Stratifying Lymphoma Risk in Sjögren's Disease

Key Takeaways:

• Cohort and meta-analytic estimates converge on a 15- to 19-fold excess of non-Hodgkin lymphoma in Sjögren's disease; population-based and self-report studies yield lower estimates for different reasons.
• Predictors differ by subtype. Cryoglobulinemia predicts all 3, low C3 tracks with mucosal disease, and low C4 with non-mucosal B-cell disease.
• Risk increases with the duration of Sjögren's disease, which supports periodic rather than one-time assessment.

Lymphoma shapes how Sjögren's follow-up should be structured. It is uncommon in absolute terms, but the relative risk is high enough and the predictors specific enough to stratify rather than watch everyone identically – and risk accrues with time rather than being fixed at diagnosis.^1,2

How large the risk is

Three estimates of the same quantity land close together: A meta-analysis of 5 cohorts pooled a standardized incidence ratio of 18.8 (95% CI 9.5–37.3) on 30 events, a Swedish cohort reported 15.57 (7.77–27.85), and a Spanish cohort 15.41 (11.58–20.51) across the 3 lymphoma coding categories – its headline non-Hodgkin figure of 6.04 is deflated because mucosa-associated lymphoma codes into the myeloma category instead.^2-5

Case-control studies give lower figures. In US Surveillance, Epidemiology, and End Results–Medicare data covering 44,350 lymphoid malignancy cases against 122,531 controls, Sjögren's carried an odds ratio of 1.9 (1.5–2.3) for non-Hodgkin lymphoma. A pooled analysis of 12 international studies, 4 US, gave 6.56 (3.10–13.9).^6,7

Two mechanisms explain the spread, and they differ. Sampling accounts for part: Risk was not significant in a population-based study but was in the hospital-based cohorts, a contrast the Spanish authors draw explicitly.³

Ascertainment accounts for the rest – the pooled analysis rests on self-reported diagnoses, which those authors identify as a limitation of that literature and which attenuates estimates independently of sampling. Which figure applies depends on the patient in front of you; a referred population sits nearer the cohort end.^3,6

Absolute risk is more reassuring: Cancer-free survival in the Spanish cohort was 94% at 5 years, 88% at 10, 77% at 20, including solid tumors.³

Histology resolves better than the counts suggest. Marginal zone disease dominates proportionally – around 65% of cases in the review literature, roughly 60% of B-cell lymphomas in the Spanish cohort – and site specificity is striking.^3,4

Salivary gland marginal zone lymphoma carried an odds ratio of 71 (40–120) in the US data, and parotid mucosa-associated lymphoma 996 (216–4596) in the pooled analysis.^6,7

But diffuse large B-cell lymphoma is genuinely elevated too – 8.92 (3.83–20.7) pooled and 2.0 (1.5–2.8) in US data – corroborating the Swedish cohort's 58% diffuse large B-cell histology. The subtype question is not either/or. The 2 case-control analyses disagree on follicular lymphoma, so no claim is made here.^2,6,7

Which findings predict which subtype

Predictors differ by subtype – the Spanish cohort was the first to report it. Cryoglobulins predicted all 3: 6.32 (2.47–16.15) for mucosa-associated, 3.34 for non-mucosal B-cell, 4.36 (1.12–16.99) for non-B-cell disease.³

Low C3 was specific to mucosa-associated disease at 3.25 (1.33–7.93). Non-mucosal B-cell disease was marked instead by low C4 at 3.83, monoclonal gammopathy at 3.45, and anemia at 2.58; thrombocytopenia at 4.85 (1.64–14.34) marked non-B-cell disease. In multivariate analysis, anemia and cryoglobulins were independently associated with hematological cancer.³

The Swedish cohort adds a predictor list without subtype resolution, on 12 events: purpura 4.64, low C3 6.18, low C4 9.49, CD4 lymphocytopenia 8.14, CD4:CD8 ratio ≤0.8 at 10.92.2 It did not measure cryoglobulins.²

Rheumatoid factor is contested on cohort evidence, not assertion. A multicentre cohort reported it as an independent predictor, separable from cryoglobulinemia and easier to measure. Yet it was non-significant across all 3 subtypes in the Spanish cohort – 0.94, 1.16, 1.28 – and in the Swedish one. Lymphopenia was likewise non-significant.^2-4

Salivary gland enlargement carries more weight than any single cohort suggests. Parotid enlargement appears among independent predictors in cohort work and in published risk-factor lists, and the ESSDAI glandular domain carried a hazard ratio of 7.35 (3.36–16.10) for mucosa-associated lymphoma. The Swedish cohort tested it directly and found it non-significant at 2.02 (0.62–6.61) — on 12 events total. Unexplained persistent enlargement warrants evaluation rather than reassurance.^1-4

What surveillance can and cannot claim

Prediction tools exist. One 7-factor score – gland enlargement, lymphadenopathy, Raynaud phenomenon, anti-Ro/SSA or anti-La/SSB, rheumatoid factor, monoclonal component, low C4 – carries a 3.8% lymphoma probability at 2 or fewer factors and approaches 100% at all 7. What does not exist is a validated score or a guideline-endorsed protocol.⁴

What it supports is periodic reassessment of the predictors: lymphocyte count, protein electrophoresis, rheumatoid factor, C3, C4, and cryoglobulins every 1 to 2 years, moving to every 6 months in higher-risk patients – with the explicit caveat from the same source that data showing improved outcomes from closer surveillance are limited.¹

The case for periodicity is the duration effect. In the Swedish cohort, ratios rose from 6.4 in the first 5 years to 11.1 at 6 to 10 and 20.8 at 11 to 15. A patient assessed as low risk at diagnosis has not been assessed for the period when risk is highest.²