Background/Purpose: There is a need for screening
questionnaires to identify patients with psoriasis who have undiagnosed
psoriatic arthritis but the optimal method is unknown.  We developed the
CONTEST questionnaires using the most discriminative items from 3 existing
tools: CONTEST consists of 8 yes/no items and CONTESTji the same items with the
addition of a joint manikin. Our aim was to test the new proposed CONTEST
questionnaires alongside the validated PEST questionnaire in a primary care
setting.

Methods: Patients with psoriasis were identified from
5 GP surgeries in Yorkshire and a random sample were invited to participate in
the study.  Patients had to be over 18 years old, have a diagnosis of psoriasis
and no diagnosis of PsA, RA or AS.  Consenting patients were asked to complete
the PEST and CONTEST questions and were then assessed by a dermatologist and
rheumatologist.  Diagnosis of PsA was made by the assessing rheumatologist as
lab and radiographic data were unavailable to assess the CASPAR criteria.  Receiver
operator characteristic (ROC) curve analysis was used to assess the PEST and
CONTEST questionnaires using physician diagnosis as the gold standard.
Sensitivity and specificity of potential cut points were examined.

Results: A total of 932 packs were sent out to
recruit 191 (20.5%) participants attending for assessment.  Of these, 169
(88.5%) had current or previous psoriasis.  Using physician diagnosis 17 (10.1%)
were found to have previously undiagnosed PsA, 90 (53.3%) were found to have
another musculoskeletal complaint and 62 (36.7%) had no musculoskeletal
problems. 

Using ROC curve analysis, all of the questionnaires showed a
significant ability to identify PsA.  The area under the curve (AUC) for the
CONTEST and CONTESTjt questionnaires was slightly higher than that of PEST (see
table 1) but there was no significant difference between any of the
questionnaires. Examining the sensitivities and specifities for the different
cut points, suggested that a PEST score of ≥2 would perform better in
this dataset, and optimal scores for CONTEST and CONTESTjt seemed to be 3 and 4
respectively.

Conclusion: The accuracy of the PEST and CONTEST
questionnaires to identify PsA appeared similar with slightly higher AUC for
the CONTEST questionnaires.  Optimal cut points in this study appeared lower
than in previous studies (PEST≥2, CONTEST≥3, CONTESTjt≥4).