Multi-Threshold Volumetric 18F-FET PET for Tumor Detection and Prognosis in Primary Glioma - A multivariate analysis

Author

Lu Mao

Published

April 13, 2026

Statistical analysis

Unadjusted analysis

Using cutoffs 11.7mL, 2.1mL, 1.4mL, 0.16mL for TBR1.6, TBR2.0, TBR2.5, TBR3.0 respectively, and 2.8 for TBRmax, binary variables are indicating high vs low for each metric. Table 1 shows the results of univariate Cox proportional hazards models for overall survival with each of these binary predictors.

Table 1: Univariate Cox proportional hazards models for overall survival.
Term HR CI Lower CI Upper p-value Predictor
tbr_16 3.59 1.95 6.63 0.0000 tbr_16
tbr_20 4.45 2.16 9.17 0.0001 tbr_20
tbr_25 3.10 1.73 5.54 0.0001 tbr_25
tbr_30 3.07 1.72 5.47 0.0001 tbr_30
tbrmax 2.30 1.18 4.47 0.0143 tbrmax

Adjusted analyses

Adjusting for age, sex, tumor_type, days_surgery_to_fet, dayssincescan1, and cycles, Table 2 shows the results of multivariate Cox proportional hazards models for overall survival with each of these binary predictors.

  • the HRs are reduced after adjustment.
Table 2: Multivariate Cox proportional hazards models for overall survival adjusting for clinical covariates.
term estimate conf.low conf.high p.value predictor
tbr_16 1.354278 0.6343909 2.891070 0.4331622 tbr_16
tbr_20 1.887654 0.7643492 4.661791 0.1683976 tbr_20
tbr_25 1.507983 0.7397351 3.074091 0.2583145 tbr_25
tbr_30 1.566779 0.7707320 3.185020 0.2147848 tbr_30
tbrmax 1.296852 0.5576257 3.016047 0.5460859 tbrmax

Table 3 shows the results of multivariate Cox proportional hazards models for overall survival adjusting for TBRmax. Each of TBR1.6, TBR2.0, TBR2.5, and TBR3.0 is adjusted for TBRmax in separate models.

  • The HRs remain strong and significant after adjusting for TBRmax, indicating that these volumetric metrics provide prognostic information beyond TBRmax alone.
Table 3: Multivariate Cox proportional hazards models for overall survival adjusting for TBRmax.
term estimate conf.low conf.high p.value predictor
tbr_16 3.690522 1.667351 8.168618 0.0012770 tbr_16
tbr_20 5.560238 2.175467 14.211317 0.0003392 tbr_20
tbr_25 3.207281 1.345610 7.644601 0.0085433 tbr_25
tbr_30 3.186174 1.337498 7.590071 0.0088815 tbr_30

Incremental value of TBR

Figure 1 shows the ROC curves comparing TBRmax alone vs TBRmax + volumetric metrics for predicting mortality. Table 4 compares the AUCs for these models, showing whether adding volumetric metrics significantly improves predictive performance over TBRmax alone.

Figure 1: ROC curves comparing TBRmax alone vs TBRmax + volumetric metrics for predicting mortality.
Table 4: Comparison of AUCs for TBRmax alone vs TBRmax + volumetric metrics.
predictor auc_tbrmax auc_combined p_value
tbr_16 0.6155395 0.7000805 0.0094013
tbr_20 0.6155395 0.6906200 0.1348971
tbr_25 0.6155395 0.6706924 0.0346961
tbr_30 0.6155395 0.6797504 0.2175744