You are on page 1of 7

„„ Knee

Intra-­articular corticosteroid
injections increase the risk of requiring
knee arthroplasty
a multicentre longitudinal observational study using
S. R. W. Wijn,
data from the Osteoarthritis Initiative
M. M. Rovers,
T. G. van Tienen,
G. Hannink Aims
Recent studies have suggested that corticosteroid injections into the knee may harm the
From Radboud joint resulting in cartilage loss and possibly accelerating the progression of osteoarthritis
Institute for Health (OA). The aim of this study was to assess whether patients with, or at risk of developing,
Sciences, Radboud symptomatic osteoarthritis of the knee who receive intra-­articular corticosteroid injections
University Medical
have an increased risk of requiring arthroplasty.
Center, Nijmegen, The
Netherlands
Methods
We used data from the Osteoarthritis Initiative (OAI), a multicentre observational cohort
study that followed 4,796 patients with, or at risk of developing, osteoarthritis of the knee
on an annual basis with follow-­up available up to nine years. Increased risk for symptomat-
ic OA was defined as frequent knee symptoms (pain, aching, or stiffness) without radiolog-
ical evidence of OA and two or more risk factors, while OA was defined by the presence of
both femoral osteophytes and frequent symptoms in one or both knees. Missing data were
imputed with multiple imputations using chained equations. Time-­dependent propensity
score matching was performed to match patients at the time of receving their first injec-
tion with controls. The effect of corticosteroid injections on the rate of subsequent (total
and partial) knee arthroplasty was estimated using Cox proportional-­hazards survival
analyses.

Results
After removing patients lost to follow-­up, 3,822 patients remained in the study. A total of
249 (31.3%) of the 796 patients who received corticosteroid injections, and 152 (5.0%) of
the 3,026 who did not, had knee arthroplasty. In the matched cohort, Cox proportional-­
hazards regression resulted in a hazard ratio of 1.57 (95% confidence interval (CI) 1.37 to
1.81; p < 0.001) and each injection increased the absolute risk of arthroplasty by 9.4% at
nine years’ follow-­up compared with those who did not receive injections.

Conclusion
Corticosteroid injections seem to be associated with an increased risk of knee arthroplasty
in patients with, or at risk of developing, symptomatic OA of the knee. These findings sug-
gest that a conservative approach regarding the treatment of these patients with cortico-
steroid injections should be recommended.

Cite this article: Bone Joint J 2020;102-B(5):586–592.

Correspondence should be Introduction of pain, although the effect is often temporary.2,3


sent to S. R. W. Wijn; email:
​Stan.​Wijn@​radboudumc.​nl The routine treatment of patients with osteo- The safety of recurrent corticosteroid injections,
© 2020 The British Editorial arthritis (OA) of the knee currently includes a however, remains unknown. Studies dealing with
Society of Bone & Joint Surgery combination of physiotherapy, analgesia, and the long-­ term complications of intra-­ articular
doi:10.1302/0301-620X.102B5.
BJJ-2019-1376.R1 $2.00 surgery where necessary.1 Intra-­articular cortico- injections have reported contradictory findings.4-6
Bone Joint J steroid injections may be used for the management It thus remains unclear whether these injections
2020;102-B(5):586–592.

586 The Bone & Joint Journal


Intra-­articular corticosteroid injections increase the risk of requiring knee arthroplasty 587

can be safely used in these patients. Corticosteroids have a time- OA, the presence of osteophytes in one or both knees but not
and dose-­dependent effect on articular cartilage. High doses, in combination with frequent symptoms in the same knee, or at
frequent injections, or both, such as > 3 mg per injection or a least two additional risk factors for OA (knee symptoms in the
cumulative dose of 18 to 24 mg, have been shown to be associ- previous 12 months, obesity, a history of injury to the knee or
ated with gross cartilage damage and chondrocyte toxicity. Low of knee surgery, family history of OA-­related arthroplasty of
doses, such as < 2 mg to 3 mg per injection or a cumulative dose the knee, Heberden’s nodes, repetitive daily flexion of the knee,
of 8 to 12 mg, have been shown to increase cell growth and and age between 70 and 79 years).
recovery from cartilage damage.7 Two randomized controlled The other inclusion criteria were patients aged between 45
trials have investigated the long-­term effect of intra-­articular and 79 years, living in the USA. Exclusion criteria included
injections on articular cartilage. Raynauld et al5 reported no inflammatory arthritis, contraindications to 3-­Tesla MRI,
long-­term harmful effects of corticosteroid injections on the and bilateral end-­stage OA of the knee. The OAI protocol is
anatomical structure of the knee when injections were repeated available at http://​data-​archive.​nimh.​nih.​gov/​binaries/​content/​
every three months for two years. McAlindon et al,4 however, documents/​ndacms/​resources/​oai/​oai-​study-​protocol/​oai-​study-​
reported that the same pattern of repeated injections resulted in protocol/​ndacms%​3Aresource. The cohort consisted of 4,674
loss of cartilage volume without a long-­term effect on knee pain patients, with a progression cohort of 1,390 (30%) and an
after two years, compared with saline injections. The adverse incidence cohort of 3,284 (70%). At the final follow-­up, 2,938
effect on cartilage volume was, however, small and possibly completed nine years of follow-­up, 844 had sporadic missing
not clinically relevant. Others have indicated that corticosteroid follow-­up visits, and 852 were lost to follow-­up with structural
injections are mainly used for acute symptoms and that repeated missing data (more than half of follow-­up was missing) due to
injections at three-­monthly intervals do not reflect routine clin- patient withdrawal and/or consent not having been obtained at
ical practice, reducing the generalizability of the findings.8 the time of extension of the study after the initial 48 months of
More recently, Zeng et al6 studied the effects of intra-­articular follow-­up. The latter were excluded from the analyses. After
corticosteroid injections on the radiological Kellgren and removing patients lost to follow-­up, 3,822 patients remained
Lawrence (KL) grade of OA9 and the width of the joint space in with a mean age of 61 years (45 to 79) at baseline and 2,233
patients with grade 2 or 3 OA. They concluded that injections (58%) were female.
increase the risk of worsening the grade of OA and joint space Variables. At the baseline and annual follow-­up visits, between
narrowing by three-­fold compared with patients who did not one and nine years, a large set of variables were collected,
receive injections. Although the radiological assessment of the from which we extracted the following details at each visit:
progression of OA using the KL grade or joint space narrowing age, sex, physical activity, body mass index (BMI), education
can be valuable, radiological features of OA do not always (categorized in ‘less than high school graduate’, ‘high school
correlate with symptoms and vice versa.10 Knee pain is ulti- graduate’, ‘some college’, ‘college graduate’, ‘some graduate
mately the indication for arthroplasty. The aim of this study was school’, ‘graduate degree’), occupational activities, access to
to assess whether patients with, or at risk of developing, symp- health care, health insurance cover, treatment centre, family
tomatic OA of the knee and have corticosteroid injections have history of knee arthroplasty surgery, clinical variables (medi-
an increased risk of undergoing (total or partial) arthroplasty. cation use, history of knee surgery, knee symptoms, number of
hyaluronic acid injections, radiological signs of OA (according
Methods to the KL grading scale of the knee and hand), quality of life
Study design. This study used data from the Osteoarthritis measurements (12-­Item Short Form Survey (SF-12),13 func-
Initiative (OAI) database which is available for public access tional scores (Knee injury and Osteoarthritis Outcome Score
at https://​data-​archive.​nimh.​nih.​gov/​oai/. It is a multicentre, (KOOS),14 and Western Ontario and McMasters Osteoarthritis
longitudinal cohort study which included patients with, or at index (WOMAC).15
risk of developing, symptomatic OA of the knee. Patients were Whether the patients had received an intra-­articular injection
enrolled between February 2004 and May 2006 in four centres or undergone (total or partial) arthroplasty was recorded at each
and were followed up for nine years. The present study was visit. For those who had received bilateral injections, the first
conducted and reported according to STROBE guidelines.11 knee that was treated was selected for inclusion. For all other
Patients. The patients in the OAI cohort were divid- patients, the left or right knee was selected randomly.
ed into a progression cohort and an incidence cohort. The Statistical analysis. Because of the annual follow-­up visits,
progression cohort consisted of patients with symptomatic the exact time of both the injections and arthroplasty were un-
OA of the knee, defined as having tibial-­femoral osteophytes known, so this was approximated by taking a random day of the
(OARSI (Osteoarthritis Research Society International) atlas year (by random sampling a value between one and 365) indi-
grade 1 to 3;12 definite osteophytes and mild to severe joint cating the day between the follow-­up visit at which the injection
space narrowing) at baseline and pain, aching, or stiffness on or arthroplasty was reported and the preceding visit.
most days of the month during the previous year. The inci- Patient characteristics at baseline had a small percentage
dence cohort consisted of patients without symptomatic OA of of missing values (< 1%). The KOOS Function in Sport and
the knee (OARSI atlas grade 0;12 no osteophytes or joint space Recreation (KOOS FSR) subdomain that had 944 (25%)
narrowing), but with an increased risk of developing sympto- missing values. All missing values (as described in Table I)
matic OA in one or both knees. The increased risk was defined were imputed using multivariate imputation by chained equa-
as either frequent symptoms without radiological evidence of tions creating ten independent imputed datasets. We performed
VOL. 102-B, No. 5, May 2020
588 S. R. W. Wijn, M. M. Rovers, T. G. van Tienen, G. Hannink

Table I. Characteristics of patients in the overall cohort and the propensity matched datasets at time of the first intra-­articular corticosteroid
injection.
Characteristic Missing at Entire cohort at baseline (n = 3,822) Matched cohort at time of first injection*
baseline, n (%) (n = 1,366)
Treated (n = 796) Control (n = 3,026) SMD Treated (n = 683) Control (n = 683) SMD
Mean age (SD) 0 (0) 62.4 (8.9) 60.9 (9.1) 0.00 68.4 (9.6) 68.4 (9.8) 0.00
Gender, female, % 0 (0) 63.4 57.1 0.11 62.7 62.1 0.01
Mean BMI, kg/m2 (SD) 71 (1.9) 29.8 (4.8) 28.3 (4.7) 0.00 29.7 (4.8) 29.7 (5.3) 0.00
Mean physical activity (PASE)† (SD) 0 (0) 159.9 (83.8) 164.7 (82.2) 0.00 141.2 (75.2) 141.9 (81.2) 0.00
Health care access and health 0 (0) 97.7 96.6 0.06 98.8 98.6 0.01
insurance coverage, %
Education (> college graduate),‡ % 0 (0) 56.0 62.9 -0.11 56.3 57.1 -0.01
Family history of knee arthroplasty 58 (1.5) 16.3 14.6 0.04 15.8 15.8 0.00
surgery, %
Occupational activities, % 0 (0) 72.4 72.4 0.00 72.6 72.2 0.01
Medication use, % 0 (0) 69.8 51.1 0.32 87.3 88.5 -0.03
Hand OA at baseline, % 0 (0) 23.8 21.1 0.05 23.7 23.9 0.00
History of knee surgery at 0 (0) 21.8 12.1 0.21 21.2 20.8 0.01
baseline, %
Knee symptoms at baseline, % 15 (1.0) 64.0 43.6 0.34 62.2 61.3 0.02
Radiological OA at baseline,§ % 25 (1.0) 53.0 38.5 0.24 52.0 51.8 0.00
Mean KOOS Knee related QoL§ 0 (0) 57.7 (22.3) 68.9 (20.9) 0.00 50.5 (20.3) 50.5 (22.5) 0.00
(SD)
Mean KOOS Knee symptoms§ (SD) 0 (0) 79.2 (18.2) 87.9 (13.9) 0.00 71.7 (18.8) 71.8 (20.1) 0.00
Mean KOOS Pain§¶ (SD) 0 (0) 76.0 (19.8) 86.4 (16.1) 0.00 66.7 (19.4) 66.7 (22.1) 0.00
Mean KOOS FSR§¶ (SD) 944 (24.7) 58.2 (27.0) 73.5 (24.8) 0.00 49.4 (26.5) 49.0 (27.3) 0.00
Mean WOMAC stiffness§, ** (SD) 0 (0) 2.2 (1.8) 1.4 (1.5) 0.05 2.6 (1.8) 2.6 (1.9) 0.00
Mean WOMAC pain§, †† (SD) 0 (0) 3.9 (3.8) 2.1 (3.0) 0.00 5.7 (4.0) 5.6 (4.4) 0.00
Mean WOMAC disability§, ‡‡ (SD) 0 (0) 11.5 (12.7) 6.9 (10.2) 0.00 16.9 (12.6) 17.1 (14.2) 0.00
Mean SF-12 Physical scale (SD) 0 (0) 46.5 (9.4) 50.1 (8.4) 0.00 42.2 (9.9) 42.2 (10.3) 0.00
Mean SF-12 Mental scale (SD) 0 (0) 53.8 (8.5) 53.7 (7.7) 0.00 54.1 (9.0) 54.0 (9.2) 0.00
*Time-­dependent propensity score matching resulted in 1,366 matched patients.
†0 to 400.
‡Education is categorized in ‘less than high school graduate’, ‘high school graduate’, ‘some college’, ‘college graduate’, ‘some graduate school’,
and ‘graduate degree’.
§Knee-­specific variable available for both left and right knee. Standardized mean difference < 0.1 is considered appropriate balance.18
¶0 to 100.
**0 to 8.
††0 to 20.
‡‡0 to 68.
BMI, body mass index; FSR, function in sport and recreation; IAC, intra-­articular corticosteroid; KOOS; Knee injury and Osteoarthritis Outcome
Score; OA, osteoarthritis; QoL, quality of life; SF-12, 12-­Item Short Form Survey; SMD, standardized mean difference; WOMAC, Western Ontario
and McMasters Osteoarthritis index.

time-­dependent propensity score matching (tdPSM) to match combined was evaluated by calculating the average standard-
patients at the time of their first injection with available ized absolute mean difference (ASAMD).18
controls at any given timepoint using a 1:1 matching algorithm The effect of intra-­articular injections on the risk of under-
and nearest neighbour matching with a caliper of 0.05.16 The going (total or partial) arthroplasty was estimated using Cox
propensity score was defined as the probability of receiving proportional hazards survival analyses with cumulative injec-
intra-­articular injections based on patient characteristics (age, tions as a time-­dependent covariate and time to (total or partial)
sex, BMI, physical activity, health care access, treatment centre, arthroplasty as the dependent variable. The Cox proportional
education, family history with OA, occupation), clinical vari- hazards survival analyses were performed for each imputation
ables (medication use, grade of knee and hand OA at baseline, dataset and the estimates were pooled using Rubin’s Rule.19
knee symptoms at baseline), quality of life (SF-12 subscales), Statistical significance was set at p < 0.05.
and functional scores (KOOS and WOMAC). Covariate In order to assess the clinical relevance of the estimated
balance was assessed by calculating standardized mean differ- effect, the increase in absolute risk per injection was derived
ences (SMDs) between patients and controls. Patients with a from the Cox proportional hazard model.20 All analyses were
balance < 0.10 SMD were assumed to have appropriate balance performed in R (version 3.5.2, The R Foundation for Statistical
and were visually inspected using balance plots.17 Balance was Computing, Vienna, Austria), using packages ‘mice’, ‘MatchIt’,
assessed for every annual follow-­up visit for all imputed data- and ‘cobalt’. As a sensitivity analysis, the effect of injections on
sets. The overall balance of all visits and the imputation sets the risk of arthroplasty was also estimated in the unmatched
Follow us @BoneJointJ The Bone & Joint Journal
Intra-­articular corticosteroid injections increase the risk of requiring knee arthroplasty 589

Fig. 1

Flowchart showing patient selection. Of the 796 patients who received intra-­articular corticosteroid (IAC) injections, 683 were matched at their first
injection with 683 controls. The dotted line represents 113 patients treated with injections who would receive treatment later during follow-­up but
were matched as controls. KAS, knee arthroplasty surgery; tdPSM, time-­dependent propensity score matching.

cohort. The progression and incidence cohorts were analyzed balanced with SMDs < 0.10 (Figure 2). The overall covariate
separately to detect potential differences between them. balance was appropriate with an ASAMD of 0.072 (SD 0.005).
The hazard ratio (HR) for arthroplasty was estimated to be
1.57 (95% confidence interval (CI) 1.37 to 1.81; p < 0.001) for
Results
each cumulative injection. Each injection increased the absolute
At follow-­up of nine years, 796 (20.8%) had received one or
more intra-­articular injection(s). The mean number of injec- risk of arthroplasty by 9.4% at nine years’ follow-­up compared
tions was (1.55 (1 to 8)) while 401 (10.5%) underwent (total with patients who did not receive injections. Sensitivity analysis
or partial) arthroplasty. Of those who received injections, 249 of the unmatched cohort showed a HR of 2.15 (95% CI 1.96 to
(31.3%) underwent arthroplasty, and of the 3,026 patients who 2.38; p < 0.001) for each cumulative injection. The pooled HRs
did not receive injections, 152 (5.0%) underwent arthroplasty of the incidence and progression sub-­cohorts were 1.78 (95%
(Figure 1). The baseline characteristics are shown in Table I. CI 1.42 to 2.23; p < 0.001) and 1.41 (95% CI 1.19 to 1.67; p <
Additionally, 288 patients received a hyaluronic acid injection. 0.001), respectively (Figure 3).
Time-­dependent propensity score matching resulted in 1,366
matched patients in total for every imputed dataset; 683 patients
at the time of their first injection were matched to 683 control
Discussion
patients at different time points that were at risk of receiving We found that intra-­ articular corticosteroid injections
injections. Because the annual follow-­up visits were matched seemed to be associated with an increased risk of (total or
sequentially, 113 treated patients (of the 796) were matched partial) arthroplasty of the knee in patients with, or at risk of
as controls before they received their first injection. Some developing, symptomatic OA of the knee. The HR was 1.57
controls had multiple time points matched to two different (95% CI 1.37 to 1.81) for each subsequent injection. The results
treated patients. Thus, at one-­ year follow-­up, two treated suggest that by nine years of follow-­up, each injection increased
patients received their first injection and were matched to a the absolute risk of requiring arthroplasty by 9.4% compared
single control at four- and eight-­year follow-­up resulting in 23 with similar patients who did not receive injections. This is in
double matches. In total, 1,343 unique patients were matched line with previous studies showing that intra-­articular corti-
(Figure 1). At every time point, most characteristics were well costeroid injections have a detrimental effect on the articular
VOL. 102-B, No. 5, May 2020
590 S. R. W. Wijn, M. M. Rovers, T. G. van Tienen, G. Hannink

Fig. 2

Covariate balance at each timepoint before and after propensity score matching, pooled for the imputation sets. The dark grey lines indicate the
standardized mean difference between the intervention and control groups before matching. The light grey lines indicate the standardized mean
difference between the intervention and control groups after matching. Dotted vertical lines represent -0.10 and 0.10 standardized mean difference;
a standardized mean difference between these lines indicates good balance between patients who received intra-­articular corticosteroid (IAC)
injections and those who did not. BMI, body mass index; FSR, function in sport and recreation; KOOS; Knee injury and Osteoarthritis Outcome
Score; OA, osteoarthritis; QoL, quality of life; SF-12, 12-­Item Short Form Survey; WOMAC, Western Ontario and McMasters Osteoarthritis index.

cartilage of the knee, possibly accelerating the progression of which are injected with corticosteroids, have been shown to be
OA.4,6,7 chondrotoxic, especially in an already inflamed or damaged
Zeng et al6 studied the association between intra-­articular knee, contributing to cartilage loss.21 Secondly, although we
injections and KL grade and joint space narrowing using OAI adjusted for several patient characteristics such as education,
data and reported a HR of 3.02 (95% CI 2.19 to 4.16) and 4.67 occupation, health care access, and health insurance cover,
(95% CI 2.92 to 7.47) for deteriorating grade of OA and 2.93 patients who received injections may be more likely to receive
(95% CI 2.13 to 4.02) and 3.26 (95% CI 1.78 to 5.96) for joint treatment because they experience pain differently, lowering
space narrowing for initial and continuous intra-­articular injec- their threshold to seek attention.22,23 Consequently, they may
tions, respectively, which is higher compared with our results. be more likely to undergo arthroplasty than patients who have
This difference can be explained by differences in population, similar levels of pain and function. Thirdly, the reduction of pain
treatment, outcome, and duration of follow-­up. Zeng et al only might increase the range of movement of the knee, provoking
included patients with a KL score of 2 or 3 with a maximum peak stress and increasing the strain on the joint. This may lead
follow-­up of four years, did not estimate the risk per cumulative to increased degradation of cartilage.
injection, and used the progression of radiological OA as the The strength of our study is the large number of patients,
outcome of interest. who were followed for nine years in an observational setting
Our results may be explained by several independent using data from the OAI. This database was developed to docu-
or combined hypothetical reasons. Corticosteroids induce ment the natural history of OA of the knee across a wide range
chondrocyte toxicity and gross cartilage loss and inflamma- of patients with different grades of OA and characteristics,
tion.7 Loss of cartilage can increase symptoms, increasing the using regular follow-­up and standardized questionnaires. This
probability that patient and physician will opt for arthroplasty. enabled us to analyze and compare patient characteristics, clin-
Local anaesthetic agents, such as bupivacaine or lidocaine, ical variables, and standardized quality of life and functional
Follow us @BoneJointJ The Bone & Joint Journal
Intra-­articular corticosteroid injections increase the risk of requiring knee arthroplasty 591

model. In both cases, hyaluronic acid did not alter the effect found
in our analysis. A portion of the OAI had incomplete follow-­up
of patients who received an injection, therefore we excluded
852 patients from the analysis. This was essential because the
missing values were, in those with incomplete follow-­up, struc-
tural and could not be assumed to be missing at random. In
order to check if the analysis had an impact on the results, we
additionally performed a last observation carried forward impu-
tation as sensitivity analysis to impute the subsequent missing
values. This method showed a similar result compared with our
analysis. Lastly, we used arthroplasty as the ultimate endpoint of
OA progression. Ideally, in order to assess the association between
intra-­articular injections and arthroplasty, radiological evidence
of cartilage deterioration such as KL grade or cartilage volume
should be included in the matching algorithm to improve the
balance between the two groups. Unfortunately, the OAI did not
routinely acquire MRIs and radiographs annually, and if available,
a large portion were missing. Nevertheless, we do not expect that
the inclusion of radiological evidence would change our conclu-
sions, as we included a wide range of variables relating to the
Fig. 3
progression of disease such as pain, and functional and quality
Hazard ratios with 95% confidence intervals. Values > 1 indicate that
of life scores in our model to account for potential imbalances
intra-­articular corticosteroid (IAC) injections had an increased risk of between the two groups.
arthroplasty compared with those who did not receive injections. The In conclusion, intra-­articular corticosteroid injections seem to
results are shown for the primary analysis, the unmatched cohort, the
be associated with an increased risk of knee arthroplasty, for each
incidence cohort, and the progression cohort. KAS, knee arthroplasty
surgery. cumulative injection in patients with, or at risk of developing,
symptomatic OA of the knee. These findings suggest that a conser-
vative approach to injections of corticosteroids into the knee joint
scores, both at baseline and at the time of treatment. Those
to treat symptoms related to OA should be recommended.
who received injections had, in general, higher pain scores
and lower functional scores compared with those who did not Take home message
receive injections. We used tdPSM to create matches of patients -- Corticosteroid injections seem to be associated with an
increased risk of knee arthroplasty in patients with, or at risk
at the time of their first injection, instead of regular propensity of developing, symptomatic osteoarthritis of the knee.
score matching that achieves a covariate balance at baseline.16,24 -- These results suggest that a conservative approach regarding the
This resulted in comparable patients at the time of treatment, treatment of these patients with corticosteroid injections should be
eliminating changes that occurred between their inclusion at the recommended.

start of the study and their first injection. Thus, a patient who
reported their first injection at the fifth follow-­up visit might Twitter
have a lower functional score compared with a control whose Follow S. R. W. Wijn @Stan_Wijn
function was similar at baseline. Follow M. M. Rovers @MaroeskaRovers
Follow T. G. van Tienen @TonyvanTienen
Some potential limitations should be highlighted. Time-­
Follow G. Hannink @gerjonhannink
dependent propensity score matching was used to account for
bias, but there remains a risk of residual confounding due to vari-
ables which were either not measured or not included. However, References
the relationship between injections and arthroplasty appears to 1. McAlindon TE, Bannuru RR, Sullivan MC, et  al. OARSI guidelines for
be robust. The sensitivity analyses that were performed showed the non-­ surgical management of knee osteoarthritis. Osteoarthritis Cartilage.
2014;22(3):363–388.
similar effects. Treatment regimens for intra-­articular injec-
2. He W-­W, Kuang M-­J, Zhao J, et al. Efficacy and safety of intraarticular hyaluronic
tions and arthroplasty might have changed during the study acid and corticosteroid for knee osteoarthritis: a meta-­ analysis. Int J Surg.
period, which might reduce the generalizability of the findings. 2017;39:95–103.
We noticed a small increase in the rate of injections and arthro- 3. Liu S-­H, Dubé CE, Eaton CB, et  al. Longterm effectiveness of intraarticular
plasty with the passage of time during the study period of nine injections on patient-­ reported symptoms in knee osteoarthritis. J Rheumatol.
2018;45(9):1316–1324.
years. However, this could also be due to the increasing age of 4. McAlindon TE, LaValley MP, Harvey WF, et  al. Effect of intra-­ articular
the cohort and progression of OA. A proportion of the patients triamcinolone vs saline on knee cartilage volume and pain in patients with knee
who were treated with injections also received hyaluronic acid osteoarthritis: a randomized clinical trial. JAMA. 2017;317(19):1967–1975.
injections into the knee; of the 288 patients that received a hyal- 5. Raynauld J-­P, Buckland-­Wright C, Ward R, et al. Safety and efficacy of long-­
uronic acid injection, 218 (76%) also received corticosteroid term intraarticular steroid injections in osteoarthritis of the knee: a randomized,
double-­blind, placebo-­controlled trial. Arthritis Rheum. 2003;48(2):370–377.
injections. We analyzed the effect of hyaluronic acid on our 6. Zeng C, Lane NE, Hunter DJ, et  al. Intra-­ Articular corticosteroids and the
results by including it once as a covariate in the propensity score risk of knee osteoarthritis progression: results from the osteoarthritis initiative.
model and once as a variable in the Cox proportional hazard Osteoarthritis Cartilage. 2019;27(6):855–862.

VOL. 102-B, No. 5, May 2020


592 S. R. W. Wijn, M. M. Rovers, T. G. van Tienen, G. Hannink

7. Wernecke C, Braun HJ, Dragoo JL. The effect of intra-­articular corticosteroids on


Author information:
articular cartilage. Orthop J Sports Med. 2015;3(5):232596711558116. S. R. W. Wijn, MSc, PhD candidate, Evidence-­Based Surgery
8. Maloney WJ, Jevsevar DS, Shea KG. Long-­Term intra-­articular steroid injections G. Hannink, PhD, Senior scientist, Evidence-­Based Surgery
and knee cartilage. JAMA. 2017;318(12):1184–1185. Department of Operating Rooms, Radboud Institute for Health Sciences,
9. Kellgren JH, Lawrence JS. Radiological assessment of osteo-­arthrosis. Ann Radboud University Medical Center, Nijmegen, the Netherlands.
Rheum Dis. 1957;16(4):494–502. M. M. Rovers, PhD, Professor, Evidence-­Based Surgery, Department
10. Guermazi A, Niu J, Hayashi D, et al. Prevalence of abnormalities in knees detected of Operating Rooms, Radboud Institute for Health Sciences, Radboud
by MRI in adults without knee osteoarthritis: population based observational study University Medical Center, Nijmegen, the Netherlands; Radboud University
Medical Center, Radboud Institute for Health Sciences, Department of
(Framingham osteoarthritis study). BMJ. 2012;345:e5339.
Health Evidence, Nijmegen, the Netherlands.
11. von Elm E, Altman DG, Egger M, et  al. The strengthening the reporting of
observational studies in epidemiology (STROBE) statement: guidelines for reporting T. G. van Tienen, MD PhD, Orthopaedic Surgeon, Laurentius Hospital,
Roermond, The Netherlands; Chief Medical Officer, ATRO Medical BV,
observational studies. Int J Surg. 2014;12(12):1495–1499. Department of Orthopedics, Radboud Institute for Health Sciences,
12. Altman R, Hochberg M, Murphy W, Wolfe F, Lequesne M. Atlas of individual Radboud University Medical Center, Nijmegen, the Netherlands.
radiographic features in osteoarthritis. Osteoarthritis and Cartilage. 1995;3(A):3–70.
13. Ware J, Kosinski M, Keller SD. A 12-­Item short-­form health survey: construction of Author contributions:
scales and preliminary tests of reliability and validity. Med Care. 1996;34(3):220–233. S. R. W. Wijn: Conceived the study, Collected, validated, and analyzed the
14. Roos EM, Roos HP, Lohmander LS, Ekdahl C, Beynnon BD. Knee Injury and data, Drafted and gave final approval of the manuscript.
Osteoarthritis Outcome Score (KOOS)--development of a self-­administered outcome M. M. Rovers: Conceived the study, Drafted and gave final approval of the
measure. J Orthop Sports Phys Ther. 1998;28(2):88–96. manuscript.
15. Bellamy N, Buchanan WW, Goldsmith CH, Campbell J, Stitt LW. Validation T. G. van Tienen: Conceived the study, Drafted and gave final approval of
study of WOMAC: a health status instrument for measuring clinically important patient the manuscript.
relevant outcomes to antirheumatic drug therapy in patients with osteoarthritis of the G. Hannink: Conceived the study, Collected, validated, and analyzed the
hip or knee. J Rheumatol. 1988;15(12):1833–1840. data, Drafted and gave final approval of the manuscript.
16. Lu B. Propensity score matching with time-­ dependent covariates. Biometrics.
Funding statement:
2005;61(3):721–728.
Junior Research project (2018) grant provided by the Radboud Institute
17. Austin PC. An introduction to propensity score methods for reducing the effects of
for Health Sciences, Radboud University Medical Centre, Nijmegen, The
confounding in observational studies. Multivariate Behav Res. 2011;46(3):399–424. Netherlands. The funding source had no role in the design or conduct of
18. Girman CJ, Gokhale M, Kou TD, et al. Assessing the impact of propensity score the study; collection, analysis, or interpretation of the data; or writing of
estimation and implementation on covariate balance and confounding control within the report.
and across important subgroups in comparative effectiveness research. Med Care. No benefits in any form have been received or will be received from a
2014;52(3):280–287. commercial party related directly or indirectly to the subject of this article.
19. Rubin DB. Multiple Imputation for Nonresponse in Surveys. New York: Wiley, 2004.
20. Austin PC. Absolute risk reductions and numbers needed to treat can be obtained Acknowledgements:
from adjusted survival models for time-­ to-­
event outcomes. J Clin Epidemiol. The OAI is a public-­private partnership composed of five contracts (N01-­
2010;63(1):46–55. AR-2 to 2258; N01-­AR-2 to 2259; 01-­AR-2 to 2260; N01-­AR-2 to 2261;
21. Kreuz PC, Steinwachs M, Angele P. Single-­Dose local anesthetics exhibit a N01-­AR-2 to 2262) funded by the National Institutes of Health, a branch
type-, dose-, and time-­dependent chondrotoxic effect on chondrocytes and cartilage: of the Department of Health and Human Services, and conducted by the
a systematic review of the current literature. Knee Surg Sports Traumatol Arthrosc. OAI Study Investigators. Private funding partners include Merck Research
Laboratories; Novartis Pharmaceuticals Corporation, GlaxoSmithKline; and
2018;26(3):819–830.
Pfizer, Inc. Private sector funding for the OAI is managed by the Founda-
22. Rosemann T, Laux G, Szecsenyi J, Wensing M, Grol R. Pain and osteoarthritis in
tion for the National Institutes of Health. This manuscript was prepared
primary care: factors associated with pain perception in a sample of 1,021 patients.
using an OAI public use data set and does not necessarily reflect the
Pain Med. 2008;9(7):903–910.
opinions or views of the OAI investigators, the NIH, or the private funding
23. Langley P, Müller-­Schwefe G, Nicolaou A, et al. The societal impact of pain in
partners.
the European Union: health-­related quality of life and healthcare resource utilization.
J Med Econ. 2010;13(3):571–581. Ethical review statement:
24. Groenwold RHH, Hak E, Hoes AW. Quantitative assessment of unobserved This study did not require ethical approval.
confounding is mandatory in nonrandomized intervention studies. J Clin Epidemiol.
This article was primary edited by P. Walmsley and first proof edited by J.
2009;62(1):22–28.
Scott.

Follow us @BoneJointJ The Bone & Joint Journal

You might also like