Planned analysis
What I would like to do is a (hierarchical) cluster analysis for the following metrics (obesity and frailty): - Age - BMI - Abdominal Circumference - Visceral adipose - Subcutaneous adipose - Total adipose - 1/(muscle mass/body weight) - 1/(bone density/body weight) - Diabetes
I have used the ratios of body weight/muscle mass and body weight /bone density as metrics of fitness because both bone density and muscle mass (putative measures of fitness) are inversely correlated with frailty but must be normalized to body weight because heavier individuals would be expected to have increased muscle mass and bone density independent of fitness.
My prediction is that this will generate significantly different clusters that are defined by parameters of obesity, fitness, age and diabetes. If this prediction holds, then we would examine the correlation of the following metrics of prostate phenotype and bladder contractility with each of those clusters.
Metrics of prostate phenotype and bladder contractility
- Prostate total Inflammation score
- Prostate collagen content.
- Qmax
- BCI
- Bladder power
- Detrusor overactivity
- BVE
How does change in urethral caliber and angle during voiding affect flow?
Hierarchical clustering
The following variables are used for clustering
- Age
- BMI
- Abdominal Circumference
- Visceral adipose
- Subcutaneous adipose
- Total adipose
- (Body weight)/(Muscle area)
- Diabetes
Fig. 1 Dendrogram of 53 patients.
We find two clusters with \(n_1=\) 24 and \(n_2=\) 29 subjects.
Comparison between the two clusters
| Cluster 1 (N=24) | Cluster 2 (N=29) | p-value | |
|---|---|---|---|
| - Obesity and frailty | |||
| Age (years) | 63.5 (52, 73) | 65 (62, 72) | 0.189 |
| BMI | 26.4 (24.4, 28) | 29.3 (26, 33.2) | 0.008 |
| Abdominal circumference | 1000 (917.8, 1059.4) | 1058.8 (989.3, 1093.1) | 0.119 |
| Visceral adipose | 105.8 (52.8, 134.4) | 221.6 (185.6, 264.3) | <0.001 |
| Subcutaneous adipose | 82.8 (61.8, 103.9) | 138.5 (103.5, 200.1) | <0.001 |
| Total adipose | 194.1 (125.1, 234.2) | 399.6 (329.7, 454.9) | <0.001 |
| Weight (kg) | 86.2 (70.6, 89.1) | 91.5 (85, 101) | 0.004 |
| Muscle area | 124.6 (107.6, 137) | 135.6 (124.7, 158.1) | 0.021 |
| Weight/Muscle area | 0.6 (0.6, 0.7) | 0.7 (0.6, 0.7) | 0.4 |
| Diabetes - No | 21 (87.5%) | 23 (79.3%) | 0.487 |
| Diabetes - Yes | 3 (12.5%) | 6 (20.7%) | |
| - Prostate phenotype and bladder contractility | |||
| Total Inflammation score | 6.5 (4, 8.2) | 9 (6, 13) | 0.021 |
| Collagen content | 60 (52.6, 67.4) | 58.1 (53.2, 63.5) | 0.825 |
| Qmax | 6 (3, 7) | 5 (4, 6) | 0.808 |
| BCI | 93 (82.8, 102.6) | 108.1 (88.6, 127.9) | 0.033 |
| Bladder power | 257 (166.5, 427.5) | 335 (210, 492) | 0.362 |
| BVE | 47 (25.2, 78.3) | 59.2 (31.2, 87.4) | 0.519 |
| Detrusor overactivity - No | 5 (20.8%) | 4 (13.8%) | 0.715 |
| Detrusor overactivity - Yes | 19 (79.2%) | 25 (86.2%) |
Categorical variables are summarized by N (%) and quantitative variables by median (IQR). p-value is based on Wilcoxon rank sum test for quantitative variables and on Chi-square test for categorical variables.
By Table 1, main drivers of the two clusters include BMI, visceral adipose, subcutaneous adipose, and total adipose, which separate the more obese patients (cluster 2) from the less obese ones (cluster 1). Although patients in cluster 2 also have significantly heavier weight and larger muscle area, the “weight/muscle area” ratio does not appear to be much different between the clusters. \(\square\)
Subgroup analysis of correlations
The following shows the correlations of visceral, subcutaneous and total adipose with prostate volume and total inflammation score within each cluster.
Cluster 1
Cluster 2
