New prognostic model for survival in patients with R/R DLBCL

Patients with diffuse large B-cell lymphoma (DLBCL) who relapse within five years of frontline immunochemotherapy (IC) have very variable outcomes. In some cases, a second long-term remission can be achieved with high-dose chemotherapy and autologous stem cell transplant (auto-SCT), though patients who are refractory to therapy tend to have poor outcomes.

Being able to predict response and risk has both clinical and research applications. In December 2019, at the 61st American Society of Hematology (ASH) Meeting & Exposition, Matthew J. Maurer, Mayo Clinic, Rochester, US, presented results from a study which aimed to evaluate clinical predictors of outcome at first relapse or progression of DLBCL after frontline IC and subsequently build and validate a risk model for this patient setting.

Study design

Different cohorts were used for modeling and validation. All cohorts were formed of patients enrolled at initial diagnosis and treated with rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisolone (R-CHOP) or similar IC. All patients were prospectively followed for progression or relapse:

• Model cohort
  • Surrogate Endpoint in Aggressive Lymphoma (SEAL) cohort comprised 13 frontline, multicenter, randomized clinical trials in patients with DLBCL
• Validation cohorts
  • University of Iowa/Mayo Clinic SPORE Molecular Epidemiology Resource (MER) cohort
  • Danish National Lymphoma Registry (LYFO) population-based cohort

The primary outcome measured was overall survival (OS) from date of first progression/relapse.

Model cohort

• Of 5,112 patients enrolled in the SEAL cohort trials, 1,125 patients with progression were used for the analysis
• Median age at progression: 68 years (interquartile range [IQR], 60–73)
• Median follow-up from progression: 8 months (range, 0–164)
• International Prognostic Index (IPI) score 3–5: 42% 
• Median OS: 11 months
• Two-year OS: 35%
• Two-year OS was associated with time to progression (TTP):
  • TTP < 6 months after starting IC (refractory to IC): 18%
  • TTP > 3 years after starting IC: 69%
• Age at time of progression was also associated with OS:
  • No significant association in patients < 50 years old
  • Risk of death increased from age 50–80 years

The final prognostic calculator was developed based on the associations between age at progression and TTP from frontline IC. The final modeling set included 1,011 patients aged 40–80 years with a TTP of 0–60 months and a concordance (C)-statistic of 0.67. When assessed for calibration using the SEAL dataset, the model predicted a mean 2-year OS of 34%, which was the same as the original 2-year OS predicted by Kaplan-Meier curves (34%).

Validation cohort

The model was then validated with the MER and LYFO cohorts in patients who initiated second-line therapy for relapsed/refractory (R/R) DLBCL. In both the MER and LYFO cohort, the model underestimated survival, as shown in Table 1.

Table 1. Model validation in MER and LYFO cohorts

DLBCL, diffuse large B-cell lymphoma; IC, immunochemotherapy; IQR, interquartile range; LYFO, Danish National Lymphoma Registry; MER, Molecular Epidemiology Resource; OS, overall survival; R/R, relapsed/refractory
	MER cohort	LYFO cohort
Patient population	Patients who initiated second-line therapy for R/R DLBCL	Patients with DLBCL who progressed after frontline IC
N	290	559
Median age, years (IQR)	62 (57–70)	67 (60–73)
C-statistic	0.67	0.67
Mean 2-year OS (predicted vs actual)	33% vs 47%	32% vs 38%

Limitations of model

• Model underestimated survival in the validation cohorts, which the investigators hypothesize may be because:
  • patients in the MER and LYFO cohorts were younger on average
  • more patients were treated with a curative intent in validation cohorts 
• Management may differ regionally for R/R DLBCL — the model may not be applicable in all local practices
• This analysis was based on historical data and before the era of immunotherapies, such as chimeric antigen receptor (CAR) T-cell therapy

Key take-home messages

• Age and TTP from frontline IC are associated with survival in patients with R/R DLBCL
• A risk model was developed based on large international cohorts of patients with DLBCL who were followed from diagnosis
• In the validation data sets, the model underestimated survival
• This calculator is available online at https://qxcalc.app.link/dlbcl  
• Future directions: the study investigators aim to re-fit the model to allow for patients treated with curative intent. This will make the model more applicable for patients receiving treatments such as auto-SCT and CAR T-cell therapy

Video