ORIGINAL RESEARCH ARTICLE

The Magnitude and Duration of the Effect of Intra-articular

Corticosteroid Injections on Pain Severity in Knee Osteoarthritis
A Systematic Review and Meta-Analysis
Mikhail Saltychev, MD, PhD, Ryan Mattie, MD, Zachary McCormick, MD, and Katri Laimi, MD, PhD

Objectives: The aims of the study were to clarify the evidence on the
magnitude and duration of treatment effect of intra-articular corticosteroid What Is Known
injections for knee osteoarthritis compared with placebo, to evaluate a treat- • Corticosteroid intra-articular injections are effective to
ment effect by steroid type, and to describe the reported adverse effects. reduce the severity of pain in knee osteoarthritis. The
Design: Cochrane Controlled Trials Register, Medline, Embase, CINAHL, evidence on the duration and the size of this effect
Scopus, and Web of Science databases were searched. The risk of sys- is uncertain.
tematic bias was assessed according to the Cochrane Collaboration’s What Is New
domain-based evaluation framework.
• When pooling the results reported by eight randomized
Results: The final sample included eight randomized controlled stud-
controlled trials, intra-articular corticosteroid injections
ies with follow-ups from 1 to 26 wks. The risk of systematic bias was
demonstrated mild to moderate effect on pain severity
considered low in five and high in three studies. The pooled standard-
and the effect seems to last up to 3 mos—much longer
ized mean difference was −0.58 (95% confidence interval = −0.88 to than previously suggested.
−0.27) and number needed to treat 5.1 (95% confidence interval =
10.0 to 3.7). The heterogeneity was considerable. The pooled effect
size approached the level of statistical insignificance at 4 mos. The
pooled risk ratio of adverse effects was insignificant 0.95 (95% confi-
dence interval = 0.34 to 2.55). Although it is commonly accepted that IAC injections have
Conclusions: The intra-articular corticosteroid had a mild to moderate some effect on pain reduction in knee OA, the magnitude and
effect on pain severity up to 3 mos after the injection—much longer duration of that effect, as well as the rates of adverse effects,
than it had previously been reported. The effect may vary substantially are not well described.12,15,16
in different patient groups and appropriate patient selection is impor- A Cochrane review reported uncertain clinically important
tant. The risk of adverse effects was low. benefits associated with IAC injections of the knee.17 Other
reviews suggest that IAC injections for knee OA might have
Key Words: Anti-inflammatory Agents, Nonsteroidal, a clinically meaning short-term effect, lasting 1 to 2 wks.11 A
Glucocorticoids, Osteoarthritis, Knee, systematic review and meta-analysis by Bjordal et al.18
Meta-Analysis [Publication Type], Intra-articular Injection reported that IAC injections for knee OA are associated with
(Am J Phys Med Rehabil 2020;99:617–625) an effect for up to 3 wks, but that the magnitude of effect
was not clinically significant (11.6 [95% confidence interval
{CI} = 7.4 to 15.7] points on a 0–100 visual analog scale) with
ntra-articular corticosteroid (IAC) injections were first intro- the number needed to treat (NNT) of 38. Another Cochrane re-
IPennsylvania
duced by Dr. J. L. Hollander from the University of
in 1953 for patients with rheumatoid arthritis. 1
view by Bellamy et al.14 also reported only a 1-wk, clinically
insignificant difference in pain reduction between IAC and pla-
Since then, IAC injections have become a widely used method cebo, though with a surprisingly small NNT of approximately
of treating pain in patients with knee osteoarthritis (OA) as three (variance of NNT not reported). Similarly, a systematic
well, and previous reviews on the topic have generally found review and meta-analysis by Godwin et al.19 reported that
them to be an effective treatment.2–12 Intra-articular corticoste- IAC injections for knee OA are associated with a 1-wk effect
roid injections for knee OA have been shown to reduce the se- on pain reduction, yet of clinically insignificant magnitude.
verity of pain and may also enhance the effects of other Prior studies have also reported low rates of adverse effects
injectable medications such as hyaloranate.13 Intra-articular with IAC injections for knee OA.20–23
corticosteroid injections for knee. Osteoarthritis, however, have The objective of this review was to clarify the evidence on
not shown significant effect on physical function or stiffness.14 the magnitude and duration of treatment effect of IAC

From the Department of Physical and Rehabilitation Medicine, Turku University PROSPERO Registration Number: CRD42019145117.
Hospital and University of Turku, Turku, Finland (MS, KL); Department of In- Financial disclosure statements have been obtained, and no conflicts of interest have been
terventional Pain Medicine, Providence Medical Institute, Providence Cedars- reported by the authors or by any individuals in control of the content of this article.
Sinai Tarzana Hospital, Los Angeles, California (RM); and Department of Phys- Supplemental digital content is available for this article. Direct URL citations appear
ical and Rehabilitation Medicine, University of Utah School of Medicine, Salt in the printed text and are provided in the HTML and PDF versions of this article
Lake City, Utah (ZM). on the journal’s Web site (www.ajpmr.com).
All correspondence should be addressed to: Mikhail Saltychev, MD, PhD, Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.
Department of Physical and Rehabilitation Medicine, Turku University Hospital, ISSN: 0894-9115
PO Box 52, FI-20521, Turku, Finland. DOI: 10.1097/PHM.0000000000001384

American Journal of Physical Medicine & Rehabilitation • Volume 99, Number 7, July 2020 www.ajpmr.com 617

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Saltychev et al. Volume 99, Number 7, July 2020

injections for knee OA, specifically compared with placebo. texts of potentially relevant articles (Fig. 1). Disagreements between
The main question was if IAC injections have longer therapeu- the reviewers were resolved by consensus or by a third reviewer.
tic effect than described by previous research. The secondary
objectives included evaluating a treatment effect by steroid Extraction Strategy
type, as well as analysis of the data on the adverse effects re- The data needed for a quantitative assessment were
ported by the studies selected for review. extracted using a standardized form based on recommendations
by the Cochrane Handbook for Systematic Reviews of Interven-
METHODS tions.24 The form included the following: the first author name,
the year of publication, a country, group sizes at randomization
This systematic review and meta-analysis were part of a
and the end of follow-up, a sex distribution, an average age of
larger study concerning the effectiveness of IAC injections
patients within groups, inclusion and exclusion criteria, main
for knee and hip OA.
conclusions, corticosteroid dosage and form, and the estimates
of main outcomes.
Inclusion and Exclusion Criteria (PICO)
This study conformed to all Preferred Reporting Items for
The criteria for considering studies for this review were Systematic Reviews and Meta-analyses guidelines and reported the
based on the PICO (Population, Intervention, Comparison, required information accordingly (see Supplemental Checklist,
and Outcome) framework as follows: Supplemental Digital Content 1, http://links.lww.com/PHM/A942).
• Patients: Adults (≥18 yrs) with primary or secondary OA of
the knee joint Assessment of the Methodological Risks of
• Intervention: Intra-articular corticosteroid injection Systematic Bias
• Comparison: Placebo Two independent reviewers rated the methodological qual-
• Outcome: Difference between groups in change of pain ity of the included trials using the Cochrane domain-based
severity quality assessment tool.24 Each study was rated as having
• Papers: Randomized controlled studies (RCTs) published “low,” “high,” or “unclear” risk of systematic bias in seven do-
in peer-reviewed academic journals; English language; ab- mains. Domains were assessed in the following sequence: (a)
stract available; no restrictions on the date of publication. selection bias (randomized sequence generation and allocation
• Data sources and search strategy: The Medline, Central, concealment); (b) allocation concealment; (c) performance bias
Embase, CINAHL, Web of Science, and Scopus databases (blinding of participants and personnel); (d) detection bias
were searched in June 2019. The search clause for the (blinding of outcome assessment); (e) attrition bias (incomplete
Medline search was: outcome data, eg, due to dropouts); ( f ) reporting bias (selective
((“Glucocorticoids/administration and dosage”[Mesh] OR reporting); and (g) other sources of bias. Disagreements between
“Glucocorticoids/adverse effects”[Mesh] OR “Glucocorticoids/ the reviewers were resolved by consensus or by a third reviewer.
pharmacology”[Mesh] OR “Glucocorticoids/therapeutic use”[Mesh])
OR (glucocorticoid*[TIAB] OR corticosteroid*[TIAB] OR cortisol* Statistical Analysis (Meta-Analysis)
[TIAB] OR hydrocortison*[TIAB] OR cortisone[TIAB] OR pred-
nisone*[TIAB] OR prednisolon*[TIAB] OR methylprednisolon*
Statistical Model, Heterogeneity, and Publication Bias
[TIAB] OR dexamethason*[TIAB] OR betamethason*[TIAB] A random-effects model was used to quantify the pooled
OR triamcinolon*[TIAB] OR fludrocortison*[TIAB] OR effect size of the included studies, which was a more fitting
deoxycorticosterone*[TIAB])) AND (inject*[TIAB] OR intra- choice than a fixed-effect model considering the context of
articul*[TIAB]) AND (osteoarthriti*[TIAB] AND (knee[TI] OR medical decisions-making and generalizing the results beyond
hip[TI]) OR (“Osteoarthritis, Knee”[Mesh] OR “Osteoarthritis, the selected samples. The results were accompanied by 95%
Hip”[Mesh])) NOT (arthroplast*[TI] OR *operativ*[TI] OR pi- CIs and two-tailed P value (significant if ≤0.05) when appro-
lot[TI] OR protocol[TI] OR rheumat*[TIAB]) AND (Controlled priate. The test for heterogeneity was conducted using the Q
Clinical Trial[ptyp] OR Randomized Controlled Trial[ptyp]) test; heterogeneity was deemed present if Q was greater than
AND (hasabstract[text] AND English[lang]) the degree of freedom (number of studies – 1).25 The I2 statistic
describes the percentage of the variability in effect estimates
To avoid missing potentially relevant studies, the use of that is due to heterogeneity rather than sampling error (chance).25
other limiters and filters was restricted and the authors relied As the correlation between pre- and postestimates within
instead on manual selection. Similar clauses were used when groups was not reported, the coefficient of precorrelation/
searching the other databases. The references of identified arti- postcorrelation was set at 0.6. Publication bias was not assessed
cles and reviews were also checked for relevance. as the number of studies in the model was 10 or more.

Selection Strategy Data Extraction Issues

The records identified from the data sources were stored In some cases, the estimates were adopted from fig-
using Endnote software (Endnote X7.8, Thomson Reuters). ures.26,27 Reported standard errors (SEs) and SEs of mean
Using a built-in search engine of the Endnote software, dupli- (SEMs) were converted into standard deviations (SDs) as
cates, conference proceedings, theses, reviews, and case re- SD = SE (or ≈SEM)  √n.26,27 Along with 95% CI, prediction
ports were deleted. Two independent reviewers screened the intervals (PIs) were calculated as: 95% PI = pooled estimate
titles and abstracts of the remaining articles and assessed the full ±1.96  τ. Although 95% CI defines an average effect

618 www.ajpmr.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Volume 99, Number 7, July 2020 Intra-articular Corticosteroids in Osteoarthritis

FIGURE 1. Search flow.

expected to be seen in different populations, 95% PI defines size at different time points, weeks were rounded to whole
the boundaries of true effect expected to be seen in 95% of months. In that case, all reported estimates up to a particular
cases in different populations. whole number of months were averaged, eg, estimates at 1, 3,
and 4 wks were rounded into an estimate of “1 mo.” The sen-
sitivity tests included the following: (a) setting the pre/post cor-
Outcome Measures Used in the Meta-Analysis
relation coefficient to 0.8 and (b) removing one study at a time
The spectrum of outcome measures of pain severity used to assess the consequence of that removal on a pooled estimate.
in the included studies was relatively uniform using the follow-
ing three main scales: The Western Ontario McMaster Osteoar-
Converting SMD Into Meaningful Units
thritis Index, The Knee injury and Osteoarthritis Outcome
Score, and visual analog scales 0–10 or 0–100. In this analysis, An SMD is not an intuitively understandable type of effect
the Western Ontario McMaster Osteoarthritis Index pain score size, as it does not have any meaningful unit. However, it can
was primarily used when available. The Western Ontario easily be converted into meaningful estimates by using an on-
McMaster Osteoarthritis Index pain score is a sum of five items line calculator (eg, https://rpsychologist.com/d3/cohend/) or
measured on a Likert-like scale from 0 to 4 points with a max- manually using Microsoft Excel. The included studies did
imum of 20 points indicating the worst severity of pain. Thus, not report the control groups’ event rates (CL = proportions
converting such different scales into one was considered inap- of patients with moderate to high pain reduction in placebo
propriate and a standardized mean difference (SMD) between groups). In this review, the rate was set at 0.6 as it seemed to
groups in change of pain level was calculated for each included be a reasonably high supposition. In other words, the authors
study as well as for the pooled sample. The effect size was con- assumed that at least 60% of patients in placebo group might
sidered small when SMD was 0.2, medium when SMD was achieve a substantial pain reduction. In reality, this rate proba-
0.5, and large when SMD was 0.8 or higher.28 bly had been even lower meaning that the effect of IAC was
possibly even higher than the one based on that speculation.
An NNT approximately 2.5 was considered an indication of ef-
Subgroup and Sensitivity Analysis fective treatment with moderate improvement over controls,
The analysis was conducted in several different ways: (a) and an NNT approximately 10 was considered an indication
pooling the means of all available time points assuming simi- of mild effect.29 In this study, five additional estimates along
larity of all corticosteroids used; (b) pooling the means of all with their asymmetrical 95% CIs were calculated based on
available time points, stratified by corticosteroid type, and (c) the control groups’ event rates approximated from the included
pooling effect size at different time points. When pooling effect RCTs.30,31 These five additional estimates were as follows:

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.ajpmr.com 619

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Saltychev et al. Volume 99, Number 7, July 2020

1) Cohen’s U3, which shows how a large proportion of the Pooling All of the Available Time Points Assuming
IAC group will be above the mean of the control group. Similarity of Corticosteroids (From All Included
U3 = Φ(SMD). Φ is a cumulative distribution function of Studies)
the standard normal distribution.
For all the included studies, the pooled SMD was −0.58
2) Percentage of two groups’ overlapping. Overlap = 2Φ(−|
(95% CI = −0.88 to −0.27) with 95% PI between −1.32 and
SMD|/2). 0.16 (Fig. 2). There was considerable heterogeneity: Q = 33
3) Probability of superiority, which estimates a chance that a
against seven degrees of freedom. Most of the heterogeneity
patient selected at random from the IAC group will have a
was related to true effect rather than to chance: I2 = 79%.
better score than a patient selected at random from the pla- Cohen’s U3 was 72 (95% CI = 61 to 81), percentage of overlap
cebo group. Probability = Φ(SMD/√2).
77% (95% CI = 89% to 66%), probability of superiority 66
4) Number Needed to Treat, which defines the number of
(95% CI = 58 to 73), NNT 5.1 (95% CI = 10.0 to 3.7), and per-
patients that should be treated to observe one favorable out-
centage of patients with a positive outcome was expected to be
come. NNT = 1/(Φ(SMD + Ψ(CER)) − CER). CER is con-
23% (95% CI = 11% to 33%). When pooling only studies with
trol group’s event rate and Ψ is inverse of Φ.
low risk of systematic bias, the pooled SMD was smaller −0.41
5) Proportion of those with more favorable outcomes (com-
(95% CI = −0.62 to −0.21) (Fig. 2). There was considerable
pared with the placebo group) out of 100 patients treated
heterogeneity: Q = 6.7 against 4 degrees of freedom. However,
with IAC. Proportion = 100/NNT.
less than half of the heterogeneity was related to true effect
All calculations for the meta-analysis were performed rather than to chance: I2 = 40%.
using the Comprehensive Meta-Analysis CMA software, Ver-
sion 3.0, available at http://www.meta-analysis.com and Pooling All of the Available Time Points Grouped
Microsoft Excel 2016. by Different Corticosteroids
The effect sizes of different corticosteroids varied from a
RESULTS very large effect with betamethasone −1.46 (95% CI = −2.25
The search resulted in 458 records (Fig. 1). Of these, 261 to −0.68), to statistically insignificant effects with cortivazol
were duplicates. Using the Endnote software search engine, and low doses (20 mg) of triamcinolone (Fig. 4). When pooling
189 records were excluded as conference proceedings, edito- all the corticosteroids separately in the same model, the total
rials, theses, and papers on subjection conditions other than pooled effect size was −0.47 (95% CI = −0.63 to −0.32) with
knee OA. Thirty records were screened based on titles and ab- 95% PI between −2.15 and −0.77. The overall heterogeneity
stracts; the agreement between reviewers was good (κ = 0.76 was substantial: Q = 23 against five degrees of freedom.
[95% CI = 0.45 to 1.0]). After further exclusion, 17 records
were screened based on their full texts; of these, eight RCTs
Pooled Effect Size at Different Time Points After
were included for further analysis.26,27,32–37 The studies were the Injection
excluded if they did not fulfill the inclusion criteria. Four stud- The pooled effect size remained statistically significant up
ies concluded that corticosteroids are effective for short-term to 4 mos after the injection (Fig. 3). However, at 4 mos, the
pain relief.32,33,36,37 Three RCTs reported corticosteroid injections 95% CI was close to the level of statistical insignificance. There
to be effective for a longer-term duration (up to a half year).26,27,35 was some difference between the effect size obtained from
Only one study reported injections to be ineffective.34 pooling all of the available studies compared with those with
The pooled sample was composed of 645 cases and 428 low risk of systematic bias. This difference was observed up to
controls (Table 1). The mean age of patients varied from 60 3 mos from the time of the IAC injection. For the studies with
to 68 yrs. There were more women than men. The sample sizes low risk of systematic bias, the pooled effect size was −0.51 (95%
at the time of randomization varied from 25 up to 162. The in- CI = −0.67 to −0.36), −0.46 (95% CI = −0.62 to −0.3), and
clusion and exclusion criteria were similar across all studies. −0.29 (95% CI = −0.44 to −0.13) at 1, 2, and 3 mos, respectively.
Four studies excluded patients with maximal severity of
pain.26,27,35,37 The follow-up durations varied widely from 1 Pooling Removing One Study at a Time
to 26 wks. The included studies used different corticosteroids The pooled effect sizes varied between −0.38 (95% CI =
in different dosages: triamcinolone acetate 16, 20, 32, and −0.52 to −0.23) and −0.63 (95% CI = −0.99 to −0.26; Fig. 4).
40 mg; cortivazol 3.75 mg; betamethasone disodium phos- Removing the study by Yavuz et al.37 resulted in an especially
phate 3 mg; and methylprednisolone acetate 40 mg. Two stud- high impact on the pooled estimate, dropping from −0.58 (95%
ies used the long-acting form of triamcinolone acetate.26,27 In CI = −0.88 to −0.27) down to −0.38 (95% CI = −0.52 to −0.23)
one study, the variance of data on 16 mg triamcinolone acetate with more negative estimate meaning stronger effect of IAC.
was reported incompletely, and thus, only data on the 32 mg When comparing a microsphere-based extended-release
dosage were employed in the analysis.26 The placebo injections triamcinolone acetonide formulation used by Conaghan et al.26,27
were similar across the studies—seven RCTs used a saline so- with a regular triamcinolone (excluding other cortisones), there
lution while one combined anesthetic and saline.34 In some were not substantial differences in the effect sizes (Supplemental
cases, the procedure included the aspiration of synovial fluid.33 Digital Content 2, http://links.lww.com/PHM/A943).
The risk of systematic bias was considered low in five and
high in three studies (Table 2). The most frequent sources of Reported Adverse Effects
potential bias were substantial drop-out rates and a direct rela- Most of the included studies did not describe the ob-
tionship with the pharmaceutical industry. served adverse effects in details (Table 3). Thus, here, the

620 www.ajpmr.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Volume 99, Number 7, July 2020 Intra-articular Corticosteroids in Osteoarthritis

TABLE 1. Characteristics of the included studies

Cases/Controls
Control Age, Women
Study Country Intervention Intervention nb Years % Inclusion Exclusion
Chao et al., 2010, 40 mg (1 ml) TA 1-ml saline 40/ 65/63 6/6 ACR-dgc Systemic corticosteroids, inflammatory
United States32 39 connective tissue disease, history of
intra-articular corticosteroids <3 mos
Conaghan et al., 32 mg (5 ml) TA 104/ 59/60 49/61 ≥40 yrs, ACR-dg,
5-ml saline Hip OA, inflammatory connective tissue
2018 (1),a United prolonged 104 Kellgren-Lawrence disease, unstable knee, history of
Kingdom26 grade 2 or 3, pain >15 intra-articular corticosteroids <3 mos,
Conaghan et al., 32 mg (5 ml) TA 5-ml saline 161/ 62/62 64/59 d/mo, pain 5–9/10, hyaluronic acid injections <6 mos,
2018 (2), United prolonged 163 systemic corticosteroids <2 wks
Kingdom27 40 mg (1 ml) TA 162/ 62/62 60/59
163
Gaffney et al., 1995, 20 mg (1 ml) TA 1-ml saline 42/ 66/68 79/64 Pain and functional n/r
United Kingdom33 42 impairment >6 mos
Henriksen et al., 40 mg (1 ml) 1-ml saline 50/ 61/66 56/66 ≥40 yrs, pain >4/10, History of intra-articular corticosteroids
2015, Denmark34 TA + 4 ml + 4 ml lidocaine 50 BMI ≤ 35 <3 mos, history of exercise therapy
lidocaine hydrochloride <3 mos, systemic corticosteroids,
hydrochloride inflammatory arthritis, radicular
pain and fibromyalgia
Mendes et al., 40 mg (2 ml) TA 2-ml saline 35/ 66/65 89/94 >50 yrs, ACR-dg, Kellgren- Secondary OA, history of intra-articular
2019, Brazil35 35 Lawrence grade 2 or 3, corticosteroids <3 mos, systemic
pain >6 mos, pain 3–8/10 corticosteroids, knee surgery, systemic
disease, fibromyalgia, pregnancy,
inability to work
Ravaud et al., 3.75 mg 1.5-ml saline 25/ 67/63 72/64 ACR-dg, pain 40–100/100, Secondary OA, systemic disease,
1999, France36 (1.5 ml) CVZ 28 Kellgren-Lawrence scheduled knee surgery, history of
grade >2 intra-articular corticosteroids <3
mos, systemic corticosteroids <2 wks
Yavuz et al., 3 mg (1 ml) 1-ml saline 30/ 60/60 67/63 ACR-dg, pain ≤5/10, Secondary OA, history of intra-articular
2012, Turkey37 BMZ 30 Kellgren-Lawrence corticosteroids <3 mos, systemic
40 mg (1 ml) TA 30/− 60/60 63/63 grade >2 disease, systemic corticosteroids
40 mg (1 ml) MA 30/− 61/60 60/63
a
As reported at 4-wk follow-up (n = 67).
b
Group sizes at the time of randomization.
c
Diagnosed according to the American College of Rheumatology (ACR) criteria.
BMZ, betamethasone disodium phosphate 3 mg; CVZ, cortivazol; MA, methyl-prednisolone acetate; TA, triamcinolone acetonide.

observed adverse effects were defined as “serious adverse ef- OA. These injections seem to be effective for up to 3 mos after
fects.” Overall, six RCTs described adverse effects or their the procedure. Taking into account the CIs, the effect size was
absence.26,27,34–37 Thirteen adverse effects were observed small, with an NNT up to 10 patients for the primary synthesis
in the corticosteroid groups (eight of them in the study by model. Based on six reports, the pooled risk of serious adverse
Conaghan et al.27) and seven in placebo groups. The pooled effects associated with IAC injection was not significant when
risk ratio was insignificant 0.95 (95% CI = 0.34 to 2.55) with compared with placebo.
low heterogeneity between the RCTs: Q = 1.9 against two de- Several limitations of this study should be noted. First, a
grees of freedom and I2 = 0%. The result was also insignificant meta-analysis is always an approximation. The number of stud-
when calculating unweighted relative risk: 1.12 (95% CI = ies included in this review was too small to assess a potential
0.45 to 2.78). The risk of adverse effect varied from a number publication bias, and the corticosteroids used for injections as
needed to harm of 53 to a number needed to harm of 68. well as their dosages varied across the included trials. In addi-
tion, the injected volumes and the techniques used (eg, syno-
vial fluid aspiration) were different. Of eight RCTs, three
DISCUSSION were considered to have a high risk of systematic bias. The het-
This systematic review was evaluating the effectiveness of erogeneity of the effect size was substantial. The synthesis of
IAC injections for knee OA using advanced statistical methods studies shown in Figures 3 and 4 was based on a very small
to separate pooled effect sizes based on study variability (with number of studies at each time point. It is self-evident that gen-
regard to either the steroid used or the measured time points). eralization of the results obtained from only a few studies of
The meta-analysis of eight RCTs found moderate evidence modest size is questionable. Saline solution with or without
for the effectiveness of IAC injections to relieve pain in knee local anesthetic was used in the included studies as placebo.

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.ajpmr.com 621

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Saltychev et al. Volume 99, Number 7, July 2020

TABLE 2. Risk of systematic bias of the included studies

Blinding of Blinding of Overall

Random Sequence Allocation Participants Outcome Incomplete Selective Risk of
Generation Concealment and Personnel Measurement Outcome Data Reporting Systematic
(Selection Bias) (Selection Bias) (Performance Bias) (Detection Bias) (Attrition Bias) (Reporting Bias) Other Bias
Chao et al., U H L L H L H H
201032
Conaghan et al., L L L L L L H L
201826
Conaghan et al., L L L L L L H L
201827
Gaffney et al., L H H L H L L H
199533
Henriksen et al., L L L L L L L L
201534
Mendes et al., L L L L H L L L
201935
Ravaud et al., U L L L H L L L
199936
Yavuz et al., U H H H H L L H
201237

However, some previous reports questioned such a consideration, consistent with previous reports showing low rates of adverse
suggesting that saline solution may have its own therapeutic ef- effect from IAC injections.20–23 The mild or, at most, moderate
fect.38 Despite these weaknesses, most included studies had low size of IAC effect seen in the present analysis is in line with
risk of systematic bias and the results of the synthesis seemed to previous research as well.14,18,19 This meta-analysis, however,
be robust in sensitivity analyses. On contrary to previous meta- differed from previous studies regarding the duration of IAC
analyses on the topic, the present synthesis use all the available effect. Previous reviews have mostly suggested that IAC injec-
time points grouped by different corticosteroid types used for tions have only short-term effect.11,18,19 The present analysis
injections. Such an approach might avoid the potential loss of points toward a longer duration of effect up to 3 mos. The rea-
valuable data. That way, the confidence level of the results be- sons for this discordance may relate to differences in the sets of
came more robust. included studies and/or a difference in statistical approach.
These results are consistent with previous reviews on the Osteoarthritis is the most common articular disease of the
topic, which have concluded that corticosteroid injections are developed world and a leading cause of chronic disability,
an effective treatment for knee OA.2–11,21 These results are also mostly due to knee and/or hip OA. With an aging population,

FIGURE 2. Forest plots for all available time points and all corticosteroids. Top, All of the included studies. Bottom, Only studies with low risk of
systematic bias.

622 www.ajpmr.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Volume 99, Number 7, July 2020 Intra-articular Corticosteroids in Osteoarthritis

FIGURE 3. Pooled effect size (SMD) at different time points after the injection.

the incidence of knee OA is steadily increasing.39,40 This pro- an imbalance of physiological processes, and obesity leading
gressive disease develops from a combination of risk factors to increased biomechanical stress. The pathogenesis of OA in-
including advanced age, trauma, knee malalignment, genetics, volves abnormal cellular activities in cartilage and synovium

FIGURE 4. Forest plot for all available time points grouped by different corticosteroid types (A) and influence of removing one study at a time on the
pooled effect size (B).

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.ajpmr.com 623

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Saltychev et al. Volume 99, Number 7, July 2020

TABLE 3. Adverse effects reported by the included studies CONCLUSIONS

Intra-articular corticosteroid injections have a mild to
Cases, n (%) Controls, n (%) moderate effect on pain severity in patients with knee OA. This
32 effect might last up to 3 mos after the injection. The effect sizes
Chao et al., 2010 Not reported Not reported
of different corticosteroids varied from a very large effect with
Conaghan et al., 201826a
betamethasone to statistically insignificant effects with
Arthralgia 8 (8) 16 (16)
cortivazol and low doses of triamcinolone. The risk of ad-
Back pain 2 (2) 2 (2)
verse effects compared with placebo was low. Broad PIs
Bronchitis 2 (2) 2 (2)
and substantial variance in NNT suggest that IAC injections
Headache 4 (4) 2 (2)
may behave very differently in particular patient groups, and
Joint swelling 5 (5) 5 (5)
thus, appropriate patient selection for the IAC injections is
Ligament sprain 4 (4) 2 (2)
important. Factors affecting the results of IAC injections
Nasopharyngitis 2 (2) 4 (4)
are unknown and deserve investigation.
Neck pain 3 (3) 0 (0)
Sinusitis 2 (2) 1 (1) REFERENCES
Toothache 3 (3) 1 (1) 1. Hollander JL: Intra-articular hydrocortisone in arthritis and allied conditions; a summary of
Conaghan et al., two years’ clinical experience. J Bone Joint Surg Am 1953;35-A:983–90
201827 (triamsinolone)b 2. Arroll B, Goodyear-Smith F: Corticosteroid injections for osteoarthritis of the knee:
Arthralgia 12 (8) 20 (12) meta-analysis. BMJ 2004;328:869
3. Wang F, He X: Intra-articular hyaluronic acid and corticosteroids in the treatment of knee
Headache 15 (9) 13 (8) osteoarthritis: a meta-analysis. Exp Ther Med 2015;9:493–500
Back pain 12 (8) 9 (6) 4. Hepper CT, Halvorson JJ, Duncan ST, et al: The efficacy and duration of intra-articular
Conaghan et al., 201827 corticosteroid injection for knee osteoarthritis: a systematic review of level I studies. J Am
Acad Orthop Surg 2009;17:638–46
(prolonged triamsinolone)b
5. McCrum C: Therapeutic review of methylprednisolone acetate intra-articular injection in the
Arthralgia 23 (14) 20 (12) management of osteoarthritis of the knee – part 1: clinical effectiveness. Musculoskeletal Care
Headache 14 (9) 13 (8) 2017;15:79–88
Back pain 5 (3) 9 (6) 6. Tian K, Cheng H, Zhang J, et al: Intra-articular injection of methylprednisolone for reducing
pain in knee osteoarthritis: a systematic review and meta-analysis. Medicine (Baltimore)
Gaffney et al., 199533 Not reported Not reported 2018;97:e0240
Henriksen et al., 201534c 1 (2) 1 (2) 7. Richards MM, Maxwell JS, Weng L, et al: Intra-articular treatment of knee osteoarthritis: from
Mendes et al., 201935 0 (0) 0 (0) anti-inflammatories to products of regenerative medicine. Phys Sportsmed 2016;44:101–8
8. Bannuru RR, Schmid CH, Kent DM, et al: Comparative effectiveness of pharmacologic
Ravaud et al., 199936 0 (0) 0 (0)
interventions for knee osteoarthritis: a systematic review and network meta-analysis. Ann
Yavuz et al., 201237 0 (0) 0 (0) Intern Med 2015;162:46–54
9. Ayhan E, Kesmezacar H, Akgun I: Intraarticular injections (corticosteroid, hyaluronic acid,
a
Adverse effects with incidence ≥2 were reported. platelet rich plasma) for the knee osteoarthritis. World J Orthop 2014;5:351–61
b
Adverse effects with incidence ≥5 were reported. 10. Alshami AM: Knee osteoarthritis related pain: a narrative review of diagnosis and treatment.
c Int J Health Sci (Qassim) 2014;8:85–104
Not reported in further details.
11. Rodriguez-Merchan EC: Intra-articular injections of hyaluronic acid and other drugs in the
knee joint. HSS J 2013;9:180–2
12. Cushman DM, Bruno B, Christiansen J, et al: Efficacy of injected corticosteroid type, dose,
that result in the activation of intracellular and extracellular and volume for pain in large joints: a narrative review. PM R 2018;10:748–57
13. Smith C, Patel R, Vannabouathong C, et al: Combined intra-articular injection of
proinflammatory mediators. Currently, there is no clear con- corticosteroid and hyaluronic acid reduces pain compared to hyaluronic acid alone in the
sensus regarding a superior steroid type or dose for efficacy/ treatment of knee osteoarthritis. Knee Surg Sports Traumatol Arthrosc 2019;27:1974–83
effectiveness and safety. One study found that a low-dose and 14. Bellamy N, Campbell J, Robinson V, et al: Intraarticular corticosteroid for treatment of
osteoarthritis of the knee. Cochrane Database Syst Rev 2006;CD005328
short-acting steroid showed a protective effect on the joint,
15. Goldman DT, Piechowiak R, Nissman D, et al: Current concepts and future directions of
whereas a high dose caused chondrotoxicity and cell dam- minimally invasive treatment for knee pain. Curr Rheumatol Rep 2018;20:54
age.41 The present analysis contributes further details to this 16. Nguyen C, Lefevre-Colau MM, Poiraudeau S, et al: Evidence and recommendations for use
debate by examining the pooled effect sizes removing one study of intra-articular injections for knee osteoarthritis. Ann Phys Rehabil Med 2016;59:184–9
17. Juni P, Hari R, Rutjes AW, et al: Intra-articular corticosteroid for knee osteoarthritis. Cochrane
at a time (Fig. 4). The pooled effect size remains overall robust Database Syst Rev 2015;CD005328
in that sensitivity test with one exception—excluding a study with 18. Bjordal JM, Klovning A, Ljunggren AE, et al: Short-term efficacy of pharmacotherapeutic
high risk of systematic bias by Yavuz et al.37 resulted in substan- interventions in osteoarthritic knee pain: a meta-analysis of randomised placebo-controlled
trials. Eur J Pain 2007;11:125–38
tially milder effect than the pooled effect of all eight RCTs. 19. Godwin M, Dawes M: Intra-articular steroid injections for painful knees. Systematic review
Overall, the evidence remains inconsistent and controver- with meta-analysis. Can Fam Physician 2004;50:241–8
sial for the use of IAC injection for knee OA.16 A previous 20. Nguyen C, Rannou F: The safety of intra-articular injections for the treatment of knee
review by a Cochrane group questions whether there are clini- osteoarthritis: a critical narrative review. Expert Opin Drug Saf 2017;16:897–902
21. Cushman DM, Christiansen J, Clements ND, et al: Infections after large joint or bursa
cally important benefits with intra-articular corticosteroid in- injection: a survey of 554 sports medicine physicians. Am J Phys Med Rehabil 2019;98:1106–9
jections,17 but the present meta-analysis provides some 22. Pal B, Morris J: Perceived risks of joint infection following intra-articular corticosteroid
evidence to contest that conclusion. This discrepancy high- injections: a survey of rheumatologists. Clin Rheumatol 1999;18:264–5
23. Charalambous CP, Tryfonidis M, Sadiq S, et al: Septic arthritis following intra-articular steroid
lights the need for additional RCTs in different settings, with injection of the knee—a survey of current practice regarding antiseptic technique used during
specified target populations, and defined corticosteroids and intra-articular steroid injection of the knee. Clin Rheumatol 2003;22:386–90
dosages used for the injections. Regarding the duration of ef- 24. The Cochrane Collaboration, Higgins JP, Green S (eds): Cochrane Handbook for Systematic
Reviews of Interventions Version 5.1.0 [updated March 2011], in 2011. Available at:
fect, well-designed randomized studies with multiple repeated
https://handbook-5-1.cochrane.org/. Accessed February 5, 2020
measures and follow-up time points of up to several months 25. Borenstein M, Higgins JP, Hedges LV, et al: Basics of meta-analysis: I(2) is not an absolute
are needed. measure of heterogeneity. Res Synth Methods 2017;8:5–18

624 www.ajpmr.com © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.

Volume 99, Number 7, July 2020 Intra-articular Corticosteroids in Osteoarthritis

26. Conaghan PG, Cohen SB, Berenbaum F, et al: Brief Report: a Phase IIb trial of a novel 34. Henriksen M, Christensen R, Klokker L, et al: Evaluation of the benefit of corticosteroid
extended-release microsphere formulation of triamcinolone acetonide for intraarticular injection before exercise therapy in patients with osteoarthritis of the knee a randomized
injection in knee osteoarthritis. Arthritis Rheumatol 2018;70:204–11 clinical trial. JAMA Intern Med 2015;175:923–30
27. Conaghan PG, Hunter DJ, Cohen SB, et al: Effects of a single intra-articular injection of a 35. Mendes JG, Natour J, Nunes-Tamashiro JC, et al: Comparison between intra-articular
microsphere formulation of triamcinolone Acetonide on knee osteoarthritis pain: a botulinum toxin type a, corticosteroid, and saline in knee osteoarthritis: a randomized
double-blinded, randomized, placebo-controlled, multinational study. J Bone Joint Surg Am controlled trial. Clin Rehabil 2019;33:1015–26
2018;100:666–77 36. Ravaud P, Moulinier L, Giraudeau B, et al: Effects of joint lavage and steroid injection in
28. Lakens D: Calculating and reporting effect sizes to facilitate cumulative science: a practical patients with osteoarthritis of the knee: results of a multicenter, randomized, controlled trial.
primer for t-tests and ANOVAs. Front Psychol 2013;4:863 Arthritis Rheum 1999;42:475–82
29. Citrome L, Ketter TA: When does a difference make a difference? Interpretation of number 37. Yavuz U, Sökücü S, Albayrak A, et al: Efficacy comparisons of the intraarticular steroidal
needed to treat, number needed to harm, and likelihood to be helped or harmed. Int J Clin agents in the patients with knee osteoarthritis. Rheumatol Int 2012;32:3391–6
Pract 2013;67:407–11 38. Altman RD, Devji T, Bhandari M, et al: Clinical benefit of intra-articular saline as a
30. Furukawa TA, Leucht S: How to obtain NNT from Cohen’s d: comparison of two methods. comparator in clinical trials of knee osteoarthritis treatments: a systematic review and
PLoS One 2011;6:e19070 meta-analysis of randomized trials. Semin Arthritis Rheum 2016;46:151–9
31. IntHout J, Ioannidis JP, Rovers MM, et al: Plea for routinely presenting prediction intervals in 39. Theis KA, Roblin DW, Helmick CG, et al: Prevalence and causes of work disability among
meta-analysis. BMJ Open 2016;6:e010247 working-age U.S. adults, 2011-2013, NHIS. Disabil Health J 2018;11:108–15
32. Chao J, Wu C, Sun B, et al: Inflammatory characteristics on ultrasound predict poorer 40. Centers for Disease Control and Prevention (CDC): Prevalence of doctor-diagnosed arthritis
longterm response to intraarticular corticosteroid injections in knee osteoarthritis. and arthritis-attributable activity limitation—United States, 2010-2012. MMWR Morb Mortal
J Rheumatol 2010;37:650–5 Wkly Rep 2013;62:869–73
33. Gaffney K, Ledingham J, Perry JD: Intra-articular triamcinolone hexacetonide in knee 41. Wernecke C, Braun HJ, Dragoo JL: The effect of intra-articular corticosteroids on articular
osteoarthritis: factors influencing the clinical response. Ann Rheum Dis 1995;54:379–81 cartilage: a systematic review. Orthop J Sports Med 2015;3:2325967115581163

© 2020 Wolters Kluwer Health, Inc. All rights reserved. www.ajpmr.com 625

Copyright © 2020 Wolters Kluwer Health, Inc. All rights reserved.