Psoriatic arthritis (PsA) presents significant challenges, with many patients experiencing inadequate treatment response. While clinical aspects of suboptimal disease control have been characterized, the emotional burden of treatment failure remains poorly understood. Our objective was to systematically characterize the emotional patterns expressed by PsA patients when describing treatment failures and management challenges. A multi-national online survey with PsA patients was conducted through GRAPPA (Group for Research and Assessment of Psoriasis and Psoriatic Arthritis). The survey was conducted to explore their perspectives on how to define complex-to-manage (C2M)-PsA and treatment-refractory (TR-) PsA, and how these conditions impact their lives. Open-ended responses were analyzed using a locally deployed Meta Llama 3.2 3B Instruct large language model, applying Ekman's six basic emotion framework and the valence-arousal dimensional model. Sentiment analysis revealed predominantly negative emotional patterns (mean valence -0.57 ± 0.25) with moderate-to-high arousal (0.65 ± 0.10). Sadness was most prevalent (54.27%), followed by fear (43.59%). Disease duration (≥ 5 years) showed similar valence scores to shorter durations. The number of reported challenging disease aspects negatively correlated with valence (r = -0.16, p = 0.02) and positively with fear (r = 0.15, p = 0.02). Axial involvement and comorbidities were significantly associated with fear as the dominant emotion. Hierarchical clustering identified four distinct emotional profiles, with high multi-domain burden clusters showing the most negative valence. Natural language processing-based sentiment analysis characterizes emotional patterns in PsA patient narratives. Negative emotions, particularly sadness and fear, predominated, with specific disease features linked to these emotional states, underscoring the need for approaches addressing both physical and emotional dimensions in PsA.
Keywords: Difficult to treat; Emotional patterns; Large language models; Patient narratives; Psoriatic arthritis; Sentiment analysis.
© 2026. The Author(s).