The utility of prognostic indices, early events, and histological subtypes on predicting outcomes in non-follicular indolent B-cell lymphomas

2.1 Study population

Patients were selected from the Molecular Epidemiology Resource (MER), of the University of Iowa/Mayo Clinic, Lymphoma Specialized Program of Research Excellence (SPORE; CA97274), from September 2002 through December 2012. Full details of this prospective cohort study of Hodgkin Lymphoma (HL), Non-Hodgkin Lymphoma (NHL) and CLL/SLL outcomes have been previously published.14 Briefly, enrollment into the MER is offered to consecutive patients newly diagnosed with HL, NHL or CLL/SLL who are evaluated at Mayo Clinic Rochester or the University of Iowa. Patients were within 9 months of diagnosis, age 18 years or older, and US residents at time of diagnosis.

The MER collects clinical, pathologic, demographic and epidemiologic data at enrollment. Central pathology review in accordance with current WHO classification is performed by MER-affiliated hematopathologists. Patients are systematically followed every sixmonths for the first three years and annually thereafter. Events assessed are death, disease progression and retreatment which, initially patient reported, are verified through medical record review. For decedents, copies of the death certificate and/or medical records associated with death are obtained. Study physicians assign a cause of death using definitions developed for the Eastern Cooperative Oncology Group (ECOG) Intergroup trial 4494.15 This study is reviewed and approved by the Human Subjects Institutional Review Board at the Mayo Clinic and the University of Iowa. Written informed consent is obtained from all participants in accordance with the Declaration of Helsinki.

2.2 Statistical analysis

The IPI, FLIPI, and MALT-IPI scores were assigned according to the algorithms defined by Shipp et al., Solal-Celigny et al., and Thieblemont et al., respectively, as the sum of each of the included risk factors at the time of diagnosis.1, 2, 6

Overall survival (OS) was defined as time from diagnosis until death due to any cause. Event free survival (EFS) was defined as time from diagnosis until relapse or progression, unplanned retreatment of lymphoma after initial management, or death due to any cause. EFS indicators at predefined timepoints (ie, EFS at 12 months (EFS12) or EFS at 24 months (EFS24)) were defined based on EFS status at the indicated timepoint after date of diagnosis. Kaplan-Meier curves were used to display survival curves. Expected survival accounting for age and sex was generated in R using the general US (survexp.us) population as reference groups.16 Observed vs expected OS was plotted using a conditional approach and summarized using standardized mortality ratios (SMR) of observed to expected deaths.17-19 Concordance statistics (C-statistics) were generated for each respective index to measure their ability to discriminate EFS and OS. C-statistics quantify the area under receiver operator characteristic (ROC) curves, with C = 0.5 indicating the model is no better than chance and C = 1.0 indicating a perfect prediction rate. All analyses were performed using SASv9.4 and Rv3.4.2.

3.1 Patient characteristics and patterns of clinical presentation

Six-hundred twenty-three patients with newly diagnosed NFIBL were enrolled in the MER from 2002-2012. The following types of lymphoma were included in our analysis: extranodal marginal zone (EMZL; N = 258), gastric MALT lymphomas (MALT; N = 41), unclassifiable low-grade B-cell lymphoma (B-NOS; N = 121), ymphoplasmacytic lymphoma (LPL; N = 78), nodal marginal zone (NMZL; N = 51), and splenic marginal zone (SMZL; N = 74). Demographic and baseline clinical characteristics are demonstrated in Table 1. Gastric MALT lymphomas were considered separate from EMZL for analysis of baseline characteristics and initial treatment approaches, but thereafter were pooled with EMZL for survival analyses. Median age was 63 years (range 18-92 years) without notable variance among subtypes. For the majority of the analyzed subsets, gender was evenly distributed with males representing 52% of the population. There was, however, a stronger male predominance (74%) noted for patients with LPL.

Some notable differences were identified in clinical presentation among the subtypes that affected prognostic index scoring. Patients with EMZL presented more often with localized disease and thus presented with a lower Ann Arbor stage (60% stage I-II) as compared to NMZL (22% stage I-II) or SMZL (8% stage I-II). NMZL and SMZL also presented with the highest frequencies of abnormal LDH (24% and 32%, respectively), and anemia was most common among patients with SMZL or LPL (63% and 57%, respectively). Adverse performance status (≥ 2) occurred infrequently, but more often in patients presenting with NMZL and SMZL (8%; 12%).

3.2 Choices for initial management

Initial management and treatment strategies for patients with NFIBL are described in Table 1. Looking at the full cohort of 623 patients, 235 (38%) were initially followed with active surveillance, with no initial treatment given. Eighty-seven (14%) patients received radiation therapy, while 107 (17%) received alkylator therapy, and 86 (14%) received rituximab monotherapy. Only 5% of aggregate patients, and less than 10% in each subtype, received an anthracycline as part of their initial treatment strategy. Initial treatment patterns varied substantially based on lymphoma subtype. Patients with SMZL had the highest rates of surgical resection (n = 17, 23%) while patients with EMZL had the highest frequency of radiation as their initial treatment strategy (n = 65, 23%). Alkylator based chemotherapy was given with the highest frequency to patients with LPL (n = 24, 31%) or NMZL (n = 16, 31%).

3.3 Patient outcomes based on histology

Differences in baseline characteristics were reflected in patient's prognostic indices at the time of presentation. As demonstrated in Figure 1, essentially half of all patients presented with low-risk disease as assessed by FLIPI or IPI, while MALT-IPI differentiated patients into more equivalent groups of low, intermediate, and high risk. Only 8 of 623 patients presented with high risk IPI. Differentiated by histologic subtype, patients with SMZL most frequently presented with high risk disease while patients with EMZL most frequently presented with low risk categorization. 

At a median follow-up of 83 months (range 5-168), 264 (42%) patients had had an event and 127 patients (20%) had died. EFS and OS, stratified by performance index, are illustrated in Figure 1, and stratified by histological subtype in Figure 2. Eight year OS for all subtypes ranged from 66% (LPL) to 85% (MALT). Of the 127 deaths, 106 had records available to verify their cause of death. Among this subset the cause of death was lymphoma (n = 38) or therapy-related (n = 6) for 42% of patients, while 47 patients (45%) died from other causes. Death was attributed to subsequent malignancy in 13% of patients.

3.4 Patient outcomes based on prognostic indices

The IPI, FLIPI and MALT-IPI all demonstrated utility in differentiating outcomes among the cohort patients as assessed by EFS or OS based on initial prognostic index scores. For the outcome of EFS, the c-statistic was 0.593 for the IPI, 0.598 for the FLIPI and 0.612 for the MALT-IPI. All indices performed somewhat better when evaluating OS as the outcome, with c-statistics of 0.689 for the IPI, 0.683 for the FLIPI, and again MALT-IPI having the highest value of 0.714.

3.5 EFS12 and subsequent overall survival

Because of the prevalence of non-lymphoma mortality over time, we sought to explore more precise predictors of premature, lymphoma-related mortality. For the entire cohort, survival was slightly lower than that of the comparable age-matched and sex- matched United States general population, with a standardized mortality ratio (SMR) of 1.37 (95% CI 1.14-1.63; Figure 3A). Kaplan-Meier estimates of patients achieving EFS12 and EFS24 were 85% (95% CI 82-88%) and 77% (95% CI 74-80%), respectively. Patients who attained event-free survival for the first 12 months following diagnosis (EFS12) subsequently experienced overall survival that was not statistically different from the general population (SMR 1.19; 95% CI 0.95-1.46; Figure 3B). Conversely, patients who did not attain EFS12 experienced significantly worse subsequent overall survival than the general population (SMR 3.14; 95% CI 2.05-4.59). Similar outcomes were observed for patients who did not attain EFS24 (SMR 2.31; 95% CI 1.59-3.24).

4.1 Survival outcomes in NFIBL and the utility of existing prognostic indices

The three prognostic indices used in this report, the IPI, FLIPI, and MALT-IPI harbor subtle differences in their included risk factors and stratification schemes. Briefly, all three indices assign a point for elevated LDH or stage III-IV disease, as well as for elevated age (>60 years for the IPI and FLIPI, and >70 for the MALT-IPI). The IPI also assigns a point each for ≥2 extranodal sites or ECOG PS >2, while the FLIPI assigns a point each for ≥4 nodal sites or hemoglobin <12 g/dL.

Each index has been previously evaluated in subsets of NFIBL. Among 144 German patients with MZL (96 extra-nodal, 32 nodal, and 16 splenic), the FLIPI was able to distinguish cohorts based upon 5 year PFS and OS, for patients with extra-nodal or splenic subtypes, but not nodal subtypes.22 This was valid only when such patients were dichotomized to either low/intermediate or high risk groups. In a subsequent study of 143 Austrian patients with EMZL, both IPI and FLIPI scores significantly correlated with time to relapse, but the FLIPI was again able to differentiate patients when dichotomized as in the German study.23 The MALT-IPI has been validated in two different studies of MALT lymphomas, although the most recent of these required high-risk patients be merged with intermediate-risk patients in order to discriminate outcomes.24, 25

While this manuscript was in preparation, another retrospective study was published, which examined the occurrence of high-grade transformation in MZL and its effects on subsequent PFS and OS.26 The authors demonstrated that dichotomized IPI, FLIPI, and MALT-IPI scores are predictive of shorter PFS and OS. Our study contributes to the field by demonstrating that several non-follicular indolent lymphomas can be considered in aggregate for prognostic purposes, and by validating EFS12 as a key early endpoint that identifies those patients most in need of more aggressive therapies. Unlike the studies described above, the FLIPI was able to distinguish OS between low, intermediate, and high risk groups with no need for dichotomization. This may reflect the larger panel, longer follow up, or the expanded analysis of several subtypes in addition to MZL. In our series, IPI effectiveness is limited by the relative infrequency of poor performance status (4%) and multifocal extranodal disease (8%) resulting in very infrequent high risk scores, while FLIPI is limited primarily by infrequent multifocal nodal disease (9%). MALT-IPI performs optimally in this collection of patients with a robust distribution of prognostic scores across risk categories. Given the relatively marginal differences in C-statistics between the 3 indices in this study, each appears to have comparable utility.

Survival, whether measured by EFS or OS, was found to be similar regardless of subtype, with somewhat worse trends for splenic MZL, LPL or Low-grade B NOS subtypes. As a whole, diagnosis of NFIBL conveys a predicted overall survival only slightly distinguishable from the general population, likely reflecting the later age at diagnosis and typically indolent course.

4.2 Existing prognostic indices and alternative prognostic factors

Independent predictive factors in various NFIBL subsets have been previously proposed. A French series identified the presence of a serum M-protein or an immunological event as prognostic for shortened PFS in splenic MZL, while noting those same factors plus elevated ß2-microglobulin and leukocytosis were predictive for shortened OS.27 Additionally, expression of proliferative marker Ki67 or the plasmacytic differentiation marker IRF4 (Mum1), have been found to correlate with a poorer prognosis among patients with lymphoplasmacytic or marginal zone lymphomas.28 Additionally, p53 abnormalities, while rare, have been found to be associated with poor survival among patients with splenic marginal zone lymphomas.29

In addition, prognostic models unique to marginal zone subsets have been previously described. A marginal zone prognostic index (MZLIPI) incorporating nodal disease status, ECOG PS, and stage, was applied to a cohort of 205 Korean patients with non-gastric marginal zone lymphoma, and successfully discriminated PFS and OS across low, intermediate, and high risk groups.30 The Intergruppo Italiano Linformi proposed a three parameter model of hemoglobin <12 g/dL, serum albumin <3∙5 g/dL and LDH > ULN, which accurately predicted cause-specific survival among 309 patients with splenic marginal zone lymphoma.7 Future efforts may focus on comparing the performance of the above described prognostic indices with the MALT-IPI in NFIBL possibly with incorporation of novel prognostic factors, although our data suggests value in reassessing prognosis at the 12-month mark, regardless of estimates at the time of diagnosis. Other approaches include the incorporation of gene expression profiling, which has shown promise in follicular lymphoma.31

4.3 Utility of the category “B-NOS”

The differential diagnosis of the NFIBLs requires integration of clinical, morphological, immunophenotypic and genetic data, yet even then diagnostic reproducibility is poor. There is only 56% concordance for LPL, and 63% for NMZL, among experts in the Non-Hodgkin's Lymphoma Classification Project.1 Small B-cell lymphomas with plasmacytic differentiation frequently present diagnostic challenges,32, 33 and marginal zone lymphoma may be mimicked closely by a variety of indolent B-cell neoplasms.34 This appropriately leads to use of terms such as “low-grade B-cell lymphoma not otherwise specified (NOS)” or “small B-cell lymphoma with plasmacytic differentiation” as a diagnosis. This entity is reported in modern series with an incidence of 1-3% of NHL.35, 36 In our series, low-grade B-cell lymphoma NOS represents 18% of the NFIBL's and 1∙5% of NHL cases enrolled in the MER. Notably the median age is slightly higher for these patients than in the remainder of the NFIBL cohort, and initial treatment is deferred in 60% of cases. Distribution of FLIPI, IPI, and MALT-IPI scores is similar to other NFIBL patients, but overall rates of death were slightly higher for this category. Among the classifiable LGL's, these patients have clinical presentations most closely resembling splenic MZL, except for the use of splenectomy in management.

4.4 Strengths and weaknesses

Strengths of this study include the prospective cohort design of consecutively enrolled a.d newly diagnosed lymphoma patients. Also included are central pathology review, systematically collected clinical data, virtually complete follow-up of the cohort for disease progression and death, and medical record validation of these events. The collection of 623 patients is a large series, and all patients were managed in the current immunochemotherapy era - largely prior to incorporation of b-cell receptor targeting agents. The major limitations include the observational design, treatment and clinical follow-up based on routine practice, without prescribed re-biopsy criteria. Our follow-up, while longer than most previous series at a median of 83 months, remains modest for lymphoma subtypes with a long natural history, and continued follow-up will be necessary to understand long-term outcomes. While the MER is not a population-based sample, the vast majority of our patients are from the local region, and the clinical characteristics of our patients parallel those of population-based data with the exception of fewer very elderly patients.