Patient characteristics and survival

The characteristics of 101 patients with gliomas are summarized in Table 1. The results of mean ADC measurement were 1.186 (ranging from 0.679 to 1.770) (× 10− 3 mm2/s) in parenchymal area with 101 patients; 1.306 (ranging from 0.816 to 1.552) (× 10− 3 mm2/s) in non-enhancing peritumoral area with 70 patients; 1.793 (ranging from 1.065 to 2.483) (× 10− 3 mm2/s) in N/C area with 75 patients. The median follow-up time for the patients was 51.8 months, ranging from 30.23 to 87.03 months. Out of the 101 patients, 62 experienced a relapse, with a median disease-free survival time of 18.8 months (ranging from 0.63 to 57.50 months). Among them, 50 patients with glioblastoma relapsed and 12 patients with non-glioblastoma relapsed. Additionally, 70 patients died, with a median survival time of 24.7 months (ranging from 1.73 to 68.57 months). Among them, 58 patients with glioblastoma died and 12 patients with non-glioblastoma died.

Table 1 Patient characteristicsThe subgroup of the patients who had parenchymal area

All the patients included in the study exhibited parenchymal area. The cut-off value for ADC in the parenchymal area was determined using X-tile software (V.3.6.1). An ADC value more than 1.123 × 10–3 mm2/s was classified as high, while the remaining values were classified as low.

A wide range of factors were evaluated using Kaplan-Meier curves and cox regression analysis to determine prognostic factors for postoperative survival. Univariate and multivariate cox regression analysis indicated that Ki67, P53, IDH, and high or low ADC value were independent prognostic factors for DFS (Fig. 2A-D; Table 2).

Table 2 Univariable and multivariable cox regression analysis for DFS

The same approach was used to identify OS-associated features, such as age, presence of glioblastoma, grade, Ki67, P53, IDH, NEPSA, status of N/C, and high or low ADC value per univariate cox regression analysis. Multivariate Cox regression analysis revealed that Ki67, IDH, and high or low ADC value were independent prognostic factors for postoperative OS (Fig. 2E-G; Table 3).

Fig. 2

Disease-free survival (DFS) and Overall survival (OS) analysis. (A) Kaplan–Meier DFS curves of IDH. (B) Kaplan–Meier DFS curves of Ki67. (C) Kaplan–Meier DFS curves of P53. (D) Kaplan–Meier DFS curves of high or low ADC value. (E) Kaplan–Meier OS curves of IDH. (F) Kaplan–Meier OS curves of Ki67. (G) Kaplan–Meier OS curves of high or low ADC value

Table 3 Univariable and multivariable cox regression analysis for OSThe subgroup of the patients who had non-enhancing peritumoral area

A total of 70 patients with non-enhancing peritumoral area were included in the study. Univariate COX regression analysis revealed that the ADC value of the non-enhancing peritumoral area did not show any statistically significant association. However, the ADC value of the parenchymal area showed a statistically significant association with DFS and OS when analyzed. Therefore, the P/N ratio was incorporated in the subsequent analysis.

The study analyzed various factors such as age, gender, presence of glioblastoma, location, grade, Ki67, P53, MGMT, IDH, tumor maximum diameter, status of N/C, and the P/N ratio to determine their association with DFS. To assess these characteristics, univariate and multivariate cox regression analyses were performed. The results showed that Ki67 (P = 0.013), P53 (P = 0.043), IDH (P = 0.042), and the P/N ratio (P = 0.030) were independent prognostic factors for DFS. Similarly, the study identified that Ki67 (P = 0.015), IDH (P = 0.009), and the P/N ratio (P = 0.047) were independent prognostic factors for OS using the same method.

Ki67, P53, IDH, and the P/N ratio were utilized to develop a nomogram model for predicting DFS and the DFS rates at 1, 2, and 3 years (Fig. 3A). Similarly, Ki67, IDH, and the P/N ratio were employed to construct a nomogram for predicting postoperative OS and the OS rates at 1, 2, and 3 years (Fig. 3D).

The study conducted a comprehensive evaluation of nomograms for discrimination, calibration, and clinical utility. This evaluation involved the use of various measurements such as C-index, receiver operating characteristic (ROC), calibration plot, and decision curve analysis (DCA). Specifically, the nomogram for DFS demonstrated a C-index of 0.734 (95% confidence interval: 0.664–0.817). Additionally, the area under the curve (AUC) for the predictions of DFS at 1-, 2-, and 3-year intervals were found to be 0.823, 0.839, and 0.875, respectively (Fig. 3B). Similarly, the nomogram for postoperative OS exhibited a C-index of 0.760 (95% confidence interval: 0.692–0.821). The AUCs for the predictions of OS at 1-, 2-, and 3-year intervals were determined to be 0.811, 0.864, and 0.827, respectively (Fig. 3E).

Calibration plots were created to compare nomogram-predicted outcomes with actual outcomes for 1-year, 2-year, and 3-year DFS/OS rates, demonstrating the high quality of the nomogram (Fig. 3C and F). The calibration curves graphically represented the predicted DFS or OS incidence on the x-axis and the observed actual DFS or OS incidence on the y-axis, ranging from 0 to 1, which represented the event incidence. The reference line, shown as a grey diagonal line, represented the predicted value equaling the actual value. The curve fitting line, which closely followed the grey diagonal line, indicated higher accuracy, and the colored area on both sides represented the 95% confidence interval.

The decision curve analysis demonstrated the value of the two models. The net benefit of 3-year DFS was consistently higher than that at other time points across a wide range of reasonable threshold probabilities, as illustrated in Fig. 3G-L.

Fig. 3

Construction of the model, time-dependent ROC curves, calibration plots of the nomogram, and decision curve analysis. (A) Nomogram for predicting the 1-, 2- and 3-year DFS. (B) Time-dependent ROC curves for predicting 1-, 2- and 3-year DFS. (C) Calibration curves of the model for predicting DFS at the 1-year, 2-year, and 3-year time points. (D) Nomogram for predicting the 1-, 2- and 3-year OS. (E) Time-dependent ROC curves for predicting 1-, 2- and 3-year OS. (F) Calibration curves of the model for OS at the 1-year, 2-year, and 3-year time points. (G) Decision curve analysis of 1-year DFS. (H) Decision curve analysis of 2-year DFS. (I) Decision curve analysis of 3-year DFS. (J) Decision curve analysis of 1-year OS. (K) Decision curve analysis of 2-year OS. (L) Decision curve analysis of 3-year OS

The subgroup of the patients who had N/C area

A total of 75 patients with N/C areas were included in the study. Various factors such as age, gender, presence of glioblastoma, location, grade, Ki67, P53, MGMT, IDH, tumor maximum diameter, NEPSA, and the ADC value of N/C area were analyzed using univariate cox regression analysis to determine their association with DFS. The results of cox regression analyses revealed that only Ki67 (P = 0.004) was an independent prognostic factor for DFS, while the ADC value of the N/C area was not.

The same method was applied to assess Ki67 (P = 0.007) and the ADC value of N/C area (P = 0.047) as independent prognostic factors of OS.