Old variables, new value: a refined IPI for DLBCL

In this issue of Blood, Zhou and colleagues present a refined prognostic index for patients with de novo diffuse large B-cell lymphoma (DLBCL), addressing the need for a tool that can better separate risk groups in the rituximab era.¹ 

The International Prognostic Index (IPI) was originally devised in the era before rituximab and before DLBCL was consistently separated from other large cell lymphoma entities.²  The use of rituximab has improved survival in DLBCL, but has also narrowed the outcome differences between the IPI risk groups. With current first-line therapies, even the highest risk group identified by the original IPI has a 5-year overall survival of 50%.^3,4  Many have sought to identify new prognostic tools using molecular markers, gene expression signatures, or interim positron emission tomography response assessment, but all of these approaches have limitations and none are ready to be broadly adopted into clinical practice in the immediate future. At present, prognostic models that rely on widely available clinical parameters are more easily applied than models that include variables requiring highly specialized or nonstandardized measures. Thus, revisiting the IPI among patients with DLBCL who were treated with rituximab-containing regimens using well-standardized clinical variables is timely and appropriate.

Zhou and colleagues have devised a new prognostic model: the National Comprehensive Cancer Network–IPI (NCNN-IPI). The model is straightforward and looks similar in many ways to the original IPI. The authors initially constructed the index using data from an NCCN cohort of patients with DLBCL and then externally validated the approach using an independent cohort from the British Columbia Cancer Agency registry. Statistical efforts gleaned additional information from familiar prognostic variables, including age, serum lactate dehydrogenase (LDH), and extranodal sites of disease. The authors demonstrate that the effect of age on survival is linear, while the effect of LDH on survival plateaus when normalized LDH is more than threefold the upper institutional limit of normal. The continuous nature of these data are taken into consideration in the NCCN-IPI and it is not surprising that additional prognostic information was gained by moving away from dichotomizing these variables. The authors also refined extranodal disease as a prognostic variable, finding that specific sites of involvement (bone marrow, liver/gastrointestinal, central nervous system, or lung) performed better in the model than a variable based solely on the number of extranodal sites of disease.

This enhanced NCCN-IPI compares favorably with the original IPI insofar as it better differentiates those with the best and worst prognoses. Using the NCCN-IPI, the outcomes among risk groups spans a larger range, with the low-risk group having an estimated 5-year overall survival (OS) of 96% and the high-risk group having an estimated 5-year OS of 33%. This is in contrast to 90% and 54% for low- and high-risk groups, respectively, using the IPI. The ability of the NCCN-IPI to better discriminate risk groups can be illustrated by examining the prognostic scores and associated outcome predictions for 2 hypothetical patients with DLBCL (see table). Although both patients have the same 5-year OS estimate by the IPI (62%), the NCCN-IPI indicates that patient 1 has a much better chance of a good outcome with a 5-year OS estimate of 82% compared with patient 2 with a 5-year OS estimate of 33%. Taken together, it seems likely that this report will set a new standard for prognostication in DLBCL. From a practical standpoint, the NCCN-IPI can be easily applied in academic and community-based practice settings alike. In the age of web-based calculators and smartphone apps, incorporating the NCCN-IPI into daily practice can be simply achieved with a software update.

DLBCL remains a heterogeneous disease with biologic mediators of treatment response and outcome that are not fully elucidated, much less incorporated into a prognostic model. In Hodgkin lymphoma, the International Prognostic Score similarly relies on routine clinical and laboratory parameters amid growing appreciation that more specific biologic characteristics in these tumors appear to carry prognostic significance. Analogous situations exist for most other hematologic malignancies as well. More complex prognostic markers such as gene-expression–based signatures and molecular biomarkers should be evaluated for their ability to improve upon the NCCN-IPI and contribute additional risk prediction on an individual level.⁵  The NCCN-IPI can serve as the new framework upon which to evaluate the clinical utility of these novel prognostic markers.