BackgroundGiant cell arteritis (GCA) is the most prevalent vasculitis in the elder. Nearly 20% of patients experience transient or permanent visual loss (PVL). It has been reported that erythrocyte sedimentation rate (ESR), haemoglobin (Hb), constitutional syndrome (CS) and fever are prognostic factors that predict PVL but models have shown poor diagnostic performance.ObjectivesTo evaluate if clinical signs, symptoms and blood tests can predict PVL at GCA diagnosis.MethodsWe retrospectively included patients from the Spanish Vasculitis Registry (REVAS) from 2005 to 2009. Clinical and blood tests data were obtained from medical records. We randomly split the cohort using shrinkage function to create a derivation and a validation cohort. In the derivation set we compared data and we built a multivariable logistic regression model to predict PVL. Internal validity was evaluated with 1000 bootstrap. External validity was evaluated using the validation set of data. Performance of the model was determined using the area under the curve (AUC) with 95% confidence interval. Calculations were done using StataBE 17.0.ResultsWe included 620 patients (derivation cohort: 397 patients). Clinical signs, symptoms and blood tests results according to the presence or absence of PVL (Table 1). Mean age at diagnosis was 76.3 years and PVL was present in 86 (21.7%) patients. Significant predictors at baseline were age (p=0.000), hypertension (p=0.04), fever (p=0.001), jaw claudication (0.000), transient visual loss (TVL, p=0.000) and decreased temporal artery (TA) pulse (p=0.004). Multivariable logistic regression showed that age older than 75 years (OR 2.7, p=0.000), jaw claudication (OR, 2.75; p=0.000) and TVL (OR 7.2, p=0.000) were risk factors for PVL. CS was the only protective factor (OR 0.57, p=0.017). Hypertension (OR 1.4, IC95%: 0.88 – 2.3) and diabetes (OR 1.63, IC95%: 0.94 – 2.8) were not statistically significant. Our model showed an AUC 0.8 (IC 95%: 0.75 – 0.84). A 1000 bootstrap analysis showed good internal validity (AUC 0.79, IC95%: 0.74 – 0.83). Validation cohort comprised 223 patients and the AUC of the model in this dataset showed an AUC 0.81. We compared our model to previously published models and we found that our model had a higher AUC (AUC 0.8, IC 95%: 0.75-0.84 vs. AUC 0.65, IC95%: 0.6 – 0.7; p < 0.0001).Table 1.Baseline date according to the presence or absence of permanent visual loss.Permanent Visual LossNo Permanent Visual LossVariableMean/ProportionSDMean/ProportionSDSignificanceFemale69.8%72.0%0.68Age >75 y.o.72.1%53.4%0.000Hypertension64.3%51.6%0.04Diabetes25.9%16.9%0.06Fever18.6%36.8%0.001Constitutional syndrome42.4%53.2%0.075Polymyalgia40.7%39.7%0.87Headache79.1%79.2%0.987Jaw claudication68.2%39.7%0.000Tenderness of the TA38.6%31.4%0.22Transient visual loss39.0%10.5%0.000Stroke3.5%3.9%0.86Transient ischaemic attack0.0%4.2%0.053Decreased TA pulse66.7%48.0%0.004TA enlargement55.1%50.9%0.51Haemoglobin11.11.211.41.40.37Erythrocyte sedimentation rate95.026.296.426.80.67C Reactive protein9.76.210.48.60.8SD: Standard deviation. TA: Temporal artery.ConclusionAge > 75 years, jaw claudication and TVL can predict PVL, being the CS a protective factor for this complication. Blood test data are not good PVL predictive factors.References[1]Nesher G. J Autoimm. 2014;48-49:73-75.[2]Cid MC et al. Arthritis Rheum. 1998;41:26-32.Acknowledgementson behalf of the Spanish Resgistry of Systemic Vascuitis (REVAS)Disclosure of InterestsNone declared