57 [89Zr]Zr-DFO-girentuximab and [18F]FDG PET/CT to predict WW in mccRCC patients (95%CI: 0.503-0.728; p=1.0 for improvement in C-index) for adding [89Zr]Zr-DFO-girentuximab gm SUVmax. An exploratory LASSO model considering the two PET-scans, IMDC, number of organ sites, lung only disease, and sum of RECIST measurable lesion diameters resulted in a model that performed less than the above best model with an overall C-index of 0.711 (95%CI: 0.599-0.824; see Supplementary data for more details about the LASSO model). We also evaluated the discriminative ability to predict prolonged indolent disease (>12 months) using time-dependent ROC curves (Figure 6). In our data, a subgroup of 12 patients (30%) could be identified based on a [18F]FDG gm SUV max <3 or a negative [ 18F]FDG PET/CT, who all had a WW-period of >12 months (100% (95%CI 76-100%) positive predictive value and 64% (95%CI 46-79%) negative predictive value; data-driven threshold maximizing sensitivity at 100% specificity). The prediction of rapid progression (<2 months after study enrolment) was not analyzed due to too few events. Figure 6. Time-dependent ROC curves at 12 months WW time. These figures show the discriminating value of the two variables underlying the “W&W criteria” together and geometric mean [18F]FDG SUVmax (A) and [89Zr]Zr-DFO-girentuximab (B), respectively, individually and when combined. To allow further external validation and translation of our results towards clinical practice, we internally validated the predictive model with IMDC scores, the number of organ sites and the [18F]FDG gm SUV max. A formula based on this over-optimism-corrected model yields predicted probabilities of remaining on WW at 12 months between 9 and 83%, depending on a particular patient’s IMDC score, the number of organ sites and [18F]FDG-PET/CT results. An easy-to use score chart showing these predicted probabilities and the underlying formula is provided in Supplementary Figure S3 3
RkJQdWJsaXNoZXIy MTk4NDMw