Fig. 3
Survival tree for incidence of CKD events in men. Squares represent terminal nodes; numbers (n) in squares denote sample size (top line), and curves inside the squares show the Kaplan–Meier estimated survival of subpopulations. Circles represent the most significant variable based on log-rank (LR) and permutation P for splitting population to smaller groups. Six variables were entered as the most important predictors of the occurrence of CKD over a 12.4 year period. Node 1 at the top of figure shows that eGFR is the most important variable for first split (p < 0.001). The best cutoff value for eGFR was 69.85 ml/min/1.73 m2; accordingly, male populations were divided into two groups: left (≤ 69.85) and right (> 69.85) branches. This procedure was applied recursively until the tree grew to an optimal number of terminal nodes. Therefore, 12 groups were identified by the survival tree algorithm. Each path from the first node to a terminal node specifies a combination of predictors and their cutoff values leading to a terminal node forms an interaction pattern. Each interaction pattern specifies a subgroup of individuals with similar survival probability. For example, node 4 shows the survival probability for a group of men with eGFR between 63.39 and 69.85 and aged ≤ 50 years which is worse than nodes 15 and 16. SBP systolic blood pressure, FPG fasting plasma glucose, eGFR estimated glomerular filtration rate