Treatments and Outcomes in Stage I Extranodal Marginal Zone Lymphoma in the United States

Simple Summary Extranodal marginal zone lymphoma (EMZL) is a rare disease commonly diagnosed at an early stage and remains localized for prolonged periods of time. This unique characteristic makes the use of local therapies, such as radiation therapy (RT), the preferred approach. Excellent results were previously reported implementing RT; however, majority of these studies included a small number of patients, and treatment patterns in the United States are presently unknown. Furthermore, EMZL may arise in various organs, and whether the survival is similar at different locations is unclear. In the present study, we assessed the Surveillance, Epidemiology, and End Results (SEER) database aiming to examine management and survival of localized EMZL. While differences in survival were observed by primary disease location, similar survival was observed in RT-treated stage I EMZL patients and general U.S. population matched by sex, age, and calendar year. Abstract A considerable number of patients with extranodal marginal zone lymphoma (EMZL) are diagnosed with stage I disease. Information on treatments and survival by primary location remains limited. We extracted data from the Surveillance, Epidemiology, and End Results (SEER) database to assess treatment, primary location, and survival of patients with stage I EMZL. Results show that 7961 patients met inclusion criteria. Observation (no treatment) was the most common approach (31%) followed by radiation therapy (RT, 23%). The median overall survival (OS) was 17.3 years (95%CI 16.3 to 18.3). Shorter survival was observed in patients with stage I EMZL compared to expected survival in a cohort derived from the general U.S. population matched by sex, age, and calendar year at diagnosis. However, similar survival was observed in RT-treated patients. We identified age ≥ 60 years (SHR = 4.00, 95%CI 3.10–5.15; p < 0.001), higher grade transformation (SHR = 4.63, 95%CI 3.29–6.52; p < 0.001), and primary lung EMZL (SHR = 1.44, 95%CI 1.05–1.96; p = 0.022) as factors associated with shorter lymphoma-specific survival (LSS). Conversely, primary skin location (SHR = 0.50, 95%CI 0.33–0.77; p = 0.002) was associated with longer LSS. Our results support the use of RT as the preferred approach in localized EMZL.


Introduction
Extranodal marginal zone lymphomas (EMZLs) arise in a wide variety of mucosal sites, most commonly stomach, ocular adnexa, lung, and salivary gland [1]. EMZL generally develops in a background of chronic inflammatory disorder. The cause of chronic

Methods
The SEER database of the National Cancer Institute (NCI) in the United States was used to identify patients with EMZL (International Classification of Diseases for Oncology (ICD-O-3) codes 9689/3) [24]. The study data were acquired from the April 2019 release of the SEER database using the SEER*STAT software version 8.3.6, selecting EMZL subjects diagnosed between 1995 and 2015 with follow-up through 2016 from 18 SEER registries encompassing the following states: California, Georgia, Connecticut, Illinois, Hawaii, Iowa, New Mexico, Washington, Utah, Kentucky, and Louisiana. Patients with unknown survival data, recorded survival time of zero (0) days, without histologic confirmation of EMZL, unknown cause of death, unknown primary site, non-stage I, age < 18 years, and EMZL not listed as a first primary malignancy were excluded from the analysis (Figure 1). Staging modality, indications for therapy, tumor recurrences, or treatments used at the time of progression are not recorded. A decision to pursue watchful waiting and refusal of treatment or antibiotic therapy is recorded as no initial therapy. Available data do not contain information about H. pylori eradication, RT dose, field, and fraction schedule. There is also no information on response to treatment, toxicities, or duration of remission. In the SEER database, rituximab is considered chemotherapy for diagnosis years 1999 to 2012, and it is excluded from chemotherapy variable after 2012. The treatment given is capturing treatment not only at diagnosis but throughout the course of EMZL. When chemotherapy is identified as "yes", SEER does not separate rituximab from conventional chemotherapy or the number of therapy lines. This study was conducted according to the guidelines of the Declaration of Helsinki, and because SEER data are publicly available, no approval was requested from the University of Miami IRB or Ethical Committee.

Statistical Analyses
Descriptive statistics were used to summarize demographic, tumor, and treatment characteristics. Treatment included chemotherapy, RT or surgery only, and combinations of two (chemotherapy and RT, chemotherapy and surgery, surgery and RT) or three treatment modalities (surgery, radiation, and chemotherapy). OS and lymphoma-specific survival (LSS) were the primary endpoints of this study. Subset analyses were performed for RT only and RT and surgery subgroups. OS was estimated by the Kaplan-Meier method; confidence intervals (CIs) for survival rate at specific times were calculated based on the log-log transformation method. The logrank test was used for comparison of survival curves. The Cox proportional hazards regression model was used to assess predictors of OS. Unadjusted and adjusted hazard ratios (HRs) and corresponding 95% CIs and p-values were reported. Competing risk analysis was used to evaluate LSS. The cumulative incidence method was used to estimate cumulative incidence rate (CIR) of lymphoma-specific death (LSD); cumulative incidence curves were compared using Gray's test, and the Fine and Gray subdistribution hazard model was used to assess the effect of demographic, tumor, and treatment-related variables on the risk of LSD. Unadjusted and adjusted subdistribution hazard ratios (SHRs) with corresponding 95% CIs and p-values are reported.
We also compared survival of this SEER patient cohort to expected survival in a cohort of the general U.S. population matched by sex, age, and calendar year at diagnosis. The expected survival curve in the matched U.S. population cohort was generated in R using the survexp.us population in the "survival" R package and the Ederer approach, which assumes complete follow-up of 20 years [25,26]. We used the one-sample logrank test to test whether survival in the SEER stage I EMZL cohort is similar to expected survival in the matched U.S. population cohort. The one-sample logrank test is a test of whether the standardized mortality ratio (SMR) is equal to 1 (similar mortality), that is, equivalent to test whether survival in the SEER cohort is similar to expected survival in a matched cohort. Thus, p > 0.05 indicates equality of mortality, or equivalently, similar survival in both cohorts. The standardized mortality ratio (SMR) is the ratio of the number of observed deaths to expected number of deaths in the study population under the assumption that the mortality rates for the study population are the same as those for the matched population [27,28].

Discussion
This study represents the largest analysis evaluating prognostic factors, treatment strategies, and treatment-related outcomes in stage I EMZL. Patients with EMZL experience excellent survival with a 10-year OS of 70.2%. However, we observed previously not reported shorter OS compared to matched-U.S. population irrespective of treatment modalities. This difference became not significant when RT was implemented, confirming this approach as the preferred treatment in stage I EMZL. Similar to prior reports in FL [22], we observed that RT was used in only 23% of patients indicating underuse of the most effective therapy. A previous study in FL demonstrated longer progression-free survival combining RT and chemotherapy [29]. We did not observe better LSS and OS survival in EMZL patients treated with this combination.
RT has been associated with excellent disease control across studies in EMZL [6,9,30]. However, we found its underutilization in our analysis.  [23]. EMZL-related clinical variables were not among the factors driving the decision not to use RT. Similar findings were reported in the National Cancer Base Analysis on early-stage FL. In that study, the following factors were associated with decreased use of RT: increasing age, female sex, African American race, increase comorbidity score, treatment at an academic/research program, stage II disease, presence of B symptoms, and more recent years of diagnosis [31].
In our study, RT was associated with excellent outcomes in all primary locations with a 10-year OS of 78.7% and lower risk of lymphoma-related death (SHR = 0.31). We identified age ≥ 60 years (HR = 4.00) and HGT (HR = 4.63) as risk factors for shorter LSS. Importantly, we did not observe significant survival differences in patients treated with an RT-based approach across all primary locations. Based on these results RT-based therapy should remain the frontline therapy in stage I EMZL, with exception of gastric EMZL with H. pylori infection, which should be first treated with antibiotics. In gastric EMZL, RT should be used only in patients who failed this treatment or in whom lymphoma is not associated with this infection. We acknowledge that RT may not be feasible in some anatomic locations such as GI-nongastric, liver, kidney, or salivary gland, where long-term toxicity may outweigh benefits. Further, previous reports on different outcomes in patients with stage I EMZL might result from differences in therapeutic approaches across the studies. Paucity of data exists regarding long-term survival in patients with localized EMZL compared to general population. A small retrospective study (n = 49) in Norway found similar median life expectancy for patients with EMZL compared to sex-and age-matched controls (79 vs. 83.6 years) [17]. Contrary to this observation, our analysis demonstrated for the first-time shorter survival in patients with stage I EMZL compared to general U.S. population which can be overcome by the implementation of RT as initial therapy. EMZL harbors a persistent risk of disease relapse requiring long-term follow-up. Our group previously demonstrated a continuous risk of lymphoma relapse with an estimated cumulative relapse of 31% at 10 years in patients with primary ocular adnexal EMZL [30]. Similarly, Raderer et al. reported a relapse rate of 37% with a median time of relapse of 47 months (range 14 to 307 months) after achieving initial complete response in EMZL [14]. However, the presence of effective second-line therapies in these patients may decrease patients' mortality resulting in similar survival.
The clinical course of patients with EMZL is characterized by an increased risk of HGT, which is an independent risk factor for poor outcomes. HGT is a rare event occurring in 4% to 8% of EMZL [32][33][34]. The incidence of HGT in this analysis was only 1.2%, however, associated with shorter OS and LSS in all multivariable models. Moreover, HGT was the most significant risk factor for shorter LSS. The true incidence of HGT in stage I EMZL remains unclear, but in a large study analyzing transformation, this event occurred in 7% of patients with a limited stage. In this study, the risk of HGT by 15 years was 4% compared to 16% in patients with advanced-stage disease [34]. These results resemble FL data where advanced-stage disease is associated with a higher risk of HGT; however, the incidence of transformation in EMZL is markedly lower [35,36].
Thyroid and gastric EMZL treated with RT have been associated with a lower risk of disease relapse and long-term survival. Tsang et al. did not observe relapses in these locations after a median follow-up of 5.1 years [37]. Similarly, the International Extranodal Lymphoma Study Group reported a 10-year freedom from treatment failure and OS of 88% and 70%, respectively, in patients with gastric EMZL [38]. We observed better survival in skin and thyroid primary locations; however, gastric primary location was associated with shorter survival, contradicting prior studies [5,10]. Factors influencing geographical differences are unclear, but we hypothesize that underuse of RT and H. pylori infection incidence are the explanation of such discrepancies. A population-based study carried out in the north of Italy demonstrated decreasing incidence in H. pylori-related gastric EMZL from 65% in the 1997-2001 period to 19% in the 2002-2007 period (p = 0.007) [39]. This decrease in association with H. pylori infection may result in different disease biology of gastric EMZL. In our study, 3031 patients with gastric EMZL were included. Of those patients, no therapy was recorded in the SEER database for 1543 (50.9%) patients. SEER database does not record the presence of H. pylori infection and H. pylori antibiotic therapy, but a significant proportion of these patients might be classified under the observation group. However, the exact number of these patients cannot be estimated from the SEER database. We also cannot assume that all these patients were treated with antibiotics, since survival of gastric EMZL stage I patients in the observation group was inferior compared to patients treated with RT. Absence of this data represents one of the major caveats in presented analysis. If this data would be available, we might better understand the findings of inferior outcome of gastric EMZL in the SEER database.
Further, the SEER database also does not collect data on diagnostic procedures used to establish the lymphoma stage. Extensive staging work-up was postulated by Raderer et al. in EMZL. The authors recommended that staging should include ophthalmologic examination, ear, nose, and throat investigation including sonography or magnetic resonance imaging of the salivary glands and lacrimal glands, endoscopies with multiple biopsies of the GI tract, computed tomography of thorax and abdomen, sonography of cervical, inguinal, and axillary lymph nodes, and bone marrow biopsy. By applying this approach, 25% of their patients with gastric EMZL had multiorgan involvement beyond the GI tract, and 46% of patients with extragastric EMZL also presented with advanced disease. Survival was not affected by extension of disease and this approach has not been generally accepted [40]. Furthermore, our group demonstrated the absence of prognostic value of staging bone marrow biopsy in patients with clinical/radiological localized EMZL treated with frontline RT [41]. In this study, staging bone marrow biopsy did not affect lymphomaspecific survival providing further rationale against extensive work up in EMZL.
The present study carries limitations of large epidemiological studies using the SEER database such as lack of information on staging modality, indications for therapy, use of antibiotic therapy in patients with gastric EMZL, H. pylori eradication, relapses, progressionfree survival, radiation doses, chemotherapy regimens, and treatment-related adverse events. Clinical and laboratory variables are not available in the SEER database preventing the confirmation of previously described predictors of increased risk of transformation.
Despite these limitations, the SEER database allows statistically powered analysis of a large number of patients with a rare and heterogeneous disease such as EMZL.

Conclusions
In the absence of randomized studies, this analysis supports the use of RT as the preferred approach in stage I EMZL. We do not have information on RT doses and approaches used in this cohort. Implementing the smallest required dose that can produce persistent local control is needed, but what this dose is remains unknown. Randomized trials of different RT doses are needed in EMZL patients to address this important question. In addition, studies are needed to discover novel, potentially less toxic therapies that will improve survival of these patients. Further studies addressing EMZL biology at different geographical locations are needed to elucidate survival characteristics at specific primary locations (e.g., gastric) observed in studies originating in U.S. and non-U.S. populations.
Supplementary Materials: The following are available online at https://www.mdpi.com/article/ 10.3390/cancers13081803/s1, Table S1. Univariable and multivariable analyses for OS and LSS in stage I EMZL that was treated with surgery and radiation, Table S2. Non-lymphoma-related cause of death. Figure S1. Overall survival by treatment strategy within EMZL primary location subgroups: gastric (A), skin (B), ocular adnexa (C), salivary gland (D), and lung (E).
Author Contributions: J.P.A., I.M.R. and I.S.L. conceptualized and designed the study, analyzed the data, and wrote the manuscript; J.A.F. and W.Z. collected and analyzed the data and wrote the manuscript. All authors have read and agreed to the published version of the manuscript.