A randomized, triple-blinded controlled clinical study with a novel disease-modifying drug combination in equine lameness-associated osteoarthritis

Objective This study aimed to test a novel treatment combination (TC) (equivalent to sildenafil, mepivacaine, and glucose) with disease-modifying properties compared to Celestone® bifas® (CB) in a randomized triple-blinded phase III clinical study in horses with mild osteoarthritis (OA). Joint biomarkers (reflecting the articular cartilage and subchondral bone remodelling) and clinical lameness were used as readouts to evaluate the treatment efficacy. Methods Twenty horses with OA-associated lameness in the carpal joint were included in the study and received either TC (n = 10) or CB (n = 10) drug intra-articularly-twice in the middle carpal joint with an interval of 2 weeks (visit 1 & 2). Clinical lameness was assessed both objectively (Lameness locator) and subjectively (visually). Synovial fluid and serum were sampled for quantification of the extracellular matrix (ECM) neo-epitope joint biomarkers represented by biglycan (BGN262) and cartilage oligomeric matrix protein (COMP156). Another two weeks later clinical lameness was recorded, and serum was collected for biomarkers analysis. The overall health status was compared pre and post-intervention by interviewing the trainer. Results Post-intervention, SF BGN262 levels significantly declined in TC (P = 0.002) and COMP156 levels significantly increased in CB (P = 0.002). The flexion test scores improved in the TC compared to CB (P =0.033) and also had an improved trotting gait quality (P =0.044). No adverse events were reported. Conclusion This is the first clinical study presenting companion diagnostics assisting in identifying OA phenotype and evaluating the efficacy and safety of a novel disease-modifying osteoarthritic drug.


Introduction
Osteoarthritis (OA) is a whole joint disease with a multifactorial aetiology involving biomechanics, joint overload, low-grade chronic systemic inflammation and immune system activation [1].
The racehorses develop OA spontaneously, therefore can serve as an excellent research animal model and is superior to other induced models [2]. Given their genome similarities with humans, the U.S. Food and Drug Administration (FDA) has approved horses as a translational model to study human OA [3,4].
Currently available pharmacological OA treatments only improve pain or symptoms and are frequently associated with side effects [2].
There are no approved disease-modifying osteoarthritic drugs (DMOADs) on the market yet. A DMOAD typically must show improvement in joint structure, and slow down the progression of cartilage, bone and synovium destruction, with or without an efficacy on joint pain [2,5,6]. Therefore, molecular tools are crucial for diagnosing OA and monitoring intervention efficacy [7].
Both native cartilage oligomeric matrix protein (COMP) and biglycan (BGN) play a structural and functional role in cartilage and bone homeostasis [8,9]. We have previously defined two novel soluble neo-epitopes that reflect ongoing ECM degradation in osteoarthritic articular cartilage and bone. (i) A soluble neo-epitope of biglycan (BGN 262 ), elevated in SF from OA horses and positively associated with increased subchondral bone sclerosis (SCBS) [10] (ii) A soluble neo-epitope of cartilage oligomeric matrix protein (COMP 156 ), associated with articular cartilage degradation in equine early OA [11,12].
Our proof of concept preclinical research has identified that low concentrations of sildenafil and mepivacaine, when combined with glucose, target basic cellular mechanisms by affecting the inflammatory system [13]. A patented combination of these three substances addressed as treatment combination (TC) in this study exerted an anti-inflammatory effect while restoring Ca 2þ signalling in gap junction coupled cells [13,14]. We speculate the effect exerted by low concentrations of sildenafil and mepivacaine for OA treatment in vivo differs from their normal pharmacological effect.
This study is a randomized, phase III clinical trial, testing TC drug on lame OA horses, with Celeston® bifas® (betamethasone) (CB) as a control substance.
We hypothesised the lameness would improve faster in TC, and the levels of the respective joint biomarkers (BGN 262 & COMP 156 ) in SF and serum should reflect the intervention efficacy. A reduction in SF biomarkers after treatment should indicate a slowdown of the disease progression.
In this study, we have validated BGN 262 & COMP 156 for use as companion diagnostics, to identify the OA phenotype and simultaneously monitor the efficacy and safety of a DMOAD in lame horses with mild OA.

Study design
The study was performed as a randomized triple-blind (blinded for; the veterinarian assessing the outcome, the horse owners/trainers, researchers and the statistician) phase III trial with a parallel group design and equal allocation ratio, and was approved by the Ethics Committee, Lund, Sweden (D.nr:5.8.18-06590/2020) and the Swedish Medical Product Agency, Uppsala, Sweden (D.nr 5.1-2020-31501). All horse owners/trainers signed a study consent and were allowed to opt-out any time without providing a reason. A computer-generated randomization list was made by the statistician Claudia von Br€ omssen (SLU, Uppsala, Sweden) using the interactive webpage 'Randomization table for clinical trials (https://aurora.shinyapps.io/random_gen/)! by A. Baluja.
The CONSORT (Consolidated Standards of Reporting Trials) checklist has been used when reporting the clinical trial 2010 [15].

Sample size
A sample size calculation was performed with a 95% confidence level and 80% power, considering a 20% reduction in COMP 156 levels in SF after treatment (based on an earlier unpublished pilot study). From there, it was intended to include a total of 10 horses in each study group.

Inclusion and exclusion criteria
Standardbred trotters (STBs) (age: 2-9 years), with lameness originating from the carpal joint, were recruited at Kungsbacka Horse Clinic, Kungsbacka, Sweden. Inclusion criteria were a positive response to diagnostic intra-articular anaesthesia of the affected carpal joint and mild radiographic signs of SCBS (only mild bone sclerosis in sky projection were accepted) with no remodelling of the affected joint. Horses that received intra-articular medications such as corticosteroids or hyaluronic acid within the previous three to six months, were excluded. Bilateral carpal lameness was accepted. The data collected throughout the trial was logged individually for each horse.

Visit 1-clinical examination and treatments
At visit 1, the horses were subjected to a complete lameness examination (by veterinarian 1). The initial lameness was evaluated objectively with Lameness locator® and subjectively following the flexion test by visual grading of lameness (scale 0-5) (suppl. Table 1). Horses fulfilling inclusion criteria were assigned to the respective treatment group according to randomization list by the assisting veterinarian (veterinarian 2).
Injections were administered the same day horses arrived at the clinic, after confirming eligibility, source of lameness and evaluating the radiographs. Treatments were double-blinded except for the assisting veterinarian who prepared the syringes according to the randomization list and administered the injections. Owners remained blinded during the whole study period and the veterinarians until after the statistical analyses were performed.

Outcomes and follow-up
Lameness evaluation (objectively and subjectively) (see suppl. data material-methods and suppl. Table 1).

Sampling of SF and serum for biomarkers
SF samples were collected from both left and right forelimb carpal joint (middle and upper joint compartments) before administering the local anaesthetic at visits 1 and 2 and serum samples at visits 1, 2 and 3 (sample handling described in suppl. data).

Rehabilitation after visit 1
Post-first treatment, the horses were allowed to rest in a box, thereafter, were allowed 30 min hand-walking per day with free access to a small paddock during the first week. During second week, they were allowed 60 min hand-walking per day.

Visit 2 after 14 days
After 14 days, the evaluating veterinarian performed a second lameness evaluation on horses with Lameness Locator® and flexion. SF and serum sampling were performed as described earlier. After the second clinical examination, the double-blinded trial was continued by injecting a second intra-articurlar treatment by the attending veterinarian.

Rehabilitation after visit 2
Post-second treatment, the rehabilitation was same as in visit 1 for a week and thereafter the horses were allowed to jogg and trot during the second week.

Visit 3 after 28 days
After 28 days, the veterinarian performed a third lameness evaluation and serum was collected.

Interview at visit 1 and follow-up after 60 days (visit 4)
The trainers (professionals and amateurs) were interviewed with the questionnaire (suppl. data).

Diagnostic accuracy for biomarkers
The STARD (Standards for Reporting of Diagnostic Accuracy Studies) checklist has been used to report the diagnostic part of the clinical trial [16].

Safety data
Full blood profile for drug safety were analysed at visits 1, 2 and 3.

Anamnestic evaluation of clinical side effects
During visits 2 and 3, trainers were asked if the horses had experienced any side effects, such as swelling of the joints or flares following the treatment.

Study outcomes
The study was blinded until the primary and secondary outcomes were analysed. The code was broken after the full statistical analysis was reported. In parallel to this study, our group has validated BGN 262 as a bone marker [10]. Therefore, the BGN 262 quantification has been included in primary outcomes through an amendment sent to the Medical product agency, Uppsala, Sweden. Comparison of the number of non-lame (lameness ¼ 0 after flexion test) horses at both visits 2& 3.

Secondary outcomes
Blood sample analysis reports.

Longterm follow-up results
Interview with the trainers.

Statistical analysis
Assessment of the normal distribution assumption for biomarkers concentrations (BGN 262 and COMP 156 in SF) were made for both nonlogarithmic and logarithmic values. The normal distribution assumption was evaluated in two ways: 1) analysing the distribution of the error terms for linear mixed models. 2) analysing the distribution for differences in concentration between visits 1 and 2.
Shapiro Wilk's test was made for both cases. In short, the analyses show that data deviate from the normal distribution, which means that a non-parametric method was primarily chosen to perform the analyses on the concentration of the biomarkers in SF.
The statistical methods used to replace linear mixed models were the Wilcoxon signed rank test and the Wilcoxon rank sum test. For these methods, the same results were obtained for logarithmic and nonlogarithmic values, as the tests were rank based. Estimates and CI for original (non-logarithmic) values are also presented. BGN 262 and COMP 156 values for the individual horses are presented as the mean of the duplicates.
For the follow-up interview questions, Fisher's exact test was used. All results are presented as mean AE SD (demographic data and Q-score), for biomarker data 95% confidence interval [CI]. A significance level of 5% was used. R (version 4.0.0) has been used for all analyses (full report in suppl. data).

Study limitations
At visit 3 the SF was not sampled because most of the horses were sound (due to ethical reasons and risk for intra-articular infection). Therefore, the SF COMP 156 and BGN 262 data are not available for this visit.

Data availability
The study data generated and recorded during the trial is not publicly available and resides with the corresponding author (E.S) and can be provided upon request.

Results
Twenty STBs fulfilling the inclusion criteria were included in the study between December 2020 to December 2021 (Fig. 1). For demographic and baseline data see Table 1.

Clinical lameness
At the inclusion, the baseline lameness score with Lameness Locator® (Q-score) and flexion did not differ between the groups (suppl. Table 6). Lameness post flexion test at visits 1,2 and 3 for the individual horses are also reported (suppl. Table 2).
TC became free of lameness faster (score 0) from visit 1 to visit 3, compared to CB (p ¼ 0.033). In both groups at visit 3, there were two non-responders (Fig. 2).
Biomarkers concentrations of COMP 156 and BGN. 262 The SF biomarkers concentrations did not differ between the upper or middle compartment of the carpal joint hence middle carpal joint results were used.

Biomarker levels at inclusion
The SF BGN 262 levels were above the reference range (!265 ng/ ml) for all horses except one (in CB). For COMP 156 , (TC: n¼6 and  At visit 2, TC showed a significant decrease in SF BGN 262 levels (p¼ 0.002) which was not found in CB. Instead, the latter showed an increase in BGN 262 concentration, however, not statistically significant ( Fig. 3a and b) ( Table 2).

BGN 262 and COMP 156 in serum
For either group, there were no changes in serum concentrations of BGN 262 or COMP 156 between the visits (suppl. Fig. 1a-1d, Table 5).

Follow-up evaluation by the trainer after 60 days
Two horses in group A were not assessed at visit 4 regarding trotting quality. The trainer couldn't start to train the horse at the time for the interview due to bad winter weather. The horses was therefore given a prolonged rest. The data is treated as missing. At visit 1, the mean trainerassessed trotting quality was 1.6 in TC and 1.2 in CB, and at follow-up, the mean was 3.5 in TC and 2.2 in CB. The trotting quality improved significantly in TC between visits 1 and 4 but not in CB (p¼0.044) (suppl. Table 7).
Other parameters such as mood, appetite and fur quality did not differ in the groups between visits 1 and 4.

Discussion
This study aimed to evaluate the efficacy of a new DMOAD combination (TC) with companion diagnostics in relieving joint pain and reducing ECM destruction in both cartilage and SCB against a CB-treated group (positive control) in STBs. Our results are the first to show the efficacy of TC treatment on clinical lameness and reduction in the soluble SF BGN 262 levels, a novel biomarker reflecting SCB degradation. An increase in SF COMP 156 (cartilage-derived biomarker) in CB can be addressed as a severe side effect. Both biomarkers meet the companion diagnostics requirements [17].
Predictive soluble biomarkers can be classified according to BIPEDS; Burden of disease (B), Investigative (I), Prognostic (P), Efficacy of intervention (E) Diagnostic (D) and Safety (S) [18]. The reduced BGN 262 levels in TC indicate impediment of the SCB destruction and are comparable to that of healthy trained horses [10]. BGN 262 potentially meets the requirement to be an ideal predictive soluble candidate biomarker according to BIPEDs classification; (B) (E) (S) and (D).
The dramatic post-treatment increase of SF COMP 156 in CB indicates severe cartilage degradation, a serious side effect of corticosteroid therapy. COMP 156 potentially meets the requirement to be a predictive soluble biomarker according to BIPEDs classification; (B) (S) and (D).
In this study, the OA-predictive biomarker neoepitopes were found to be capable of defining both clinical phenotypes and their corresponding molecular endotypes (cartilage-driven, bone-driven and mechanical injury-driven) [10,19]. All the study horses had similar demographics (age, sex, breed and athletic profile), targeted OA phenotypes (mild OA in the same joints) and joint pain at inclusion. Additionally, by quantifying soluble BGN 262 and COMP 156 , we present the possibility of accurately phenotyping the incipient OA in a horse according to the recommeneded/proposed endotypes.
In agreement with the STARD guidelines pertaining to the diagnostic accuracy of a candidate biomarker, we can conclude that the horses in the trial at inclusion could be grouped into endo-types (a) high BGN 262 (n¼19) or (b) high BGN 262 & COMP 156 (n¼8) ( Table 2) with latter reflecting a more severe incipient OA. Table 2 Concentration of BGN 262 (ng/ml) and COMP 156 (μg/ml) in SF from middle carpal joint at visit 1 and 2 for TC and CB group. As a reference, the value for healthy horses BGN 262 265 ng/ml (173AE92) [10] and for COMP 156  The cleavage site within the neo-epitopes (BGN 262 and COMP 156 ) is conserved across the species including humans thus making them ideal biomarker candidates for drug development as they come with a proofof-concept, safety, surrogate-end-point, and companion diagnostics.
Race horses are more vulnerable in developing OA, especially in highmotion joints such as carpal joint. Therefore, lameness originating from this joint has targeted for treatment. The results clearly indicate that one dose of TC injection was equivalent to 2 doses of CB in subsiding joint pain. At 60 days post-intervention (visit 4 and still blinded) the trotting assessment by the trainer received a better score in TC, suggesting a prolonged treatment benefit.
For several decades, intra-articular treatment with corticosteroids is in use for OA-related pain and remains to be the most used drug in highmotion joints in horses and large joints in humans. Despite their efficacy in releiving joint pain, their repeated use has been reported to have shortcomings. A study with triamcinolone resulted in a greater cartilage volume loss when compared to intra-articular saline injection at two-year follow up [22]. The pros and cons of corticosteroids are unclear [20][21][22]. However, our study data sides with the negative effect of such joint treatment. A dramatic increase in SF COMP 156 in CB group indicates cartilage degradation and a severe post-intervention side effect. In an exploratory approach, joints were grouped by low concentrations of biomarkers (BGN 262 and COMP 156 ) at visit 1. This was done to assess the response of such joints to the respective treatments. Both markers showed a significant increase in SF following CB treatment, which was not seen in TC, clearly supporting the dual potential of the biomarkers in rightly diagnosing the affected joint and monitoring the side effects.
The TC drug primarily acts by restoring cell-to-cell signalling which is essentially impaired in OA. Studying basic cellular mechanisms, which is mostly similar in many inflamed cell types (chondrocytes, fibroblast and astrocytes) has been crucial in understanding how different drugs can restore the derailed cellular networks [14,[25][26][27][28]. Gap junction-coupled cells forming networks in different organs in the body are another focus point [29,30]. Various cell types i.e., astrocytes, osteoblasts, osteocytes, osteoclasts and chondrocytes express connexin 43 (Cx43) and communicate via Ca 2þ waves [31][32][33][34]. Intracellular Ca 2þ release is controlled by different signalling pathways that can be stimulated by various neurotransmitters, such as ATP, glutamate and 5-HT [35][36][37].
Local anaesthetic agents, such as bupivacaine and carbocaine®, are widely used clinical agents inducing analgesia by blocking voltage-gated Na þ channels when used for neuro-axial blockades. Lower concentrations of bupivacaine unlike clinical doses, evoked Ca 2þ transients and blocked nerve impulse propagation, in turn may have a pain-relieving effect via targeting G protein-coupled receptors and binding sites on immune cells [38][39][40].
Lower concentrations of bupivacaine (<10 À8 M) were found to evoke intracellular Ca 2þ transients that were inositol trisphosphate (IP 3 ) receptor-dependent in astrocytes. The concentration-dependent curve for bupivacaine did not follow a Gaussian curve. At higher concentrations, >10 À8 M, bupivacaine blocked the Ca 2þ release. The absence of such a response at higher concentrations could be due to Na þ channel inhibition [37,41,42]. In addition, we found low concentrations of bupivacaine decreased the inflammation-induced interleukin-1β (IL-1β) release in astrocytes [39]. The same results were seen in OA chondrocytes [14]. These results support the anti-inflammatory properties of bupivacaine in low concentrations.
Sildenafil (clinical dose) is a potent and selective PDE-5 inhibitor, which induces cyclic GMP accumulation [43]. This drug at lower concentrations reduces the Ca 2þ response intensity and induces a more organized actin fiber pattern in inflamed cells [44]. Santillo et al., 2018 [43] studied the dose-response curves for sildenafil and other PDE-5 inhibitor analogues and reported the inhibition dose range to be between 0.1 and 100 ng/ml. In our study, the Sildenafil dose was below that dose range, where PDE-5 inhibition is completely abolished. Taken together, the beneficial effect exerted by low doses thus is by balancing Ca 2þ responses and maintaining the cytoskeleton integrity by direct interaction with Na þ /K þ -ATPase as a target.
Glucose, a primary substrate for ATP production (in glycolysis) and for matrix molecule synthesis, such as hyaluronan: crucial in articular cartilage assembly [45]. Inflamed chondrocytes consume more glucose, downregulates the GLUTs thus disrupting the glucose balance [45][46][47]. Metformin exposure increases cellular glucose consumption and enhances glycolytic flux across the cell membrane [48]. Compared to normal chondrocytes, in vitro OA chondrocytes exhibit increased intracellular Ca 2þ release, associated with TLR4 induction and other inflammatory mediators [27]. Glucose or metformin in combination with an anaesthetic agent and sildenafil restored elevated 5-HT-and ATP-evoked intracellular Ca 2þ signalling between the cells through gap junctions in vitro [13,14]. TLR4 has been identified as a potential drug target for the treatment of inflammatory diseases including OA therefore inhibitors modulating TLR4 signalling in joint tissues have been proposed as DMOADs [49,50]. The prominent downregulation of TLR4 in chondrocytes treated with metformin and bupivacaine further indicates anti-inflammatory properties of this drug combination [14].
We, therefore, aimed to modify the underlying OA pathophysiology by restoring basic cellular parameters to normal physiological state, ie. intracellular Ca 2þ signalling, increase Na þ /K þ -ATPase activity, enhancing glucose uptake and downregulating TLR4. The cartilage and bone cells thereby alleviate the inflammatory mediators expression associated with structural damage in cartilage and SCB.

Conclusions
This study presents sensitive and specific neo-epitope biomarkers to monitor TC efficacy as well as corticosteroid side effects. TC exhibits DMOAD properties by slowing down OA-associated SCB destruction, revoking lameness and joint pain.
Since validated neo-epitope biomarkers (COMP 156 & BGN 262 ) used in this study are highly conserved across the species makes it possible to put our current study findings into human OA perspective.

Declarations of interest
ES and AL are stakeholders of SGPTH Life Science AB holding the patent for BGN 262 and COMP 156 neo-epitopes.
EH, ES, KA-A and AL are stakeholders of ARTROA AB holding, the patent for pharmaceutical drug combination. The other co-author has no conflicts of interest to declare.

Fundings
The study was funded by the Grants and Innovation Office at Gothenburg University, Sweden. It had no role in study design, data collection, interpretation or manuscript writing.

Author contributions
ES and EH designed the study, ES, EH, KAA and AL the overall direction and study implementation. SA, CL, SN, and KB set up and validated the ELISAs, analysed all the samples. ES, SA, CL and EH wrote the manuscript. All authors contributed to and approved the final manuscript. Table 3 Biomarker-assisted monitoring of side effects. Five of 10 horses that received treatment with Celeston® Bifas® showed increased concentrations of COMP 156 (μg/ml) and BGN 262 (ng/ml) in SF post-intervention at visit 2. Additional two horses showed increased concentration of BGN 262 in SF. CB¼Celeston® Bifas®. As a reference, the value for healthy horses BGN 262 265 ng/ml (173AE92) [4] and for COMP 156 289  8  6487  1484  9  6560  5000  11  717  103  19  3871  5000  20 2463 5000 statistical evaluation (v) Lindgård, Gothenburg University, Grants and Innovation Office for auditing the clinical trial.