Background Many screening questionnaires have been developed to identify PsA amongst patients with psoriasis but their performance in head to head studies has been diappointing.
Objectives To combine most discriminatory questions from established PsA screening questionnaires to create a more sensitive and specific screening test for PsA.
Methods Data from the CONTEST study, a head-to-head study of the PEST, PASE and ToPAS screening questionnaires in secondary care dermatology clinics, were used to establish which questions from any of the three questionnaires had a Youden’s index of >0.1 using CASPAR criteria for PsA as the gold standard. Optional weighting was included based on logistic regression of individual questions. The number of painful joint areas on the PEST mannequin predictive of PsA was identified using ROC analysis. CART analysis was used to identify other questions that showed good differentiation. Data from comparative head-to-head studies based in Dublin and Utah were used to independently validate the new proposed questionnaires.
Results In this individual analysis, only 2 of 25 questions showed significant differentiation of PsA cases using Fishers exact test: ToPAS 2A “Have you ever noticed any of these changes in your fingernails – pits in the nails?” and PEST 4 “Have you had pain in your heel?”. These were the only questions identified using logistic regression with an OR of 5 and 2 respectively. A further 10 questions had a Youden score of ≥0.1 and these were considered for inclusion into the first candidate questionnaire. Where overlapping questions about similar symptoms (eg nail psoriasis) were identified, the most discriminatory question was chosen. One question (ToPAS 7A) was excluded as many patients did not complete that question. This resulted in the CONTEST questionnaire of 8 yes/no questions. The weighted CONTEST questionnaire used the OR identified above and weighted all other questions as 1 giving a score of 0-13. CONTEST-joint included the presence of joint areas on the PEST mannequin with a cut off of ≥6 areas being positive. CART analysis identified a further 5 questions for inclusion giving a total of 13 questions for CONTEST-tree.
In the original CONTEST data the AUC for the CONTEST, CONTEST-weighted, CONTEST-joint and CONTEST-tree were 0.69, 0.74, 0.70, 0.59 respectively with p<0.01 for all except CONTEST-tree which was not analysed further. The remaining candidates were tested in the validation cohorts and the results are shown in table 1.
Conclusions The CONTEST questionnaires provide a new tool which could improve sensitivity and specificity of identifying PsA compared to existing questionnaires. Further research in independent populations is required to identify the optimal cut point for referral.
Disclosure of Interest None Declared