Factors affecting outcomes of intra-articular injection of autologous, culture-expanded, adipose-derived mesenchymal stem cells in knee osteoarthritis

Abstract

Background

This study aims to assess the outcomes of intra-articular injection of adipose-derived mesenchymal stem cells (ADMSC) in osteoarthritis (OA) of knee joint up to 3-year follow-up and evaluate possible factors related to the outcomes.

Methods

This was a prospective observational follow-up of participants from the ADMSC treatment arm of a prior phase III clinical trial. A total of 104 patients were assessed through the follow-up study for 3 years after ADMSC injection. We assessed the WOMAC and VAS for pain at 1-, 2-, and 3-year post–injection and compared with baseline values. Minimal clinically important difference (MCID) was calculated, and the proportion of patients achieving MCID was evaluated at each time point. Finally, we assessed which factors may affect the outcome.

Results

Significant improvements in WOMAC and VAS scores were observed after ADMSC injection, with these benefits sustained throughout the 1-, 2-, and 3-year follow-up. The proportion achieving MCID slightly decreased over time without statistical significance. Significant factors affecting the favorable outcomes were lower body mass index (BMI) (<25), higher baseline clinical scores up to 3 years after the injection. Other variables did not show the significant relation.

Conclusions

Intra-articular injection of autologous, culture-expanded ADMSC in patient with knee OA was sustained improvements observed up to 3 years in terms of WOMAC and VAS values. Moreover, patient BMI and baseline WOMAC were related to the outcomes after a single injection. Patient condition including severity of pain may affect the effect of ADMSC injection in OA knee. Clinical trial registration number: NCT04427930. This study was retrospectively registered.

Graphical Abstract

Graphical Abstract.

Significance statement.

This study explores how individual patient characteristics affect the results of stem cell injection for knee osteoarthritis (OA). By examining factors such as body mass index and baseline condition, the research identifies key elements that can influence treatment success. The findings suggest that personalizing stem cell therapy based on these factors may enhance the outcomes, offering a more effective approach for managing OA.

Introduction

Mesenchymal stem cell (MSC)-based therapies are gaining significant attention as a disease-modifying treatment option for knee osteoarthritis (OA).^1,2 MSCs can be utilized in the treatment of knee OA through the approaches such as a surgical intervention or an intra-articular injection.^3–5 Intra-articular injection of MSCs is particularly appealing for patients who are reluctant to undergo surgery, as it combines efficacy with a lower risk profile. Among the methods exist for harvesting MSCs,⁶ the adipose-derived MSCs (ADMSC) have gained prominence due to their easy accessibility and abundant availability.^7–9 There have been many studies elaborating on the efficacy of an intra-articular ADMSC injection in knee OA patients.^9–12 However, most studies reported limited sample sizes and relatively short follow-up periods.^10,13–16 Moreover, though a few study has been conducted to explore the factors influencing the efficacy of the ADMSCs treatments, existing literature only focused on surgical MSC implantation.³ Intra-articular injection of ADMSC has been reported successful short to midterm outcomes in the treatment of OA of the knee joint. However, there has been paucity regarding the factors affecting the outcomes after the injection. This study aims to assess the outcomes of ADMSC injection in patients with knee OA with a 3-year follow-up on previous phase III study evaluate the possible factors related to the outcomes.^12,17

Materials and methods

Study design and outcome measurements

This study was prospectively conducted in the patients with knee OA who completed the 6-month follow-up of previous phase III, a randomized and double-blinded clinical trial regarding the intra-articular autologous ADMSCs injection.¹² The present analysis performed in participants who received ADMSCs in the original trial (treatment arm) and no control group was included in this study. At the time of previous phase III clinical trial, autologous adipose tissue was harvested from the abdominal subcutaneous fat using the tumescent technique, as previously described.¹² Three weeks after the lipo-aspiration, an intra-articular injection containing autologous ADMSCs (1 × 10⁸ ADMSCs; normal saline, 2.1 mL; autologous serum, 0.9 mL) was administered under the ultrasound guidance in the outpatient clinic. Eligible participants were required to be over 20 years old, with K-L grade 3 knee OA, a pain intensity score of at least 50 on a 100-mm visual analog scale (VAS), and a functional impairment score of at least 40 on the total Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) at the time of screening. Among these patients, we excluded the control group who received placebo injection in the phase III clinical trial and included only the study group who received autologous ADMSCs injection. The study received approval from the Institutional Review Board at our institution (approval number: KHNMC 2024-12-004). Additional follow-up from the 6 month of phase III study was performed at 1, 2, and 3 years post–injection, measuring WOMAC and VAS scores. WOMAC score was set up to measure VAS scores on a scale of 0-100, resulting in a total score of 2400. Of the total 125 patients in the treatment group from the previous phase III clinical trial,¹² 104 patients consented to participate in the long-term follow-up and were monitored for an additional 3 years. Among 104 patients, 90 and 89 patients were followed up at 2 and 3 years after the injection, respectively. During the follow-up period, patients were allowed to receive conservative treatments such as oral analgesics or nonsteroidal anti-inflammatory drugs (NSAIDs) if they experienced intermittent knee pain. These treatments were permitted as part of routine symptom management and were not considered as study interventions. However, detailed information on concomitant medication use (eg, frequency, dose, or duration) was not systematically collected for quantitative analysis. We utilized the WOMAC to evaluate clinical outcomes, categorizing the WOMAC into total WOMAC, WOMAC function, and WOMAC pain as the index phase III study.¹² Alternative pain assessment was conducted using the VAS. VAS scores were obtained at baseline (prior to the injection), at 6 months post–injection, and at 1, 2, and 3 years follow-up. The possible factors influencing the changes in WOMAC and VAS score after the entire follow-up periods were categorized into body mass index (BMI), sex, symptom duration, cartilage defect on magnetic resonance imaging (MRI), age, height, and baseline WOMAC score. BMI was classified as <25 versus ≥ 25 kg/m² in a subgroup analysis, which is the standard for “obesity” in Asian population by the WHO Asia-Pacific classification, which also corresponds to the WHO’s standard for “overweight” by Western population standards.¹⁸ The detailed MRI protocol was described previously.¹¹ Cartilage defect assessment was performed using sagittal and coronal 3.0-mm T2-weighted Dixon in-phase sequences on MRI. The defect area in the medial compartment was quantified by multiplying the maximum anteroposterior (sagittal plane) and mediolateral (coronal plane) diameters, as defined for grade 3 or 4 lesions based on the modified Outerbridge classification, following previously validated imaging protocols.^10,11 Aim of current study was to assess the 3-year follow-up after the intra-articular injection of autologous ADMSCs with relatively large cohort and identify which factors significantly influenced the changes in WOMAC and VAS scores throughout the follow-up periods. Initially, we performed univariable analysis to determine the impact of each factor on changes in WOMAC and VAS scores. Subsequently, we confirmed the statistical significance of these factors through the multivariable analysis. And a further analysis of the clinical efficacy of the treatment was conducted to evaluate the number of patients who achieved the minimal clinically important difference (MCID) for the primary outcome measures, using 2 different approaches. According to previous literature, the MCID for the 100-mm VAS was defined as an absolute improvement of 14 points and a relative improvement of 20%, while the MCID for the total WOMAC score was defined as an absolute improvement of 9 points and a relative improvement of 33%.^12,14 Because WOMAC total score was analyzed on a 0-2400 scale in the current study, the published WOMAC MCID thresholds were rescaled (from 9 points to 216 points) to the equivalent values on our scoring system and applied accordingly.

Statistical analysis

The normality of endpoints was assessed using the Shapiro–Wilk test and normality plots. Due to non-linearity with endpoints, symptom duration and baseline cartilage defect area (AVAL_b) were log-transformed (log(AVAL_b + 1)) for analysis. A value of 1 was added prior to log transformation due to the presence of zero values. Variables included in the multivariable analysis were selected based on univariable analysis results, with factors showing a P-value <0.2 at any time point included. This ensured consistent variable inclusion across different time points. Baseline values of the endpoint were adjusted in all multivariable analyses. Multicollinearity was assessed using the variance inflation factor (VIF), with a VIF <4 indicating no significant multicollinearity. Although multiple testing was conducted across 4 time points (6 months, 1, 2, and 3 years), the Bonferroni correction, which multiplies each P-value by 4 to control for increased type I error, was not applied in this analysis. To account for within-patient correlation across repeated follow-up visits, longitudinal linear mixed-effects model was utilized with change-from-baseline as the dependent variable, time as a repeated factor, and BMI/baseline severity as fixed effects. Missing WOMAC and VAS scores at 2 and 3 years were imputed using the last observation carried forward (LOCF) method to minimize data loss and preserve statistical power for longitudinal comparisons. To determine whether a sample size of 104 patients was sufficient for this study, we conducted a post–hoc power analysis. In this study, the primary outcomes were the changes in WOMAC total at each follow up periods (1, 2, 3 years) from baseline. For each outcome, the post–hoc power to detect differences between two BMI groups (BMI ≥ 25 and BMI < 25) was calculated using multiple linear regression, adjusting for additional covariates. For WOMAC total change, with adjustment for 5 covariates (sex, symptom duration, age, height, and baseline WOMAC total), the power to detect the mean difference between BMI groups was 81.2%, indicating sufficient statistical power.

Results

WOMAC total

The WOMAC total score showed significantly improvement at 6 months compared to baseline (P < 0.0001), and this improvement was sustained at a significant level during the follow-up at 1 (P < 0.0001), 2 (P < 0.0001), and 3 years (P < 0.0001) (Figure 1A). Uni- and multivariable analysis revealed that the patients with a lower BMI (<25) and higher baseline WOMAC scores experienced significant changes in WOMAC scores across all follow-up periods (Table 1, Figure 2). Other factors did not reach the significant relationship. Standardized regression coefficients analysis for WOMAC total scores (Figure 2) revealed that lower BMI (<25) was significantly associated with greater improvements in WOMAC total scores. Similarly, higher baseline WOMAC scores were found to be a strong predictor of greater improvements over time. These results indicate that patients with more severe symptoms at baseline and lower BMI tend to experience the most substantial improvements in WOMAC total scores following intra-articular ADMSC injections. With respect to functional improvement as measured by the total WOMAC score (0-2400 scale), MCID was defined as an absolute improvement of ≥216 points (equivalent to 9 points on a 0-100 WOMAC scale) and/or a relative improvement of ≥33%.¹² Using this definition, 57.7% of patients achieved MCID at 6 months post–treatment; MCID achievement rates were 56.7% at 1 year, 51.9% at 2 years, and 52.9% at 3 years (P = 0.763).

Figure 1.

Changes in clinical outcome scores following intra-articular injection of autologous adipose-derived mesenchymal stem cells (ADMSCs) in 104 patients with knee osteoarthritis. (A) WOMAC total scores, (B) WOMAC function scores, (C) WOMAC pain scores, (D) WOMAC stiffness scores, and (E) VAS showed statistically significant improvements at 6 months, 1 year, 2 years, and 3 years compared to baseline (all P < .001).

Table 1.

Univariable and multivariable analysis—change of total WOMAC, WOMAC function, WOMAC pain, WOMAC stiffness, and VAS.

N = 104	6 months		1 year		2 years		3 years
Total WOMAC	βa	P-value	βa	P-value	βa	P-value	βa	P-value
Univariate variables
BMI (≥25, <25 [ref])	229.98	.0099	279.58	.0064	299.59	.0035	292.27	.0067
Sex (F, M [ref])	133.12	.1747	131.7	.2448	14.04	.9018	89.69	.4518
Log (duration)	78.31	.0686	64.9	.1923	−56.26	.2596	41.39	.43
Age (years)	6.55	.2984	12.57	.0823	15.06	.037	11.4	.1344
Height	−9.86	.0509	−8.59	.142	−11	.0595	−10.03	.1023
Weight	1.64	.6651	3.13	.472	4.15	.3414	4.97	.2763
Log (Cartilage defect size)	−0.54	.9886	−8.6004	.8177	−3.4788	.9259	−8.4525	.8299
Baseline WOMAC_total	−0.35	.0089	−0.71	<.0001	−0.5	.001	−0.71	<.0001
Multivariate variables
BMI (≥25, <25 [ref])	241.23	.0066	279.26	.0035	293.98	.0029	274.86	.0073
Sex (F, M [ref])	22.98	.8692	97.73	.5153	−214.95	.1663	−51.64	.7484
Log (duration)	81.97	.0438	59.21	.1734	−55.17	.2182	39.85	.392
Age (years)	3.17	.6278	10.54	.1368	12.39	.09	7.43	.3269
Height	−9.29	.2242	−2.33	.7758	−16.02	.0603	−10.28	.2441
Baseline WOMAC_total	−0.27	.0349	−0.61	<.0001	−0.41	.0044	−0.62	<.0001
WOMAC function
Univariate variables
BMI (≥25, <25 [ref])	171.11	.0078	209.71	.0054	212.89	.0043	203.96	.0096
Sex (F, M [ref])	99.62	.1591	135.54	.1021	2.18	.9789	54.92	.5274
Log (duration)	53.61	.0843	64.7	.0758	−40.33	.2658	33.82	.376
Age (years)	4.55	.3175	13.42	.0108	10.67	.0421	8.4	.1299
Height	−7.69	.0346	−8.1	.0585	−8.44	.0462	−7.16	.1094
Weight	1.2	.6596	0.96	.7645	2.52	.4264	3.49	.2944
Log (Cartilage defect size)	−11.38	.5377	−18.74	.3872	−26.62	.2137	−18.58	.411
Baseline WOMAC_function	−0.32	.0218	−0.68	<.0001	−0.52	.0011	−0.73	<.0001
Multivariate variables
BMI (≥25, <25 [ref])	182.16	.0046	233.87	.0007	205.58	.004	190.02	.0109
Sex (F, M [ref])	10.97	.9135	131.36	.2233	−186.02	.0991	−47.59	.6856
Log (duration)	56.96	.0529	58.29	.0626	−39.23	.2269	31.59	.353
Age (years)	2.2	.6412	12.96	.0113	8.43	.1105	5.82	.2915
Height	−7.57	.1722	0.1072	.9854	−13.26	.0323	−7.39	.2514
Baseline WOMAC_function	−0.23	.082	−0.55	.0001	−0.42	.0043	−0.64	<.0001
WOMAC pain
Univariate variables
BMI (≥25, <25 [ref])	35.42	.0645	45.31	.0456	54.06	.0139	61.17	.0076
Sex (F, M [ref])	13.33	.5254	27.56	.2606	−5.62	.817	21.67	.3919
Log (duration)	18.42	.0441	12.2	.2635	−14.5	.1723	6.68	.5488
Age (years)	1.4	.2964	2.62	.099	2.29	.1385	2.06	.2031
Height	−0.94	.387	−1.32	.3038	−1.53	.2202	−1.88	.1507
Weight	0.5	.5324	0.71	.4584	0.93	.3147	1.09	.2592
Log (Cartilage defect size)	5.31	.4412	3.03	.711	5.12	.5198	1.65	.8431
Baseline WOMAC_pain	−0.09	<.0001	−0.77	<.0001	−0.61	<.0001	−0.76	<.0001
Multivariate variables
BMI (≥25, <25 [ref])	32.85	.0648	40.91	.0364	51.1	.0113	56.19	.005
Log (duration)	18.56	.0265	11.73	.1971	−14.67	.1165	0.12	.4395
Age (Years)	0.58	.6634	1.56	.2825	1.82	.2225	0.91	.5391
Height	−1.1	.295	−1.21	.2927	−1.22	.2979	−2.01	.0856
Baseline WOMAC_pain	−0.46	<.0001	−0.73	<.0001	−0.56	<.0001	−0.72	<.0001
WOMAC stiffness
Univariate variables
BMI (≥25, <25 [ref])	26.47	.0047	16.44	.0984	24.05	.0162	23.42	.0277
Sex (F, M [ref])	18.42	.0734	20.65	.0556	14.63	.1827	15.27	.1903
Log (duration)	5.39	.2362	6.61	.1656	−5.38	.2658	1.74	.7349
Age (years)	0.82	.2179	1.51	.029	1.29	.0664	0.75	.32
Height	−0.1	.061	−1.23	.2065	−1.47	.0089	−1.22	.0411
Weight	0.12	.7611	−0.24	.5634	−0.07	.8756	0.19	.6714
Log (Cartilage defect size)	0.67	.845	3.01	.3986	2.51	.4886	0.48	.9005
Baseline WOMAC_stiffness	−0.6	<.0001	−0.68	<.0001	−0.54	.0002	−0.74	<.0001
Multivariate variables
BMI (≥25, <25 [ref])	26.67	.0025	17.28	.0578	23.75	.1028	20.05	.0429
Sex (F, M [ref])	10.95	.4311	7.66	.5988	−10.12	.5038	−7.14	.6513
Log (duration)	5.33	.1829	6.17	.1412	−5.11	0.24	1.96	.6656
Age (years)	0.45	.4868	0.93	.1768	0.72	.3083	−0.03	.9725
Height	−0.41	.5881	−0.56	.4813	−1.55	.0612	−1.35	.1183
Baseline WOMAC_stiffness	−0.51	<.0001	−0.59	<.0001	−0.46	.001	−0.68	<.0001
VAS
Univariate variables
BMI (≥25, <25 [ref])	13	.0078	13.35	.0114	17.21	.0006	14.76	.0054
Sex (F, M [ref])	−5.88	.2748	1.49	.7981	−6.22	.2642	−4.91	.4021
Log (duration)	1.94	.4137	0.6	.814	−4.38	.0757	0.01	.9973
Age (years)	0.23	.514	0.59	.1138	0.68	.0538	0.45	.2276
Height	−0.14	.6238	−0.38	.2096	−0.2	.4822	−0.09	.7724
Weight	0.34	.0947	0.22	.3198	0.39	.0704	0.46	.0377
Log (Cartilage defect size)	0.79	.6671	−0.76	.6908	0.55	.76	0.38	.8443
Baseline VAS_pain	−0.79	<.0001	−0.79	<.0001	−0.76	<.0001	−0.88	<.0001
Multivariate variables
BMI (≥25, <25 [ref])	8.08	.132	11.4	.0588	13.53	.0131	7.14	.2137
Log (duration)	2.96	.1356	1.49	.495	−3.65	.0689	0.95	.6548
Age (years)	0.15	.6277	0.5	.1514	0.81	.0114	0.5	.1382
Weight	0.03	.8834	−0.1	.7098	0.09	.6889	0.24	.3298
Baseline VAS_pain	−0.75	<.0001	−0.72	<.0001	−0.63	<.0001	−0.8	<.0001

^aChange-from-baseline was used as the dependent variable; BMI group and time were included as fixed effects; baseline outcome was included as a covariate; and a random intercept was specified for each participant. Time was treated as a repeated factor (6 months, 1, 2, and 3 years).

*WOMAC (Western Ontario and McMaster Universities Osteoarthritis Index), VAS (visual analog scale), BMI (body mass index).

Figure 2.

Standardized regression coefficient and 95% confidence interval from multivariable analysis for factors affecting changes of WOMAC total (A) at 6 months from baseline (N = 104), (B) at 1 year from baseline (N = 104), (C) at 2 year from baseline (N = 104), (D) at 3 year from baseline (N = 104).

WOMAC function

The WOMAC function score showed significant improvement at 6 months compared to baseline (P < 0.0001), and this improvement was sustained at a significant level during follow-up at 1 (P < 0.0001), 2 (P < 0.0001), and 3 years (P < 0.0001) (Figure 1B). For WOMAC function, uni- and multivariable analysis showed that patients with a BMI (<25) experienced significant decreases in WOMAC function scores (1 year: P = 0.0007, 2 years: P = 0.004, 3 years: P = 0.0109). Moreover, baseline WOMAC function scores significantly influenced scores at all time points (1 year: P < 0.0001, 2 years: P = 0.0043, 3 years: P < 0.0001) (Table 1, Figure 3). Other factors did not reach the significant relationship. Standardized regression coefficients analysis for WOMAC function scores (Figure 3) confirmed that lower BMI (<25) consistently correlated with more significant improvements in function scores over time. Additionally, baseline WOMAC function scores were identified as a key predictor of functional improvements, with higher baseline scores leading to more substantial gains. These findings emphasize the role of baseline severity and BMI in predicting functional recovery following ADMSC injections.

Figure 3.

Standardized regression coefficient and 95% confidence interval from multivariable analysis for factors affecting changes of WOMAC function (A) at 6 months from baseline (N = 104), (B) at 1 year from baseline (N = 104), (C) at 2 year from baseline (N = 104), (D) at 3 year from baseline (N = 104).

WOMAC pain

For WOMAC pain, the results were similar to the previous analyses. The WOMAC pain score showed significant improvement at 6 months compared to baseline (P < 0.0001), and this improvement was sustained at a significant level during follow-up at 1 (P < 0.0001), 2 (P < 0.0001), and 3 years (P < 0.0001) (Figure 1C). Uni- and multivariable analysis showed that the patients with a BMI below 25 related to significant decreases in WOMAC pain scores (1 year: P = 0.0364, 2 years: P = 0.0113, 3 years: P = 0.005), although it showed only tendency without significance at 6 months (P = 0.0648). Furthermore, baseline WOMAC pain scores significantly influenced WOMAC pain improvement at all time periods (1 year: P < 0.0001, 2 years: P < 0.0001, 3 years: P < 0.0001) (Table 1, Figure 4). Other factors did not reach the significant relationship. Standardized regression coefficients analysis for WOMAC pain scores (Figure 4) showed that lower BMI was significantly associated with greater reductions in pain scores over time. Additionally, higher baseline WOMAC pain scores were strongly predictive of more substantial pain improvement, further highlighting the importance of baseline severity and BMI in shaping the pain reduction outcomes.

Figure 4.

Standardized regression coefficient and 95% confidence interval from multivariable analysis for factors affecting changes of WOMAC pain (A) at 6 months from baseline (N = 104), (B) at 1 year from baseline (N = 104), (C) at 2 year from baseline (N = 104), (D) at 3 year from baseline (N = 104).

WOMAC stiffness

The WOMAC stiffness score showed a significant reduction at 6 months compared to baseline (P < 0.0001) and remained improved throughout the follow-up period at 1 year (P < 0.0001), 2 years (P < 0.0001), and 3 years (P < 0.0001) (Figure 1D). Uni- and multivariable analyses demonstrated that patients with a lower BMI (<25) exhibited greater reductions in WOMAC stiffness scores across all time points. Notably, the improvement in stiffness was significantly more pronounced in the BMI <25 group compared to BMI ≥ 25 at 3 years (P = 0.0429) (Table 1). Other variables, including sex, age, symptom duration, and height, did not demonstrate a statistically significant relationship with stiffness improvement. Standardized regression coefficient analysis further confirmed that lower BMI (<25) was strongly associated with greater reductions in stiffness scores over time. Additionally, higher baseline WOMAC stiffness scores were significantly correlated with greater improvement in stiffness at all time points (1 year: P < 0.0001, 2 years: P = 0.001, 3 years: P < 0.0001) (Figure 5). These findings indicate that patients with more severe stiffness at baseline and a lower BMI derive the greatest benefit in stiffness reduction following intra-articular ADMSC injections. This suggests that baseline stiffness severity and BMI may serve as key prognostic factors in predicting response to treatment.

Figure 5.

Standardized regression coefficient and 95% confidence interval from multivariable analysis for factors affecting changes of WOMAC stiffness (A) at 6 months from baseline (N = 104), (B) at 1 year from baseline (N = 104), (C) at 2 year from baseline (N = 104), (D) at 3 year from baseline (N = 104).

VAS pain scores

The VAS pain score showed significant improvement at 6 months compared to baseline (P < 0.0001), and this improvement was sustained at a significant level during follow-up at 1 (P < 0.0001), 2 (P < 0.0001), and 3 years (P < 0.0001) (Figure 1E). Both uni- and multivariable analyses revealed that lower BMI was significantly associated with greater improvements in VAS pain scores at 2 years (P = 0.0131). Although the associations at 6 months (P = 0.132), 1 year (P = 0.0588) and 3 years (P = 0.2137) were not statistically significant, the trend still indicated that lower BMI correlated with a greater reduction in pain (Table 1). Baseline VAS pain scores consistently emerged as the most robust predictor of VAS pain changes across all time points, with higher baseline scores associated with more substantial pain improvement (Table 1, Figure 6). Other factors did not reach the significant relationship. At 6 months following treatment, 70.2% of patients who received autologous ADMSC injection, achieved MCID for the VAS for pain, based on both the absolute improvement threshold of ≥14 points and the relative improvement threshold of ≥20%.¹² This proportion slightly decreased over time, with 67.3% achieving the MCID at 1 year, 56.7% at 2 years, and 61.5% at 3 years (P = 0.290). Standardized regression coefficients analysis for VAS pain scores (Figure 6) confirmed that lower BMI (<25) was consistently associated with greater reductions in pain, particularly at earlier follow-up points. Baseline VAS pain scores also played a crucial role in predicting pain improvement, with higher baseline scores leading to more pronounced reductions in pain over time.

Figure 6.

Standardized regression coefficient and 95% confidence interval from multivariable analysis for factors affecting changes of VAS (A) at 6 months from baseline (N = 104), (B) at 1 year from baseline (N = 104), (C) at 2 year from baseline (N = 104), (D) at 3 year from baseline (N = 104).

Linear mixed effects model

In the longitudinal mixed-effects models, BMI group was a significant fixed effect for all WOMAC total, function, pain, and stiffness (Table 2). Meanwhile, VAS pain score showed only tendency associate with BMI group (P = 0.0599). Baseline endpoint values were significant predictors across all outcomes. The effects of time as a repeated factor were not statistically significant. BMI and time interaction were also not statistically significant (all P > 0.05).

Table 2.

Linear mixed effects model.

Outcome	BMI (P)	Baseline outcome (P)	Time (P)	BMI × time (P)
WOMAC total	.0017	<.0001	.7402	.9481
WOMAC function	.0019	<.0001	.8132	.968
WOMAC pain	.0027	<.0001	.8165	.7834
WOMAC stiffness	.0104	<.0001	.9355	.8665
VAS pain	.0599	<.0001	.8862	.8962

Discussion

In the current study, 2 key factors—BMI and baseline symptom severity—were associated with clinical outcomes. Patients with lower BMI (<25) showed greater improvements over follow-up suggesting that over-weight may be associated with a reduced therapeutic response to ADMSC therapy. These findings provide valuable insights into patient selection criteria for intra-articular ADMSC treatment. Numerous studies have demonstrated the promising clinical efficacy of intra-articular MSCs injection for knee OA.^{11,12,19–22} However, most studies have reported follow-up period up to 1 year, with only a few extending to 24 months.^10,13–16 Previous meta-analyses²³ have reported limited or no long-term benefits of MSC therapy. In contrast, our study demonstrated that improvements in VAS and WOMAC scores were maintained consistently over a 3-year period. To the best of our knowledge, this study would be the first to assess the factors affecting the outcomes after intra-articular injections of ADMSCs in a relatively large cohort of 104 patients based on observational extension, 3 years follow-up of the treatment arm from a prior randomized, double-blind phase III trial. This study utilized various statistical analyses to compare and analyze patient characteristics (age, gender, BMI, existing cartilage defects, baseline WOMAC and VAS scores, etc.) and their impact on clinical outcomes. Our findings confirm that intra-articular ADMSC injections lead to sustained clinical improvements over 3 years, as evidenced by significant reductions in WOMAC and VAS scores.

Sustained improvements in pain and function observed up to 3 years after ADMSC administration may be attributable to the known biological properties of MSCs, including immunomodulatory effects and regenerative activity. Preclinical and clinical studies have shown that MSCs can secrete soluble factors that reduce inflammation and promote cartilage repair.^24–27 The specific ADMSC product used in this study has demonstrated such mechanisms in in vitro and in vivo experiments conducted by the manufacturer. In a prior proof-of-concept trial using this formulation, histological analysis confirmed cartilage regeneration via Safranin O and collagen staining after 6 months of treatment.¹⁰ Additionally, a 5-year follow-up MRI study revealed a decrease in cartilage defect area, supporting the long-term regenerative effect.²⁸ These mechanistic insights may help explain the durability of clinical benefits seen in our cohort.

Obesity is well-established the risk factor for knee OA.^29–32 Moreover, previous report indicated that MSCs from the obese patients exhibit reduced proliferation rates, increased cellular senescence, and decreased differentiation potential, including chondrogenesis.³³ Additionally, 1 study has shown poor outcomes in patients with a BMI above 27 who received MSCs without fibrin glue as a scaffold.³⁴ Also, the relationship between BMI and surgical treatment outcomes in OA has been extensively investigated in previous studies.^35–38 Yang et al. found that patients with lower BMI who underwent high tibial osteotomy experienced significant clinical improvement.³⁹ This aligns with our finding that the intra-articular injection of autologous ADMSCs yielded better WOMAC improvements in patients with lower BMI (<25) compared to those with a BMI of 25 or higher (Supplementary Material 1). This significant clinical improvement was observed not only in the total WOMAC scores but also in the functional, pain, stiffness scores and VAS of patients with a BMI below 25 (Supplementary Material 1). However, improvements in VAS scores were observed, but statistical significance was not consistently achieved across analyses and time points (Table 1). Although both instruments assess pain, they differ in their measurement properties. The VAS is a single-item measure of overall pain intensity, whereas the WOMAC pain subscale is a multi–item instrument capturing pain across multiple functional activities. The multi–item structure of the WOMAC pain scale may reduce random measurement error and improve responsiveness, thereby increasing statistical power to detect changes. This difference may explain why statistically significant changes were observed in WOMAC pain but not in VAS.

The relationship between baseline knee condition before the treatment and the response to the intra-articular injection remains a subject of debate.⁴⁰ However, Pendleton et al.⁴¹ reported that higher baseline pain levels were predictive of a better response to the intra-articular corticosteroid injections. Similarly, Wu et al.⁴⁰ found that the WOMAC pain score was an independent predictor of response to the intra-articular steroid injections in patients with knee OA, even after adjusting for variables such as age, BMI, stiffness, disability, and concomitant analgesic use. Their study demonstrated that the patients with WOMAC pain scores of ≥5 had a 2.54 times higher positive response rate to intra-articular corticosteroid injections (44.8% vs. 17.6%, P < 0.001) compared to those with WOMAC pain scores <5.⁴⁰ These findings align with our results, which show that the patients with higher baseline WOMAC and VAS scores, indicating more severe initial symptoms, experienced greater improvement in WOMAC scores following the intra-articular injection of autologous ADMSCs. Similarly, in the context of surgical intervention, it has been observed that lower baseline WOMAC scores are associated with greater difficulty in achieving the MCID following the high tibial osteotomy for knee OA.⁴⁰ This suggests that, like a surgical treatment, intra-articular injection of autologous ADMSCs may yield more favorable outcomes in patients with worse knee scores before the treatment. However, the association between higher baseline WOMAC/VAS scores and greater improvement may partially reflect regression to the mean. Overall, patients with higher baseline WOMAC/VAS scores showed greater improvements following ADMSC injection in our cohort. While regression to the mean may partly contribute, the consistent direction of this association across outcomes supports baseline symptom severity as a relevant factor to consider when interpreting clinical change.

Age would be the another risk factor for knee OA and may influence on the quality of autologous MSCs.⁴² Chang et al. compared the number and function of MSCs in articular cartilage among fetuses, healthy adults (28-45 years), and elderly adults (60-75 years), finding significant differences in the MSC prevalence and chondrogenic differentiation potential.⁴³ They reported that MSCs accounted for 94.69%, 4.85%, and 6.33% of cells in these groups, respectively (P < 0.001), with lower chondrogenic differentiation in elderly patients potentially contributing to OA development.⁴³ Other studies have shown age-dependent effects on the characteristics of adipose-derived MSCs.^44,45 Choudhery et al. found that MSCs from the older donors exhibited more senescent features and reduced viability, proliferation, and differentiation potential compared to those from younger donors.⁴⁴ Wu et al. reported significantly higher yields of adipose-derived MSCs in younger rats compared to older ones (P < 0.02).⁴⁵ However, in our study, age did not consistently affect the improvements in WOMAC total and functional scores at the 3-year follow-up. This inconsistency would be resulted from the limited age distribution of our patient cohort (average: 64.6 years old, range: 50-80 years). With a more diverse age distribution, age could potentially be a more significant prognostic factor.

This study has some limitations. First, the study population primarily consisted of Asian women with relatively low BMI, which may limit the generalizability of our findings to populations with different demographic characteristics. Previous studies have indicated that BMI may influence the therapeutic efficacy of ADMSC treatment, and populations with higher BMI distributions may exhibit different treatment responses. Therefore, future studies are warranted to investigate how the effectiveness of intra-articular ADMSC injections may vary according to ethnicity, BMI range, and sex distribution. Second, because this was a single-arm observational extension of the treatment arm without a concurrent control group, sustained improvements cannot be conclusively attributed to ADMSC injection alone, and may be influenced by natural history, placebo effects, and/or concomitant treatments. In addition, conservative treatments (eg, oral analgesics or NSAIDs) during the follow-up periods were permitted. These medications might have had a marginal effect on the observed clinical improvement. Third, given that multiple outcomes, follow-up time points, and predictors were evaluated, there is a potential risk of inflated type I error; thus, particularly borderline findings should be interpreted cautiously. Fourth, baseline meniscal structural abnormalities on MRI (eg, meniscal root tears) were not systematically collected in a manner that allowed inclusion as covariates in our analyses. Therefore, we could not assess the impact of these lesions; however, serial MRI during follow-up was used to evaluate weight-bearing cartilage defects as an indicator of OA severity. Finally, there were some missing data during follow-up periods. The number of missing observations at each follow-up time point were 14 at 2 years and 15 at 3 years. To address this, missing WOMAC and VAS scores at 2 and 3 years were imputed using the LOCF method. While LOCF has certain limitations, particularly in long-term follow-up studies, its application in this context helped minimize data loss and allowed for consistent longitudinal analyses.

Conclusion

Intra-articular injection of autologous, culture-expanded ADMSC in patient with knee OA was sustained improvements observed up to 3 years in terms of WOMAC and VAS values. Moreover, patient BMI and baseline WOMAC were related to the outcomes after a single injection. Patient condition including severity of pain may affect the effect of ADMSC injection in OA knee.

Supplementary material

Supplementary material is available at Stem Cells Translational Medicine online.

Funding

This study was supported by a research grant from R-Bio Co., Ltd. (Clinical trial registration: NCT04427930).

Conflicts of interest

K.I.K. declares that he has received research grants from R-Bio Co., Ltd. J.B.P., K.Y.C., H.L., and G.Y.J. declare that they have no competing interests.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to patient privacy restrictions, but are available from the corresponding author on reasonable request.

Ethics statement

This study was conducted in accordance with the Declaration of Helsinki and was approved by the Institutional Review Board of Kyung Hee University Hospital at Gangdong (approval number: KHNMC 2024-12-004). Written informed consent was obtained from all participants prior to their inclusion in the study.