Evaluation of Risk Factors for Histological Transformation in Patients with Marginal Zone Lymphoma (MZL): Results from a Multicenter Cohort Study

Introduction
Although MZL is an indolent B-cell lymphoma typically associated with excellent overall survival (OS), a subgroup of patients (pts) experience histological transformation (HT) to an aggressive lymphoma, significantly compromising their OS. Prior studies evaluating risk factors for HT largely lacked MZL-specific clinicopathological factors such as the presence of monoclonal (M-) protein, large B-cell (LC) content in the tumor, and Ki67 expression. Attainment of complete response (CR) to first-line therapy was associated with a lower risk of HT in a single-center retrospective study (Alderuccio et al., JCO, 2018), but it remains unclear whether attainment of CR mitigates other baseline risk factors and whether its effect might reflect guarantee-time bias. Using a large multi-institutional cohort of pts with MZL we investigated a comprehensive set of risk factors for HT and described its outcomes.

Methods
This multicenter, retrospective cohort study included adult pts with MZL treated between 2010 and 2020 at 10 US medical centers. Patients with HT recorded within 2 months after MZL diagnosis were excluded. We used penalized LASSO regression to select variables with the strongest association with HT in competing risk models for cumulative incidence function (CIF), reporting subhazard ratios (SHR) with 95% confidence intervals (CI). Candidate variables included age, sex, performance status, B symptoms, stage, bone marrow involvement, LDH, MZL subtype, gastric location, baseline hemoglobin, albumin, presence of M-protein, Ki67 and CD5 expression, LC tumor content, and complex cytogenetics. Missing data were accounted for using multiple imputation. Association of HT with CR was examined in models accounting for guarantee-time bias using landmark analysis. OS after HT was evaluated using stratification based on the time between MZL diagnosis and HT.

Results
The study included 917 pts (median age 63 years; 56% women; 59% stage 3/4) with 58% extranodal, 22% nodal and 20% splenic MZL. LDH was elevated at diagnosis in 24%, M-protein was present in 46%, LC in 12%, and high (≥20%, cutoff based on Epperla et al., ASH 2023) Ki67 expression in 22%. Forty pts (4.4%) experienced HT at a median of 29 months from diagnosis (range, 2.7 to 106). The CIF of HT was 3.9% (95%CI, 2.7-5.5) at 5 years and 9.0% (95%CI, 5.8-13.0) at 10 years.

Of all candidate variables, only high Ki67 expression (SHR=2.82; 95%CI, 1.28-6.20) and presence of LC at diagnosis (SHR=2.46; 95%CI, 1.07-5.66) were consistently selected by LASSO as associated with HT. However, these two variables were highly correlated, and only high Ki67 remained significantly associated with HT in a bivariate model (HR=2.53; 95%CI, 1.13-5.65). The CIF of HT was 3.7% (95%CI, 1.9-6.5) for tumors with low, and 9.9% (95%CI, 4.6-17.6) for those with high Ki67.

Among first-line treatments for MZL, the most common were rituximab monotherapy (41%), bendamustine with rituximab (21%), and radiation therapy (18%). Attainment of CR with first-line therapy was associated with a significantly lower risk of HT using landmark analysis up to 4 months from the start of therapy (n=853, SHR=0.47; 95%CI, 0.23-0.95). However, the association was lost with longer landmarks. Attainment of CR partly mitigated the increased risk associated with high Ki67 (using landmark of ≥4 months). Pts with high Ki67 who did not attain CR had a 5-year risk of HT of 23.1% (95%CI, 8.7-41.4).

Ten (25%) of all HT events occurred during the first year since MZL diagnosis. OS after HT did not significantly differ between pts with HT ≤1 year from diagnosis (3-year OS, 55.6%, 95%CI, 20.4-80.4) and those with later HT events (3-year OS 47.1%, 95%CI, 25.3-66.2; log-rank P=0.84).

Discussion
Among baseline clinicopathologic variables, only high Ki67 expression was consistently associated with an increased risk for HT in MZL. Therefore, molecular markers (TBL1XR1, CCND3, TP53 alterations) should be incorporated into future predictive models. The increased risk associated with high Ki67 was partly mitigated by attainment of CR with (predominantly anthracycline-free) first-line treatment, suggesting that a deeper remission might be meaningful for MZL pts, particularly those with high baseline Ki67. Pts with early and later HT had similar prognoses. OS after HT in MZL remains suboptimal even with contemporary therapies, and both preventing and treating HT remain important goals for future research.