We read with considerable interest the article by Zhang and He examining the clinicopathological and imaging distinctions between ocular adnexal MALT lymphoma (OAML) and IgG4-related ophthalmic disease (IgG4-ROD). The authors present a comprehensive, pathologically validated analysis and propose an integrated diagnostic framework combining clinical, imaging, and artificial intelligence–based approaches for differentiation.
Their findings reaffirm key distinctions—unilateral involvement and monoclonality in OAML vs bilateral lacrimal gland disease and fibro-inflammatory pathology in IgG4-ROD—while appropriately emphasizing the limited specificity of serum IgG4, consistent with few studies in literature. ,, The high diagnostic performance of their clinical model (AUC 0.865) and multimodal AI approaches (AUC up to 0.974) is notable and reflects the growing potential of data-driven diagnostics.
From a clinical standpoint, however, the diagnostic boundary remains inherently probabilistic rather than absolute. In our experience in Indian population, lesions strongly suggestive of lymphoma on clinical and radiologic grounds may ultimately prove to be IgG4-ROD following histopathological evaluation. We would like to highlight our experience that, despite strong clinical suspicion of lymphoma, lacrimal gland biopsy revealed IgG4-related disease, with subsequent favorable response to corticosteroids. Sometimes during systemic evaluation, osteolytic lesions can also be associated in these patients which strongly raises suspicion for malignancy. Although multiple myeloma is rarely associated in this context, it can add significantly to the diagnostic dilemma. Importantly, a broad-based M spike may represent polyclonal hypergammaglobulinemia rather than monoclonality, and comprehensive myeloma workup, including bone marrow evaluation and metabolic parameters, is essential for exclusion. This underscores a key clinical principle: pre-biopsy differentiation among lymphoma, IgG4-ROD, and even plasma cell disorders may be unreliable, and premature therapeutic decisions should be avoided. Additionally, the observation of IgG4-positive plasma cell infiltration in a subset of OAML highlights a biologically relevant overlap, supporting emerging evidence that chronic antigenic stimulation and immune dysregulation may contribute to lymphomagenesis. , This intersection warrants deeper exploration through integrated molecular and immunologic profiling.
In the setting of such diagnostic ambiguity, the translational role of artificial intelligence warrants careful consideration. While the pathology-informed cross-modal framework is innovative, its real clinical value may lie not in replacing biopsy but in risk stratification and biopsy prioritization—for example, identifying cases where immediate tissue diagnosis is imperative vs those where a short therapeutic trial may be reasonable. Future models incorporating uncertainty quantification and calibrated probability outputs may further enhance clinical decision-making. The retrospective single-center design and lack of external validation limit generalizability. Prospective multicenter studies evaluating model performance in real-world, pre-biopsy settings will be essential to establish clinical utility.
CRediT authorship contribution statement
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
