Fig 1.
Microbial diversity of normal and lymphedema limbs in 30 patients with upper extremity secondary lymphedema.
(a) Boxplot of Shannon index for lymphedema and normal limbs; paired samples are connected by grey lines. P-values obtained by Wilcoxon signed-rank test for paired data. (b) Boxplot of inverse Simpson’s index for lymphedema and normal limbs; paired samples connected by grey lines. P-values obtained by Wilcoxon signed-rank test for paired data. (c) Principal component analysis (PCoA) plot of the first two principal coordinates (PC1, PC2) based on Bray-Curtis distance matrix.
Table 1.
Patient demographics and clinical characteristics.
Fig 2.
Association of microbial distance between paired limbs and clinical covariates.
(a) Swimmer plot of Bray-Curtis distance between lymphedema and normal limbs, with history of infection indicated. (b) Scatter plot of paired Bray-Curtis distance versus relative volume differential, with fitted line and confidence band from marginal linear regression, stratified by history of infection. (c) Scatter plot of paired Bray-Curtis distance versus duration of lymphedema in months, with fitted line and confidence band from marginal linear regression.
Table 2.
Point estimates, 95% confidence intervals, and corresponding p-values of patient clinical characteristics in the multivariable linear regression model of (a) paired Bray-Curtis distance and (b) paired Aitchison’s distance.
Fig 3.
Patient-specific taxa variability between normal and lymphedema limbs for the ten most variable taxa.
(a) Stacked bar chart of patient-specific taxa variability, defined as the absolute difference of taxa relative abundance. (b) Directional bar chart of patient-specific difference of taxa relative abundance. (c) Directional bar chart of patient-specific difference of taxa relative abundance, stratified by history of infection. p: phylum; g: genus; c:class.
Table 3.
Top 10 variable taxa found in microbial variability analysis.
Table 4.
Top 10 genera with smallest unadjusted p-values obtained by the linear decomposition model (LDM) test of differential abundance for paired data.