Fig. 3

Identification of the optimal variable combination for peripheral blood lymphocyte prognostic models. a represents the result plot of the bidirectional stepwise regression bootstrap resampling. The vertical axis represents the frequency of each independent variable being drawn in the 1000 resampling processes, while the horizontal axis represents three different lymphocyte prediction models. Each subgroup in the models represents the total independent variables initially included in the analysis. The red dashed line at X = 500 indicates the threshold, and independent variables with a frequency greater than 500 will be included in their respective models. From a, it can be inferred that T cells, NK cells, and the combination of CD8 + T cells, gravidity and parity are the best independent variable combinations for the T-cell model, NK cell model, and CD8 + T-cell model, respectively. The distribution plot of the AIC values after bootstrap resampling for the best variable combination represented by b with the lowest/second lowest AIC. The sampling distribution of AIC values for the four independent variable combinations in the plot is relatively tight and concentrated, indicating that the represented lymphocyte prognosis model is stable and reliable