Reduced Specificity and Positive Predictive Value of Surveillance FDG-PET/CT for Diffuse Large Cell B Cell Lymphoma in the Rituximab Era

Surveillance PET is reported to allow the detection of asymptomatic relapses in a substantial number of patients with diffuse large cell B cell lymphoma (DLBCL). Recent data suggest that rituximab (R) administration results in a higher incidence of false positive (FP)“end-therapy” scans. However, the predictive value (PV) of surveillance PET in patients receiving R versus chemotherapy has not been fully explored.

The current study compared the PV of surveillance PET in patients with DLBCL receiving CHOP-R therapy versus CHOP alone.

This retrospective study was approved by the IRB of the Rambam Health Care Campus (Haifa, Israel). The institutional database was searched for all newly diagnosed adult patients with DLBCL, treated with the CHOP or CHOP-R between January 1995–2008, who achieved complete remission (CR), had at least one follow-up (FU) FDG PET/CT during remission and were followed until relapse/death, or for at least 12 months after the last FU scan. The routine FU protocol included PET scans, performed at 3, 6, 12, 18, 24, 36, 48 and 60 months after CR. Demographic, clinical and imaging data at disease staging, during FU and at recurrence were collected and analyzed separately for patients treated with CHOP alone (group 1) and for those receiving CHOP-R (group 2). The ability of PET-FU to detect recurrence was assessed for the whole cohort, depending on rituximab administration, duration of CR, and location of suspicious findings. All scans were originally reviewed by 2 PET specialists, and positive scans were re-evaluated using the same criteria as those employed to report initial findings. PET-FU was considered positive in the presence of an uptake unrelated to physiological bio-distribution or a known benign process. PET-FU results were confirmed by biopsy or further imaging and clinical FU.

119 patients, 35 treated with CHOP and 84 with CHOP-R, were analyzed. Median age was 59 years (24–88); 59% presented with an advanced stage (III-IV) and 45% had an IPI score ≥2. There were no statistically significant differences in patient characteristics in the 2 groups, except for a shorter median FU period for patients receiving R (2.9 vs 6.4 years, p<0.0001). Within a median FU of 3.4 years (0.6–8.6), 31 patients relapsed (17 confirmed histologically), 14 in the CHOP-R group (15%) vs 17 in the CHOP cohort (47%), (p=0.02). Nine (29%) relapses were initially detected by PET-FU in asymptomatic patients, with no difference in the incidence of these relapses between the 2 groups. Relapse involved the original sites in 85% of cases, with no differences between the groups. A total of 422 PET studies were performed; 113 in group 1 and 309 in group 2. Eighty three studies were judged to be positive, 23 in group 1 and 60 in group 2. However, in the CHOP-R group, only 23% (14/60) were truly positive compared to 74% (17/23) in the CHOP group (p=0.001).The median time to FP PET was significantly longer for patients receiving CHOP-R (1.3 vs 0.6 years, p=0.03). Specificity and positive PV (PPV) were significantly lower for patients receiving CHOP-R compared to those treated with CHOP only (Table 1). An FDG uptake involving head and neck lymph nodes was more likely to be FP, especially in group 2 (88% vs 4%, p=0.0004). Furthermore, age younger than 60 years and earlier disease stage were also found to be significantly associated with an increased incidence of FP results, particularly in patients receiving R.

Surveillance PET in patients with DLBCL is highly sensitive for the detection of recurrence, providing the first indication of relapse in 29% of patients. However, comparative analysis shows that specificity of PET-FU is significantly lower in the R era, yielding a PPV of only 23%. Interestingly, late FP PET, involving nodal sites, uniquely observed in patients receiving R, is assumed to reflect lymph node “recovery” following the R-induced B cell depletion. The emerging data emphasize the limitations of surveillance PET in the R era and the need for an efficient algorithm for its use in this setting.

No relevant conflicts of interest to declare.