POS1410 DEVELOPMENT OF PREDICTION MODELS FOR SENIOR PATIENTS WITH RHEUMATOID ARTHRITIS AND COMORBIDITIES TREATED WITH CHRONIC LOW-DOSE GLUCOCORTICOIDS IN THE GLORIA TRIAL

Abstract

BackgroundRheumatoid arthritis (RA) is a systemic, inflammatory disease primarily located in the joints resulting in pain, joint damage, functional disability and reduced quality of life. Treatment of RA is essential to prevent these outcomes, but the treatment itself may also result in adverse events and comorbidity [1]. Although many investigators are working on personalized medicine [2], better models to predict harm and benefit from a certain drug need to be developed before they can be used in daily clinical practice [3].ObjectivesTo develop prediction models for individual patient harm and benefit outcomes in senior patients with RA and comorbidities treated with chronic low-dose glucocorticoid therapy or placebo.MethodsIn the GLORIA trial 451 RA patients aged 65+ were randomized to 2 years 5 mg/day prednisolone or placebo. Eight prediction models were developed from the dataset in a stepwise procedure. In preparation, to limit excessive statistical testing and false positive results, possible predictors were grouped into five predictor sets based on prior knowledge (Table 1). The first set of four models disregarded study treatment and examined general predictive factors. The second set of four models was similar but examined the additional role of study treatment, as main factor and as interaction factor with other predictive variables. In each set two models focused on harm (1: occurrence of ≥1 adverse event of special interest (AESI); 2: number of AESIs per year) and two on benefit (3: early clinical response–disease activity; 4: lack of joint damage progression). AESI comprised all serious adverse events, events leading to discontinuation of study treatment, and events related to glucocorticoid exposure (see main GLORIA study abstract). Linear and logistic multivariable regression methods with backward selection were used to develop the models. The final models were assessed and internally validated with bootstrapping techniques, and their performance was evaluated with model fit and discrimination measures.Table 1.Predictor sets.Personal factorsDisease factorsComorbiditiesAgeDAS28Active comorbidity: cont, dich,SexRA durationGC-relatedEducationRFPrior comorbidity: cont, dich,SmokingAnti-CCPGC-relatedAlcoholDamage (cont, dich)# comorbidity medicationsBMICoping RAJoint surgeryBlood pressureImpact RA# patient symptomsMedicationHealth and daily functioning# concomitant medicationsHAQPrevious use DMARD, bDMARD, GCQoLCurrent use bDMARDVAS healthAdherenceSF36 physical, mentalSwitch antirheumatic drugscont=continuous; dich=dichotomous; GC=glucocorticoid.ResultsStudy treatment (i.e. prednisolone) was highly predictive as a main factor in models 5-8, increasing the risk of both benefit and harm. In addition, a few additional variables were slightly (but not relevantly) predictive for the outcome in one of the models (Figure 1). Their association was much weaker than that of study treatment. In three instances, prednisolone interacted with another predictive factor (see Figure 1). The quality of the prediction models was sufficient, the performance low to moderate: explained variance: 12-15%, AUC 0.67-0.69.ConclusionBaseline factors are not helpful to select senior RA patients for treatment with low-dose prednisolone given their low power to predict the chance of benefit or harm.