A Methodologic Solution to Missing Deauville Scores Using Imaging Report Data to Classify Lymphoma Treatment Response in Real-World Data
Author(s)
Swain R, Zimmerman Savill KM, Klink A, Asgarisabet P, Balanean A, Hays H, Kaufman J, McAllister L, Omary C, Yu HT, Kalesan B, Laney J, Feinberg B
Cardinal Health, Dublin, OH, USA
Presentation Documents
OBJECTIVES: Outcome misclassification is an important source of bias when comparing real-world data (RWD) to clinical trial data because classification methods are inherently different. To address this, we developed a standardized real-world methodology using Lugano 2014 (i.e., rwLugano) to classify lymphoma outcomes using RWD. However, Deauville score, a necessary component in Lugano, was frequently missing, as was standardized uptake value (SUV) of background tissues, which are used to calculate Deauville. Therefore, we developed rwDeauville using charted tumor SUV compared to normal SUV as reported in literature.
METHODS: We conducted a retrospective multisite real-world study using Cardinal Health’s Practice Research Network (PRN) to identify adult patients with diffuse large B-cell lymphoma (DLBCL) treated with first line chemoimmunotherapy in US clinical practice (01JAN2015-31DEC2022). PRN sites conducted manual chart abstraction from medical records and positron emission tomography–computed tomography (PET-CT) reports at treatment initiation and first response assessment. We calculated rwDeauville comparing tumor SUV to literature-based values: background SUV=1, mediastinum SUV=3, and liver SUV=5. We assessed agreement [%] and concordance [Cohen’s kappa (κ)] between Deauville recorded in PET-CT reports and rwDeauville.
RESULTS: We identified 174 DLBCL patients [male n=103 (59%); female n=71 (41%); mean age 66 years] across 6 PRN sites. Deauville and tissue SUV were frequently missing in baseline [n=77 (44%)] and first response assessment [n=29 (17%)]. Among first response PET-CT reports with Deauville data available for comparison (n=145), rwDeauville demonstrated high concordance [86%, κ=0.81]. Consequently, we were able to calculate rwLugano for all study patients using reported Deauville, when available, and rwDeauville when Deauville was missing.
CONCLUSIONS: Deauville score and background SUV are frequently missing from lymphoma PET-CT reports but can be estimated with high accuracy using the rwDeauville algorithm. Consequently, we observed sufficient charted information to calculate rwLugano – a standardized method for lymphoma outcome classification using RWD – for all study patients.
Conference/Value in Health Info
Value in Health, Volume 27, Issue 6, S1 (June 2024)
Code
PT29
Topic
Clinical Outcomes, Methodological & Statistical Research
Topic Subcategory
Clinical Outcomes Assessment, Clinician Reported Outcomes, Confounding, Selection Bias Correction, Causal Inference, Missing Data
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, Oncology