Skip to main content

Table 4 Multivariate analysis of how well the parameters allow distinguishing between patients with BC and controlsa

From: Using Resistin, glucose, age and BMI to predict the presence of breast cancer

Variables

Figures of interest

Classifier

LR

RF

SVM

V1-V2

AUC

[0.76, 0.81]

[0.70, 0.75]

[0.76, 0.81]

Sensitivity

[0.75, 0.81]

[0.75, 0.82]

[0.81, 0.86]

Specificity

[0.73, 0.80]

[0.63, 0.70]

[0.70, 0.76]

V1-V3

AUC

[0.76, 0.80]

[0.81, 0.85]

[0.82, 0.86]

Sensitivity

[0.74, 0.81]

[0.85, 0.90]

[0.87, 0.92]

Specificity

[0.74, 0.80]

[0.72, 0.78]

[0.78, 0.83]

V1-V4

AUC

[0.79, 0.83]

[0.84, 0.88]

[0.87, 0.91]

Sensitivity

[0.72, 0.78]

[0.80, 0.86]

[0.82, 0.88]

Specificity

[0.80, 0.87]

[0.81, 0.87]

[0.84, 0.90]

V1-V5

AUC

[0.79, 0.83]

[0.82, 0.87]

[0.86, 0.90]

Sensitivity

[0.73, 0.79]

[0.79, 0.85]

[0.84, 0.90]

Specificity

[0.81, 0.87]

[0.77, 0.83]

[0.81, 0.87]

V1-V6

AUC

[0.78, 0.83]

[0.82, 0.86]

[0.83, 0.88]

Sensitivity

[0.74, 0.80]

[0.79, 0.85]

[0.81, 0.86]

Specificity

[0.79, 0.85]

[0.76, 0.82]

[0.80, 0.86]

V1-V9

AUC

[0.76, 0.81]

[0.78, 0.83]

[0.81, 0.85]

Sensitivity

[0.70, 0.76]

[0.78, 0.85]

[0.75, 0.81]

Specificity

[0.80, 0.86]

[0.70, 0.77]

[0.78, 0.84]

  1. aFor each classifier (LR logistic regression, RF random forest, SVM support vector machine), predictive models were created taking in as predictors the variables deemed more significant. The predictive capacity of each model was computed resorting to a ROC analysis and determining the pair of values of specificity and sensitivity that maximise the Youden index. Again for each model, the resulting AUC value depends on the number of variables included, as can be seen on the table below, where V1 = Glucose, V2 = Resistin, V3 = Age, V4 = BMI - body mass index, V5 = HOMA - homeostasis model assessment for insulin resistance, V6 = Leptin, V7 = Insulin, V8 = Adiponectin, V9 = MCP-1 - monocyte chemoattractant protein-1