Title: Development of a Pathologic Prognostic Index (PPI) for Newly Diagnosed Large B Cell Lymphoma Patients Treated with R-CHOP

Introduction Pathologic evaluation of large B cell lymphoma (LBCL) biopsies in clinical laboratories includes immunohistochemical staining (IHC), fluorescence in situ hybridization (FISH), and next generation sequencing, although results of these assays have not routinely been incorporated into prognostic models. We aimed to determine if a novel prognostic index incorporating features detected by these assays could predict survival outcomes in patients (pts) with newly diagnosed LBCL, and how such an index performs as compared to the Revised International Prognostic Index (R-IPI). Methods Clinicopathologic data from newly diagnosed LBCL pts obtained from 3 publicly available datasets (PMID 37851898, 37032379, and 30523719) were retrospectively compiled and analyzed. Included pts were those treated with R-CHOP with available IPI score, cell of origin (COO) and double expressor (DE) status (MYC ≥40% and BCL2 ≥50%) by IHC, MYC rearrangement and MYC/BCL2 double hit status by FISH, variant reporting of genes common to all three datasets (B2M, CD79B, CIITA, CREBBP, EZH2, GNA13, MYD88, NOTCH1, NOTCH2, SOCS1, STAT3, TET2, TNFAIP3, TNFRSF14 and TP53) and follow-up time of ≥1 year (yr) from diagnosis if not experiencing disease progression prior to that time. The above features were each analyzed by univariate Cox regression analysis to predict both disease progression and death at 2 yrs, with those demonstrating statistically significant hazard ratios (HR) for progression or death events, defined as two-tailed P value <0.05, included in multivariable Cox regression analysis. We then constructed a novel, ternary prognostic index incorporating factors predictive of progression or death events. To assess the predictive power and goodness-of-fit of our novel index against the R-IPI, we conducted Concordance Index (C-Index), Akaike Information Criterion (AIC), and Net Reclassification Index (NRI) measurements comparing the two models. Results A total of 412 pts were included with a median length of follow-up of 35.8 months. Multivariable analysis demonstrated IPI score ≥ 3 (HR, 2.26 [95% confidence interval {CI}, 1.29-3.95]; p = 0.0044), DE lymphoma (HR, 2.28 [95% CI, 1.29-4.01]; p = 0.0044), and TP53 loss of function (LOF) mutation (HR, 2.10 [95% CI, 1.16 - 3.82]; p = 0.015) were independently predictive of disease progression at 2 yrs, and IPI score ≥3 (HR, 1.86 [95% CI, 1.27-2.71]; p = 0.0014), DE lymphoma (HR, 2.21 [95% CI, 1.51-3.24]; p = 0.00005), and TP53 LOF mutation (HR, 1.66 [95% CI, 1.08-2.53]; p = 0.02) were independently predictive of death at 2 yrs. We subsequently constructed the Pathologic Prognostic Index (PPI) incorporating these 3 risk factors, where each unit increase indicates the presence of one additional factor, with the relatively small 3-risk factor pt group (n = 11) being combined with the 2-risk factor pt group (n = 72). Distribution of PPI scores was 0 in 178 pts (43.2%), 1 in 151 pts (36.7%), and 2 in 83 pts (20.1%), as compared to the R-IPI (0/Very Good in 68 pts (16.5%), 1-2/Good in 212 pts (51.5%); 3-5/Poor in 132 pts (32.0%). Kaplan Meier estimates of 2 yr progression-free survival (PFS) and overall survival (OS) for all pts were 72.7% (95% CI, 68.4% - 77.2%) and 87.2% (95% CI, 83.9% - 90.7%), and the 2 yr PFS and OS rates using the PPI (0: 85.4% and 96.1%; 1: 68.2% and 82.8%; 2: 57.8% and 80.7%) differed significantly (p < 0.0001 for both 2 yr PFS and OS). The PPI predicted lower or similar median survival rates than its R-IPI counterpart (0/Very Good: 91.2% and 98.5%; 1-2/Good: 75.0% and 89.6%; 3-5/Poor: 62.1% and 80.3%) at upper strata. The PPI (OS - C-Index: 0.67, AIC: 562.49, NRI: 0.07; PFS - C-Index: 0.63, AIC: 1250.94, NRI: 0.09) outperformed R-IPI (OS - C-Index: 0.65, AIC: 562.92; PFS - C-Index: 0.61, AIC: 12656.38). Combined with the shifts in scoring distribution, higher C-Indexes, lower AICs, and positive NRIs suggest that PPI respectively confers greater discriminative ability, better model fit, and improved risk reclassification. Conclusion In this cohort of newly diagnosed LBCL pts treated with R-CHOP, the PPI enhanced prognostication as compared to the R-IPI by identifying nearly twice the proportion of pts who achieve lower rates of 2 yr PFS and OS as compared to that of the entire population. The PPI should be validated in other retrospective or prospective cohorts of newly-diagnosed LBCL pts treated with R-CHOP or other first-line regimens.