Clinical factors predictive of POD24, a robust early clinical endpoint of poor survival in patients with FL

Follicular lymphoma (FL), the most common form of non-Hodgkin lymphoma and one that is generally associated with favorable outcomes, can have a variable clinical course for a subset of patients who may experience multiple relapses, early disease progression, or transformation to a more aggressive histology. Approximately 20% of patients have early FL relapse, defined as recurrence or progression of disease within 24 months of front-line treatment (POD24). POD24 is associated with poor outcomes and five-year overall survival (OS) rates range from 34%‒54%.

Despite ongoing research studying clinical and biologic predictors of POD24, currently no markers have been identified that can be uniformly implemented for risk-adapted therapy at diagnosis of FL. Casulo, et al. recently published in Blood¹ an analysis of patients with FL from 13 randomized prospective trials with the aim of identifying clinical factors predictive of POD24 and thus the potential for identifying patients at diagnosis who are at high risk of poor future outcomes. Their findings are summarized in this article.

Study design

This was a pooled analysis of 5,225 patients from 13 multicenter international randomized controlled trials (RCTs) using the Follicular Lymphoma Analysis of Surrogate Hypothesis (FLASH) data. FLASH data includes individual patient data from RCTs of first-line therapy in FL, including rituximab only, chemoimmunotherapy, and chemotherapy only; this pooled analysis included patients both the pre- and post-rituximab era.

The association between Follicular Lymphoma International Prognostic Index (FLIPI) risk category, gender, beta 2 microglobulin (B2M), and performance status (PS) with POD24 was evaluated with logistic regression models.

• An analysis of the association between POD24 and OS for patients who were alive at 24 months post trial registration was performed.
• To reduce potential bias, time for all patients began at trial registration, patients were grouped based on progression status as 24 months post trial registration, and any patient censored for OS before 24 months or who died before 24 months was excluded.
• The association between clinical factors and progression of disease or death within 24 months of front-line therapy (PFS24) was evaluated with a further logistic regression model, where death within 24 months of treatment was treated as an event.
• To estimate the effect of maintenance treatment with rituximab on POD24 and OS, a maintenance rituximab variable was added to the logistic regression.

Baseline characteristics

Data was analyzed from ten induction and three maintenance RCTs. Patients who were censored for clinical follow-up before 24 months were excluded from these trials. Baseline characteristics are summarized in Table 1.

Table 1. Patient baseline characteristics by POD24 status*

Results

Clinical factors predicting POD24

Using a multivariable logistic regression model, clinical factors found to increase risk of POD24 included:

• being male
• having a PS ³2  
• FLIPI risk score
  • intermediate [2]
  • high [3–5]
• elevated baseline B2M ³3mg/L
  • This association was found in a subset of patients from 9 studies that included B2M values at baseline.

Within the treatment regimen subsets (chemotherapy, rituximab-chemotherapy, and rituximab), results remained mostly consistent, but due to smaller sample sizes, the power to detect statistically significant effects was reduced. Results are summarized in Table 2.

Table 2. Results for POD24 logistic regression and treatment regimen subsets*

In the three trials which included maintenance randomization (n = 1,255), maintenance rituximab treatment was associated with a decreased risk of POD24 (OR = 0.39; p < 0.01). A high FLIPI risk score was associated with increased risk of POD24 (OR = 1.94; p < 0.01). Being male and having a PS ³2 showed evidence of increased risk of POD24; however, this was not statistically significant. Results from the PFS24 model were consistent with the POD24 model.

The association of POD24 on OS

The time-dependent Cox model, adjusted for gender and stratified by PS and FLIPI, showed POD24 was associated with poor subsequent OS (hazard ratio [HR], 4.85; p < 0.01). The results remained consistent in the treatment subgroups (all p < 0.01), with the strongest effect seen in the rituximab only treatment subset:

• rituximab-chemotherapy group (n = 2,252; HR, 5.39)
• chemotherapy (n = 2,335; HR, 4.26)
• rituximab (n = 265; HR, 6.57)

In the subset of three trials with maintenance randomization, time dependent POD24 was also associated with poor subsequent OS (HR, 4.68; p < 0.01), however there was no statistically significant effect on subsequent OS (HR, 1.18; p = 0.24) in patients receiving maintenance rituximab.

The association of POD24 on OS for patients alive at 24 months post trial registration

The multivariable Cox model, adjusted for gender, and stratified by PS and FLIPI, showed POD within 24 months of trial registration for patients alive at 24 months was associated with poorer subsequent OS (HR, 3.03; p < 0.01). The results remained consistent within the treatment subgroups (all p < 0.01), with the strongest effect in patients treated with rituximab-chemotherapy:

• rituximab-chemotherapy (n = 2,133; HR, 3.58)
• chemotherapy (n = 2,041; HR, 2.63)
• rituximab (n = 249; HR, 3.96)

Post-trial registration survival rates were predicted and are summarized in Table 3.

Table 3. Three and five-year post-registration survival probabilities*

Regarding survival outcomes for patients based on treatment group:

• Of those treated with rituximab-chemotherapy induction, 17.3% had POD24 with an OS of 73.5.
• Of the rituximab-only patients, 42.2% had POD24 with an OS of 90.9.
• Of the chemotherapy-only patients, 30.4% had POD24 with an OS of 66.9.

Conclusion

This is the largest cohort of patients with FL analyzed to date. The findings support early disease progression as a robust clinical indicator of poor survival. Whilst poor PS and high FLIPI score have previously been found to predict histologic transformation of FL, elevated baseline B2M and male gender as predictive factors for POD24 are significant findings that further help to identify patients at diagnosis who are at high risk of poor future outcomes.

POD24 is associated with poorer OS, and differences in survival after relapse were demonstrated depending on the induction treatment patients received. Disease progression after rituximab-chemotherapy had the most pronounced influence on poor subsequent outcome, presumably due to higher tumor burden/less favorable disease requiring anthracyclines in this subgroup.

While the findings from this large heterogeneous group of patients with FL provides clearly defined clinical factors, a prognostic model is not yet available for widespread use. Limitations include lack of harmonization and lack of biologic and molecular data, as well as a heterogeneously treated patient population. Despite these limitations, the data from this study validate POD24 as a robust marker that will help further the development of comprehensive prognostic models to identify patients at diagnosis who are at risk of poorer outcomes.