Within the cohort of 91 patients, a male predominance was observed (58.2%, 53/91) (Table 1). The mean age at CT examination was 65.9 ± 10.9 years. Most patients had an R-ISS score of 2 (53.2%, 41/77). A high-risk cytogenetic profile was observed in 16.0% (13 out of 81) of all patients. The mean clinical follow-up time was 15.7 ± 15.5 month (median: 9.2, IQR: 13.3).
The mean body composition parameters ± standard deviation of the cohort are depicted in Table 1. The correlation analysis revealed weak to moderate correlations between age and bone density (Spearman r −0.299, p = 0.004), EAT (Spearman r 0.211, p = 0.044) and CM (Spearman r 0.308, p = 0.003) (Supplementary Fig. 1A). Moreover, a significant positive correlation was observed between all adipose tissue compartments, as well as between adipose tissue compartments and CM and SM. In addition, the muscle volume correlated positively with TAT, IMAT, SAT, VAT, PAT, EAT, and SM. Furthermore, bone density correlated positively with muscle and negatively with CM. In addition, a Mann–Whitney U analysis revealed a significant difference between male and female study participants regarding muscle volume (W = 549, p = 0.0002), VAT (W = 611, p = 0.0015), PAT (W = 515, p = 0.0001), EAT (W = 691, p = 0.011), CM (W = 746, p = 0.036) and SM (W = 592, p = 0.0009) (Supplementary Fig. 1B).
Cardiac marker index differs between patients depending on high-risk cytogenetics
The BCA parameters were analyzed across R-ISS categories, with the results presented in Fig. 1A. No significant differences in BCA parameters were identified among the three R-ISS categories. Additionally, a comparison of BCA parameters was conducted between patients with and without high-risk cytogenetics using the Mann–Whitney U test, as shown in Fig. 1B. Patients with high-risk cytogenetic profiles exhibited a significantly elevated cardiac marker index compared to those without high-risk cytogenetics (W = 278, p = 0.0354). No statistically significant differences were observed for the remaining BCA parameters between the two groups.
Violin plots illustrating BCA parameters in patients stratified by (A) R-ISS scores (1, 2, or 3) and (B) the presence or absence of high-risk cytogenetics (no/yes). Comparisons among R-ISS categories were performed using the Kruskal–Wallis test, while differences based on cytogenetic risk were analyzed with the Mann–Whitney U test. Statistical test results are presented within the figure.
Distinct Body composition parameters differ between patients with progression-free survival and progression
Clinical follow-up data were available for all patients. Patients were followed-up for a mean duration of 15.7 ± 15.5 months (minimum follow-up 0.6 months, maximum 68.8 months). Progression-free survival was observed in 48 (52.7%) patients, whereas 43 (47.3%) patients developed disease progression. Disease progression was defined according to the response criteria established by the International Myeloma Working Group (IMWG) which was observed in 32 (35.2%) of patients, or death of all causes in 11 (12.1%) patients.
In patients who experienced disease progression or died during follow-up, the volume of subcutaneous adipose tissue (Mann–Whitney U test, W = 1325, p = 0.020) was significantly lower compared to those with progression-free long-term follow-up (Fig. 2A). The volume of total adipose tissue showed a trend to be lower in patients of the progression/death during follow-up group (Mann–Whitney U test, W = 1279, p = 0.05). Binomial logistic regression revealed an association with lower muscle volume, lower visceral adipose tissue and lower sarcopenia marker and progression-free survival. The data were adjusted to age and sex. The respective results are depicted in Fig. 2B.
Distinct body composition parameters differ between patients with progression-free survival and those with disease progression or death during follow-up. (A) Violin plots comparing BCA parameters between patients with progression-free survival (grey) and those with disease progression or death during follow-up (black), analyzed using the Mann–Whitney U test. (B) Forest plots showing the results of binomial logistic regression assessing the association between progression-free survival and disease progression or death during follow-up, adjusted for age and sex. Odds ratios with 95% confidence intervals are displayed on a logarithmic scale, with significant results (confidence intervals not crossing 1) highlighted by a star.
Clustering of patients based on BCA parameters identified groups with significantly different disease progression probabilities
K-means clustering was performed on normalized BCA parameters, with adjustments for sex and age, resulting in the identification of two distinct patient clusters. The cluster analysis revealed two groups of patients characterized by differences in body composition, as visualized in a cluster plot (n = 39 in light blue and n = 52 in dark blue) in Fig. 3A. A PCA biplot further illustrates the contribution of BCA parameters to the clustering outcome with bone density, muscle volume and SM determining one dimension and the adipose tissue compartments together with CM determining the other dimension. To explore the clinical relevance of these clusters, the distribution of patients with different R-ISS scores and high-risk cytogenetic profiles was analyzed (Fig. 3B). Fisher’s exact test indicated no significant differences between the two clusters regarding the frequencies of these clinical parameters (R-ISS: p = 0.315; high-risk cytogenetics: p = 0.761). Despite no significant differences in parameters used in clinical practice to predict disease progression and death, a respective comparison between the two clusters revealed a significant difference using log-rank test (X2 = 6.700, p = 0.010). The corresponding Kaplan–Meier survival analysis is depicted in Fig. 3C. These findings suggest that the identified clusters may reflect distinct disease trajectories.
K-means clustering based on normalized BCA parameters, with adjustments for sex and age, identified two distinct clusters. (A) The left panel presents a cluster plot dividing patients into two groups based on their body composition. These groups are depicted in light blue (n = 39) and dark blue (n = 52), with each data point representing an individual patient. The right panel features a variables-PCA biplot, illustrating the contributions of various body composition parameters. (B) Bar plots display the number of patients with R-ISS scores and high-risk cytogenetics, grouped according to the identified clusters. Fisher’s exact test was used to compare patient counts (R-ISS: p = 0.315; high-risk cytogenetics: p = 0.761). (C) A Kaplan–Meier plot compares disease progression between the two clusters. The log-rank test revealed a significant difference in disease progression probabilities (X2 = 6.700, p = 0.010).
Model based patients’ survival prediction
To evaluate the predictive capability for disease progression versus progression-free survival in MM patients, several parameters were analyzed. High-risk cytogenetics and R-ISS scores, both commonly utilized in clinical practice, were included. Additionally, a stepwise model selection process was employed to develop a model based on specific BCA parameters. The resulting BCA model incorporated bone density, muscle volume, and SM, with an AIC of 127.37. A similar approach was undertaken to construct a combined model, integrating clinical parameters such as sex, age, high-risk cytogenetics, and the R-ISS score, along with BCA-derived variables. The generated combined model included the parameters muscle volume, TAT, SAT, EAT, CM, sex and R-ISS. This combined model achieved a superior AIC of 119.36. The ROC curves for high-risk cytogenetics, the R-ISS score, the BCA model, and the combined model are presented in Fig. 4. The AUC values determined from these analyses were 0.57 for high-risk cytogenetics, 0.66 for the R-ISS score, and 0.60 for the BCA model. Notably, the combined model, integrating clinical and BCA parameters, demonstrated the highest predictive performance for disease progression vs. progression-free survival with an AUC of 0.80.
The ROC curves for high-risk cytogenetics, the R-ISS score, the BCA model, and the combined model (both developed using a stepwise model selection algorithm) are shown, comparing their ability to predict progression-free survival versus disease progression in MM patients. The BCA model, derived through the stepwise selection algorithm, included the variables bone density, muscle volume, and SM, with an AIC of 127.37. The combined model, also developed using the stepwise selection algorithm, incorporated muscle volume, TAT, SAT, EAT, CM, ISS, and sex, achieving an AIC of 119.36. The AUC values for all four models are displayed on the right side.