Recent phase 3 trials targeted to patients with adverse prognostic factors in DLBCL (non GC, double hit/expressors) have faced problems in interpretation because control groups scored clearly better than expected, (Lym2034, Pyramid, ReModl). This suggests that the application of inclusion/exclusion criteria, combined with a delay in treatment caused by their confirmation, may lead to a selection bias that excludes the patient group most in need for treatment improvement.


To characterize the magnitude and nature of exclusion bias we conducted a retrospective real world study on unselected patients diagnosed with DLBCL between 2011-2015 and followed up for at least 2 years at the University Hospital Ghent, a tertiary care centre serving an area of approximately 1.5 million inhabitants.

Four questions were asked:

- How many patients are considered fit enough in a real world population of DLBCL to receive standard treatment (R-(mini)CHOP, -like)?

- What is the PFS/OS in this group compared to the total patient population?

- How is the PFS/OS influenced by time from diagnosis to initiation of treatment?

- How is treatment choice and PFS/OS influenced by comorbidities?

Patients and methods:

A total of 162 patients with DLBCL were included. The median age was 65 years (range 16-91) with amale/female ratio 56%/44%. Four % of patients had an ECOG >2 and the distribution of the IPI-score was 0-1: 44 patients (27%), IPI 2: 27 (17%), IPI 3: 49 (30%), IPI ≥ 4: 41 patients (25%). NCCN-IPI was low (0-1) in 16 pts, low interm (2-3) in 48 pts, high-int (4-5) in 71 pts and high(>5) in 22 pts (14%). All patients had a pathological diagnosis of DLBCL (excl PMBL). In 25 patients there were signs of transformed lymphoma with DLBCL, 12 were double expressor (DEL), 3 double hit lymphoma (DHL). Cell of origin was assessed by IHC, Hans Algorythm in 131 patients (72 GC, 57 non GC). 133 patients were treated with standard R-(mini)CHOP or R-CHOP like regimens (81%) outside clinical trials (RCT). 17 patients were considered unfit for at least R-mini CHOP: 5 received R-CvP, 8 R-mono, 4 palliative care. Four young patients with aggressive presentation received more intensive regimens. 107/162 completed full treatment as planned (66%) 12 patients were entered in RCT and received at least R-CHOP like regimens. Statistical significance was tested using a log-rank based analysis on actuarial survival curves


Median follow-up was 61 mo (range 19-323).Response assessment was based on PET-CT as defined by the Lugano 2014 criteria.

Overall, 143/162 could be treated with R-(mini)-CHOP like regimens (88%), 12 in RCT and 131 out of RCT. 123/162 76% of patients responded, 18 died (11%) and 19/162 (12%) were refractory to 1st line treatment. 37 of 123 responders relapsed, 10 within 6 months, 28 others after a median of 44 months.

In the subgroup of 131 patients receiving standard R-(mini)CHOP like treatment outside RCT, 101 reached CR (77%), 4 PR, 15 PD. 9 died and 28 relapsed following CR. Two years PFS is 67% in the standard R-(mini)CHOP group outside RCT, and 83% in 12 patients entering RCT. 2 years overall survival is 71% in the 131 patients receiving standard R-(mini)CHOP and 92% in the 12 patients in RCT.

There was no difference in PFS and overall survival in patients with GC versus non GC based on IHC, (p=0.503) or by the presence of signs of transformation (p=0.131).

Time from diagnosis to initiation of treatment divided the patient population in patients receiving treatment within 16 days (n=66), 16-30 days (n=55) and beyond 30 days (n=41).Two years PFS and OS in these 3 groups were 50%, 76%, 79% (p=0,004) and 59%, 72%, 83% (p=0,017), respectively.

Significant comorbidities were observed in 114/162 patients. Two years OS in patients without comorbidity was 92% compared to 54% in patients with comorbidity. (p<0.05)


Standard treatment in DLBCL can be given to up to 88% of patients, but it selects for fitter patients with less comorbidities and better PFS/OS and in RCT it is even more so (2yr OS 67, 71 and 91%). Patients requiring early treatment who cannot be subject of central review as part of study inclusion because of potential treatment delay fare worse than patients in whom treatment can be delayed. These combined factors may explain part of the better results observed in RCT control groups compared to real world data.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.