Sarah Verhoeff

47 [89Zr]Zr-DFO-girentuximab and [18F]FDG PET/CT to predict WW in mccRCC patients from baseline CT to disease progression warranting systemic treatment, based on radiological and clinical disease progression. Local therapy was permitted while on protocol. Prespecified secondary endpoints reported in this paper are progression-free survival (PFS) and overall survival (OS). Meanwhile, Rini et al. predicted indolent disease in RCC patients with <2 IMDC risk-factors and ≤2 involved organs9. Therefore, the validation of these “W&W criteria” was added as an extra secondary aim. Finally, we report the time on 1st line treatment after WW. Statistical analyses Based on the incidence of 450 patients with mccRCC yearly in the Netherlands, it was estimated that we could enroll 80 patients in 4 years’ time, including approximately 25 (30%) patients with rapid disease progression and another 25 (30%) with indolent disease. Based on a lower accrual than anticipated, the number of patients to be included was changed into 40 patients. To accommodate this lower number of patients and still provide meaningful results, we adapted the way to evaluate the primary outcome to a continuous instead of categorical assessment, thereby improving the statistical efficiency. Of note, no analyses related to the primary outcome were performed before making this decision. Standard descriptive statistics were used to describe baseline patient characteristics, overall and according to [18F]FDG and [89Zr]Zr-DFO-girentuximab uptake. For the latter, we grouped patients into high and low uptake groups based on their gm SUVmax of RECIST-measurable PET-positive lesions, binning patients into equal groups (using the median across patients). We used gm values to account for the right-skewed SUVmax data and assigned patients without visual PET-positive lesions a gm SUVmax of zero. Standard Kaplan-Meier analyses served to describe WW-time, time to RECIST PD and OS, and univariably evaluate IMDC, “W&W criteria”, and [18F]FDG and [89Zr]Zr-DFO-girentuximab high/low uptake groups, using log-rank tests for statistical inference. We used pointwise 95% confidence intervals for Kaplan-Meier curves and for median survival times. Furthermore, Firth’s penalized Cox regression was used to obtain small-sample bias-corrected hazard ratios (HRs)28. Here, PET data (SUVmax, TLG, MTV) were analyzed dichotomously (using a median split) as well as continuously (after truncation at the 90th percentile to decrease the influence of extreme values), assuming linearity. Non-linear relations or optimal grouping thresholds beyond a median split were not assessed due to the relatively small dataset and associated risk of overfitting. The added value of predictors in multivariable analyses was assessed by likelihood-ratio tests, specifically focusing on the added value of gm SUVmax of the two PET-scans beyond the (underlying variables of) “W&W criteria” as the current clinical standard (primary analyses in view of events-topredictor ratio). In an exploratory analysis to evaluate more predictors (the two PET-scans, IMDC, number of organ sites, lung only disease, and sum of RECIST measurable lesion diameters) LASSO penalized Cox regression was used (with 5-fold cross-validation optimizing the partial likelihood)29, combined with 2500-fold bootstrap resampling to assess variable-selection robustness. 3