Advances in positron emission tomography and radiomics (ICML 2023)
AuthorNumber and patient population/trialMedian FUDiseaseTreatmentModelMethod of measurement and cut-offs as applicableStatistical relationship with survivalPFS/OSComments Ceriani et al.1910362 monthsPMBCL first line I-IV R-CHOP R-CHOP-like R-VACOP-B R-MACOP-B MTV + metabolic Heterogeneity (MH)MTV of hottest lesion using SUV ≥2.5 and AUC-CSH for MH cut–off 0.45ROC-AUC analysis5 years PFS 94% versus 73% in low and high MH groups, respectively (p = 0.0001) IELSG 26 Ceriani et al.20141 test64 monthsDLBCL first line I-IVR-CHOPMTV + MH MTV of hottest lesion using SUV ≥2.5 and AUC-CSH for MH Cut-off 931 cm3 PFS, 1149 cm3 OS 0.43 MH ROC-AUC analysis5 years PFS 83% versus 61% (p = 0.0005), 5 years OS 91% versus 65% (p = 0.0001), in low and high MTV respectivelyIn high MTV group, pts with high MH had higher risk of progression (HR, 5.6; 95% CI, 1.8–17) and death (HR, 9.5; 95% CI, 1.7–52) 113 validation SAKK 38/07 study Cottereau et al.219544 months DLBCL first line I-IV 60–80 R-CHOP R-ACVBP tMTV + Dmax tMTV using 41% SUVmax Cut-off >394 cm3 SDmax >58 cm ROC-AUC analysis4 years PFS 94% versus 73% versus 53% and 4 years OS 97% versus 88% versus 50% for pts with 0, 1 or 2 risk factors (p = 0.0003 PFS, p = 0.0011 OS) respectively LNH073B Cottereau et al.222905 years DLBCL first line I-IV 60–80 R-CHOPtMTV + SDmax (Dmax standardized by body surface area) tMTV using 41% SUVmax Cut-off 220 cm3 SDmax >0.32 m−1 ROC-AUC analysis4 years PFS 90% versus 63% versus 41% and 4 years OS 95% versus 79% versus 66% for pts with 0, 1 or 2 risk factors (p = 0.0001) respectively REMARC Driessen et al.276540 months HL R/R BV-DHAPsTARC + TLR using SUV peak lesion + SUV mean liver and sTARC after one cycle 1 of BV-DHAP and SUVpeak prior to ASCTsTARC cut-off 500 pg/mL baseline TLR ≥3.0 pre-ASCT TLR ≥1.0 using liver SUVmean and lesion SUVpeakROC-AUC analysis 3 years FFP 35% versus 95% for pts with high baseline sTARC and high baseline TLR versus either high risk factor 3 years FFP 0% versus 95% for 4 pts with high sTARC post c1 and high TLR pre-ASCT versus all pts BRAVE Durmo et al.53155 retrospective RW63 monthsHL I-IVABVDDmax + MTVMedian Dmax 20 cmMedian used to subdivide pts4 years PFS 90% versus 72% for pts with low and high Dmax respectivelyGene expression and cell populations differentially expressed in high and low Dmax groups Eertink et al.831792 months (parent trial)DLBCL first line I-IVR-CHOPRadiomics (MTV4, SUVpeak Dmaxbulk) clinical (IPI components and bulk >10 cm) tMTV4 Continuous scale ROC-AUC analysis2 years TTP, 28% versus 44% for high risk pts for radiomics + clinical model and IPI respectively Patient level features performed better than individual lesion analysis Simple radiomics model performed better than complex model HOVON-84 Eertink et al.4323DLBCL first line I-IVRadiomics (MTV4, Dmaxbulk, DSUVpeak, spread) + MYC status tMTV4 Continuous scale ROC-AUC analysis 2 years TTP 50% versus 70% 2 years PFS 51% versus 65% 2 years OS 57% versus 69% for high risk pts for radiomics + MYC model and IPI respectively HOVON-84 PETAL Mikhaeel et al.35124155 monthsDLBCL first line I-IVR-CHOPMTV-age-stage tMTV4 using SUV ≥4.0 and 3mls min volume Continuous scale Linear spline with 2 coefficients above and below median 310 cm33 years PFS for high risk pts 46.3% versus 58.0% and 3 years OS 51.5% versus 66.4% for IMPI and IPI respectivelySplit by scale into low risk 60% intermediate risk 30% high risk 10% 5 clinical trials from PETRA consortium Thieblemont et al.541825 PETAL + GOYA47.1–76.5 mDLBCL first line I-IVR-CHOPMTV + PStMTV using 41% SUVmax (PETAL, RW) and liver threshold (GOYA) PS ≥2COMBAT method used to harmonize tMTV measurements 4 years PFS 54% versus 59% (PETAL) 49% versus 58% (GOYA) 36% versus 55% (RW) for high risk pts for MTV + PS and IPI respectively p < 0.001 4 years OS 61% versus 70% (PETAL) 61% versus 72% (GOYA) 41% versus 59% (RW) for high risk pts for MTV + PS and IPI respectively p < 0.001 349 RW Vercellino et al.143015 years DLBCL first line I-IV 60–80 R-CHOPMTV + PS tMTV using 41% SUVmax Cut-off 220 cm3 PS ≥2 ROC-AUC analysis X-tile analysis 4 years PFS 82% versus 63% versus 41% and 4 years OS 94% versus 79% versus 59% for pts with 0, 1 or 2 risk factors (p = 0.0001) REMARC Abbreviations: ASCT, autologous stem cell transplant; DLBCL, diffuse large B-cell lymphoma; MH, metabolic heterogeneity; OS, overall survival; PFS, progression-free-survival; PMBCL, primary mediastinal B-cell lymphoma; R-CHOP, rituximab, cyclophosphamide, doxorubicin, vincristine and prednisone; sTARC, serum thymus and activation regulated chemokine; SUV, standardized uptake value; tMTV, total metabolic tumor volume. A model including baseline tMTV ≥220 cm3 (using 41% of the maximum SUV to delineate tumors) and performance status (PS) ≥2 was also developed in a population of patients aged 60–80 with DLBCL in the REMARC study treated with R-CHOP.14 These two risk factors were the only independent variables associated with 4-year OS from tMTV, IPI, NCCN-IPI, B2 microglobulin, albumin and treatment arm (maintenance lenalidomide vs. placebo)...A new concept of “near CMR” with >90% reduction in MTV has been suggested as a more sensitive method of response to checkpoint inhibitors than CMR, given either sequentially or in combination with doxorubicin, vinblastine and darcarbazine (AVD) chemotherapy in first-line treatment of patients with HL.62, 63 The combination of emerging blood biomarkers as well as established risk factors with early PET scans, which may provide complementary information to monitor response is appealing...In another report exploring baseline and intra-treatment blood biomarkers and PET scans in HL,27 baseline serum thymus and activation regulated chemokine (sTARC), baseline SUV peak (the “hottest” 1 cm3 of tumor) and sTARC after one cycle 1 of brentuximab vedotin (BV)-dexamethasone, cytarabine, cisplatin (DHAP) and SUVpeak prior to autologous stem cell transplant were shown to predict response in 65 patients with relapsed/refractory (R/R) disease in the transplant BRAVE study, where tMTV had low prognostic value...The prognostic value of these models will need to be tested in patients treated with polatuzumab vedotin, rituximab, cyclophosphamide, doxorubicin, prednisolone (R-CHP),64 which is rapidly becoming an alternative standard approach.65 Newer algorithms incorporating artificial intelligence40 are likely to allow even faster and reproducible measurement of tMTV...It is clear that evaluation of radiomic features, molecular markers and ctDNA at baseline and during treatment should be factored into current clinical trial designs using appropriate statistical models to explore the relationship of continuous variables with patient outcomes.35 Promising radiomic and circulating blood biomarkers should be assessed with established clinical risk factors in clinical decision models. Prospective collection of data and retrospective analysis of curated imaging data from clinical trials will enable the development of baseline and dynamic risk scores that could further advance the field to facilitate testing of novel treatments and personalized therapy in aggressive lymphomas.