Supplementary materials: A novel real-world data methodology for lymphoma outcome classification: the real-world Lugano study

These are peer-reviewed supplementary materials for the article 'A novel real-world data methodology for lymphoma outcome classification: the real-world Lugano study' published in the Journal of Comparative Effectiveness ResearchAppendix Table 1: Lugano Criteria for Response Assessment (simplified)Appendix Table 2: Study Inclusion and Exclusion CriteriaAppendix Table 3: PRN Site CharacteristicsAppendix Table 4: Comparing Agreement Between Deauville Score and rwDeauville and Between rwLugano-Derived With Deauville Versus rwDeauville ScoresAppendix Table 5: Univariate and Multivariable GLMM Analysis to Identify Predictors of Treatment Response Within 6 months from 1L Therapy Initiation: Agreement Between Physician-charted and rwLugano-derived, Physician-charted and BICR-adjudicated, and rwLugano-derived and BICR-adjudicated CR AssessmentsAppendix Table 6: Univariate and Multivariable GLMM Analysis of Predictors of Concordance Between Physician-charted and rwLugano-derived Response With BICR-adjudicated Response Within 6 Months of 1L Therapy InitiationAppendix Table 7: Subgroup Analyses of Agreement Between Physician-charted and BICR-adjudicated Response Based on BICR ConcordanceAppendix Table 8: Subgroup Analyses of Agreement Between BICR-adjudicated and rwLugano-derived Response Based on BICR ConcordanceAppendix Table 9: Subgroup Analyses of Agreement Between Physician-charted and rwLugano-derived Response Based on BICR ConcordanceAppendix Figure 1: rwDeauville MethodologyAim: In oncology trials, blinded independent central review (BICR) is the standard for treatment response classification. Real-world data methodologies that align with BICR may reduce misclassification in real world evidence (RWE) studies and enhance reproducibility, increasing value of RWE. We aimed to develop and validate a novel real-world data-based methodology – real-world Lugano (rwLugano) – for assessing lymphoma response to align with clinical trials. Materials & methods: We conducted a retrospective, multisite chart abstraction study using Cardinal Health Practice Research Network (PRN) sites to identify adults with diffuse large B-cell lymphoma initiating first-line (1L) therapy from 1 January 2015, through 31 December 2022, in US community oncology. Sites collected patient characteristics and PET/CT scans at baseline and first response. Two radiologists independently classified responses; a medical oncologist adjudicated discordances. Results: We compared initial treatment responses using three methods: physician-charted from electronic health records, rwLugano-derived per Lugano 2014 and BICR-adjudicated per Lugano 2014. Agreement was assessed via percentage concordance, kappa (κ), and multivariable generalized linear mixed modeling for assigning complete response (CR). Among 178 patients, CR rates were 63.5% (physician-charted), 81.5% (rwLugano) and 83.1% (BICR). Compared with BICR, rwLugano showed higher agreement (87.9%, κ = 0.52) than physician-charted (77.0%, κ = 0.40). The generalized linear mixed modeling analyses identified clinical factors associated with concordance: for physician charted assessments, greater numbers of extranodal sites increased agreement with BICR (OR 1.92), while MYC mutation (OR 0.38) and anemia (OR 0.37) reduced agreement. For rwLugano, nonprivate insurance was associated with higher agreement (odds ratio [OR]: 4.40), whereas MYC mutation reduced agreement (OR: 0.26). Conclusion: rwLugano improves real-world lymphoma response classification, aligning with BICR and supporting more accurate, reproducible RWE for clinical and regulatory decision-making. Using methods BICR and rwLugano may provide opportunities to minimize outcome misclassification and improve comparability of clinical trial and clinical practice approaches.