Clinicogenetic risk models predict early progression of follicular lymphoma after first-line immunochemotherapy

Abstract

Follicular lymphoma (FL) is a clinically and molecularly heterogeneous disease. Posttreatment surrogate end points, such as progression of disease within 24 months (POD24) are promising predictors for overall survival (OS) but are of limited clinical value, primarily because they cannot guide up-front treatment decisions. We used the clinical and molecular data from 2 independent cohorts of symptomatic patients in need of first-line immunochemotherapy (151 patients from a German Low-Grade Lymphoma Study Group [GLSG] trial and 107 patients from a population-based registry of the British Columbia Cancer Agency [BCCA]) to validate the predictive utility of POD24, and to evaluate the ability of pretreatment risk models to predict early treatment failure. POD24 occurred in 17% and 23% of evaluable GLSG and BCCA patients, with 5-year OS rates of 41% (vs 91% for those without POD24, P < .0001) and 26% (vs 86%, P < .0001), respectively. The m7-FL International Prognostic Index (m7-FLIPI), a prospective clinicogenetic risk model for failure-free survival, had the highest accuracy to predict POD24 (76% and 77%, respectively) with an odds ratio of 5.82 in GLSG (P = .00031) and 4.76 in BCCA patients (P = .0052). A clinicogenetic risk model specifically designed to predict POD24, the POD24-PI, had the highest sensitivity to predict POD24, but at the expense of a lower specificity. In conclusion, the m7-FLIPI prospectively identifies the smallest subgroup of patients (28% and 22%, respectively) at highest risk of early failure of first-line immunochemotherapy and death, including patients not fulfilling the POD24 criteria, and should be evaluated in prospective trials of precision medicine approaches in FL.

Figure 1

Progression of disease within 24 months (POD24) is an accurate predictor of poor overall survival (OS). (A) Distribution of patients from the GLSG and BCCA cohorts according to the POD24 classifier. (B) Kaplan Meier curves for OS from risk-defining events for patients with or without POD24 of diagnosis (dashed lines) or 24 months of treatment initiation (solid lines) from the GLSG (left) and BCCA cohorts (right). Displayed statistics refer to POD24 status calculated from time of treatment initiation (ie, modified definition).

Figure 2

Accuracy of 3 pretreatment risk models to predict POD24 status. (A) Follicular Lymphoma International Prognostic Index (FLIPI), (B) m7-FLIPI, and (C) POD24 Prognostic Index (POD24-PI).

Figure 3

Risk stratification for failure-free survival (FFS) in patients without POD24 according to 3 pretreatment risk models. Kaplan-Meier curves for FFS beyond 2 years after treatment initiation for patients without POD24 according to (A) the FLIPI, (B) the m7-FLIPI, and (C) the POD24-PI. Numbers in parentheses indicate patients with event/number of patients per subgroup. Pie charts illustrate distribution of risk status of the respective risk classifier and POD24 status.

Figure 4

The POD24 Prognostic Index (POD24-PI). (A) The POD24-PI is calculated as the sum of individual clinical and gene mutation predictor values weighted by their individual coefficients. (B) Kaplan-Meier curves for failure-free survival (FFS), and (C) overall survival (OS) for patients from the GLSG and BCCA cohorts according to POD24-PI status. Numbers in parentheses indicate patients with event/number of patients per subgroup.

Figure 5

Identification of high-risk patients by 3 pretreatment risk models and POD24 status. Venn diagrams visualize the overlap of high-risk patients as identified by the FLIPI, m7-FLIPI, POD24-PI, and POD24.