PJI. When any disagreement occurred, a third reviewer (TL) helped to reach a consensus.

Search Results

In the initial search by title, we obtained 408 results and 176 articles remained after we eliminated duplicate studies. Subsequently, after reading the titles and abstracts, we discarded 152 studies that were not relevant to our topic, and 24 studies remained. Finally, after reading the full text of these articles, we excluded 13 that did not meet the inclusion criteria. Ultimately, 11 studies with a total of 173,465 hip or knee arthroplasties were included in our meta-analysis [9, 13, 14, 16, 25, 26, 32, 34, 35, 37, 43] (Fig. 1).

Fig. 1.

This flowchart shows the selection of studies.

Characteristics of the Included Studies

The 11 included studies all were retrospective comparative studies that were published between 2005 and 2021 (Table 1). Five were conducted in the United States, three in the United Kingdom, two in Canada, and one in Belgium. Six articles evaluated hip arthroplasty and five evaluated knee arthroplasty. Among the selected articles, the sample sizes varied from 80 to 77,249 knees or hips. The proportion of patients experiencing PJI varied from 0% to 10% in the injection group and from 0% to 2.35% in the control group (Table 2). The overall infection proportion was 1.36% (2362 of 173,465), of which 1.08% (1089 of 100,416) occurred in the control group and 1.74% (1273 of 73,049) occurred in the injection group.

Table 1.

Characteristics of the 11 included retrospective comparative studies

Study	Year	Country	Prosthesis	Injection	Number of prostheses		NOS score
					Injections	Controls
Kaspar and de Beer [25]	2005	Canada	Hip	CS	40	40	7
McIntosh et al. [34]	2006	United States	Hip	CS	224	224	6
Papavasiliou et al. [37]	2006	United Kingdom	Knee	CS	54	90	6
Sreekumar et al. [43]	2007	United Kingdom	Hip	CS	68	136	7
Desai et al. [14]	2009	United Kingdom	Knee	CS	90	180	7
Meermans et al. [35]	2012	Belgium	Hip	CS	182	182	7
Croft and Rockwood [13]	2013	Canada	Hip	CS	48	48	7
Cancienne et al. [9]	2015	United States	Knee	CS	22,240	13,650	7
Khanuja et al. [26]	2016	United States	Knee	CS	302	302	7
Kurtz et al. [32] a	2021	Unites States	Knee	CS	20,743	56,506	7
Forlenza et al. [16]	2021	United States	Hip	CS	29,058	29,058	7

^aPublished online ahead of print; CS = corticosteroid; NOS = Newcastle-Ottawa Quality Assessment Scale (0-9 points); for the scale, the higher the score, the better the quality of the study.

Table 2.

Injection and infection in the included studies

Study	Time from injection to arthroplasty	Follow-up	Deep infection		Diagnostic criteria
			Injection	Control
Kaspar and de Beer [25]	0.5-29.0 months	9.9-86.2 months	10% (4 of 40)	0% (0 of 40)	Revision for deep hip infection
McIntosh et al. [34]	112 ± 81 days; within 1 year	Mean 2.6 years	1.34% (3 of 224)	0.45% (1 of 224)	No report
Papavasiliou et al. [37]	11 months	1 year	5.56% (3 of 54)	0% (0 of 90)	Deep infection involving the deep tissues, occurring within 1 year of surgery, and met at least one of the following: (1) purulent drainage; (2) microbiological culture from aseptically-aspirated fluid, a swab, or a tissue biopsy from the deep-tissue layers or pus cells on microscopy; (3) deep incision which spontaneously dehisced or was explored for a temperature above 38°C, localized pain or tenderness; (4) an abscess or other evidence of infection involving the deep incision; (5) diagnosis of a deep infection by an attending clinician
Sreekumar et al. [43]	Mean 14 months	Mean 23.3 months	0% (0 of 68)	0.74% (1 of 136)	No report, but one case of deep infection had a hip aspiration positive for Staphylococcus aureus
Desai et al. [14]	45 patients within 12 months	1 to 6 years	0% (0 of 90)	0% (0 of 180)	Defined as positive swab cultures or deep tissue biopsy; exploration and wash out of the wound with positive culture; revision surgery for infection
Meermans et al. [35]	Mean 155 days; within 1 year	12 to 131 months	0.55% (1 of 182)	0.55% (1 of 182)	Defined as (1) a sinus tract communicating with the implant; (2) the identical pathogen isolated from two or more separate tissue samples; or (3) the presence of purulence in the joint
Croft and Rockwood [13]	Mean 5.9 months	1.4 to 54.1 months	0% (0 of 48)	0% (0 of 48)	Deep infection requiring revision surgery
Cancienne et al. [9]	0 to 3 months; 3 to 6 months; 6 to 12 months	6 months	2.61% (581 of 22,240)	2.35% (319 of 13,650)	Postoperative PJI was characterized by either a diagnosis or procedure for wound or deep infection within 3 or 6 months after TKA using codes from the database
Khanuja et al. [26]	Mean 149 days (range 8 to 840)	17 to 69 months	0.99% (3 of 302)	1.99% (6 of 302)	According to the CDC/National Healthcare Safety Network definitions: A sinus tract was found communicating with the prosthesis, a pathogen was isolated by culture from two separate tissue or fluid samples obtained from the prosthetic joint, or when four of the following six underlying criteria were present: (1) elevated serum erythrocyte sedimentation rate or serum C-reactive protein concentration; (2) elevated synovial white blood cell count; (3) elevated synovial neutrophil percentage; (4) purulence in the joint; (5) isolation of a microorganism from one culture of periprosthetic tissue or fluid; and (6) greater than five neutrophils per high-power field in five high-power fields observed from histologic analysis of periprosthetic tissues at 400 times magnification
Kurtz et al. [32]	Within 12 months	2 years	0.75% (156 of 20,743)	0.75% (424 of 56,506)	The PJIs were identified using diagnosis codes in Medicare database
Forlenza et al. [16]	Within 6 months	6 months	1.80% (522 of 29,058)	1.16% (337 of 29,058)	Identified from each cohort using ICD-9/10 codes from the database

Quality Assessment of the Included Studies

The methodologic quality of the included studies was evaluated according to the items of the modified Newcastle-Ottawa Scale [46], including patient selection, study group comparability, and outcomes assessment. Observational studies were graded on a scale of 0 to 9, with higher scores indicating better study quality. The quality assessment scores of the articles ranged from 6 to 7, indicating that all studies had high methodologic quality according to the Newcastle-Ottawa Scale (Supplementary Material 1, http://links.lww.com/CORR/A664).

Statistical Analysis

All statistical analyses were performed using Review Manager software (Version 5.3., The Nordic Cochrane Centre, the Cochrane Collaboration) and Stata 12.0 software (StataCorp LP). Odds ratios were calculated with 95% confidence intervals. Chi-square and I² tests were used to assess the influence of heterogeneity on the output of the meta-analysis. If there was no heterogeneity (I² < 50% and chi-square p > 0.1), a fixed-effects model was used. Otherwise, a random-effects model was used to lower the impact of heterogeneity on the results. All 11 retrospective studies reported postoperative PJI at the final follow-up interval. Because there was heterogeneity among the studies, a random-effects model was used (I² = 67%; p < 0.05). Two studies about injections received within 3 to 6 months before arthroplasty were available for the current analysis. The random-effects model was adopted because there was heterogeneity in intrastudy comparisons (I² = 69%; p < 0.1). Because of statistical heterogeneity, three studies that evaluated injections within 3 months before arthroplasty were included in the random-effects model (I² = 85%; p < 0.05) (Supplementary Material 2, http://links.lww.com/CORR/A665). The Egger test was performed using Stata version 12.0 software to evaluate the potential publication bias. The results of the Egger test showed there was no publication bias among the studies (Supplementary Material 3, http://links.lww.com/CORR/A666).

Statistical significance was set at p < 0.05.

Results

Overall Difference in PJI Risk Among Patients Receiving Intra-articular Injections

Overall, we found no difference in the proportion of patients experiencing PJI in the injection group compared with the control group (OR 1.22 [95% CI 0.95 to 1.58]; p = 0.12) (Supplementary Material 2, http://links.lww.com/CORR/A665).

Difference in PJI Risk as a Function of Time Between Injection and Arthroplasty

We found that patients who had injections in the knee or hip within 3 months of arthroplasty had an increased PJI risk after surgery compared with patients who did not have an injection within 3 months of arthroplasty (OR 1.39 [95% CI 1.04 to 1.87]; p = 0.03). However, there were no differences in the infection risk in patients who had injections between 3 and 6 months before arthroplasty (OR 1.19 [95% CI 0.95 to 1.48]; p = 0.13) or between 6 and 12 months before arthroplasty (two studies, Cancienne et al. [9] and Kurtz et al. [32], reported on PJI in patients undergoing arthroplasty who had intra-articular injections received between 6 and 12 months before arthroplasty; both showed no increase in the likelihood of infection, but only one was suitable for data extraction for our analysis [10]).

Discussion

Although the American Academy of Orthopaedic Surgeons’ clinical practice guidelines indicate that intra-articular corticosteroid injections may provide temporary relief in patients with symptomatic knee arthritis (based on only moderate evidence) [2], the use of the injection before hip or knee arthroplasty remains widespread [5, 10, 30]. PJI is a serious complication after total joint arthroplasty and remains a serious challenge for individual patients and the healthcare system; some evidence suggests the risk of PJI is increased by intra-articular injections before arthroplasty [7, 9, 16, 26, 37, 40], though the evidence is mixed [13, 19, 20, 21, 26], including systematic reviews on the topic that have disagreed with one another [11, 45, 48]. Therefore, there is an urgent need for better information on this important topic to develop efficacious preventive strategies to decrease PJI rates after TJA [18, 39]. In our meta-analysis, we found that injections of corticosteroid performed within 3 months of total joint arthroplasty (knee or hip) are associated with an increased PJI risk. Based on this, we recommend not performing the injections within 3 months of arthroplasty, and if an injection has been performed, surgery should be delayed at least that long.

Limitations

This study has several limitations. First, the results in this paper depended principally on findings from three large databases, which provided more than 90% of the weight in our meta-analysis. According to PRISMA guidelines, all studies that met the inclusion criteria should be included for analysis, and so we included the smaller studies as well, but the larger studies principally drove our findings. The sample sizes of included studies varied, which obviously is a factor we could not control, but we employed a random-effects model to lower the impact of heterogeneity on the results, which is good practice in a meta-analysis. In addition, we performed a subgroup analysis based on sample size, in which the three large databases and the other eight studies were analyzed independently; this resulted in no changes to our findings by subgroup (Supplementary Material 2, http://links.lww.com/CORR/A665). Second, owing to the limited number of included studies and the fact that almost all the hip patients came from one study, we did not conduct separate subgroup analyses of hip and knee papers; instead, we performed a combination analysis between hip and knee studies, similar to the approach used by other authors [11, 45, 48]. Nonetheless, we caution readers to interpret the findings considering the fact that the knee analyses came from a more diverse population of studies than did the hip analyses, and that the overall findings were driven somewhat disproportionately by findings in one large study. Third, as all the studies included were retrospectively designed, confounding factors, such as injection procedure, patient characteristics, surgical experience, operating room conditions, and the time of follow-up, could not be well controlled for in our analysis; particularly concerning in this regard was patients’ comorbidities, such as hypertension and diabetes, which are known PJI risk factors [7, 40]. However, in 9 of 11 studies, the control groups were matched by age and gender [9, 13, 14, 16, 25, 26, 34, 35, 43] (the other two studies did not [32, 37]); we believe this at least partly should mitigate that concern. Nonetheless, readers should bear in mind that studies of this design generally suffer from the same kinds of biases—the most important ones being selection bias, transfer bias, and assessment bias—and if anything, those tend to make the results appear better than they are. Clinical Orthopaedics and Related Research® usually does not permit meta-analysis (data pooling) of articles about treatments unless the source studies are randomized; the journal makes exceptions when the topic being studied cannot practically be approached with randomized trials. Given the rarity of PJI, this exception was made here.

Another limitation was that because of sparse source material, we only classified the time from the injection to arthroplasty to any injection or injection within 1 year (0-3 months, 3-6 months, and 6-12 months); although it would have been preferable to treat time as a continuous variable, the available evidence did not allow this, and so we were forced to use larger time windows for our analyses.

Discussion of Key Findings

In this large meta-analysis, we found no increased PJI risk overall among patients who had an intra-articular injection before major lower-extremity arthroplasty, but in one subgroup—those who underwent arthroplasty within 3 months of an injection—the PJI risk was increased. Previous studies have disagreed with regard to the infection risk associated with preoperative intra-articular corticosteroid injections [13, 19, 20, 21, 25, 26, 37]. The sample sizes of these studies ranged from 80 to 1317, which is clearly inadequate given that PJI occurs only in 1% to 3% of patients. The results of previous meta-analyses [11, 45, 48] were inconsistent because of differences in their methodology and interpretation; two analyses [11, 48] reached opposite conclusions because different methods were used to control for heterogeneity. On the other hand, previous studies have not explored the relationship between the injection timing and infection risk, which we found to be associated with infection risk. The period of apparently increased risk has differed among studies [7, 9, 16, 40]. In our study, we included findings from several large database–based studies and divided the timing of the injection into the following groups: any injection, injection between 0 and 3 months, between 3 and 6 months, and between 6 and 12 months before arthroplasty. Based on this approach, we believe our finding of increased PJI risk for injections given in the 3 months before surgery is a reliable conclusion. This extends what Bhattacharjee et al. [7] found; they concluded that TKA performed within 4 weeks of a corticosteroid injection might be at a higher risk of postoperative infection. We believe our findings support delaying lower extremity arthroplasty surgery for at least 3 months after intra-articular injections of corticosteroids to mitigate the PJI risk.

The use of joint injections in patients with hip or knee arthritis calls for a thoughtful risk-benefit analysis. In general, the evidence in support of intra-articular corticosteroid injections is weakly favorable [2, 3, 22]. Intra-articular corticosteroid injections may achieve short-term efficacy to relieve pain, but long-term efficacy is doubtful. It also has been reported that corticosteroid injections have adverse effects on articular cartilage; they result in osteoarthritis progression, increase the risk of subchondral insufficiency fracture, and can cause complications like osteonecrosis, rapid joint destruction, and premature arthroplasty [1, 29, 47]. Considering these concerns, surgeons should deliberate carefully whether the benefits of corticosteroid injections justify the risk.

Given the relative rarity of PJI, it is extremely unlikely that prospective studies will be able to determine convincingly whether injections before surgery increase the PJI risk afterward. However, there are plausible mechanisms for this complication. Potential etiologies include direct inoculation of the joint or immunosuppression resulting from corticosteroid [26, 40, 41]. Sterile technique, therefore, is essential if these injections are to be performed. But even if sterile technique is used, PJI might nonetheless follow an arthroplasty in a joint that was recently injected. Corticosteroids have local and systemic immunosuppressive effects. Intra-articular corticosteroids do not dissolve quickly [26], and they may persist and lead to local immunosuppression after TJA. Patients with chronic corticosteroid use have a higher likelihood of surgical infection than corticosteroid-naïve patients; this increased likelihood may be caused by corticosteroid suppression of the proliferation of fibroblasts and an inhibition of the ability to remodel collagen [28]. Future studies will likely need to come from registries, national databases, or insurance company data.

Conclusion

Our meta-analysis of the best-available evidence suggests ipsilateral intra-articular corticosteroid injections within 3 months of knee or hip arthroplasty are associated with an increased PJI risk after surgery. We recommend against performing total joint arthroplasty on a patient who has received an intra-articular corticosteroid injection within 3 months of the proposed surgical date. Further high-quality studies on this topic from registries, national databases, or insurance company data are still required to confirm and extend our findings.