Results

Participant characteristics

Participant baseline characteristics are presented in Table 1. When comparing participants with muscle gain and muscle loss to those with stable muscle, no significant differences between the groups were found regarding sex, mean age (with a SMD of −0.13 for muscle gain and SMD = 0.09 for muscle loss), and BMI at baseline (for muscle gain SMD = −0.12; for muscle loss SMD = 0.15). Of the 2298 eligible participants, 1715 (74.6%) had complete WORMS data and were included in the analysis, whereas 583 (25.4%) were excluded because of missing WORMS scores (Supplementary Table S1). Compared with included participants, those with missing WORMS data were slightly older (61.3 ± 9.3 vs. 60.0 ± 8.9 years; SMD = 0.15) and had a lower BMI (26.6 ± 5.0 vs. 28.8 ± 4.3 kg/m2; SMD = 0.49). Physical activity (PASE score; SMD = 0.04) was similar between groups.

 | Baseline Characteristics | All participants (n = 1715) | Muscle gain group (n = 431) | Stable muscle group (n = 864) | Muscle loss group (n = 420)
Mean WORMS sum score | Mean age, y (SD) | 60.0 (8.9) | 59.0 (8.2) | 60.1 (8.9) | 60.8 (9.4)
Female sex (%) | 931 (54.3) | 242 (56.1) | 462 (53.5) | 227 (54.0)
Mean BMI, kg/m2 (SD) | 28.8 (4.3) | 28.3 (4.1) | 28.8 (4.3) | 29.0 (4.3)
Baseline muscle volume, cm3 (SD) | 745.2 (188.6) | 702.4 (179.6) | 759.5 (183.0) | 759.9 (202.3)
Menisci (SD) | 3.4 (3.8) | 3.4 (3.9) | 3.4 (3.9) | 3.4 (3.8)
Ligament (SD) | 0.4 (0.9) | 0.4 (1.0) | 0.4 (0.9) | 0.4 (1.0)
Tendon (SD) | 0.2 (0.7) | 0.3 (0.8) | 0.3 (0.7) | 0.2 (0.6)
Cartilage (SD) | 6.9 (5.3) | 6.5 (4.9) | 6.9 (5.4) | 7.6 (5.6)
BMELL (SD) | 2.2 (2.2) | 2.2 (2.2) | 2.1 (2.2) | 2.4 (2.3)
Mean WOMAC Subscores | Pain (SD) | 2.1 (2.9) | 2.0 (2.7) | 2.0 (2.8) | 2.2 (3.1)
Disability (SD) | 6.5 (9.0) | 6.2 (8.5) | 6.3 (8.6) | 7.0 (10.1)
Stiffness (SD) | 1.4 (1.5) | 1.3 (1.5) | 1.4 (1.5) | 1.5 (1.5)

Baseline clinical and structural characteristics by muscle change group

No significant differences were observed between the groups for WOMAC scores (pain: SMD = −0.01 for muscle gain, SMD = 0.06 for muscle loss; stiffness: SMD = −0.05 for muscle gain, SMD = 0.04 for muscle loss; functional limitation: SMD = −0.01 for muscle gain, SMD = 0.07 for muscle loss) or WORMS sum scores (menisci: SMD = 0.03 for muscle gain, SMD = −0.01 for muscle loss; ligaments: SMD = 0.01 for muscle gain, SMD = −0.01 for muscle loss; tendons: SMD = 0.05 for muscle gain, SMD = −0.10 for muscle loss; cartilage: SMD = −0.06 for muscle gain and 0.13 for muscle loss). Comparing the mean cartilage WORMS sum score at baseline between the muscle gain group and muscle loss group, the muscle gain group showed a significantly lower WORMS cartilage sum score (6.5 ± SD 4.9),p = 0.02). The muscle loss group had a significantly higher WORMS cartilage sum score than the stable group (7.6 ± 5.6,p = 0.04). Baseline TMV was significantly lower in the muscle gain group (702.4 ± 179.6 cm3,p < 0.01) compared to the muscle loss (759.9 ± 202.3 cm3,p < 0.01) and stable muscle group (759.5 ± 183.0 cm3,p < 0.01), with no significant difference between the muscle loss and stable muscle group (p = 0.99).

Longitudinal changes in muscle and WORMS sum scores

No significant differences were observed for adjusted mean changes in meniscus (muscle gain: 0.87 [95%CI: 0.48, 1.25], stable muscle: 0.85 [0.49, 1.22], muscle loss: 1.01 [0.62, 1.39]) or cartilage (muscle gain: 1.94 [1.38, 2.49], stable muscle: 1.98 [1.46, 2.51], muscle loss: 2.14 [1.58, 2.69] scores across the muscle groups. For BMELL scores, the muscle gain group (0.16 [−0.16, 0.48]) showed significantly lower adjusted mean changes compared to the stable muscle group (0.45 [0.14, 0.75],p = 0.03), while the comparison between muscle gain and muscle loss (0.47 [0.15, 0.80],p = 0.11) and muscle loss and stable muscle (p = 1.00) was not significant. Imaging examples of changes in BMELL are shown in Fig. 2.

Fig. 2: Imaging examples of changes in bone marrow edema-like lesions (BMELL). Left: Axial T1-weighted MR images of the right thigh with artificial intelligence-enabled muscle segmentation at baseline and after 48 months. Right: Sagittal intermediate-weighted fast spin-echo fat-suppression sequence of the right knee at baseline and after 48 months. The patient with muscle gain (predominantly in the quadriceps muscle, indicated in red, with an approximately 11% increase in total muscle volume over 48 months) was a 65-year-old man with mild knee pain at baseline (WOMAC pain subscale of 1) and no knee pain at 48 months (WOMAC pain subscale of 0). Despite the clinical improvement, new BMELLs developed in the medial femoral condyle (grey arrowhead), with baseline WORMS BMELL grades of 0 and 1 for medial femoral condyle at 48 months. The patient with muscle loss (predominantly in the quadriceps muscle (red) and hamstrings (green) with an approximately 14% decrease in total muscle volume over 48 months) was a 64-year-old woman with mild knee pain at baseline (WOMAC pain subscale of 1) and an increase of knee pain at 48 months (WOMAC pain subscale of 11). New BMELL developed in the medial tibial plateau and in the medial femoral condyle (white arrowheads), with baseline WORMS BMELL grades of 0 and 3 for both medial tibial plateau and medial femoral condyle at 48 months.

Significant differences were found for ligament and tendon scores: for ligament score changes, the muscle gain group (0.02 [−0.15, 0.19]) showed significantly lower adjusted mean changes compared to the muscle loss group (0.14 [−0.03, 0.31],p = 0.04), while the comparison between the muscle gain and stable muscle group (0.14 [−0.02, 0.29],p = 0.08) and muscle loss and stable muscle group (p = 0.61) was not significant. Fig. 3 shows imaging examples of ligament pathologies. For tendon score changes over 48 months (Fig. 4), the muscle gain group (0.05 [−0.07, 0.16]) showed significantly smaller adjusted mean changes compared to the muscle loss group (0.15 [0.03, 0.26],p = 0.02). Comparisons between the muscle gain and stable muscle group (0.07 [−0.04, 0.18],p = 0.87) and muscle loss and stable muscle group (p = 0.12) were not significant (Table 2).

Fig. 3: Imaging examples of changes in anterior cruciate ligament (ACL). Left: Axial T1-weighted MR images of the right thigh with artificial intelligence-enabled muscle segmentation at baseline and after 48 months. Right: Sagittal intermediate-weighted fast spin-echo fat-suppression sequence of the right knee at baseline and after 48 months. The patient with muscle gain (predominantly of the adductors (yellow) and quadriceps (red), with an approximately 6% increase in total muscle volume over 48 months) was a 53-year-old woman without knee pain at baseline and at 48 months (WOMAC pain subscores of 0) showed no change of the ACL (WORMS ACL subscore of 1 for both at baseline and at 48 months). The patient with muscle loss (predominantly of the quadriceps (red), with an approximately 9% decrease in total muscle volume) was a 62-year-old man with mild knee stiffness at baseline and at 48 months (both with a WOMAC stiffness subscore of 2) showed a new partial tear of the ACL after 48 months (white arrowhead), with baseline WORMS ACL subscore of 0 and 3 at 48 months.

Fig. 4: Imaging examples of changes in the patellar tendon. Left: Axial T1-weighted MR images of the right thigh with artificial intelligence-enabled muscle segmentation at baseline and after 48 months. Right: Sagittal intermediate-weighted fast spin-echo fat-suppression sequence of the right knee at baseline and after 48 months. The patient with muscle gain (predominantly of the quadriceps (red) and hamstrings (green), with an approximately 8% increase in total muscle volume over 48 months) was a 78-year-old man with a slight decrease in knee pain over 48 months (WOMAC pain subscore at baseline of 3 and after 48 months a score of 2) showed a new signal abnormalities around the patellar tendon at 48 months (black arrowhead), with a baseline WORMS patellar tendon subscore of 0 and 1 at 48 months). The patients with muscle loss (predominantly of the quadriceps (red), with an approximately 8% decrease in total muscle volume) was a 66-years-old man mild knee pain at baseline (WOMAC pain subscore of 1) and an increase of knee pain over 48 months (WOMAC pain subscore of 7 at 48 months) showed an increased tendinopathy of the patellar tendon (white arrowhead), with baseline WORMS patellar tendon subscore of 0 and 2 at 48 months.

Change in WORMS over 48 months | Muscle gain group | Stable muscle group | Muscle loss group | P value for muscle gain vs. muscle loss | P value for muscle gain vs. stable muscle | P value for muscle loss vs. stable muscle
Δ mean Menisci (95% CI) | 0.87 (0.48, 1.25) | 0.85 (0.49, 1.22) | 1.01 (0.62, 1.39) | 0.50 | 0.65 | 0.75
Δ mean Ligament (95% CI) | 0.02 (−0.15, 0.19) | 0.14 (−0.02, 0.29) | 0.14 (−0.03, 0.31) | 0.04 | 0.08 | 1.00
Δ mean Tendon (95% CI) | 0.05 (−0.07, 0.16) | 0.07 (−0.04, 0.18) | 0.15 (0.03, 0.26) | 0.02 | 0.87 | 0.12
Δ mean Cartilage (95% CI) | 1.94 (1.38, 2.49) | 1.98 (1.46, 2.51) | 2.14 (1.58, 2.69) | 0.06 | 0.75 | 0.07
Δ mean BMELL (95% CI) | 0.16 (−0.16, 0.48) | 0.45 (0.14, 0.75) | 0.47 (0.15, 0.80) | 0.11 | 0.03 | 1.00

Multivariable regression analyses revealed significant associations between muscle group and changes in WORMS sum score over 48 months. Muscle gain was significantly associated with less WORMS progression in the ligaments (β = −0.12 [−0.22, −0.02],p = 0.01), while muscle loss was significantly associated with higher WORMS progression of the tendon score (β = 0.08 [0.01, 0.15],p = 0.02), suggesting worsening of tendon pathologies in the muscle loss group. Furthermore, muscle gain was significantly associated with less WORMS progression in BMELL (β = −0.29 [−0.47, −0.10],p = 0.002). Given the mean baseline BMELL score of 2.2 points, this corresponds to an estimated difference of approximately 13% of the average baseline BMELL burden. No significant associations were observed for the WORMS progression in cartilage (muscle gain vs. stable: β = −0.05 [−0.37, 0.27]; muscle loss vs. stable: β = 0.15 [−0.17, 0.48]) and menisci in the muscle groups (muscle gain vs. stable: β = 0.01 [−0.21, 0.23]; muscle loss vs. stable: β = 0.16 [−0.07, 0.38]) (see also Supplementary Table S2). After Benjamini–Hochberg false discovery rate adjustment for the prespecified primary outcomes, the associations between muscle gain and reduced BMELL progression (adjusted p = 0.008), ligament progression (adjusted p = 0.02), and WOMAC pain (adjusted p = 0.04) remained statistically significant. No significant association was observed for cartilage progression (adjusted p = 0.75) (see Supplementary Table S3). Sensitivity analyses using percentage TMV change as a continuous exposure variable showed associations in the same direction as the primary categorical analyses. Associations with ligament, tendon, and BMELL progression showed a statistical trend, yet did not reach the statistical level of significance (Supplementary Table S4). Additional adjustment for baseline physical activity (PASE score) did not alter the estimated associations between muscle group and changes in WORMS outcomes. Regression coefficients and statistical significance remained essentially unchanged (Supplementary Table S5). In a sensitivity analysis including participants with extreme TMV changes, the association between muscle gain and reduced ligament progression remained statistically significant, whereas the association with BMELL progression was attenuated and no longer reached statistical significance (Supplementary Table S6).

Longitudinal changes in muscle and clinical scores

The WOMAC pain subscore decreased slightly in all groups, with adjusted mean changes of −1.04 [−1.61, −0.48] for muscle gain, −0.69 [−1.22, −0.15] for stable muscle, and −0.68 [−1.25, −0.11] for muscle loss. Similarly, the WOMAC functional subscore showed slight reductions, with adjusted mean changes of −2.66 [−4.35, −0.96], −1.74 [−3.34, −0.14], and −1.97 [−3.66, −0.27] for muscle gain, stable muscle, and muscle loss, respectively. Changes in WOMAC stiffness subscore were minimal across all groups, with adjusted mean changes of −0.31 [−0.62, −0.01] for muscle gain, −0.26 [−0.55, 0.03] for stable muscle, and −0.25 [−0.56, 0.05] for muscle loss.

Multivariable regression analyses revealed significant association between muscle gain and the reduction in WOMAC pain score (β = −0.36 [−0.68, −0.04],p = 0.03) when compared with the stable muscle group, while no significant association was found for the muscle loss group (β = −0.004 [−0.33, 0.32]) as well as for the WOMAC stiffness (muscle gain vs. stable: β = −0.07 [−0.24, 0.11]; muscle loss vs. stable: β = −0.003 [−0.18, 0.17]) and disability subscores (muscle gain vs. stable: β = −0.88 [−1.83, 0.07]; muscle loss vs. stable: β = −0.17 [−1.12, 0.79]) (see also Supplementary Table S2).

However, no significant differences in adjusted mean CST change were observed between the muscle groups (all p > 0.05). To further evaluate functional change, an ordinal regression analysis was performed using categorized CST change (deterioration, stable, improvement) as the outcome while adjusting for age, sex, race, and BMI. In this analysis, the muscle loss group was significantly associated with a greater likelihood of functional deterioration compared with the stable muscle group (β = −0.58 [−1.08, −0.08], p = 0.02), indicating higher declines in performance over time. No significant associations were observed for the muscle gain group (β = −0.21 [−0.70, 0.29], p = 0.41).