Data Driven Prediction Lymphoma Model and 10-year Overall Survival Rates of a Large Harmonized Cohort of Patients with Primary Sjögren’s Syndrome Associated Lymphomas

Andreas Goules ¹, Michael Voulgarelis ², Loukas Chatzis², Vasilis Pezoulas ³, Francesco Ferro ⁴, Saviana Gandolfo ⁵, Valentina Donati ⁶, Luca Quartuccio ⁷, Cathryn Scott ⁸, Ginevra De Marchi ⁹, Georgios Michalopoulos ¹, Aliki Venetsanopoulou ¹, Aristea Papageorgiou ¹, Dimitris Ziogas ¹, Marina Sikara ¹, Argyropoulou Ourania ¹, Clio Mavragani ¹, Dimitris fotiadis ¹⁰, Salvatore De Vita ⁵, Chiara Baldini ⁴ and Athanasios Tzioufas ², ¹Pathophysiology Department, Athens School of Medicine, National and Kapodistrian University of Athens, Athens, Greece, athens, Greece, ²Department of Pathophysiology, Athens School of Medicine, National and Kapodistrian University of Athens, Athens, Greece, Athens, Greece, ³Unit of Medical Technology and Intelligent Information Systems, University of Ioannina, GR45110, Ioannina, Greece, Ioannina, Greece, ⁴Rheumatology Unit, Department of Clinical and Experimental Medicine, University of Pisa, Pisa, Italy, Pisa, Italy, ⁵Rheumatology Clinic, Udine University Hospital, Department of Medical Area, University of Udine, Udine, Italy, Udine, Italy, ⁶University of Pisa, Pisa, ⁷Rheumatology Clinic, Department of Medical Area, University of Udine, Udine, Italy, Udine, Italy, ⁸Institute of Anatomic Pathology, Department of Medical Area, University of Udine, Udine, Italy. *equally contributed, Udine, Italy, ⁹Rheumatology Clinic, Udine University Hospital, Udine, Italy, Udine, Italy, ¹⁰Pathophysiology Department, Athens School of Medicine, National and Kapodistrian University of Athens, Athens, Greece, Ioannina, Greece

Meeting: 2019 ACR/ARP Annual Meeting

Keywords: MALT lymphoma and salivary gland, Sjogren's syndrome

Session Information

Date: Tuesday, November 12, 2019

Title: Sjögrenʼs Syndrome – Basic & Clinical Science Poster I

Session Type: Poster Session (Tuesday)

Session Time: 9:00AM-11:00AM

Background/Purpose: Non-Hodgkin’s lymphomas (NHLs) may complicate primary Sjögren’s syndrome (pSS) with significant impact on morbidity and mortality among patients. A large cohort of SS associated lymphoma patients was constructed from 3 specializing (pSS) centers [Udine, Athens, Piza-(UPA]) aiming to a) identify predictors of lymphoma by applying innovative data driven analysis tools b) estimate outcome and survival curves of SS associated lymphoma patients.

Methods: One hundred and sixty-two patients with SS associated lymphomas who fulfilled the 2016 ACR/EULAR were included in the study. Clinical, histological and laboratory data were collected. For lymphoma prediction,
a robust decision-making machine learning algorithm was applied on the harmonized dataset using the Extreme Gradient Boosting (XGBoost) framework. Weak decision tree ensembles were combined through the gradient boosting optimization approach which reduces the prediction errors yielding a high-performance supervised learning model for predicting binary lymphoma outcomes in pSS. Two hundred and seventy-eight SS patients without lymphoma were recruited as controls for the prediction model matched according to age, gender and SS disease suration. A conventional 10-fold cross validation approach was then applied on the harmonized dataset to evaluate the sensitivity, specificity, and accuracy of the robust classifier. Kaplan-Meir survival curves for the total lymphoma population as well as for patients with mucosal associated lymphoid tissue lymphomas (MALTL), diffuse large B cell lymphomas (DLBCL) and other lymphoma types were generated. Lymphoma features were compared between Greek and Italian SS lymphoma patients including gender, age at lymphoma diagnosis, time from SS to lymphoma diagnosis, time from SS onset to lymphoma diagnosis, lymphoma follow up time and histologic lymphoma subtypes.

Results: No differences were found among the main lymphoma features mentioned above between Greek and Italian SS associated lymphoma patients. The median age at lymphoma diagnosis, time from SS diagnosis to lymphoma and lymphoma follow up time of the total population was 58 years old (range: 25-82), 4 years (range: -5, 30) and 6 years (range: 0, 30) respectively. Preliminary data analysis, revealed parotid gland enlargement >2months, neutrophils count at diagnosis, palpable purpura due to cryoglobulinemia, salivary or lachrymal gland enlargement at SS onset and hypergammaglobulinemia as main contributors of the most prominent decision tree pathways, with average area under the curve (AUC) =0.90, sensitivity=0.76, accuracy=0.83, and precision=0.84 (Figure 1). The estimated 10-year overall survival rates (OS) were 87% for the total population, 88,2% for patients with MALTL and 80% for DLBCL group (Figure 2).

Conclusion: To our knowledge, this is the largest pSS associated lymphomacohort. Preliminary harmonized pooling data, after applying novel bioinformatics tools, confirmed classical features as lymphoma predictors, pointing out the dynamic perspective of the proposed lymphoma prediction model. In addition, SS patients with NHLs display a favorable prognosis, especially those with MALTL.

The ROC curve depicting the performance of the proposed lymphoma prediction model.

Kaplan-Meier curves for overall survival -OS- of SS HNLs patients and SS NHL histologic subgroups

Disclosure: A. Goules, None; M. Voulgarelis, None; L. Chatzis, None; V. Pezoulas, None; F. Ferro, None; S. Gandolfo, None; V. Donati, None; L. Quartuccio, None; C. Scott, None; G. De Marchi, None; G. Michalopoulos, None; A. Venetsanopoulou, None; A. Papageorgiou, None; D. Ziogas, None; M. Sikara, None; A. Ourania, None; C. Mavragani, None; D. fotiadis, None; S. De Vita, None; C. Baldini, None; A. Tzioufas, None.

To cite this abstract in AMA style:

Goules A, Voulgarelis M, Chatzis L, Pezoulas V, Ferro F, Gandolfo S, Donati V, Quartuccio L, Scott C, De Marchi G, Michalopoulos G, Venetsanopoulou A, Papageorgiou A, Ziogas D, Sikara M, Ourania A, Mavragani C, fotiadis D, De Vita S, Baldini C, Tzioufas A. Data Driven Prediction Lymphoma Model and 10-year Overall Survival Rates of a Large Harmonized Cohort of Patients with Primary Sjögren’s Syndrome Associated Lymphomas [abstract]. Arthritis Rheumatol. 2019; 71 (suppl 10). https://acrabstracts.org/abstract/data-driven-prediction-lymphoma-model-and-10-year-overall-survival-rates-of-a-large-harmonized-cohort-of-patients-with-primary-sjogrens-syndrome-associated-lymphomas/. Accessed .

« Back to 2019 ACR/ARP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/data-driven-prediction-lymphoma-model-and-10-year-overall-survival-rates-of-a-large-harmonized-cohort-of-patients-with-primary-sjogrens-syndrome-associated-lymphomas/