INTRODUCTION
Primary Sjögren syndrome (pSS) is a chronic, slowly progressive autoimmune disease characterized by lymphocytic infiltration of the exocrine glands affecting preferentially the salivary and lacrimal glands, but frequently involving other exocrine sites too.34,41,49 The spectrum of the disease, which has alternatively been termed "autoimmune exocrinopathy" or "autoimmune epithelitis," ranges widely from minimal local symptoms of the eyes and oral mucosa to systemic involvement and the development of malignant lymphoma, the latter being the most worrisome complication of the disease.25,26,35,43 pSS has been estimated to affect 0.09%-0.6% of the general population.2,5,7,12,47
The disease displays a similarly vast array of hematologic abnormalities. In most cases such findings are mild and draw little attention. However, there is a plethora of reports suggesting that pSS hematologic manifestations form a spectrum ranging from mild asymptomatic laboratory abnormalities to life-threatening manifestations, such as the development of lymphoma.
Several studies have consistently demonstrated an enhanced risk of lymphoma development in the course of pSS. In fact, the relative risk of non-Hodgkin lymphoma (NHL) in SS, compared to the general population, is to our knowledge the highest among all autoimmune diseases studied to date, ranging from 6.1- to 44.4-fold, with some controversy as to whether the primary form of the disease is associated with a greater risk than the secondary form.15,26,27,29,32,42,45,48 Given that malignant lymphoma is the only complication with a considerable impact on survival in this otherwise benign disorder, it is compelling to identify certain aspects of disease, either clinical or laboratory, that, when present at diagnosis, can accurately predict subsequent progression to malignant lymphoma.26,41,46 Such aspects should ideally constitute information readily available in clinical practice rather than extractable through sophisticated and cumbersome laboratory investigations, so that they could be used as a screening tool in all pSS patients.
Indolent extranodal marginal zone B-cell lymphoma (MZBCL) of the mucosa-associated lymphoid tissue (MALT) has been indicated as the most common lymphoma subtype complicating the course of pSS patients.40,42,52 Nevertheless, recent research has suggested an association between pSS and diffuse large B-cell lymphoma (DLBCL). Depending on the cohort studied, DLBCLs may account for a small yet considerable number of NHL cases, or, less often, may be the predominant lymphoma subtype in pSS.15,42,45 Differences in both disease severity and prognosis between patients with MALT lymphoma and those with DLBCL warrant the identification of risk factors that can predict the development of distinct lymphoma subtypes. The application of such risk factors in clinical practice would enable closer follow-up of patients at higher risk for the development of an aggressive lymphoma. In addition, as NHL subtypes arise from different stages of lymphocyte differentiation, specific associations between certain pSS phenotypes and corresponding lymphoma subtypes could offer valuable insight into the mechanisms of lymphomagenesis.
In the current study we investigated the prevalence of hematologic abnormalities in pSS, their correlations with various aspects of the disease, and their evolution during the course of the disease. We also hypothesized that the presence of glandular and/or extraglandular manifestations of the disease can be combined with laboratory information easily obtainable at diagnosis, such as that included in standard complete blood counts or cryoglobulin assays, to provide prognostic information regarding the risk of lymphoma development.
PATIENTS AND METHODS
Study Cohort
The medical records of all patients diagnosed with pSS and/or followed-up consecutively in the Department of Pathophysiology, Medical School of Athens, a University-affiliated internal medicine department based at a tertiary-level general hospital, from 1981 through May 2008 were reviewed, and the cases that fulfilled the American-European Consensus Classification (AECC) criteria for pSS were considered eligible for the study.49
Patient follow-up extended from the patient's first visit in our department to the time of lymphoma diagnosis or to the patient's last follow-up visit. All patients in our cohort had been followed at regular intervals of 6-12 months or more frequently if required by their clinical condition. Both incident and prevalent cases were considered; incident cases had a diagnosis made in our department during the first or on a subsequent patient visit, whereas prevalent cases had been diagnosed with pSS before their first visit, and their diagnosis was confirmed upon referral. Separate analysis of incident and prevalent cases did not reveal any difference (data not shown); therefore, only aggregate results are presented. Limited data on 315 patients included in this report have been previously published.21
Data Collection
All data included in the analyses were extracted from the patient medical records and comprised information that had been collected meticulously at every patient visit by successive teams of experienced rheumatology fellows or internal medicine residents, always with oversight by an attending rheumatologist with a research interest in SS and autoimmune disorders. The inclusion of all clinical data of importance was ensured by the use of a preformed questionnaire, while additional information may have been retrieved and recorded according to the physicians' discretion.
Extraction of information from the case files was performed by an internal medicine resident (EB), and all extraction was reviewed by a hematologist experienced in pSS-related lymphoproliferation (MV) and a rheumatologist experienced in pSS management (HMM). Data extracted from medical records included the following: incident or prevalent case, sex, date of first visit, age at first visit, date of lymphoma diagnosis, lymphoma subtype and stage, and date of last visit. In addition, the following features of disease were included in analyses: oral symptoms, ocular symptoms, parotid gland enlargement, ocular signs (Schirmer test, rose bengal staining, tear break-up time), minor labial salivary gland biopsy, palpable purpura, lymphadenopathy, splenomegaly, fatigue, fever, arthritis, Raynaud phenomenon, pulmonary involvement22 (dyspnea, small airway disease, interstitial pneumonitis), renal involvement6 (interstitial nephritis, glomerulonephritis, renal tubular acidosis, proteinuria), hepatic involvement1 (autoimmune hepatitis, primary biliary cirrhosis, sclerosing cholangitis), peripheral nervous system involvement19 (peripheral neuropathy, sensory or motor deficits, cranial palsies), autoimmune thyroiditis, and results of laboratory tests obtained routinely or according to the physician's discretion within the first 6 months of follow-up (hemoglobin; white blood cell [WBC] count and differential; platelet count; rheumatoid factor; antinuclear antibodies; anti-Ro/SSA, anti-La/SSB antibodies; complement C3 and C4 levels; serum protein electrophoresis, quantification of serum immunoglobulins and immunofixation; and cryoglobulins).16,25,49 Hepatitis C was actively excluded in all patients. Histologic reports of patients diagnosed with malignant lymphoma before 2001 were reviewed, and all lymphoma cases in our cohort were classified according to the World Health Organization classification for tumors of the hematopoietic and lymphoid tissues.23
Statistical Analysis
Descriptive data are presented as frequencies for categorical variables, or means ± standard deviation (SD) and medians, range of observations (minimum-maximum) for continuous variables. Proportions are presented as frequencies with their corresponding 95% confidence intervals (CI).36 To compare frequencies of categorical variables we used the Pearson chi-square test, or the Fisher exact test when appropriate. We used the Mann-Whitney test to compare mean age and disease duration between lymphoma subtypes.
The lymphoma-free survival time was measured as the time since pSS diagnosis, and patients were censored at the time of either lymphoma diagnosis or last follow-up visit. Time-to-event analyses were conducted using the Kaplan-Meier method.24
Cox regression analysis with a backward stepwise selection method by likelihood ratio criteria was applied to identify potential risk factors for lymphoma development.11 We tested proportional hazards assumptions with the methods proposed by Grambsch and Therneau.20 Features of the disease previously identified as risk factors for the development of lymphoproliferative disease or as predictive of mortality in pSS were included in the multivariate analysis as possible prognostic variables; the major hematologic manifestations of the disease were also included in the analysis as possible risk factors. Thus, the following parameters at diagnosis were taken into consideration in the multivariate model: parotid gland enlargement, palpable purpura, lymphadenopathy, splenomegaly, anemia (hemoglobin <12 g/dL and <13.5 g/dL for female and male patients, respectively), neutropenia (neutrophil count <1500/μL), lymphocytopenia (lymphocyte count <1000/μ/L), thrombocytopenia (platelet count <140,000/μL), cryoglobulinemia, low C3 levels (C3 <75 mg/dL), and low C4 levels (C4 <10 mg/dL). These analyses were also repeated to identify risk factors for specific lymphoma subtypes (MZBCLs vs. non-MZBCLs).
To construct a prognostic model for the development of lymphoma in the course of pSS, we split the cohort into 2 groups based on the number of risk factors present at diagnosis: Group A, no risk factors ("low-risk" group); and Group B, at least 1 risk factor ("high-risk" group). We plotted Kaplan-Meier lymphoma-free survival curves for both groups, and compared them using the log-rank test. The HR with 95% CI was calculated; we also used a bootstrap approach to estimate the variance of the HR.14
In an attempt to validate our model, we measured its intrinsic prognostic information using the statistics proposed by Altman and Royston3 specifically for quantifying the performance of a prognostic model. These authors proposed the consideration of an index of separation, PSEP, which is calculated as the difference (pworst − pbest); pworst expresses the probability for a "high-risk" patient to develop lymphoma, while pbest expresses the probability of lymphoma in a "low-risk" patient. When just 2 prognostic groups are considered, pbest and pworst are related to the familiar concepts of positive predictive value (PPV) and negative predictive value (NPV) by the relations pworst = PPV and pbest = 1 − NPV. Thus PSEP = PPV + NPV − 1. The importance of the prognostic information contained in a model is supposed to correlate with the PSEP value; thus, the greater the PSEP the more likely the model can in fact serve as a prognostic tool. For demonstration purposes, we also applied the prognostic model proposed by Ioannidis et al21 to our cohort.
To cope with a problem inherent in retrospective studies, that is, patients with missing data, we conducted analyses in 3 ways: (a) we excluded patients with missing data, (b) we included them assuming that they did not carry the risk factors in question, (c) we created multiple imputed data sets, we analyzed the statistical model on each of them, and combined the analyses to yield the final set of results.4 Because all analyses yielded qualitatively similar estimates (data not shown), we present only the results of the analyses that excluded patients with missing data. For all data-points collected, completeness of data exceeded 90%.
Analyses were conducted using SPSS for Windows, version 13.0 (SPSS, Chicago, IL) and STATA, version 10/SE (StataCorp, College Station, TX). Statistical significance was defined as a p value of less than 0.05 for all comparisons; p values were 2-tailed.
RESULTS
Study Population
Five hundred thirty-six pSS patients followed consecutively in our department fulfilled the AECC criteria and were thus included in the study.49 Overall, median follow-up was 31 months (range, 0-317 mo; mean, 50.9 ± 55.8 mo). As expected, there was a preponderance of women in the cohort, and the male:female ratio was 1:12. The major clinical characteristics of our cohort at presentation are shown in Table 1.
;)
TABLE 1: Clinical Characteristics of 536 Patients With pSS at Diagnosis
Hematologic Features at Diagnosis
The hematologic manifestations of pSS already present at diagnosis and the spectrum of abnormal laboratory parameters during the follow-up are summarized in Table 2. Notably, the most common hematologic abnormalities encountered in pSS patients were anemia, especially of chronic disease, and hypergammaglobulinemia.
TABLE 2: Hematologic and Immunologic Characteristics of 536 pSS Patients at Time of Diagnosis and During Follow-Up
One hundred fifty-three (29%; 95% CI, 28.5%-32.6%) patients had anemia at diagnosis. In 40 of these patients anemia was secondary to a number of causes: β-thalassemia trait (n = 16), drug toxicity (n = 11), iron deficiency (n = 5), and vitamin B12 and/or folate deficiency in the context of atrophic gastritis (n = 2) and postgastrectomy (n = 1). Three cases of autoimmune hemolytic anemia were detected, and in 2 cases, anemia was the consequence of renal failure. In another 2 cases, the cause of anemia had not been thoroughly investigated. In the remaining 111 patients anemia could be attributed to no other cause than chronic inflammation.
The hemogram of 75 (14%; 95% CI, 4.2%-17.2%) patients at diagnosis revealed a WBC count <4,000/μL. In 24 of these patients, leukopenia was induced by medications. The WBC differential count disclosed 11 cases of neutropenia and 35 cases of lymphocytopenia. Thrombocytopenia was detected in only 20 patients (3.7%; 95% CI, 2.3%-5.7%), of whom 8 were on medication known to cause a decrease in the platelet count. Univariate analysis between cytopenias and glandular manifestations revealed a statistically significant correlation between lymphocytopenia and parotid gland enlargement (p = 0.002), as well as between neutropenia and xerostomia (p = 0.019). Furthermore, anemia, lymphocytopenia, thrombocytopenia, hypergammaglobulinemia, the presence of monoclonal serum proteins, and cryoglobulinemia correlated significantly with the presence of extraglandular symptoms such as palpable purpura and splenomegaly.
Serum protein electrophoresis and serum immunoglobulin quantification demonstrated the presence of hypergammaglobulinemia (γ-globulin fraction >22% of total serum proteins or γ-globulins >20 g/L) in 141 (26.3%; 95% CI, 22.6%-30.3%) patients, while only 16 (3%; 95% CI, 1.7%-4.8%) patients presented with hypogammaglobulinemia (γ-globulins <8 g/L). Monoclonal immunoglobulins were detected in 21 (3.9%; 95% CI, 2.4%-5.9%) patients. Patients presenting with hypergammaglobulinemia had a higher prevalence of peripheral nervous system involvement (p = 0.045), palpable purpura (p < 0.001), splenomegaly (p < 0.001), and lymphadenopathy (p = 0.003), while those with a detectable monoclonal component at diagnosis were more likely to have presented with symptoms or signs indicative of pulmonary involvement (p = 0.035), as well as with palpable purpura (p < 0.001), lymphadenopathy (p = 0.004), splenomegaly (p = 0.001), or histologically proven vasculitis (p < 0.001).
Forty-four (8.2%; 95% CI, 6.0%-10.9%) of the patients in the cohort had detectable cryoglobulins at diagnosis. Cryoglobulinemia was polyclonal (type III) in 19 cases and mixed monoclonal (type II) in 16 cases, whereas in 9 cases the type of cryoglobulins had not been determined. No cases of monoclonal (type I) cryoglobulinemia were detected. Cryoglobulinemia at presentation was associated with parotid gland enlargement (p = 0.013), Raynaud phenomenon (p = 0.001), palpable purpura (p < 0.001), histologically proven vasculitis (p = 0.003), lymphadenopathy (p = 0.006), and splenomegaly (p < 0.001).
Other statistically significant correlations between the major hematologic manifestations of pSS patients at diagnosis and clinical and immunologic aspects of their disease are presented in Table 3.
TABLE 3: Correlations Between Major Hematologic Manifestations of pSS Patients and Clinical and Immunologic Characteristics*
Hematologic Features During Follow-Up
Eighty-eight (16%) patients with a normal hemoglobin value at diagnosis developed anemia at some point during follow-up. In 14 of these patients a cause for the anemia could be detected, such as drug toxicity (n = 7), autoimmune hemolytic anemia (n = 3), renal failure (n = 3), and gastrectomy (n = 1), whereas the remaining 74 cases were attributed to the disease itself.
The occurrence of leukopenia proved to be relatively common during the course of the disease, with the WBC count of 64 (11.9%) patients falling below the cutoff limit of 4000 cells/μL. A considerable number of these cases (n = 17) was induced by medications. Neutropenia and lymphocytopenia occurred in 6 (1.1%) and 14 (2.6%) patients with a normal differential count at diagnosis, respectively.
Hypergammaglobulinemia was found in 53 (9.9%) patients with a normal baseline electrophoresis, while a low concentration of γ-globulins was detected in an additional 15 (2.8%) patients, thereby doubling the cumulative incidence of hypogammaglobulinemia.
Cryoglobulinemia developed during follow-up in 19 (3.5%) patients with no detectable cryoglobulins at diagnosis. Cryoglobulins were of type II and III in 8 and 9 cases, respectively, whereas the type of cryoglobulinemia had not been assessed in 2 patients.
Lymphoma
Forty cases of malignant lymphoma occurred in our cohort. The median age at lymphoma diagnosis was 54 years (interquartile range, 43-61 yr), while the mean time between pSS diagnosis and the development of lymphoma was 6.8 years (SD ± 5.9; range, 0-26.2 yr).
Only 1 of the 40 patients was male. Thirty-nine of the lymphomas were of the non-Hodgkin subtype, 38 of which were of B cell origin. Review of the histologic reports revealed 26 cases of MZBCL (of which 21 were of the extranodal MALT type, and 5 were of the nodal MZ type), 7 cases of DLBCL, 3 cases of lymphoplasmacytic lymphoma, 1 case of follicular lymphoma, 1 case of polymorphic B-cell lymphoma, 1 case of peripheral T-cell lymphoma, and 1 case of Hodgkin disease. The MZBCL:DLBCL ratio was 3.7:1. Median age at MZBCL diagnosis was 51 years (range, 30-60 yr; mean, 50.2 ± 11.1 yr) and median time from pSS diagnosis to MZBCL development was 1.5 years (range, 0-26 yr; mean, 4.5 ± 6.0 yr). In contrast, median age at DLBCL diagnosis was 57 years (range, 41-75 yr; mean, 58.4 ± 15.1 yr; Mann-Whitney test p value = 0.23), and median time from pSS diagnosis to DLBCL development was 7 years (range, 5-15 yr; mean, 8.3 ± 3.7 yr; Mann-Whitney test p value = 0.06).
According to histologic reports, the lymphoma was indolent in the majority of cases (30 of 40 total cases, 75%; 95% CI, 58.8%-87.3%) and only 9 patients were diagnosed with aggressive disease. The patient with Hodgkin disease had the nodular sclerosis subtype. Staging at the time of diagnosis disclosed disease limited to the primary site (stage I and II) in 25 cases, while in 15 patients the disease had already spread (stage III and IV) by the time it was diagnosed. The main clinical and histologic features of the lymphoma cases are presented in Table 4.
TABLE 4: Clinical and Histologic Features of 40 pSS Patients who Developed Lymphoma
Predictors of Lymphoma Development
In multivariate analysis, neutropenia (HR, 8.97; 95% CI, 1.10-73.30; p = 0.04), cryoglobulinemia (HR, 2.91; 95% CI, 1.15-6.44; p = 0.008), splenomegaly (HR, 3.97; 95% CI, 1.49-10.62; p = 0.006), lymphadenopathy (HR, 2.62; 95% CI, 1.15-5.94; p = 0.021), and low C4 levels (HR, 3.31; 95% CI, 1.35-8.12; p = 0.009) were independent risk factors for the development of lymphoma.
Predictors of Lymphoma by Subtype
Multivariate analysis retained cryoglobulinemia (p < 0.0001), neutropenia (p = 0.001), low C4 levels (p = 0.052), lymphadenopathy (p = 0.024), and splenomegaly (p = 0.012) as independent predictors for the development of MZBCL. Conversely, only lymphocytopenia (p = 0.044) was retained as a risk factor predicting the development of non-MZBCL, the majority of which were DLBCL.
Prognostic Model
We constructed a prognostic model for the development of lymphoma in pSS by allocating patients into 2 groups in the following way: patients with none of the risk factors for lymphoma suggested by our multivariate analysis were designated as Group A (low-risk), while patients with at least 1 risk factor present at diagnosis formed Group B (high-risk). The Kaplan-Meier curves comparing lymphoma-free survival are presented in Figure 1. The probability of a high-risk patient developing lymphoma during follow-up was 20.6%, compared with 3.6% for a low-risk patient. Thus, the patients in Group B had about a 5.4-fold (HR, 5.43; 95% CI, 2.79-10.64; p < 0.0001) higher risk of lymphoma compared with the patients in Group A; the model had a PSEP value of 0.17. Bootstrap estimation of the HR variance did not alter the results. In our cohort, the proportion of patients who developed lymphoma by number of risk factors was 3.62%, 11.96%, 34.78%, 80%, and 100%, for patients with 0, 1, 2, 3, and 4 risk factors, respectively. No patient had all 5 risk factors. The 3-factor model (low C4 levels, palpable purpura, and parotid gland enlargement) proposed by Ioannidis et al21 had a PSEP value of 0.115 when applied in our cohort, with 14.7% of the high-risk patients and 3.2% of the low-risk patients developing lymphoma during follow-up.
FIGURE 1: Kaplan-Meier lymphoma-free curves drawn for each risk group in the cohort. Patients presenting with no clinical or laboratory features suggesting adverse prognosis at diagnosis make up Group A, while patients with at least 1 of the risk factors identified in the multivariate analysis (neutropenia, cryoglobulinemia, splenomegaly, lymphadenopathy, or low C4 levels) form Group B. The 2 groups display distinct probabilities for the development of lymphoma; Group B patients are 5.4 times more likely to develop lymphoma compared to Group A patients. Thus, patients with any risk factor at diagnosis should be regarded as high-risk patients and followed more closely than patients with none.
DISCUSSION
In the current retrospective study, we focused on delineating the range of hematologic abnormalities encountered in pSS and their correlations with the disease's protean manifestations. Furthermore, we attempted to illuminate the phenotypic characteristics of pSS most closely associated with the development of NHL, not only aiming to stratify patients according to their risk of such an adverse outcome, but also hoping to unveil some aspects of the complex pathways mediating lymphomagenesis.
A variety of hematologic abnormalities accompanies pSS and may be already present at diagnosis or may evolve during the course of the disease. In keeping with previous studies, we report the predominance of hemocytopenias, usually mild, and the increased prevalence of hypergammaglobulinemia.39 Anemia of chronic disease was by far the most common hemocytopenia encountered and was associated with systemic involvement and circulating antibodies, possibly reflecting, thereby, a state of generalized inflammation rather than disease limited to glandular sites. The high prevalence of hypergammaglobulinemia and its strong correlation with circulating autoantibodies are congruent with the polyclonal B-cell activation implicated in the pathogenesis of the disease (Table 3).
The well-documented close association between pSS and the development of lymphoma is reflected in the current study: 7.3% of our patients received an NHL diagnosis.15,32,42,52 The majority (65%) of lymphomas were MZBCLs, while DLBCLs accounted for almost 18%; the remaining cases were of miscellaneous subtypes that could not be grouped into any major category. This distribution of lymphoma subtypes is in accordance with previous pooled analyses and case-control studies assessing the risk of lymphoma by subtype in various autoimmune conditions, but contradicts others that have identified a preponderance of DLBCLs in pSS.15,38,42,45,50,52 Such inconsistency among studies may be indicative of a diversity of triggering stimuli of lymphomagenesis in different geographical regions, although no such difference in the distribution of lymphoma subtypes was evident in a large, multicenter European analysis of pSS-associated NHL cases.50 Alternatively, the newly reported association of DLBCL with pSS may reflect the emergence of a more slowly evolving type of lymphoma in patients who have been followed for longer periods. Irrespective of the possible interpretations, our findings contribute to the growing body of evidence suggesting that the association of pSS with NHL is less subtype-specific than previously described; in line with other studies, we report that the pSS patients who eventually develop lymphoma are 4 times more likely to have an MZBCL diagnosis than a DLBCL diagnosis.15,42 Thus, while the association between pSS and MZBCL, especially of the MALT type, is indisputable, the increased occurrence of DLBCL in pSS patients cannot be overlooked. The mechanisms underlying the pathogenesis of DLBCL in the context of pSS have not yet been defined; however, it has been demonstrated that a proportion of DLBCLs arises from the transformation of indolent lymphomas, and that MZBCLs have the potential to progress into a more aggressive histologic type.18,53 The identification of an aberrant somatic hypermutation mechanism in DLBCLs arising from MALT lymphomas and to a lesser extent in MALT lymphomas offers additional molecular evidence for the linkage between these 2 NHL-subtypes.13 In light of these facts, the emergence of DLBCL in the course of pSS can be considered as equally attributable either to the transformation of the prevalent MZBCLs or to the de novo development of DLBCLs.
In the current cohort, the risk of lymphoma development was maintained at high levels even 10 years after pSS diagnosis (Figure 1), in accordance with previous studies.15,32,45 The sustained risk of lymphoma over time is an argument in favor of the hypothesis that the chronic stimulation of RF-secreting B cells in pSS plays a pivotal role in generating and maintaining the multistep process resulting in lymphomagenesis.31
Our multivariate analysis suggests that the development of lymphoma in the course of pSS can be predicted by the presence of any of the following features of the disease at diagnosis: neutropenia, cryoglobulinemia, splenomegaly, lymphadenopathy, or low C4 levels. Almost all of these disease characteristics have been identified in the past as risk factors for lymphoma in pSS. However, to the best of our knowledge, no study to date has identified risk factors for NHL subtypes in the course of the disease. While we found the development of non-MZBCLs (most of which were DLBCLs) to be predicted by the presence of lymphocytopenia at diagnosis, the risk factors for MZBCLs were shown to overlap greatly with the risk factors for lymphomas overall. Such an overlap is barely surprising, given the preponderance of MZBCLs in our cohort.
Neutropenia is a relatively uncommon hematologic manifestation in pSS patients, but is well documented through a number of case reports.10,17 In the current cohort, only 2.1% of patients presented with a low neutrophil count; the increased prevalence of neutropenia in pSS reported elsewhere is likely attributable to the high cutoff level used to define this entity.8 Both peripheral antibody-mediated destruction and immune-mediated inhibition of neutrophil production have been implicated as possible mechanisms for the occurrence of neutropenia in pSS.17 The current univariate analysis demonstrated a statistically significant correlation between neutropenia and splenomegaly. Nonetheless, basic research is warranted to delineate the pathogenesis of neutropenia in the context of the disease. Although to our knowledge neutropenia has never before been associated with the development of NHL in pSS patients, it has been identified as 1 of the laboratory factors able to predict the transformation of IgM-related disorders into lymphoid malignancy.33 The term "IgM-related disorders" was introduced at the Second International Workshop on Waldenström's Macroglobulinemia37 and encompasses patients with clinical features attributable to the presence of an IgM monoclonal protein, such as symptomatic cryoglobulinemia or peripheral neuropathy, but with no overt evidence of lymphoma. Patients with pSS and mixed monoclonal cryoglobulinemia or evidence of a monoclonal component clearly fit this distinct entity, which demonstrates a probability of evolution to malignant lymphoma of up to 15% at 5 years.9,44
Mixed monoclonal cryoglobulinemia has been demonstrated to confer an increased risk of progression to lymphoma to the affected pSS patients, heralding the onset of NHL by months to years.51 The literature is replete with studies addressing the pathogenetic role of cryoglobulinemia in the development of overt lymphoma. Mixed cryoglobulinemia has been thought to demonstrate the transition from polyclonal B-cell hyperactivity, a hallmark of SS, to monoclonal expansion of B cells producing an IgMκ immunoglobulin with RF activity.31
To our knowledge, Ioannidis and colleagues21 were the first to postulate that specific clinical features at diagnosis of pSS (namely, the presence of parotidomegaly, palpable purpura, and low C4 levels) distinguish high-risk patients from patients with an uncomplicated disease course. Based on a large retrospective study held at 2 university institutes in Greece, they concluded that the presence of low C4 levels, palpable purpura, or parotid gland enlargement at diagnosis could predict the development of any lymphoproliferative disorder including lymphoma.21 It is very likely, though, that the palpable purpura present in the high-risk patients of the aforementioned study reflected their cryoglobulinemic status; the latter, however, had not been included as a possible predictive variable in the analyses. When both variables were included in our multivariate analysis, palpable purpura could not be retained as an independent risk factor in the presence of cryoglobulinemia. Low C4 levels have also been suggested to reflect underlying cryoglobulinemia. In fact, it has been proposed that circulating immune complexes, most notably cryoglobulins, activate the complement cascade that is, thereby, consumed.21,46 In the study by Skopouli et al,41 palpable purpura, mixed monoclonal cryoglobulinemia, and low C4 levels were the most important risk factors for the development of lymphoproliferative disease, but only the latter was shown to be an independent predictor of lymphoma. Although it is tempting to attribute low C4 levels to cryoglobulinemia, there are arguments favoring the suggestion that the C4 fraction of complement and cryoglobulins have distinct roles in the development of lymphoma. In our multivariate model, both low C4 levels and cryoglobulinemia were retained as independent risk factors for lymphoma and for MZBCL in particular. Similarly, 2 studies conducted in Sweden on large Nordic cohorts, in which cryoglobulinemia has been reported only as a rare finding, demonstrated an association between hypocomplementemia and subsequent lymphoma development.45,46 Insufficient production of complement factors could be suggested as an alternative mechanism underlying the low C4 levels in pSS patients at risk of lymphoma. An additional point worth mentioning is that in our cohort about 42% of all low C4 cases presented at some point during follow-up, while the corresponding values at diagnosis were within the normal range (Table 2). The retrospective design of the current study precluded the investigation of any potential association between C4 kinetics and risk of lymphoma. However, given the strong predictive value of low C4 levels at baseline, the demonstration of a dynamic association between oscillations of C4 levels and progression to overt lymphoma is an intriguing possibility and a challenge for future prospective studies.
In our multivariate analysis, lymphocytopenia was the only independent variable predicting the development of any type of lymphoma other than MZBCL. Because of the predominance of DLBCLs in the group of lymphomas that we defined as non-MZBCLs, it is very likely that lymphocytopenia could serve in fact as a predictor of the development of DLBCL. Due to the lack of routinely performed flow cytometry in the samples of our patients, we were unable to investigate the distribution of peripheral lymphocyte subtypes. Nevertheless, it is highly likely that our results reflect, at least in part, the conclusions of Theander et al.45 In their study of a cohort of pSS patients in which the DLBCL cases outnumbered the MZBCL cases, they identified CD4+ T lymphocytopenia and a low CD4+/CD8+ T-cell ratio as the strongest predictors of lymphoma. Research has also demonstrated that the low CD4+ T-cell counts frequently seen in pSS patients reflect the elimination of CD4+ T populations by sustained antigenic activation-induced CD4+ apoptosis.51 Prolonged antigenic stimulation (for example, viral) has been regarded as a key component in the pathogenesis of MZBCL, especially of the MALT type, and possibly can be extrapolated to the etiopathogenesis of DLBCL. Indeed, the DLBCLs are known to arise from B cells that display somatic hypermutations in the variable-region immunoglobulin genes; such mutations are indicative of T-cell-dependent antigen exposure.23 Alternatively, a number of case reports have implied an association between idiopathic CD4+ T lymphocytopenia (ICL), not rarely seen in pSS patients, with an increased risk of lymphomagenesis.28,30 Whatever the actual mechanism inducing the depletion of lymphocytes may be, it should be admitted that lymphocytopenia, and especially low CD4+ T-lymphocyte levels, reflect a state of decreased immune surveillance. In this context, the probability of the progression from autoimmunity to a more aggressive type of lymphoma, such as DLBCL, is greater. In hindsight, it is worth underscoring the predictive value of lymphocytopenia in relation to the risk of DLBCL only, and not in relation to the risk of MZBCL.
To identify pSS patients at high risk for NHL, we propose a model for pSS classification according to the underlying risk of lymphoma.21 Patients presenting with neutropenia, cryoglobulinemia, splenomegaly, lymphadenopathy, or low C4 levels have a much greater probability of developing lymphoma and are considered high-risk patients, as opposed to patients with none of the above adverse factors at diagnosis, whose risk of NHL is 3.6%. The PSEP value of our model, an overall indicator of prognostic information, is 0.17. When applying the simple 3-factor model proposed by Ioannidis et al21 to our cohort, its ability to identify a low-risk group with minimal probability of lymphoma development is strong, with only 3.2% of patients in the low-risk group developing lymphoma during follow-up. However, the same does not hold true with regard to high-risk patients; only 14.7% of the patients in the high-risk group developed lymphoma. The ability of our model to better predict the development of lymphoma in the high-risk group is reflected in its greater PSEP value and, hence, in the PSEP difference between the 2 studies. Because over-fitting may be a factor inflating our model's prognostic value, prospective validation in an independent cohort is warranted.
In conclusion, pSS patients may present with a variety of hematologic abnormalities, most commonly anemia and hypergammaglobulinemia, some of which correlate with certain disease manifestations and with a predisposition to the development of NHL. The latter is predominantly of the MZBC type and less often of the DLBC type. Neutropenia, cryoglobulinemia, splenomegaly, lymphadenopathy, and low C4 levels are simple clinical tools for the identification of patients at high risk of developing NHL; the presence of any of these risk factors at diagnosis should justify closer follow-up of these patients. The above disease characteristics comprise a pSS phenotype with a predisposition to MZBCL; by contrast, patients presenting with lymphocytopenia are more likely to develop a non-MZBCL lymphoma, most importantly a DLBCL.
REFERENCES
1. Abraham S, Begum S, Isenberg D. Hepatic manifestations of autoimmune rheumatic diseases.
Ann Rheum Dis. 2004;63:123-129.
2. Alamanos Y, Tsifetaki N, Voulgari PV, Venetsanopoulou AI, Siozos C, Drosos AA. Epidemiology of primary Sjogren's syndrome in north-west Greece, 1982-2003.
Rheumatology (Oxford). 2006;45:187-191.
3. Altman DG, Royston P. What do we mean by validating a prognostic model?
Stat Med. 2000;19:453-473.
4. Ambler G, Omar RZ, Royston P. A comparison of imputation techniques for handling missing predictor values in a risk model with a binary outcome.
Stat Methods Med Res. 2007;16:277-298.
5. Andrianakos A, Trontzas P, Christoyannis F, Dantis P, Voudouris C, Georgountzos A, Kaziolas G, Vafiadou E, Pantelidou K, Karamitsos D, Kontelis L, Krachtis P, Nikolia Z, Kaskani E, Tavaniotou E, Antoniades C, Karanikolas G, Kontoyanni A. Prevalence of rheumatic diseases in Greece: a cross-sectional population based epidemiological study. The ESORDIG Study.
J Rheumatol. 2003;30:1589-1601.
6. Bossini N, Savoldi S, Franceschini F, Mombelloni S, Baronio M, Cavazzana I, Viola BF, Valzorio B, Mazzucchelli C, Cattaneo R, Scolari F, Maiorca R. Clinical and morphological features of kidney involvement in primary Sjogren's syndrome.
Nephrol Dial Transplant. 2001;16:2328-2336.
7. Bowman SJ, Ibrahim GH, Holmes G, Hamburger J, Ainsworth JR. Estimating the prevalence among Caucasian women of primary Sjogren's syndrome in two general practices in Birmingham, UK.
Scand J Rheumatol. 2004;33:39-43.
8. Brito-Zeron P, Soria N, Munoz S, Bove A, Akasbi M, Belenguer R, Siso A, Ramos-Casals M. Prevalence and clinical relevance of autoimmune neutropenia in patients with primary Sjogren's syndrome.
Semin Arthritis Rheum. 2009;38:389-395 (Epub 2008 Apr 18).
9. Cesana C, Barbarano L, Miqueleiz S, Lucchesini C, Ricci F, Varettoni M, Filippini D, Lazzarino M, Morra E. Clinical characteristics and outcome of immunoglobulin M-related disorders.
Clin Lymphoma. 2005;5:261-264.
10. Coppo P, Sibilia J, Maloisel F, Schlageter MH, Voyer AL, Gouilleux-Gruart V, Goetz J, Desablens B, Tribout B, Lassoued K. Primary Sjogren's syndrome associated agranulocytosis: a benign disorder?
Ann Rheum Dis. 2003;62:476-478.
11. Cox D. Regression models and life tables.
J R Stat Soc. 1972;34(Series B):187-220.
12. Dafni UG, Tzioufas AG, Staikos P, Skopouli FN, Moutsopoulos HM. Prevalence of Sjogren's syndrome in a closed rural community.
Ann Rheum Dis. 1997;56:521-525.
13. Deutsch AJ, Aigelsreiter A, Staber PB, Beham A, Linkesch W, Guelly C, Brezinschek RI, Fruhwirth M, Emberger W, Buettner M, Beham-Schmid C, Neumeister P. MALT lymphoma and extranodal diffuse large B-cell lymphoma are targeted by aberrant somatic hypermutation.
Blood. 2007;109:3500-3504.
14. Efron B. Bootstrap methods: another look at the jackknife.
Ann Stat. 1979;7:1-26.
15. Ekstrom Smedby K, Vajdic CM, Falster M, Engels EA, Martinez-Maza O, Turner J, Hjalgrim H, Vineis P, Seniori Costantini A, Bracci PM, Holly EA, Willett E, Spinelli JJ, La Vecchia C, Zheng T, Becker N, De Sanjose S, Chiu BC, Dal Maso L, Cocco P, Maynadie M, Foretova L, Staines A, Brennan P, Davis S, Severson R, Cerhan JR, Breen EC, Birmann B, Grulich AE, Cozen W. Autoimmune disorders and risk of non-Hodgkin lymphoma subtypes: a pooled analysis within the InterLymph Consortium.
Blood. 2008;111:4029-4038.
16. Fox RI. Sjogren's syndrome.
Lancet. 2005;366:321-331.
17. Friedman J, Klepfish A, Miller EB, Ognenovski V, Ike RW, Schattner A. Agranulocytosis in Sjogren's syndrome: two case reports and analysis of 11 additional reported cases.
Semin Arthritis Rheum. 2002;31:338-345.
18. Ghesquieres H, Berger F, Felman P, Callet-Bauchu E, Bryon PA, Traverse-Glehen A, Thieblemont C, Baseggio L, Michallet AS, Coiffier B, Salles G. Clinicopathologic characteristics and outcome of diffuse large B-cell lymphomas presenting with an associated low-grade component at diagnosis.
J Clin Oncol. 2006;24:5234-5241.
19. Goransson LG, Herigstad A, Tjensvoll AB, Harboe E, Mellgren SI, Omdal R. Peripheral neuropathy in primary sjogren syndrome: a population-based study.
Arch Neurol. 2006;63:1612-1615.
20. Grambsch PM, Therneau TM. Proportional hazards tests and diagnostics based on weighted residuals.
Biometrika. 1994;81:515-526.
21. Ioannidis JP, Vassiliou VA, Moutsopoulos HM. Long-term risk of mortality and lymphoproliferative disease and predictive classification of primary Sjogren's syndrome.
Arthritis Rheum. 2002;46:741-747.
22. Ito I, Nagai S, Kitaichi M, Nicholson AG, Johkoh T, Noma S, Kim DS, Handa T, Izumi T, Mishima M. Pulmonary manifestations of primary Sjogren's syndrome: a clinical, radiologic, and pathologic study.
Am J Respir Crit Care Med. 2005;171:632-638.
23. Jaffe EH, Stein H, Vardiman JW.
Pathology and Genetics of Tumors of Hematopoietic and Lymphoid Tissues. Lyon: IARC Press; 2001.
24. Kaplan E, Meier P. Nonparametric estimation from incomplete observations.
J Am Stat Assoc. 1958;53:457-481.
25. Kassan SS, Moutsopoulos HM. Clinical manifestations and early diagnosis of Sjogren syndrome.
Arch Intern Med. 2004;164:1275-1284.
26. Kassan SS, Thomas TL, Moutsopoulos HM, Hoover R, Kimberly RP, Budman DR, Costa J, Decker JL, Chused TM. Increased risk of lymphoma in sicca syndrome.
Ann Intern Med. 1978;89:888-892.
27. Kauppi M, Pukkala E, Isomaki H. Elevated incidence of hematologic malignancies in patients with Sjogren's syndrome compared with patients with rheumatoid arthritis (Finland).
Cancer Causes Control. 1997;8:201-204.
28. Kirtava Z, Blomberg J, Bredberg A, Henriksson G, Jacobsson L, Manthorpe R. CD4+ T-lymphocytopenia without HIV infection: increased prevalence among patients with primary Sjogren's syndrome.
Clin Exp Rheumatol. 1995;13:609-616.
29. Lazarus MN, Robinson D, Mak V, Moller H, Isenberg DA. Incidence of cancer in a cohort of patients with primary Sjogren's syndrome.
Rheumatology (Oxford). 2006;45:1012-1015.
30. Mandl T, Bredberg A, Jacobsson LT, Manthorpe R, Henriksson G. CD4+ T-lymphocytopenia-a frequent finding in anti-SSA antibody seropositive patients with primary Sjogren's syndrome.
J Rheumatol. 2004;31:726-728.
31. Mariette X. Lymphomas complicating Sjogren's syndrome and hepatitis C virus infection may share a common pathogenesis: chronic stimulation of rheumatoid factor B cells.
Ann Rheum Dis. 2001;60:1007-1010.
32. Mellemkjaer L, Pfeiffer RM, Engels EA, Gridley G, Wheeler W, Hemminki K, Olsen JH, Dreyer L, Linet MS, Goldin LR, Landgren O. Autoimmune disease in individuals and close family members and susceptibility to non-Hodgkin's lymphoma.
Arthritis Rheum. 2008;58:657-666.
33. Morra E, Cesana C, Klersy C, Barbarano L, Varettoni M, Cavanna L, Canesi B, Tresoldi E, Miqueleiz S, Bernuzzi P, Nosari AM, Lazzarino M. Clinical characteristics and factors predicting evolution of asymptomatic IgM monoclonal gammopathies and IgM-related disorders.
Leukemia. 2004;18:1512-1517.
34. Moutsopoulos HM. Sjogren's syndrome. In: Fauci AS, Braunwald E, Kasper DL, Hauser SL, Longo DL, Jameson JL, Loscalzo J, eds.
Harrison's Principles of Internal Medicine. 17th ed. New York: McGraw-Hill; 2008:2107-2109.
35. Moutsopoulos HM. Sjogren's syndrome: autoimmune epithelitis.
Clin Immunol Immunopathol. 1994;72:162-165.
36. Newcombe RG. Two-sided confidence intervals for the single proportion: comparison of seven methods.
Stat Med. 1998;17:857-872.
37. Owen RG, Treon SP, Al-Katib A, Fonseca R, Greipp PR, McMaster ML, Morra E, Pangalis GA, San Miguel JF, Branagan AR, Dimopoulos MA. Clinicopathological definition of Waldenstrom's macroglobulinemia: consensus panel recommendations from the Second International Workshop on Waldenstrom's Macroglobulinemia.
Semin Oncol. 2003;30:110-115.
38. Pariente D, Anaya JM, Combe B, Jorgensen C, Emberger JM, Rossi JF, Sany J. Non-Hodgkin's lymphoma associated with primary Sjogren's syndrome.
Eur J Med. 1992;1:337-342.
39. Ramos-Casals M, Font J, Garcia-Carrasco M, Brito MP, Rosas J, Calvo-Alen J, Pallares L, Cervera R, Ingelmo M. Primary Sjogren syndrome: hematologic patterns of disease expression.
Medicine (Baltimore). 2002;81:281-292.
40. Royer B, Cazals-Hatem D, Sibilia J, Agbalika F, Cayuela JM, Soussi T, Maloisel F, Clauvel JP, Brouet JC, Mariette X. Lymphomas in patients with Sjogren's syndrome are marginal zone B-cell neoplasms, arise in diverse extranodal and nodal sites, and are not associated with viruses.
Blood. 1997;90:766-775.
41. Skopouli FN, Dafni U, Ioannidis JP, Moutsopoulos HM. Clinical evolution, and morbidity and mortality of primary Sjogren's syndrome.
Semin Arthritis Rheum. 2000;29:296-304.
42. Smedby KE, Hjalgrim H, Askling J, Chang ET, Gregersen H, Porwit-MacDonald A, Sundstrom C, Akerman M, Melbye M, Glimelius B, Adami HO. Autoimmune and chronic inflammatory disorders and risk of non-Hodgkin lymphoma by subtype.
J Natl Cancer Inst. 2006;98:51-60.
43. Strand V, Talal N. Advances in the diagnosis and concept of Sjogren's syndrome (autoimmune exocrinopathy).
Bull Rheum Dis. 1979;30:1046-1052.
44. Sutcliffe N, Inanc M, Speight P, Isenberg D. Predictors of lymphoma development in primary Sjogren's syndrome.
Semin Arthritis Rheum. 1998;28:80-87.
45. Theander E, Henriksson G, Ljungberg O, Mandl T, Manthorpe R, Jacobsson LT. Lymphoma and other malignancies in primary Sjogren's syndrome: a cohort study on cancer incidence and lymphoma predictors.
Ann Rheum Dis. 2006;65:796-803.
46. Theander E, Manthorpe R, Jacobsson LT. Mortality and causes of death in primary Sjogren's syndrome: a prospective cohort study.
Arthritis Rheum. 2004;50:1262-1269.
47. Trontzas PI, Andrianakos AA. Sjogren's syndrome: a population based study of prevalence in Greece. The ESORDIG study.
Ann Rheum Dis. 2005;64:1240-1241.
48. Valesini G, Priori R, Bavoillot D, Osborn J, Danieli MG, Del Papa N, Gerli R, Pietrogrande M, Sabbadini MG, Silvestris F, Valsecchi L. Differential risk of non-Hodgkin's lymphoma in Italian patients with primary Sjogren's syndrome.
J Rheumatol. 1997;24:2376-2380.
49. Vitali C, Bombardieri S, Jonsson R, Moutsopoulos HM, Alexander EL, Carsons SE, Daniels TE, Fox PC, Fox RI, Kassan SS, Pillemer SR, Talal N, Weisman MH. Classification criteria for Sjogren's syndrome: a revised version of the European criteria proposed by the American-European Consensus Group.
Ann Rheum Dis. 2002;61:554-558.
50. Voulgarelis M, Dafni UG, Isenberg DA, Moutsopoulos HM. Malignant lymphoma in primary Sjogren's syndrome: a multicenter, retrospective, clinical study by the European Concerted Action on Sjogren's Syndrome.
Arthritis Rheum. 1999;42:1765-1772.
51. Zeher M, Szodoray P, Gyimesi E, Szondy Z. Correlation of increased susceptibility to apoptosis of CD4+ T cells with lymphocyte activation and activity of disease in patients with primary Sjogren's syndrome.
Arthritis Rheum. 1999;42:1673-1681.
52. Zintzaras E, Voulgarelis M, Moutsopoulos HM. The risk of lymphoma development in autoimmune diseases: a meta-analysis.
Arch Intern Med. 2005;165:2337-2344.
53. Zucca E, Conconi A, Pedrinis E, Cortelazzo S, Motta T, Gospodarowicz MK, Patterson BJ, Ferreri AJ, Ponzoni M, Devizzi L, Giardini R, Pinotti G, Capella C, Zinzani PL, Pileri S, Lopez-Guillermo A, Campo E, Ambrosetti A, Baldini L, Cavalli F. Nongastric marginal zone B-cell lymphoma of mucosa-associated lymphoid tissue.
Blood. 2003;101:2489-2495.
