Materials and methods

This was a retrospective, multicenter, single-arm cohort study that included data from patients who underwent osteoblast cell therapy for ONFH between April 2017 and April 2022 at 21 tertiary care centers across India. The participating centers were selected based on the investigators' expertise in the management of ONFH and hip reconstruction procedures, geographic spread across India, inclusion of both public and private healthcare centers, and willingness to contribute retrospective patient data for the study. The study was approved by the Regrow Biosciences Independent Ethics Committee (letter number: REG/IEC/2025/006), and was conducted in accordance with Good Clinical Practice (GCP) and the Declaration of Helsinki.

Patient population

This study included patients aged 18-52 years, diagnosed with stages I, II, and IIIA of ONFH of multifactorial etiology, according to the 2019 revised version of the Association Research Circulation Osseous (ARCO 2019) classification [19]. Additionally, patients included were those who had previously been treated with osteoblast cell therapy and were followed up for two to seven years. Patients who had undergone any treatment for cancer over the last two years or had an active infection were excluded. Patients with neurological or psychiatric conditions, cardiac abnormalities, genetic disorders, or hemoglobinopathies were also excluded.

No predefined sample size calculation was performed. Instead, all patients who met the inclusion criteria within the defined study period were included.

Study procedure

The diagnosis of ONFH and disease stage were confirmed based on findings of X-ray and magnetic resonance imaging (MRI). After osteoblast cell therapy, patients were followed up clinically at three, six, and 12 months, and then annually to assess pain, functional outcomes, and the need for additional surgeries. Retrospective data of the patients were extracted from progress notes, discharge summaries, and radiology reports. The coronavirus disease 2019 (COVID-19)-associated ONFH subgroup was identified retrospectively from hospital medical records. Patients with a history of steroid exposure were excluded from this subgroup to avoid overlapping with steroid-associated ONFH. The interval between COVID-19 infection/recovery and ONFH diagnosis was calculated from the recorded dates. During the COVID-19 pandemic (early 2020 to mid-2021), patients were followed up telephonically.

Surgical procedure and rehabilitation

Osteoblast cell therapy involves a two-stage surgical procedure performed under spinal anesthesia. The details of the surgical procedure and rehabilitation have been described previously [16]. In the first stage, 4-8 mL of bone marrow was aspirated and collected from the posterior superior iliac crest and transported to a Good Manufacturing Practices (GMP)-certified facility (Regrow Biosciences Pvt. Ltd., Lonavala, Pune) for ex vivo expansion.

Thereafter, MSCs from the bone marrow were centrifuged, resuspended in a growth medium containing L-ascorbic acid and other growth factors, allowed to differentiate into osteoblastic lineage cells, and expanded over three to four weeks. Cells were stained with anti-human bone alkaline phosphatase antibodies (BioLegend, San Diego, California, United States) and characterized using flow cytometry (Attune NXT; Thermo Fisher Scientific Inc., Waltham, Massachusetts, United States). Additionally, cells were subjected to Alizarin Red staining to confirm osteoblastic phenotype. After passing release testing criteria (negative for endotoxin, bacterial, mycoplasma, and other viable impurities), a minimum of 48 million viable osteoblast cells were transported back for implantation, maintaining the cold chain.

In the final operative step, patients were placed supine on a fracture table, with the affected limb internally rotated by 15 degrees. The lesion was marked in both anteroposterior and lateral planes using 2 mm K-wires (Kirschner wires) under C-arm guidance. Core decompression was performed using an 8 mm cannulated drill over the guide wires, and the sclerotic bone at the lesion site was thoroughly removed using curettage. A total of 4.8 × 107 autologous live cultured osteoblasts were delivered into the decompressed area using a certified fibrin glue (TISSEEL; Baxter International Inc., Deerfield, Illinois, United States) and an 18-gauge spinal needle via an eight mm interference screw. Finally, the tract was sealed with cancellous allograft obtained from a certified bone bank (Figure 1). A standardized surgical procedure manual was provided to all participating centers in the study.

Figure 1: Pre- and intraoperative images of osteoblast cell therapy(A) Preoperative X-ray of a 47-year-old male patient with ONFH; (B) Incision and opening for passing the guide wire; (C, D) Core decompression, a standard procedure, was performed under C-arm guidance, and angular curettes were used to remove sclerotic necrotic bone, creating space for new bone formation; (E) Allograft plug; (F) Allograft plug placed onto the long spinal needle (CE-marked; 18G/20 cm); (G) Long spinal needle positioned for final product implantation; (H, I) Final osteoblast cell–gel mixture delivered to the necrotic site for bone regenerationCE: Conformité Européenne (indicating compliance with European requirements that ensure safety, health, and environmental protection standards); ONFH: osteonecrosis of the femoral head

After implantation, patients were instructed to perform partial weight-bearing exercise for four weeks using a walker. Patients were allowed to walk using a stick by Week 6 and then to bear full weight by Week 8. However, in the case of bilateral ONFH, a walker was needed until Week 6. Passive lower-limb exercises to regain muscle strength and hip joint movements were performed as early as possible.

Study outcomes

The primary outcomes included improvement in pain intensity and functional capacity at follow-up from baseline, and the secondary outcome was treatment failure.

Improvements in pain intensity at the patient level were assessed using the patient-reported Visual Analogue Scale (VAS) (range, 0-10, with 10 indicating the worst imaginable pain) [20]. Functional capacity was assessed by the treating physician at the patient level using the Harris Hip Score (HHS) (0-100 points, with > 90: excellent, 80-90: good, 70-80: okay, and < 70: poor outcomes) [21]. The minimal clinically important difference (MCID) thresholds were predefined as two points for VAS and 10 points for HHS. Treatment failure was defined as the patients undergoing surgical treatment, including but not limited to THA, as assessed by the surgeon. The follow-up time for hip-level analyses, such as THA conversion and time-to-THA, was defined as the interval from osteoblast cell therapy to THA conversion or censoring at the last known follow-up; bilateral hips were treated as separate observations with patient-level clustering. Patients were diagnosed using an X-ray or MRI, which revealed a fracture, arthritis, collapse beyond 2 mm, or progression to grade IIIb/IV; these patients generally progressed to THA surgery.

Statistical analysis

Data were summarized using descriptive statistics. Continuous data were represented as mean ± standard deviation (SD), median (interquartile range (IQR)), and were compared using a Student’s t-test or a non‑parametric test, as applicable. Categorical data were presented as frequencies and percentages, and the chi-square test was used for comparative analysis. A p-value of < 0.05 was considered statistically significant. To account for within-patient correlation, additional analyses using cluster-robust standard errors and shared-frailty models were performed. The normality of VAS and HHS scores was assessed using the Shapiro-Wilk and Kolmogorov-Smirnov tests. Failure-free survival for THA was evaluated using Kaplan-Meier (KM) curves, and median follow-up duration was estimated using the reverse KM method. The log-rank test was used to compare different groups based on etiology and stage. The primary analysis for hip-level outcomes was performed using an adjusted Cox proportional hazards model to evaluate THA conversion with robust standard errors clustered by patients. Results were presented as hazard ratios (HRs) and 95% confidence intervals (CIs). Proportional hazards assumptions were evaluated using Schoenfeld residuals. The Cox model adjusted for age, sex, stage, etiology, bilaterality, baseline HHS and VAS, and center effect (modeled via robust variance/frailty), where ARCO stage was modeled as a three-level categorical variable (Stage I reference). Etiology was modeled as a multi-level categorical variable, with alcohol used as the reference category. As a secondary analysis, THA conversion was evaluated using multiple logistic regression with a generalized linear model. All analyses were performed using Python v3.12.12 (Python Software Foundation, Wilmington, Delaware, United States) in Google Colab (Google LLC, Mountain View, California, United States).