Study design and setting
This was a placebo-controlled, randomized, double-blind, parallel-group study conducted at 44 sites including university hospitals, clinics and general hospitals in Japan (JapicCTI-173,678, 21 August 2017). The study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice (GCP) guidelines after approval by the central institutional review board (IRB) or the respective IRBs. Written informed consent was obtained from all patients.
Subjects
OA patients aged ≥20 years who had pain in the hip, ankle, shoulder, or elbow joint for at least 6 months, with a diary-based mean 11-point (0–10) numerical rating scale (NRS) for pain of 5–9 in the target joint during the screening period were eligible to participate in this study. In patients with OA involving multiple joints, only one joint was treated with the study drug and assessed as the target joint. In addition, the following inclusion criteria were specified for OA in each joint: 1. Hip OA diagnosed as primary OA or secondary OA owing to acetabular dysplasia and classified radiographically as early (mild articular wear) or advanced OA (advanced articular wear) according to the Japanese Orthopaedic Association stage classification of hip osteoarthritis [13]. 2. Ankle OA diagnosed as primary talocrural OA or secondary talocrural OA owing to trauma and classified radiographically as stage II (partial narrowing of the joint space) or III (partial disappearance of the joint space) varus ankle OA. 3. Shoulder OA diagnosed as primary glenohumeral OA or secondary glenohumeral OA owing to trauma or grade ≥ 4 rotator cuff tear according to the Hamada classification [14, 15], a radiographic classification of massive rotator cuff tear, and classified radiographically as Kellgren–Lawrence (KL) grade 2 or 3. 4. Elbow OA diagnosed as primary OA or secondary OA owing to trauma and classified radiographically as KL grade 2 or 3. Patients with serious diseases, such as cardiac, hepatic, or renal diseases, blood dyscrasia, or immunodeficiency, with a history of hypersensitivity to HA, DF-HA, or acetaminophen, who received any HA preparation in the target joint within the specified period, or with shoulder/elbow OA or hip/ankle OA with an NRS pain score ≥ 4 in any non-target joint in the upper or lower body were excluded from the study.
Randomization and blinding
The study drugs were dynamically allocated by the minimization method using an interactive web response system incorporating a randomized algorithm, so that subjects were assigned to the DF-HA or placebo group in a 1:1 ratio for each joint [16]. The stratification factors were the site, cause of OA, stage of OA, baseline NRS for pain, sex, and age. DF-HA was more viscous than the placebo, requiring a higher pressure during injection; therefore, the difference in viscosity might have allowed the treating investigator to distinguish whether the study drug was DF-HA or placebo To maintain the study blinding, evaluations after the first injection until the end of the study were performed by an investigator who did not administer the study drug. A treating investigator administered the study drug to subjects after providing a written signed pledge not to disclose any information to the subjects or other staff that might affect the blindness of the study.
Treatment method
Patients were screened from 1 week before randomization. Patients meeting the inclusion criteria received a total of three injections of either DF-HA (syringe filled with 3 mL of sodium citrate buffer including 30 mg DF-HA; Seikagaku Corporation, Tokyo, Japan) or placebo (syringe filled with 3 mL of sodium citrate buffer without DF-HA; Seikagaku Corporation) in the target joint cavity every 4 weeks (Weeks 0, 4, and 8). Because a 3-mL injection could be administered to all joints, the same dose regimen was selected as in previous studies of knee OA [11, 12]. The study drug was administered by an orthopedic surgeon according to the prescribed procedure for injection into each joint (Supplementary Table 1, Additional file 1). While ultrasound- or fluoroscopy-guided injection was recommended, blind injection was allowed for the ankle, shoulder, and elbow joints if administered by specialists (members of designated academic societies) who carried out the practices in daily medical practice. Oral NSAIDs, steroids, opioids, psychotherapeutic agents, and anesthetics, as well as intra-articular agents, including steroids, were prohibited during the study. Acetaminophen was used as rescue medication. During the study, existing physical therapy could be continued at the same frequency or intensity, but the initiation of new physical therapies was prohibited. Similarly, patients were instructed not to change the frequency or intensity of exercise in their daily activities. Removal of joint effusion from the target joint was prohibited unless the investigator judged it necessary before study drug injection.
Evaluation methods
The primary outcome measure was the diary-based 11-point NRS for pain in the target joint, and the primary endpoint was the mean change from baseline in NRS for pain over 12 weeks after the first injection. Secondary endpoints comprised joint-specific endpoints for the hip, ankle, shoulder, and elbow joints over 12 weeks after the first injection using the Western Ontario and McMaster Universities Osteoarthritis 3.1 index (WOMAC) [17], Self-Administered Foot Evaluation Questionnaire (SAFE-Q) [18], Shoulder36 [19], and Patient-Rated Elbow Evaluation (Japanese Version) (PREE-J) [20], respectively. In addition, the proportion of responders, patient and physician global assessment scores (100-mm visual analog scale [VAS]), Medical Outcomes Study 36-Item Short Form Health Survey (SF-36) [21,22,23], EuroQol 5 Dimensions (EQ-5D) [24], joint range of motion, and acetaminophen consumption were assessed.
Safety was evaluated according to treatment-emergent adverse events (TEAE) discovered after the initial injection, which were reported in accordance with the definitions in Supplementary Table 2, Additional file 2. The relationship of TEAE to the study drug was assessed by a blinded investigator. TEAEs of special interest were TEAEs relating to the injection site, gastrointestinal disorders, cardiovascular disorders, renal dysfunction, hypersensitivity, and anaphylactic reaction. To confirm safety, the target joint was radiographed before the first injection of the study drug and at Week 12 or discontinuation. Joints were assessed by the investigator for morphological changes (osteophyte, joint space narrowing, osteosclerosis, and epiphyseal deformity) according to the study protocol (Supplementary Table 3, Additional file 3). In addition, manual joint examination, laboratory testing (hematology, blood biochemistry, and urinalysis), and vital signs measurement were performed, and the investigator determined whether any abnormal change in these four tests corresponded to a TEAE.
Statistical analysis
According to the results of a phase 2 study in patients with knee OA [11], the between-group difference for each joint was assumed to be − 0.55, with a standard deviation of 2.00. To demonstrate that DF-HA was as effective for OA in each joint as for knee OA, the target sample size was calculated to be 45 subjects per group for each joint, with a 90% probability of the point estimate in the DF-HA group being above that in the placebo group. Given the small number of patients with ankle or elbow OA and the feasibility of study, the target sample sizes were calculated to be 30 and 20 patients per group, respectively, with 85 and 80% probabilities, respectively, of the point estimate in the DF-HA group being above that in the placebo group.
Efficacy was evaluated for each target joint in the full analysis set (FAS), which was a population comprising all subjects who had at least one post-treatment efficacy data value. The primary analysis was to compare the mean change from baseline in NRS for pain over 12 weeks after the first injection in each target joint between the DF-HA and placebo groups using a mixed model for repeated measures (MMRM) analysis. The fixed effects were treatment, time, treatment-by-time interaction, baseline NRS for pain, type of OA, stage of OA, age, and sex. The correlation between time points in subjects was assumed to be unstructured. The Kenward–Roger method was used to calculate the degrees of freedom. Secondary analyses comprised comparisons of the change from baseline in NRS for pain at each time point using the same MMRM analysis. In addition, the mean change from baseline over 12 weeks after the first injection was compared between the two groups for the following secondary endpoints using an MMRM analysis: joint-specific endpoints, and patient and physician global assessment scores. The proportion of responders, defined as patients with at least 30% improvement in NRS for pain from baseline, was calculated at each time point. In addition, the odds ratio was calculated using a generalized estimating equation (GEE). The explanatory variables in the model were treatment, time, treatment-by-time interaction, type of OA, stage of OA, diary-based baseline pain score in the target joint, age, and sex. An analysis of covariance was performed for the changes in SF-36 and EQ-5D from baseline to Week 12. The covariates were the baseline value of each outcome measure, type of OA, stage of OA, age, and sex. For joint range of motion and the mean daily consumption of acetaminophen, the change from baseline at each time point was summarized. Acetaminophen consumption was considered to be zero in subjects who did not take it. SAS software, version 9.4 (SAS Institute Japan Ltd., Tokyo, Japan) was used to perform the analyses.
Safety was evaluated in the safety set, a population consisting of all subjects who received at least one injection of the study drug. The incidences of TEAEs were tabulated for all joints combined and by joint. TEAEs were coded using the Medical Dictionary for Regulatory Activities (MedDRA) ver. 22.0. TEAEs of special interest other than those at the injection site were classified using the standardized MedDRA query. Other safety endpoints were also tabulated for all joints combined and by joint.