Interobserver Agreement Rates on CXCR4-Directed PET/CT in Patients with Marginal Zone Lymphoma

C-X-C motif chemokine receptor 4 (CXCR4)-directed molecular imaging provides excellent read-out capabilities in patients with marginal zone lymphoma (MZL). We aimed to determine the interobserver agreement rate of CXCR4-targeted PET/CT among readers with different levels of experience.

Methods: 50 subjects with MZL underwent CXCR4-targeted PET/CT, which were reviewed by four readers (including two experienced and two less experienced observers). The following 8 parameters were investigated: overall scan result, CXCR4 density in lymphoma tissue, extranodal organ involvement, No. of affected extranodal organs and extranodal organ metastases, lymph node (LN) involvement and No. of affected LN areas and LN metastases. We applied intraclass correlation coefficients (ICC; < 0.4, poor; 0.4-0.59, fair; 0.6-0.74, good and > 0.74 excellent agreement rates).

Results: Among all readers, fair agreement was recorded for No. of affected extranodal organs (ICC, 0.40; 95% confidence interval [CI], 0.25-0.68), overall scan result (ICC, 0.42; 95%CI, 0.28-0.57), CXCR4 density in lymphoma tissue (ICC, 0.52; 95%CI, 0.38-0.66), and No. of extranodal organ metastases (ICC, 0.55; 95%CI, 0.41-0.61) and LN involvement (ICC, 0.59; 95%CI, 0.46-0.71). Good agreement rates were observed for No. of LN metastases (ICC, 0.71; 95%CI, 0.60-0.81) and No. of LN areas (ICC, 0.73; 95%CI, 0.63-0.82), while extranodal organ involvement (ICC, 0.35; 95%CI, 0.21-0.51) achieved poor concordance. On a reader-by-reader comparison, the experienced readers achieved significantly higher agreement rates in 4/8 (50%) investigated scan items (ICC, range, 0.21-0.90, P < / = 0.04). In the remaining 4/8 (50%), a similar trend with higher ICCs for the experienced readers was recorded (n.s.).

Conclusion: CXCR4-directed PET/CT mainly provided fair to good agreement rates for scan assessment, while a relevant level of experience seems to be required for an accurate imaging read-out.