Background/Purpose:

Sjogren’s syndrome (SS) is a chronic, progressive disease characterized by dry eyes and dry mouth resulting from lymphocytic infiltration of the lacrimal and salivary glands, respectively. Although SS is defined by keratoconjunctivitis sicca and xerostomia (dry eyes and mouth), the full spectrum of disease can involve a complex myriad of systemic symptoms similar to other rheumatic diseases.  The American-European Consensus Group (AECG) classification criteria for SS require serologic and hematologic tests as well as the expertise of at least two clinical specialties for diagnosis.  As such, proper diagnosis of SS is often challenging.

Methods: In order to develop a model to better predict SS, we first assessed the predictive value of a commonly used screening tool, the AECG six-question phone interview, in 379 European American (EA) individuals with dry eyes and mouth using structural equation modeling (SEM) and logistic regression (LR).  Then, using responses to 440 general health and medical history questions in addition to the six-question phone interview from the same participants, we developed a novel predictive model using genetic algorithm (GA), k-nearest neighbor and forward selection model.

Results:

The six-question phone interview, which in our experience results in ~36% classification of SS, yielded an ill-fitting model with low prediction accuracy (~56% under violated assumptions) in silico. Our novel model generated from the general health, medical history questionnaire, and six-question phone interview data obtains an accuracy of ~61% in the test set and ~64% in the unknown set. The generated model considered clinically substantive consisted of 9 questions relating to autoimmunity, difficulty swallowing, mouth/tongue dryness, aching, and disorientation.

Conclusion:

The six-question phone interview, which in our experience results in ~36% classification of SS, yielded an ill-fitting model with low prediction accuracy (~56% under violated assumptions) in silico. Our novel model generated from the general health, medical history questionnaire, and six-question phone interview data obtains an accuracy of ~61% in the test set and ~64% in the unknown set. The generated model considered clinically substantive consisted of 9 questions relating to autoimmunity, difficulty swallowing, mouth/tongue dryness, aching, and disorientation.

ACR Meeting Abstracts - https://acrabstracts.org/abstract/a-novel-screening-tool-indicative-of-primary-sjogrens-syndrome-versus-sicca-symptoms/