Local Review Versus (vs) Central Review of Fluorodeoxyglucose Positron Emission Tomography (FDG-PET) in Diffuse Large B-Cell Lymphoma (DLBCL): Results from the CALGB 50303 Trial [Alliance]

Background

FDG-PET is used in staging and response assessments in patients (pts) with DLBCL. An analysis of 158 pts with DLBCL receiving R-CHOP or DA-EPOCH-R on CALGB/Alliance 50303 failed to show an association between the interim PET (iPET, after cycle 2 of therapy) visual 5-point scale score (5-PS) by central review and progression-free survival (PFS) or overall survival (OS), but did show a significant association between percent change in FDG uptake from baseline to after cycle 2 (∆SUV) and PFS/OS (Schoder et al. 2020). As central review of FDG-PET is not feasible in routine practice, we retrospectively compared local vs central readings, and determined associations from local imaging data with PFS/OS.

Methods

126 pts at baseline and 109 at iPET had local imaging data submitted across 21 institutions. Visual scores 1-3 defined PET- and scores 4-5 defined PET+ disease. ∆SUV was defined as the percent change in the maximum SUV at any site (maxSUV) from baseline to iPET. Continuous ∆SUV and categorized ∆SUV (<66% vs >66%) were analyzed. Differences in quantitative measures were calculated as the local minus central result. The Kappa statistic estimated agreement in PET and ∆SUV groups between local and central review. PFS and OS distributions for local PET and ∆SUV groups were landmarked at iPET, estimated with the Kaplan-Meier method, and compared using the log-rank test.

Results

At baseline, maxSUV was reported locally and centrally in 118 pts. Median maxSUV was 24.3 (range: 5.9-77.3) by local review and 24.3 (range: 5.9-77.0) by central review. The median difference in maxSUV was 0 (range: -12.9 to 21.9), with 50% of differences between -1.1 and 0.1, and 90% of differences between -7.8 and 4.1.

In 106 pts with visual iPET 5-PS results, 52 (49.1%) were PET+ by local review and 37 (34.9%) by central review. Agreement in local and central review was moderate (kappa=0.53), occurring in 81 pts (76.4%; 32 PET+ and 49 PET-). Disagreement occurred in 25 patients, 20 with local PET+ but central PET- and 5 with local PET- but central PET+. iPET maxSUV was reported locally and centrally in 90 pts. Median maxSUV was 3.4 (range: 0-18.3) by local review and 3.3 (range: 1.2-18.2) by central review. The median difference in maxSUV was 0 (range: -8.6 to 2.4), with 50% of differences between -0.3 and 0.4, and 90% of differences between -4.4 and 1.8.

∆SUV was calculated in 87 pts. Median ∆SUV was 84.6% (range: -3.0% to 95.9%) by local review and 85.1% (range: -34.9% to 95.8%) by central review. The median difference in ∆SUV was -0.3 (range: -19.6 to 31.9), with 50% of differences between -2.1 and 1.1, and 90% of differences between -7.1 and 12.3. ∆SUV was <66% in 12 (13.8%) pts by local review and 12 pts by central review. Agreement in local and central review was high (kappa=0.81), occurring in 83 pts (95.4%; 10 with ∆SUV <66%, 73 with ∆SUV >66%). Disagreement occurred in 4 pts, 2 in each direction.

Using local data, PFS and OS estimates were numerically lower in pts with iPET+ vs iPET- disease, but did not reach statistical significance (p=0.12 and p=0.15). Two-year estimates for PFS were 79% (95% CI 68-91%) and 89% (95% CI 81-98%), and 2-year estimates for OS were 84% (95% CI 75-95%) and 96% (95% CI 91-100%) in pts with iPET+ and iPET-, respectively. In contrast, ∆SUV groups were significantly associated with PFS and OS (p=0.03 and p=0.002). Two-year estimates for PFS were 56% (95% CI 34-94%) and 87% (95% CI 79-95%) and 2-year estimates for OS were 56% (95% CI 34-94%) and 93% (95% CI 88-99%) in pts with ∆SUV <66% and ∆SUV >66%, respectively.

Conclusion

Agreement in visual 5-PS at iPET was moderate; most discrepancies arose from a local PET+ result when central review called PET-. Differences in maxSUV and ∆SUV tended to be small. Grouping ∆SUV resulted in only 4 discrepancies (4.6%). Similar to the previous analysis using central data, the separation in PFS and OS curves using local data for ∆SUV was greater than by visual 5-PS. Encouragingly, the agreement in ∆SUV of <66% vs ≥66% was high between local and central review. Overall agreement however may be unduly influenced by the large proportion of pts with ∆SUV ≥66%, with less certainty in agreement for pts with ∆SUV <66%. Larger prospective studies comparing local and central imaging data are warranted to best determine the utility of local imaging in clinical trials.

Support: P30 CA008748, U10CA180821, U10CA180882; https://acknowledgments.alliancefound.org. ClinicalTrials.gov Identifier: NCT00118209 (CALGB 50303)