Twenty-six global orthopaedic centres were involved in the RCT. The study was registered on clinicaltrials.gov (NCT: 03299959), approved by the United States of America’s Food and Drug Administration and each site’s Ethics Committees/Institutional Review Boards and was performed in line with the principles of the Declaration of Helsinki. All patients provided written informed consent. Patients were randomized to the aragonite-based scaffold (implant group) or SSoC (control group) in a 2:1 ratio (two patients in implant group vs. one patient in control group). Subjects were assigned to their treatment arm at each study site and within strata defined on the basis of osteoarthritis grade, according to the Kellgren–Lawrence (KL) classification. Details on patient selection and treatment allocation have been previously reported [15]. Enrolment was conducted from September 2017 to November 2019.

The Agili-C™ implant (Cartiheal Ltd, Israel), is a biphasic, cell-free implant made of natural calcium carbonate (aragonite) derived from purified, inorganic coral exoskeleton. This biomaterial offers a three-dimensional structure with high interconnected macro-porosity as well as robust mechanical properties, both crucial for the ingrowth of new bone and cartilage tissue. Previous studies have provided a detailed description of the implant’s physicochemical characteristics [18,19,20].

Main inclusion criteria were: (1) patients aged 21–75 years; (2) presence of up to three joint surface lesion(s), International Cartilage Repair Society (ICRS) Grade IIIa or above; (3) total treatable area 1–7 cm2; (4) patients willing and able to comply with post-operative rehabilitation protocol and follow-up schedule.

Main exclusion criteria were: (1) Baseline Knee Injury and Osteoarthritis Outcome Score (KOOS) Pain Subscale score < 20 or > 65 (maximum pain = 0, pain free = 100); (2) bony defect deeper than 8 mm on baseline imaging; (3) articular cartilage lesions in the tibia or the patella, ICRS grade IVa or above; (4) severe osteoarthritis of the index knee, KL Grade 4; (5) instability of the index knee, International Knee Documentation Committee (IKDC), Grade C (abnormal) or D (severely abnormal); (6) Malalignment more than 8 degrees varus or 8 degrees valgus; (7) Lack of functional remaining meniscus, at least 5 mm rim; (8) History of intraarticular or osseous infection of the index knee; (9) uncontained lesion, lack of vital bone wall surrounding the lesion of at least 2 mm thickness.

The aragonite-based scaffold implantation technique has been described in previous publications [15, 16]. Briefly, the surgical site was prepared using an open or mini-open approach by sequentially drilling through the articular surface and into the subchondral bone with a specific tool set. The implant was inserted using a press-fit method ensuring that the implant’s top was flush with the subchondral bone, typically 2 mm or more recessed below the cartilage surface. In cases requiring multiple implants, a minimum 5 mm bone bridge was left to prevent any overlap between biological zones of repair. Implant stability was examined by repeatedly bending the knee before and after removal of the tourniquet. The control group was treated arthroscopically with debridement or microfracture. The appropriate SSoC comparator for each patient (i.e., microfracture vs. debridement) was determined using a previously described algorithm based on patient age, concurrent level of OA, and lesion size [15]. Debridement consisted of removing damaged and unstable cartilage fragments from the articular surface. Microfracture was performed according to the established technique [21].

The rehabilitation program was the same for all study participants and included partial weight bearing for four weeks followed by progressive weight bearing to reach full weight bearing and full range of motion at eight weeks after surgery [15].

Patients were assessed for up to 48 months after surgery. All adverse events were recorded. The primary endpoint was the change in average KOOS Overall Score from baseline to last follow-up. Secondary outcomes included the KOOS subscales and IKDC Subjective Knee Evaluation score. Treatment failures were defined as any secondary intervention in the index knee, including intra-articular injection or surgery.

All patients underwent MRI at 12 and 24 months to assess the percentage of articular defect fill after surgery. The following MRI protocol was adopted: Field of view: 14 cm; slice thickness 3–3.5 mm; matrix 512 × 256 (or 384); Receiver bandwidth: 80-120Hx/pixel. Sequences: (a) Coronal IW FSE no fatsat; TR ≥ 3000ms; TE = 30-40ms; (b) Coronal PDW FSE with fatsat; TR ≥ 3000ms; TE = 10-20ms; (c) Sagittal IW FSE no fatsat; TR ≥ 3000ms; TE = 30-40ms; (d) Sagittal PDW FSE with fatsat; TR ≥ 3000ms; TE = 10-20ms; (e) Axial IW FSE no fatsat; TR ≥ 3000ms; TE = 30-40ms; (f) Axial T2W FSE with fatsat; TR ≥ 3000ms; TE = ≥ 70ms; (g) Sagittal T1W no fatsat; TR = 600–800; TE = 10-20ms; (h) Oblique PDW FSE with fatsat; TR ≥ 3000ms; TE = 10-20ms oriented perpendicularly to the scaffold. Independent, blinded radiologists with expertise in cartilage repair performed the defect fill assessments. Assessment included at least two sagittal slices and two coronal slices located within the implant/lesion. In the notch and trochlea lesions, axial cuts were used. Cartilage defect volume fill was semi-quantitatively assessed for each slice in increments of 25% fill (i.e.: 0–24%, 25–49%, 50–74% and 75–100%) and results of the slices were averaged. Thus, each score was a composite average of 4–6 slices in both sagittal and coronal orientation. For multiple defects, a single value was calculated by averaging all defects.

The sample size was determined using an adaptive design that employed Bayesian predictive probability in a “Goldilocks” strategy, as detailed in a previous paper [15]. Baseline observation carried forward was applied for patients considered treatment failures. Missing data for reasons other than treatment failure were assumed missing at random and imputed using a maximum likelihood estimation of a mixed model for repeated measures (MMRM). The MMRM included treatment group, baseline values and all available intermediate values of changes over time. This approach produces unbiased estimates of the treatment effect under the assumption that, conditional on these variables, the likelihood of missing data is independent of the distribution of the missing values. The MMRM was used to determine the 48-month treatment group contrast, 95% confidence intervals, and p-values. Baseline characteristics and outcomes were summarized using descriptive statistics. Patients were categorized into the condylar, trochlear, and mixed-lesions groups based on defect location. The mixed-lesions group consisted of patients with concomitant condylar and trochlear cartilage lesions. Outcomes for this group were compared to the “isolated” condylar and trochlear lesion groups to assess potential of the implant in complex cases. For all tests p < 0.05 was considered statistically significant.