Objectives To compare the effect of a single infusion of zoledronic acid (ZA) with placebo on knee pain and bone marrow lesions (BMLs).
Methods Adults aged 50–80 years (n=59) with clinical knee osteoarthritis and knee BMLs were randomised to receive either ZA (5 mg/100 ml) or placebo. BMLs were determined using proton density-weighted fat saturation MR images at baseline, 6 and 12 months. Pain and function were measured using a visual analogue scale (VAS) and the knee injury and osteoarthritis outcome score (KOOS) scale.
Results At baseline, mean VAS score was 54 mm and mean total BML area was 468 mm2. VAS pain scores were significantly reduced in the ZA group compared with placebo after 6 months (−14.5 mm, 95% CI −28.1 to −0.9) but not after 3 or 12 months. Changes on the KOOS scales were not significant at any time point. Reduction in total BML area was greater in the ZA group compared with placebo after 6 months (−175.7 mm2, 95% CI −327.2 to −24.3) with a trend after 12 months (−146.5 mm2, 95% CI −307.5 to +14.5). A greater proportion of those in the ZA group achieved a clinically significant reduction in BML size at 6 months (39% vs 18%, p=0.044). Toxicity was as expected apart from a high rate of acute phase reactions in treatment and placebo arms.
Conclusions ZA reduces knee pain and areal BML size and increases the proportion improving over 6 months. Treatment of osteoarthritis may benefit from a lesion specific therapeutic approach.
Clinical trial registration number ACTRN 12609000399291.
Statistics from Altmetric.com
Knee osteoarthritis (OA) is a leading cause of chronic disability. While there are therapies for symptoms, there are currently no approved disease-modifying OA drugs available which modify structural progression in OA. Bone marrow lesions (BMLs) are regions of increased signal intensity within the bone marrow and appear to be a promising target. They are the first sign of OA after experimental ligament damage and precede cartilage erosion and degeneration in an animal model of OA,1 and strongly correlate with knee pain in humans.2,–,4 Incident2 and progressing4 ,5 BMLs have been shown to be associated with development of knee pain. Further, a reduction in BML size is associated with pain improvement.5 Importantly, BMLs are also associated with structural changes. They predict joint space loss on x-ray,6 cartilage defect progression7 and cartilage loss on MRI8,–,10 as well as knee replacement surgery.5 ,11 ,12
Despite this, there are no therapies in use. Meizer et al report on uncontrolled studies of patients treated with the prostacyclin analogue iloprost,13 ,14 and a study compares iloprost with core decompression.15 BML end points have been reported in a pilot study of chondroitin sulphate on cartilage volume loss.16 However, the sample size was small and differences in pain and function between treatment and placebo groups were not significant. Bisphosphonates are a potential therapeutic candidate as there is observational evidence that BMLs are less common in persons taking alendronate.17 A study investigating the effect of risedronate on cartilage loss in knee OA suggested that risedronate 50 mg weekly may prevent an increase in BML size,18 although this was not statistically significant. While the effect of bisphosphonates on BMLs could be a class effect, bisphosphonates given intravenously appear to have greater treatment effects, at least, for osteoporosis.19
Therefore, this study aimed to compare the effect of a single infusion of zoledronic acid (ZA) 5 mg with placebo on knee pain and BMLs over 12 months in participants with a pain intensity score >40 mm on a visual analogue scale (VAS) and a prevalent knee BML.
Patients and methods
This study was a single centre, double blind, parallel group, placebo controlled, randomised trial of intravenous ZA (5 mg) versus placebo with a 1:1 allocation ratio.
Setting and participants
Participants were recruited from May to December 2009 through advertising in local print media. Participants were eligible for inclusion if they were ≥50 years of age with significant knee pain on most days (VAS ≥40 mm) and at least one BML on MRI.
Participants were screened for eligibility by a rheumatologist (GJ) to confirm clinical knee OA as defined by the American College of Rheumatology criteria20 and enrolled into the study by a nurse (PB). Participants supplied a blood specimen for serum chemistry, renal function and 25 hydroxycalciferol; provided a urine specimen; and had a semiflexed knee x-ray.
Major exclusion criteria were abnormal blood tests (serum calcium >2.75 mmol/l (11.0 mg/dl) or <2.00 mmol/l (8.0 mg/dl), or creatinine clearance <35 ml/min), prior diagnosis of cancer (metastatic cancer or cancer diagnosed <2 years with ongoing treatment), use of bisphosphonates (except according to a washout schedule), history of non-traumatic iritis or uveitis, or severe knee OA (joint space narrowing (JSN) on x-ray of grade 3 using the Osteoarthritis Research Society International atlas).21 Participants then had a screening MRI. The ‘study knee’ was either the knee in which the patient was experiencing the most pain or difficulty or the remaining knee if the most painful knee had grade 3 JSN.
Use of other medications was allowed but kept constant through the trial period where possible. Participants who had serum vitamin D levels of <50 nmol/l were prescribed vitamin D supplements. All participants provided written consent. The study was approved by the Tasmanian Human Research Ethics Committee and conducted at the Menzies Research Institute Tasmania, Hobart, Australia.
Randomisation and interventions
Participants were randomly allocated to one of two treatment arms (ZA or placebo) using computer generated block randomisation in blocks of thirty. The random allocation sequence was automatically generated, and a security protected central automated allocation procedure was used to allocate participants to treatment arm 1 or 2. This was then used by one author (LL) to dispense the allocated medication to another (EB) who administered medication to each individual patient. Participants and staff involved in patient care (PB, EB, GJ) and assessment of MRI (DD) remained blinded to treatment allocation throughout the trial.
Participants (n=59) received an identical intravenous infusion of 100 mg of fluid, containing ZA (5 mg in normal saline) or placebo (normal saline), given by one nurse at the Menzies Institute. All participants were advised to take paracetamol as a prophylaxis for any acute phase reactions; these were recorded 3 days later by phone interview.
Primary outcomes were pain intensity (assessed using a VAS) and maximal area of BML as assessed by MRI at 6 months. Secondary outcomes were pain intensity at 3 and 12 months; knee pain and symptoms at 3, 6 and 12 months (using the knee injury and osteoarthritis outcome score (KOOS) questionnaire); and BML size at 12 months and safety outcomes (including any adverse or serious adverse event, and acute phase reactions). Outcomes were assessed using change between baseline and the relevant time point. Baseline questionnaires were completed in the clinic. Subsequent questionnaires were completed by mail.
Pain and function
Knee pain intensity and function were measured on four occasions (baseline, 3, 6 and 12 months). Knee pain intensity was measured using a 100 mm VAS and subjects were asked ‘on this line, where would you rate your pain today? Zero represents no pain and 100 represents extreme pain.’.
Knee pain and symptoms were also assessed using the KOOS questionnaire.22 These two subscales have nine (pain) and seven (symptoms) questions, each with five response levels scored from 0 to 4. Subscales were transformed according to instructions in the original manuscript.22 The transformed scale had possible values from 0 to 100 with zero representing extreme knee pain or symptoms, and 100 representing no knee pain or symptoms.
MRIs of the study knee were obtained at baseline, 6 and 12 months with a 1.5T whole body MR unit (GE-Signa; GE Healthcare, Buckinghamshire, UK). using a dedicated 8-channel knee coil. All MRIs were performed using proton density-weighted fat saturation two-dimensional fast spin echo sequence in the sagittal plane (repetition time 3875 ms, echo time 42 ms, slice thickness 3–5 mm, interslice gap 0–3 mm, field of view 16 cm, 224×448-pixel matrix, number of excitations 2).
A BML was defined as an ill-defined hyperintensity in the subchondral bone. One trained observer (blinded to treatment allocation and clinical data) scored lesions using Osiris software (University of Geneva, Geneva, Switzerland). To maximise sensitivity, chronological order was known to the observer.23 ,24 The maximum size was measured in mm2 using software cursors applied to the greatest area of each lesion as previously described.5 The lesion with the highest score was used if more than one was present at the same site. The intraobserver repeatability for this method of measurement is excellent with an intraclass correlation coefficient (ICC) of 0.97.5 Participants were given a BML score (mm2) at each of the four sites (medial tibial, medial femoral, lateral tibial and lateral femoral sites) and these were summed to create a total BML score (mm2).
A clinically significant reduction in BML size was classified as a reduction of 140 mm2.5
BMLs were also scored using an ordinal scale at each of the four sites, where 0=normal; 1=mild, <25% of the region; 2=moderate, 25–50% of the region; and 3=severe, >50% of the region. If the lesion was depicted in multiple slices, the one with the largest extent was chosen.25
Sample size for change in BML size was based on preliminary results from an in-house observational study (Tasmanian Older Adult Cohort Study) in which participants with existing BMLs had a mean BML area at baseline of 83 mm2 and an SD for change in area of 16 mm2, and thus a 20% reduction (or 16 mm2) in lesion area required 16 per group with α=0.05 and β=0.20.
Sample size for pain intensity using VAS was based on demonstrating a 20 mm greater reduction compared with placebo with an SD of 25 mm,26 ,27 which required 25 per group; thus, we aimed to enrol 60 participants to allow for dropouts.
Information on medication use (particularly change in analgesia), hospital admissions and newly diagnosed medical conditions was recorded at baseline, 3, 6 and 12 months. Blood samples were taken to assess safety at 3 and 6 months.
Primary hypotheses were tested using intent to treat analysis. Statistical significance was determined using a p value ≤0·05 (two-tailed). Due to an error in treatments given where one subject randomised to placebo received ZA (figure 1), all subsequent analyses used per protocol analysis. Groups were compared for change in the pain and BML area between the two time points of interest using linear regression, ordered logistic regression for the BML ordinal scale and logistic regression for the clinical improvement analyses as log binomial models did not converge. Groups were compared for safety end points using log binomial and Poisson regression. Analyses were adjusted for factors which were clinically significantly different between the two groups: baseline medication use (glucosamine, fish oil, paracetamol and non-steroidal anti-inflammatory medications), age, sex and baseline pain score. Missing data (n=8 at 12 months) were imputed using multivariate imputation by chained equations (MICE). We used Stata V.10.0 (StataCorp LP) for statistical analyses.
A total of 88 participants were screened for the study (figure 1). Twenty-nine participants were excluded as they failed to meet inclusion criteria (n=26), failed to attend appointments (n=2) or declined to participate further (n=1). Fifty-nine participants received infusions of either ZA (n=31) or placebo (n=28). Follow-up was 100% at 6 months and 89% (n=53) for the questionnaire and 86% (n=51) for MRI at 12 months.
Table 1 and figure 1 show the characteristics of study participants at baseline by treatment received. Participants (n=59) had a mean age of 62.4 years, mean body mass index of 29.7, mean VAS score of 54.4 mm and mean total BML area of 468 mm2 at baseline. Most had radiographic OA (defined as any score ≥1 for JSN or osteophytes).
Pain and symptom scores
Pain reduced in both groups between baseline and 3 months (figure 2). Intent to treat analysis for change in pain intensity between the groups receiving ZA and placebo after 6 months was borderline (β coefficient -13.3, 95% CI -26.8 to 0.3; p=0.055), after adjustment for age, sex, baseline pain score and baseline medication use. Table 2 displays the per protocol analysis.
Effect sizes were similar in per protocol and intent to treat analyses after 6 months but the per protocol analysis reached statistical significance (β=−14.5, p=0.04). Changes in pain intensity between baseline and the other time points were not statistically significant.
There were no differences in change in pain and symptom scales from the KOOS questionnaire in either group at any time point.
Few participants reported major changes in use of pain medications; therefore, analysis was adjusted for only baseline medication data.
The intent to treat analysis for change in BML area over 6 months was significant (β coefficient −163.8, 95% CI −314.3 to −13.2, p=0.03) after adjustment for age, sex, baseline pain score and baseline medication use. Table 3 demonstrates the change over 6 months was also statistically significant in per protocol analysis, and the effect size was similar (β coefficient −175.7, p=0.02). After 12 months of follow-up, change in BML area was lower in magnitude and of borderline significance (β coefficient −146.5, p=0.07 (table 3, figure 3)). Results were consistent but of borderline significance using the ordinal scale.
In post hoc analysis, a larger proportion of participants in the ZA group had improvements in BMLs which should equate to a 1 unit improvement in pain (140 mm2),5 as measured by the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (39% vs 18%, OR 5.0, p=0.044) after 6 months, but this was no longer significant after 12 months (OR 2.7, p=0.13) (table 4).
Adverse events were common during the study period and occurred more frequently in the ZA group (table 5). The most common adverse events were acute phase reactions: primarily cold or flu-like symptoms and headaches.
Serious adverse events were primarily non-elective hospital admissions, which included admissions related to the cancer diagnoses of two patients, admissions for angina and fractures. The two cases of cancer were bladder cancer and non-Hodgkin's lymphoma. No participants died during the study.
In this proof of principle study, a single infusion of intravenous ZA was effective in reducing pain intensity and areal BML size after 6 months of observation. This study is therefore an important step in identifying treatments which may slow the progression of knee OA, as we have demonstrated improvement in pain and structure. We also demonstrate that the proportion of BMLs improving is more frequent in ZA treated patients than those receiving placebo after 6 months, whereas risedronate is reported, at best, to halt progression.18
Pain improvement compared with baseline was clinically and statistically significant in both groups (using the VAS) after 3 months, as expected from an infusion therapy. However, by 6 months the improvement was numerically and statistically greater in the ZA group. This was in addition to the use of existing pain medicines and occurred regardless of the high prevalence of co-pathologies such as effusions, meniscal tears and cartilage defects (data not shown). The effect size was smaller than the 20 mm we had hypothesised, but is comparable with effect sizes seen in pharmacotherapy28 and was in addition to the effect of existing pain medicines.
By 12 months, pain had returned to baseline levels in the placebo group but remained 10 mm better than baseline in the ZA group. This suggests some residual effect, but also indicates that more frequent infusions may be required for BMLs than for osteoporosis. There were no significant changes in the pain scale from the KOOS questionnaire, suggesting that this scale may be less sensitive to change than pain intensity from the VAS in these participants.
Previous studies have shown conflicting findings on the natural history of BMLs. In participants with existing knee OA, one study reported that <1% of patients showed a decrease in BML size over 30 months9; while, in contrast, another study found that 20% of BMLs decreased over 2 years.29 Roemer et al30 found that in participants with prevalent knee OA or at risk for OA, half of pre-existing BMLs decreased in size after 30 months follow-up,30 while Dore et al5found that in a community cohort, similar proportions both worsened and improved. The reasons behind these variations are unclear, but might be related to the use of different definitions of BMLs or imaging protocols. In our study, around one in five patients in the placebo arm experienced clinical improvement; suggesting improvement is common in this sample, but more frequent in participants receiving ZA.
Some patients' BMLs continued to enlarge despite ZA treatment and use of other medications. This could be due to normal fluctuation in BMLs, progression of the underlying clinical condition or BML resolution may occur in some histological profiles but not others. BML histology is heterogeneous and consists of several abnormalities including bone marrow necrosis, abnormal trabeculae, bone marrow fibrosis, bone marrow bleeding, bone marrow oedema and sclerosis.31 ,32 The challenge is to identify which of these types are responsive to therapy.
The proportion of our participants who had BMLs on their screening MRI was very high at 88% and more prevalent than the 43% in our community cohort with existing knee pain.25 This suggests that BMLs are almost ubiquitous in adults of this age with significant knee pain and less than grade 3 JSN. Thus, MRI has a high yield as a screening tool for BMLs in this population for deciding if targeted therapy is appropriate.
Adverse events were common in this study, and types of events observed were consistent with previous trials.19 ,33 However, the rate of acute phase reactions was markedly higher than previous studies in placebo and ZA arms, and most likely reflects advising subjects on this potential side effect but could also be due to these reactions being more common in participants taking non-steroidal anti-inflammatory drugs,34 which comprised 40% of our cohort. There was a trend to a higher rate of serious adverse events in the ZA group but these were disparate in nature and none were considered causally related to ZA.
The strengths of our study are the complete follow-up after 6 months of observation and high follow-up after 12 months (89%), and assessment of change in BMLs using several different methods. Limitations include the smaller effect size detected for pain scores than expected, reducing the actual power of the study to detect differences between groups and the dispensing error requiring primarily per protocol analyses. Residual confounding was still apparent despite the randomisation, but this was dealt with by multivariate methods. Much larger, longer studies are required to assess whether these decreases in BML size will translate to reductions in cartilage loss or rates of knee replacement over time. However, this can be hypothesised from the observational studies, given that presence and severity of BMLs predict cartilage loss and knee replacement surgery.5 ,11
In conclusion, a single infusion of ZA reduces knee pain, areal BML size, and proportion experiencing clinically significant reduction in BML size after six months.
We thank Mr Tim Albion (Senior Database Administrator Health IT, Menzies Research Institute Tasmania) for generating the random allocation sequence and Associate Professor Changhai Ding (Principal Research Fellow, Menzies Research Institute Tasmania) for his contributions to the sample size calculations.
Funding Funding is provided by Novartis Pharmaceuticals Australia; Australian Government; National Health and Medical Research Council; and Osteoporosis Australia. LLL is supported by an Australian Government Australian Postgraduate Award. DAD is supported by a Tasmanian Postgraduate Research Scholarship and a Heald Fellowship from Arthritis Australia. GJ is supported by a National Health and Medical Research Council practitioner fellowship. TMW is supported by an Osteoporosis Australia/Australian and New Zealand Bone and Mineral Society/Amgen Fellowship.
Competing interests GJ has participated in advisory boards, given talks and acted as a clinical investigator for Novartis.
Ethics approval Tasmanian Human Research Ethics Committee.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.