Article Text
Abstract
Objectives Guidelines recommend intra-articular glucocorticoid injection in patients with painful hip osteoarthritis. However, intra-articular hip injection is an invasive procedure. The efficacy of systemic glucocorticoid treatment for pain reduction in hip osteoarthritis is unknown. This randomised, double-blind, trial assessed effectiveness in hip pain reduction of an intramuscular glucocorticoid injection compared with a placebo injection in patients with hip osteoarthritis.
Methods Patients with painful hip osteoarthritis were randomised to either 40 mg triamcinolone acetate or placebo with an intramuscular injection into the gluteus muscle. The primary outcomes were severity of hip pain at rest, during walking (0–10) and WOMAC pain at 2-week postinjection. We used linear mixed models for repeated measurements at 2, 4, 6 and 12 weeks for the intention-to-treat data analysis.
Results Of the 107 patients randomised, 106 could be analysed (52 in the glucocorticoid group, 54 in the placebo group). At 2-week follow-up, compared with placebo injection, the intramuscular glucocorticoid injection showed a significant and clinically relevant difference in hip pain reduction at rest (difference −1.3, 95% CI −2.3 to −0.3). This effect persisted for the entire 12-week follow-up. For hip pain during walking, the effect was present at 4-week, 6-week and 12-week follow-ups, and for WOMAC pain the effect was present at 6-week and 12-week follow-up.
Conclusions An intramuscular glucocorticoid injection showed effectiveness in patients with hip osteoarthritis on one of the three primary outcomes at 2-week postinjection. All primary outcomes showed effectiveness from 4 to 6 weeks, up to a 12-week follow-up.
Trial registration number NTR2966.
- osteoarthritis
- treatment
- corticosteroids
Statistics from Altmetric.com
Introduction
Several international guidelines recommend intra-articular (IA) glucocorticoid injections for patients with hip osteoarthritis (OA) experiencing moderate to severe pain and not responding to oral analgesics.1–3 A systematic review on the efficacy of intra-articular steroids in moderate/severe hip OA included five randomised controlled trials (RCT) and the assessed quality of the studies was high.4 The treatment effect was large at 1 week post-injection, but declined afterwards. At 8 weeks, there were two trials that reported a reduction in pain with a moderate effect size.4
However, injection into the hip joint is challenging because the joint cannot be palpated and is adjacent to important neurovascular structures. An IA hip injection is best performed under fluoroscopic or ultrasound guidance.
A serious side effect of an IA injection is a septic arthritis. The incidence of this side effect is very low, and scarce in the available literature. A systematic review and meta-analysis of IA injection in knee OA comparing effectiveness of pharmacological interventions included 29 studies (3152 patients, 9500 IA glucocorticoid injections) and reported only 1 septic arthritis (in the IA placebo group).5
A systemic effect of glucocorticoids on joint pain has been indicated in patients with subacromial impingement shoulder pain. A double-blinded RCT showed no important differences in effectiveness on pain of ultrasound-guided subacromial glucocorticoid injection compared with gluteal injection.6 A systemic effect of glucocorticoids was also suggested in an RCT reporting the effect of local glucocorticoid injection for greater trochanteric pain syndrome: patients with concurrent hip OA or chronic low back pain had an equal or even more pronounced decrease in pain.7 8
If an intramuscular (IM) glucocorticoid injection is shown to have a clinically relevant effect on pain, this would offer a less complex alternative treatment for episodes of increased pain in hip OA. Therefore, this study assessed the efficacy of an IM glucocorticoid injection compared with an IM placebo injection on hip pain severity in patients with hip OA who were not responding to oral analgesics.
Methods
Trial design
This was a multicentre, double-blinded, randomised controlled superiority trial with two parallel groups and a follow-up period of 12 weeks: details of the study protocol were published earlier.9 10 The Medical Ethics Committee of the Erasmus University Medical Centre (EMC; Rotterdam) approved the study protocol (MEC2011-115) and all included patients provided written informed consent.
Patients
Patients with hip OA were invited to participate in the trial by general practitioners and orthopaedic surgeons located in the south-west of the Netherlands. Patients (aged >40 years) were eligible for inclusion if they met the American College for Rheumatology (ACR) clinical criteria for hip OA during clinical screening and radiological evidence of hip OA was present (Kellgren & Lawrence ‘score’ (KL) ≥2).11 12 Patients were included if they had symptomatic disease for ≥6 months, and had moderate to severe hip pain score ≥3 (scale 0–10; 0=no pain) despite the use of oral analgesics at time of inclusion.
Radiological hip OA was scored on an anterior–posterior pelvic radiograph of (at most) 6-month old. The radiological grade of hip OA was scored by two researchers (DD, PKB) independently and the interobserver reliability was κ=0.7 for KL <2 vs KL ≥2. In case of disagreement, a consensus was formed during a consensus meeting. If a patient had bilateral hip OA, the more painful hip was selected as the study hip.
Patients were excluded if they had diabetes mellitus, were using oral glucocorticoids, had local/systemic infection, had presence of inflammatory rheumatic diseases (eg, rheumatoid arthritis, psoriatic arthritis and spondylarthropathies), coagulopathy, used coumarins, had a gastric ulcer, allergy to glucocorticoids, radiological signs of osteonecrosis, had an IA injection in the hip in the previous 6 months, were on the waiting list for total hip replacement (THR) surgery or were unable to complete questionnaires in Dutch.
Interventions
Patients received either 40 mg triamcinolone acetate (1 mL) or 1 mL normal saline (placebo) with an IM injection. At the research centre, the trial nurse administered the allocated injection in the upper lateral quadrant of the gluteal musculature on the ipsilateral side of the study hip.
Randomisation
An independent pharmacy assistant allocated each included patient based on a computerised randomisation list using random blocks of 2 and 4 to either placebo (saline) injection or triamcinolone acetate 40 mg injection. Randomisation was stratified for setting (general practice and orthopaedic outpatient clinic). After randomisation, the vials for the injections were prepared, packed and sealed in an identical way for both groups by the pharmacy of the EMC. The randomisation list was available only to the pharmacy assistant.
Blinding
In this trial, the outcome assessors, patients, treating physicians, researchers (including the statistical analyses) and research assistants involved in data collection were blinded to the content of the injections. Since normal saline is transparent and triamcinolone acetate is a white solution, the independent nurse who prepared and injected the injection was non-blinded. To assure blinding of the others, the independent trial nurse prepared and administered the injection out of sight of the patient, assessors, treating physicians and the researchers. This trial nurse was not involved with follow-up measurements. After preparation and before injection, the syringe was covered with an opaque foil to assure blinding of the patient.
Outcomes
The primary outcomes were severity of hip pain at 2-week postinjection measured on an 11-point numerical rating scale (NRS: 0–10, 0=no pain) at rest and during walking, and measured with the Western Ontario and McMaster University Osteoarthritis Index pain subscale (WOMAC pain: 0–100, 0=no symptoms).13 14
Secondary outcomes were the primary outcomes at 4-week, 6-week and 12-week follow-ups. Additional secondary outcome measures were WOMAC function and stiffness, WOMAC total score, Hip disability and Osteoarthritis Outcome Score for pain (HOOS pain), quality of life (EQ-5D), Intermittent and Constant Osteoarthritis Pain (ICOAP) and patients’ perceived recovery assessed on a 7-point Likert scale.15–17 At all time points, the WOMAC and ICOAP scales are presented as normalised scores (0–100, 0=no symptoms). The HOOS subscale is presented as normalised score (0–100, 100=no symptoms). Also recorded was patients’ medical consumption, including analgesic use and adverse reactions at all time points. Patients were allowed to use escape pain medication as needed.
Another secondary outcome was the percentage of responders as defined by the OMERACT-OARSI criteria (improvement in at least two of the three following domains: ≥20% improvement in WOMAC pain, ≥20% improvement in WOMAC function and markedly improved on patients’ global assessment).16 For patients’ global assessment the 7-point Likert scale for patients’ perceived recovery was dichotomised in ‘improved’ (scores: completely recovered, almost completely recovered and slightly recovered) and ‘not improved’ (scores: no change, slightly worse, significantly worsened and worse than ever).
At baseline and at a 12-week follow-up, patients visited the research centre to undergo a physical examination of hips, spine and knees. At baseline, blood samples were collected to measure the erythrocyte sedimentation rate and C reactive protein to gain insight in the inflammatory processes.14
Sample size
Power calculations were based on a study with similar inclusion criteria.18 A 10-point difference (SD 20) on hip pain at rest and during walking (visual analogue scale: 0–100) was assumed to be the minimal clinically important difference between both groups.19 The accompanying effect size of 0.5 is from the clinical standpoint a moderate effect and in general considered clinically relevant.20 For our study, this implies one point on the NRS pain scale. With a power of 80% and an alpha of 5%, 64 patients per group were required (including 5% loss to follow-up=67 patients per group). The same sample size was needed when an 8-point difference (SD 16) on the standardised WOMAC total score and pain subscale score (0–100) was assumed as a clinically relevant difference between the groups.
Statistical analysis
The intention-to-treat principle was used. Descriptive statistics were used to describe patients’ characteristics at baseline, items of physical examination and the severity of radiologic hip OA. Linear mixed models with repeated measures were used for continuous outcomes. When patients underwent a THR, data of these patients were included up to the date of surgery. To model the covariance of repeated measures by patients, the option for data structure in the analyses was set on ‘unstructured’, because this yielded the lowest Akaike’s information criterion. Fixed effects were time and time by treatment. Analyses were adjusted for baseline variables and for the baseline value of the outcome of interest.9 Therefore, at a 2-week follow-up presented are the outcomes at 2 weeks, corrected for baseline scores. At 4 weeks, presented are the results at a 4-week follow-up time point, corrected for baseline and 2 weeks scores and so on.
Generalised estimating equations (GEE) analyses with repeated measures were performed for the dichotomous outcomes perceived improvement, and the OMERACT-OARSI responder. Before GEE analyses, multiple imputations were performed for missing values, creating 100 imputed datasets.
The Pearson χ2 test was used to analyse differences between groups concerning medical consumption, analgesic use and adverse events. An explorative, predefined, subgroup analysis was performed assessing the interaction effects between injections and setting on the primary outcomes.9
All analyses were performed using SPSS V.24.
Results
Patient flow
A total of 422 invited patients contacted the research centre and were screened for eligibility; of these, 92 refused to participate and 223 did not meet the inclusion criteria (figure 1).
Finally, 107 patients provided informed consent: 53 were randomised to the glucocorticoid injection and 54 to the placebo injection. One patient in the glucocorticoid group withdrew his consent just before the appointment for baseline physical examination and subsequent injection, because his pain had resolved spontaneously. Because this patient did not receive the allocated treatment and was not willing to send us the completed baseline questionnaire or any follow-up questionnaires, he was not included in the analyses.
Recruitment
Recruitment of patients took place between September 2011 and October 2014 and follow-up measurements were done until January 2015. Of the 107 included patients, general practitioners referred 81 patients.
Lost to follow-up
At a 6-week follow-up, one patient in the glucocorticoid group reported being scheduled for a THR; in the placebo group two patients (at 4-week and 6-week follow-up, respectively) reported being scheduled for a THR. One patient in the placebo group was not willing to participate after 6 weeks due to logistical problems.
Patient population
Of all patients, 52 received the allocated glucocorticoid injection and 54 the allocated placebo injection, and were included in the analyses. Baseline characteristics of both patient groups are presented in table 1. Of the 106 patients, 73 (68%) were women; the mean age was 64 (SD 11) years, and the duration of hip OA symptoms was ≥1 year for 74 (70%) patients.
The KL score of hip OA at baseline, ethnicity, morning hip stiffness and patients’ expected effect of the injection changed the estimates for the primary outcome ≥10%. We adjusted for these baseline variables in the statistical analyses.
Primary outcomes
At a 2-week follow-up, compared with the placebo injection, the glucocorticoid injection showed a significant association with hip pain reduction at rest (between group difference −1.3, 95% CI −2.3 to −0.3) (table 2 and figure 2). Also, at a 2-week follow-up, there were no significant associations between glucocorticoid injection and hip pain during walking (difference −0.9, 95% CI −1.9 to 0.1) and WOMAC pain (difference −6.1, 95% CI −13.4 to 1.2). The correlation between the primary endpoints ranged from r 0.64 to r 0.84. The results of the unadjusted linear mixed model analysis were similar (online supplementary table S1).
Supplementary file 1
Supplementary file 3
Secondary outcomes
At 4-week, 6-week and 12-week follow-ups, the glucocorticoid injection was associated with a significant hip pain reduction at rest and during walking (table 2 and figure 2). Moreover, at almost all follow-up measurements, the estimates showed significant differences in favour of the glucocorticoid injection on WOMAC pain, function, stiffness and total; HOOS pain; and ICOAP total, intermittent and constant. No significant differences between groups were found for quality of life (table 2). At a 2-week follow-up, perceived improvement and the OMERACT-OARSI responders showed a significant effect in favour of glucocorticoid injection: RR 1.7 (95% CI 1.1 to 2.7) and 2.0 (95% CI 1.1 to 3.6), respectively (table 3).
Adverse events and medical consumption
In the glucocorticoid group, 19 patients reported 27 adverse events. In the placebo group 13 patients reported 18 adverse events. All adverse events were classified as non-serious (table 4). Hot flushes, headache and itching were reported most frequently in the glucocorticoid group. There were no significant differences in medical consumption between the two groups (online supplementary table S2).
Supplementary file 2
Ancillary analyses
In the explorative subgroup analyses, the results of the interaction of setting on injections showed no significant differences between the two groups (NRS at rest 1.5, 95% CI −0.6 to 3.7).
Discussion
We found that an IM glucocorticoid injection showed effectiveness in patients with hip OA on one of the three primary outcomes at a 2-weeks postinjection and that the effect is probably clinically relevant. All primary outcomes and almost all secondary pain and function outcomes showed effectiveness from 4 to 6 weeks that lasted for the entire 12- weeks follow-up. The highest effects were seen at 4–12 weeks follow-up instead of at the expected 2-weeks follow-up.
In this study, three primary endpoints were used. This increases the risk of a type 1 error. However, with a bonferroni correction for the three primary outcomes (p<0.017) pain at rest was still significantly associated with pain reduction at a 2-week follow-up.
It was surprising that hip pain reduction after IM glucocorticoid injection was still present at a similar degree at 12-week follow-up. Previous studies on IA glucocorticoid injections in hip OA studies mostly showed a peak effect after 1–3 weeks, but still showed significant pain reduction at 8–12 weeks follow-up.18 21–23 In a recent Cochrane review on IA glucocorticoid injections in knee OA, the effects were moderate at 1–2 weeks after treatment (effect size 0.48), small to moderate at 4–6 weeks (effect size 0.41) and small at 13 weeks after treatment (effect size 0.22).24 Therefore, our findings should be replicated in future research.
Also, a surprising finding was that, for patients’ perceived improvement and the OMERACT-OARSI responders, there was a significant association in favour of glucocorticoid injection only at a 2-week follow-up. There are two possible explanations for this. First, for patients’ perceived improvement we dichotomised the 7-point Likert scale, which resulted in less power. Second, the answer options we provided in the questionnaire were not clearly formulated. For example, answer options were ‘completely recovered’, ‘almost completely recovered’ and ‘slightly recovered’, resulting in a large step between ‘almost completely recovered’ and ‘slightly recovered’. A better delineation would have been: ‘completely improved’, ‘markedly improved’ and ‘slightly improved’.
In our RCT, we gave a single IM glucocorticoid injection. In clinical practice, patients are sometimes offered multiple IA injections per year. A recent trial in patients with knee OA showed that three monthly injection with glucocorticoid in the knee joint during 2 years resulted in a significantly greater cartilage volume loss than did saline injection.25 However, there are concerns that even one IA glucocorticoid injection may cause toxicity to chondrocytes, possibly resulting in progression of OA. This has been confirmed in in vitro and in vivo animal studies and needs further study in humans.26 27 It is unknown whether a single IM glucocorticoid injection has a negative effect on chondrocytes.
The effect of an IA glucocorticoid injection for pain reduction in hip OA has been reported in several RCTs.18 21–23 The present study shows that systemic treatment with an IM glucocorticoid injection is effective compared with placebo on pain reduction in patients with hip OA. The administration of an IM injection is much easier than an IA hip joint injection without the need for ultrasound/radiologic guidance and can, therefore, be performed in both secondary and primary care. However, the comparative effectiveness of an IM injection compared with an IA injection is unknown.
Strengths and limitations
An important strength of our placebo RCT is that it was blinded for outcome assessors, patients, treating physicians, and researchers (including the statistical analyses). Second, we had a high follow-up rate, that is, 100% at 2 weeks in both groups, which was the primary outcome time point. At a 12-week follow-up, the follow-up rate was 98% in the glucocorticoid injection group and 94% in the placebo group.
A limitation was not including 128 participants. Nevertheless, we were still able to significantly detect a 10-point difference on a 0–100 scale that was beforehand assumed to be the minimal clinically important difference.
A total of 92 patients (22%) declined to participate, because they did not want to risk receiving a placebo. Similarly, Lambert et al found that almost 50% of their patients refused to participate to avoid the risk of placebo treatment.23 Second, although our patients reported moderate to severe pain (NRS ≥3), the main exclusion reason was that they had not used any analgesics during the past 3 weeks. It would be interesting to establish why patients with moderate to severe pain do not take analgesics. For hip OA little is known about patients’ preference and perceptions on treatment. In knee OA, although about 75% of patients use over-the-counter oral analgesics, they do not perceive this treatment as being the most effective; instead, patients perceived viscosupplementation (74.1%), narcotics (67.8%) and steroid injection (67.6%) as being the most effective.28
To exclude patients with other painful hip diseases, we set strict criteria for the presence of radiological hip OA (KL ≥2). A final point was the exclusion of patients with diabetes mellitus, a frequently occurring comorbidity in this patient population. It is well known that glucocorticoids can give rise to hyperglycaemia in diabetic patients.29 For reasons of patient safety, our medical ethics committee stipulated that we excluded diabetic patients from the present trial; this means that we cannot extrapolate our results to patients with diabetes and hip OA.
Based on the present results, we conclude that an IM glucocorticoid injection showed effectiveness in patients with hip OA in only one of the three primary outcomes at 2-weeks post-injection, but that all primary outcomes and almost all secondary pain and function outcomes showed effectiveness from 4 to 6 weeks that lasted for the entire 12-week follow-up.
References
Footnotes
Handling editor Josef S Smolen
Contributors DMJD: trial coordination, data-acquisition, data-analysis and data interpretation, writing article. PAJL: study design, participation in trial coordination, data interpretation and extensive review of article. MR, MK, JANV and PJEB: study design and review of article. PKB: study design, radiological assessment, data interpretation and review of article. SMAB-Z: conceived the study, study design, data interpretation and extensive review of article. All authors read and approved the final manuscript.
Funding Financial support was received from the Dutch Arthritis Foundation and the NutsOhra Fund.
Disclaimer There was no role of both funding sources.
Competing interests MK reports grants from Abbvie, grants from Levicept, grants from GlaxoSmithKline, grants from Arthritis Rheumatology, grants from APPROACH consortium, grants from Foreum, grants from TI-Pharma, grants from Pfizer; outside the submitted work. SMAB-Z reports grants from Dutch Arthritis Foundation, grants from Nuts Ohra, during the conduct of the study. Pending grants from Dutch Arthritis Foundation, at the Netherlands Organization for Health research and development, and EU Horizon 2020, outside the submitted work. Grants received from the Dutch Arthritis Foundation, Netherlands organization for Health research and development, Nuts-Ohra, and EU Fp7, outside the submitted work.
Patient consent Obtained.
Ethics approval Medical Ethics Committee of the Erasmus University Medical Center.
Provenance and peer review Not commissioned; externally peer reviewed.