Clinical Features
A total of 37,474 MG patients diagnosed between 2004 and 2019 were included in this study. All patients were randomly assigned to the training set (N = 26232) or the validation set (N = 11242), with 58.37% of patients being male and 41.63% being female. Overall, 51.71% of patients were between the ages of 18–64, while 48.29% were over 65 years of age. In addition, patients were married (66.12%) and white (89.25%). The site of tumors included stGM (74.9%), cGM (0.72%), bsGM (0.39%) and others (23.99%). Most cases were unilateral (97.97%), with a tumor size ≥ 3cm (70.11%). Most of the MG cases were glioblastoma (96.74%). The staging distribution was Localized (81.36%), Regional (17.23%) and Distant (13.4%). Patients who underwent no surgery accounted for 20.56% of the study population, with partial resection comprising 34.92%, gross total resection 44%, and unspecified 0.52%. Finally, 70.1% of patients received chemotherapy, and 77.22% received radiotherapy. Most patients were OP (84.3%). No statistically significant difference was found in both groups, and the results are shown in Table 1.
Table 1
Patients characteristics from the SEER database.
Characteristics | All N = 37474 | Training cohort N = 26232 | Validation cohort N = 11242 | p |
Sex | | | | 0.590 |
Male | 21874 (58.37) | 15336 (58.46) | 6538 (58.16) | |
Female | 15600 (41.63) | 10896 (41.54) | 4704 (41.84) | |
Age | | | | 0.200 |
18–64 | 19378 (51.71) | 13622 (51.93) | 5756 (51.20) | |
≥ 65 | 18096 (48.29) | 12610 (48.07) | 5486 (48.80) | |
Marital | | | | 0.853 |
Unmarried | 12696 (33.88) | 8879 (33.85) | 3817 (33.95) | |
Married | 24778 (66.12) | 17353 (66.15) | 7425 (66.05) | |
Race | | | | 0.778 |
White | 33446 (89.25) | 23418 (89.27) | 10028 (89.20) | |
Black | 2031 ( 5.42) | 1409 ( 5.37) | 622 ( 5.53) | |
Other | 1997 ( 5.33) | 1405 ( 5.36) | 592 ( 5.27) | |
Site | | | | 0.673 |
stGM | 28069 (74.90) | 19645 (74.89) | 8424 (74.93) | |
cGM | 269 ( 0.72) | 187 ( 0.71) | 82 ( 0.73) | |
bsGM | 146 ( 0.39) | 109 ( 0.42) | 37 ( 0.33) | |
Other | 8990 (23.99) | 6291 (23.98) | 2699 (24.01) | |
Laterality | | | | 0.114 |
Unilateral | 36713 (97.97) | 25679 (97.89) | 11034 (98.15) | |
Bilateral | 761 ( 2.03) | 553 ( 2.11) | 208 ( 1.85) | |
Size | | | | 0.044 |
< 3cm | 11201 (29.89) | 7923 (30.20) | 3278 (29.16) | |
≥ 3cm | 26273 (70.11) | 18309 (69.80) | 7964 (70.84) | |
Histology | | | | 0.367 |
Glioblastoma | 36252 (96.74) | 25357 (96.66) | 10895 (96.91) | |
Giant cell glioblastoma | 334 ( 0.89) | 244 ( 0.93) | 90 ( 0.80) | |
Gliosarcoma | 888 ( 2.37) | 631 ( 2.41) | 257 ( 2.29) | |
Stage | | | | 0.538 |
Localized | 30490 (81.36) | 21350 (81.39) | 9140 (81.30) | |
Regional | 6456 (17.23) | 4524 (17.25) | 1932 (17.19) | |
Distant | 528 ( 1.41) | 358 ( 1.36) | 170 ( 1.51) | |
Surgery | | | | 0.789 |
No surgery | 7706 (20.56) | 5388 (20.54) | 2318 (20.62) | |
Partial resection | 13085 (34.92) | 9194 (35.05) | 3891 (34.61) | |
Gross total resection | 16487 (44.00) | 11517 (43.90) | 4970 (44.21) | |
Unspecified | 196 ( 0.52) | 133 ( 0.51) | 63 ( 0.56) | |
Chemotherapy | | | | 0.926 |
No/Unknown | 11206 (29.90) | 7840 (29.89) | 3366 (29.94) | |
Yes | 26268 (70.10) | 18392 (70.11) | 7876 (70.06) | |
Radiation | | | | 0.926 |
No/Unknown | 8537 (22.78) | 5972 (22.77) | 2565 (22.82) | |
Yes | 28937 (77.22) | 20260 (77.23) | 8677 (77.18) | |
SN | | | | 0.549 |
OP | 31590 (84.30) | 22133 (84.37) | 9457 (84.12) | |
MP | 5884 (15.70) | 4099 (15.63) | 1785 (15.88) | |
Univariate And Multivariate Cox Regression Analysis
Univariate Cox regression models were first used in the training set to analyze and screen for factors associated with patient survival. The results showed that these factors, including age, marital status, race, tumor site, laterality, histology, stage, surgery, chemotherapy, radiotherapy and SN, could affect patient survival. Then, multivariate Cox regression models were used to screen for independent risk factors associated with CSS and OS of adult MG patients. The results revealed that age, marital status, race, site, laterality, histology, stage, surgery, chemotherapy, radiotherapy and SN were prognostic factors affecting patient CSS and OS. The results are shown in Tables 2 and 3.
Table 2
Univariate and multivariate analyses of CSS in training cohort.
| Univariate | Multivariate |
| HR | 95%CI | p | HR | 95%CI | p |
Sex | | | | | | |
Male | ref | | | | | |
Female | 0.996 | 0.969–1.023 | 0.757 | | | |
Age | | | | | | |
18–64 | ref | | | ref | | |
≥ 65 | 2.000 | 1.947–2.055 | < 0.001 | 1.800 | 1.750–1.851 | < 0.001 |
Marital | | | | | | |
Unmarried | ref | | | ref | | |
Married | 0.861 | 0.837–0.885 | < 0.001 | 0.934 | 0.908–0.961 | < 0.001 |
Race | | | | | | |
White | ref | | | ref | | |
Black | 0.894 | 0.842–0.949 | < 0.001 | 0.792 | 0.745–0.841 | < 0.001 |
Other | 0.782 | 0.735–0.833 | < 0.001 | 0.762 | 0.716–0.811 | < 0.001 |
Site | | | | | | |
stGM | ref | | | ref | | |
cGM | 0.989 | 0.839–1.165 | 0.895 | 0.916 | 0.777–1.080 | 0.297 |
bsGM | 1.054 | 0.858–1.295 | 0.616 | 0.998 | 0.812–1.228 | 0.988 |
Other | 1.325 | 1.284–1.366 | < 0.001 | 1.131 | 1.096–1.168 | < 0.001 |
Laterality | | | | | | |
Unilateral | ref | | | ref | | |
Bilateral | 1.571 | 1.437–1.717 | < 0.001 | 1.243 | 1.135–1.360 | < 0.001 |
Size | | | | | | |
< 3cm | ref | | | | | |
≥ 3cm | 1.020 | 0.991–1.050 | 0.182 | | | |
Histology | | | | | | |
Glioblastoma | ref | | | ref | | |
Giant cell glioblastoma | 0.657 | 0.569–0.760 | < 0.001 | 0.772 | 0.667–0.892 | < 0.001 |
Gliosarcoma | 0.983 | 0.902–1.070 | 0.690 | 1.065 | 0.977–1.160 | 0.151 |
Stage | | | | | | |
Localized | ref | | | ref | | |
Regional | 1.488 | 1.437–1.540 | < 0.001 | 1.390 | 1.341–1.440 | < 0.001 |
Distant | 1.632 | 1.461–1.822 | < 0.001 | 1.345 | 1.204–1.503 | < 0.001 |
Surgery | | | | | | |
No surgery | ref | | | ref | | |
Partial resection | 0.490 | 0.473–0.508 | < 0.001 | 0.658 | 0.633–0.683 | < 0.001 |
Gross total resection | 0.397 | 0.384–0.412 | < 0.001 | 0.549 | 0.528–0.570 | < 0.001 |
Unspecified | 0.401 | 0.332–0.484 | < 0.001 | 0.486 | 0.403–0.587 | < 0.001 |
Chemotherapy | | | | | | |
No/Unknown | ref | | | ref | | |
Yes | 0.371 | 0.360–0.382 | < 0.001 | 0.554 | 0.533–0.576 | < 0.001 |
Radiation | | | | | | |
No/Unknown | ref | | | ref | | |
Yes | 0.358 | 0.347–0.369 | < 0.001 | 0.635 | 0.608–0.662 | < 0.001 |
SN | | | | | | |
OP | ref | | | ref | | |
MP | 1.249 | 1.203–1.296 | < 0.001 | 1.042 | 1.004–1.082 | 0.032 |
Table 3
Univariate and multivariate analyses of OS in training cohort.
| Univariate | Multivariate |
| HR | 95%CI | p | HR | 95%CI | p |
Sex | | | | | | |
Male | ref | | | | | |
Female | 0.992 | 0.966–1.018 | 0.541 | | | |
Age | | | | | | |
18–64 | ref | | | ref | | |
≥ 65 | 2.057 | 2.004–2.112 | < 0.001 | 1.841 | 1.791–1.892 | < 0.001 |
Marital | | | | | | |
Unmarried | ref | | | ref | | |
Married | 0.843 | 0.82–0.866 | < 0.001 | 0.916 | 0.891–0.942 | < 0.001 |
Race | | | | | | |
White | ref | | | ref | | |
Black | 0.937 | 0.885–0.992 | 0.024 | 0.826 | 0.78–0.875 | < 0.001 |
Other | 0.815 | 0.768–0.865 | < 0.001 | 0.797 | 0.751–0.846 | < 0.001 |
Site | | | | | | |
stGM | ref | | | ref | | |
cGM | 1.028 | 0.88–1.202 | 0.724 | 0.943 | 0.807–1.102 | 0.462 |
bsGM | 1.074 | 0.882–1.309 | 0.478 | 1.028 | 0.843–1.254 | 0.783 |
Other | 1.331 | 1.292–1.372 | < 0.001 | 1.14 | 1.105–1.175 | < 0.001 |
Laterality | | | | | | |
Unilateral | ref | | | ref | | |
Bilateral | 1.518 | 1.391–1.656 | < 0.001 | 1.208 | 1.106–1.32 | < 0.001 |
Size | | | | | | |
< 3cm | ref | | | | | |
≥ 3cm | 1.015 | 0.987–1.044 | 0.304 | | | |
Histology | | | | | | |
Glioblastoma | ref | | | ref | | |
Giant cell glioblastoma | 0.672 | 0.585–0.771 | < 0.001 | 0.79 | 0.688–0.908 | 0.001 |
Gliosarcoma | 0.986 | 0.908–1.071 | 0.739 | 1.066 | 0.982–1.158 | 0.129 |
Stage | | | | | | |
Localized | ref | | | ref | | |
Regional | 1.469 | 1.42–1.519 | < 0.001 | 1.377 | 1.33–1.426 | < 0.001 |
Distant | 1.586 | 1.424–1.766 | < 0.001 | 1.296 | 1.163–1.444 | < 0.001 |
Surgery | | | | | | |
No surgery | ref | | | ref | | |
Partial resection | 0.49 | 0.473–0.508 | < 0.001 | 0.663 | 0.639–0.688 | < 0.001 |
Gross total resection | 0.397 | 0.383–0.41 | < 0.001 | 0.553 | 0.533–0.573 | < 0.001 |
Unspecified | 0.398 | 0.331–0.477 | < 0.001 | 0.487 | 0.406–0.585 | < 0.001 |
Chemotherapy | | | | | | |
No/Unknown | ref | | | ref | | |
Yes | 0.363 | 0.353–0.373 | < 0.001 | 0.548 | 0.528–0.569 | < 0.001 |
Radiation | | | | | | |
No/Unknown | ref | | | ref | | |
Yes | 0.35 | 0.339–0.36 | < 0.001 | 0.629 | 0.604–0.655 | < 0.001 |
SN | | | | | | |
OP | ref | | | ref | | |
MP | 1.309 | 1.263–1.355 | < 0.001 | 1.088 | 1.05–1.128 | < 0.001 |
Nomogram Development For The 3-month, 6-month, And 12-month Css And Os
Based on the multivariate Cox regression models, nomograms were constructed to predict CSS and OS in adult MG patients at 3-month, 6-month, and 12-month (Fig. 2). As illustrated in the figure, age, race, tumor site, laterality, histology, stage, surgery, chemotherapy, and radiotherapy were the most influential factors for predicting CSS and OS in adult MG patients. However, marital status and SN had little effect on patient survival. To facilitate the use of the nomogram, calculators were designed in the form of a computer or mobile phone application. The websites are as follows: https://luoshu.shinyapps.io/DynNomapp_CSS/
https://luoshu.shinyapps.io/DynNomapp_OS/
Validation Of The Nomograms
Internal cross-validation was performed to test the accuracy and discriminability of models. The C-index of the training set and the validation set for predicting OS were 0.728 (95%CI:0.725–0.732) and 0.725 (95%CI:0.719–0.730), respectively. The C-index of the training set and the validation set for predicting CSS were 0.725 (95%CI:0.722–0.729) and 0.721 (95%CI:0.715–0.726), respectively. The results indicated that nomograms for CSS and OS have good recognition ability. In the training and validation set, the calibration curve showed that the predicted value of the nomograms for CSS and OS are highly consistent with the actual observed value (Fig. 3).
The nomograms were shown to have good accuracy. The AUC in the training set for CSS at 3, 6, and 12 months was 0.841, 0.822, and 0.781, respectively, and 0.849, 0.821, and 0.775 in the validation set. In the training set for OS, the AUC at 3, 6, and 12 months was 0.841, 0.824, and 0.785, respectively, and 0.850, 0.824 and 0.781 in the validation set. The results revealed that the nomograms were highly discriminative (Fig. 4).
Clinical Application Of The Nomograms
In both the training set and the validation set for CSS and OS, DCA suggested that the nomograms had good clinical potential value (Fig. 5). The nomograms for CSS and OS were superior to the stage model of the SEER database. Based on the nomogram, each patient’s risk value was calculated, and the optimal cut-off value was determined using the ROC curve. For predicting CSS, patients were classified into the high-risk group (total score ≥ 227.778) and the low-risk group (total score < 227.778). For the OS group, patients were divided into the high-risk group (total score ≥ 190.0104) and the low-risk group (total score < 190.0104). The K-M curve showed that the CSS and OS rates of the patients in the high-risk group were significantly lower than that in the low-risk group (Fig. 6). In the training set, the 3-, 6-, and 12-month CSS rates of the patients in the high-risk group were 54.2%, 33.0%, and 16.5%, respectively. In contrast, the 3-, 6-, and 12-month CSS rates in the low-risk group were 92.7%, 81.7%, and 59.5%, respectively. The 3-, 6-, and 12-month OS rates of the patients in the high-risk group were 59.4%, 38.7%, and 20.6%, respectively. However, the 3-, 6-, and 12-month OS rates in the low-risk group were 93.7%, 84.0%, and 61.7%, respectively. Subsequently, the training and validation sets were merged.
The K-M curve showed that the CSS and OS rates of patients with no surgery were the lowest in both the high-risk group and the low-risk group, with statistical significance. The K-M curve combined with Bonferroni's test was used to compare the CSS and OS rates between different surgical methods. We found that in the high-risk group, the CSS and OS of patients with gross total resection were higher than those with partial resection (Fig. 7). The 3-, 6-, and 12-month CSS rates for patients with gross total resection were 62.1%, 39.7%, and 20.6%, respectively. The 3-, 6-, and 12-month CSS rates for patients with partial resection were 56.2%, 35.3%, and 18.3%, respectively, and the 3-, 6-, and 12-month OS rates for patients with gross total resection were 64.6%, 43.5%, and 22.9%, respectively. The 3-, 6-, and 12-month OS rates for patients with partial resection were 67.0%, 46.6%, and 26.1%, respectively.
Interestingly, no significant difference in CSS was found in the high-risk group among patients with different SN. However, the CSS in the low-risk group and the OS of both the high-risk group and the low-risk group were significantly different (Fig. 8).