Psoriatic arthritis (PsA) is a chronic inflammatory disease with an extremely heterogeneous phenotype that may include musculoskeletal, skin, and systemic involvement and several comorbidities that often complicate the clinical picture. This diverse presentation contributes to the complexity of treating PsA, necessitating optimal management of all domains that constitute its clinical picture in each patient (1).

Treatment strategies for PsA have significantly improved with the application of treat-to-target and personalized medicine approaches, as well as the development of new drug classes, such as biological and targeted-synthetic disease-modifying anti-rheumatic drugs (b/tsDMARDs) (2–4). However, there is still a high proportion of patients who do not achieve adequate disease control or who experience secondary ineffectiveness or drug-related adverse events after an initial good clinical response (5, 6). Accordingly, real-life cohorts still include a significant proportion of PsA patients refractory to multiple targeted drugs (7, 8). In this scenario, managing these difficult-to-treat (D2T) patients still represents an unmet need and a significant challenge for the rheumatology community.

In 2021, the European Alliance of Associations for Rheumatology (EULAR) defined D2T rheumatoid arthritis (RA) (9), while international scientific societies have not yet elaborated on an official definition of D2T-PsA. A first attempt was made to adapt the criteria for D2T-RA to PsA, resulting in a preliminary definition from a single-center observational cohort that identified 33.9% of patients as D2T-PsA (10). However, being designed for a different disease, the definition of D2T-RA could only partially apply to a clinical scenario as complex and multifaceted as PsA. Building on the foundational work of Perrotta et al., various authors have defined the profile of D2T PsA patients, identifying varying demographic and clinical characteristics associated with this peculiar condition (11–14). Given the complexity of psoriatic disease, where a one-size-fits-all approach does not seem feasible (15), a recent review of the literature conducted to inform a research project on this topic revealed that a consensus on the definition of active disease and the number of failed therapies to define D2T PsA patients is still lacking (6).

Furthermore, the lack of any reference to the time frame in which the failures occurred within the natural history of the disease may undoubtedly represent a limitation in properly identifying D2T patients.

The study aimed to test the use of the preliminary D2T-PsA definition (10) within a single-center observational cohort of PsA patients treated with b/tsDMARDs. In addition, we aimed to identify a numerical index defining D2T-PsA based on the number of b/tsDMARDs failed compared to the number of b/tsDMARDS cycles, considering a (worse) scenario with a high rate of therapy failure.

In our study, we aimed to assess the applicability of a proposed definition of D2T-PsA, which borrows its criteria from the EULAR D2T-RA definition. In our cohort, we observed an extremely low proportion of D2T-PsA patients, about 3% of our cohort, which is approximately 10-fold lower than that published by Perrotta et al. (10) in a similar analysis. This result was unexpected, as we anticipated a higher proportion of difficult-to-manage patients in our PsA cohort.

Hence, we aimed to study the effect of covariates on a numerical continuous D2T score, which differs from the mere presence of a condition by quantifying the association of each covariate with the degree of treatment difficulty.

Thus, we developed a statistical model to intercept prognostic factors associated with more complex management. Using this model, we identified 17.2% of patients as potential D2T-PsA. This proportion reduced to 8.4% when we applied the predicted actual-to-expected number of cycled b/tsDMARDs. The use of this model allowed us to investigate the role of demographic and clinical factors as possible markers of a D2T condition. Concomitant fibromyalgia, certain types of skin involvement, as well as corticosteroid use, were positively associated with a higher treatment difficulty. The result of our study only partially aligns with the proposed definition (10) and the results from Philippoteaux et al. (11). In detail, fibromyalgia was associated with the D2T status in both analyses, according to previous research depicting fibromyalgia as an unfavorable prognostic factor in terms of treatment response (8) and therapeutic target achievement (16–18).

On the other hand, in our study, obesity was not associated with D2T-PsA (10, 11, 14) despite evidence pointing at a high BMI as an obstacle, lowering the response to treatment with TNF inhibitors (TNFis) (19, 20) and the achievement of remission or minimal disease activity (21, 22). However, the link between obesity and poor treatment response to TNFis does not apply to other available mechanisms of action, with even an inverse association with IL-17 blockade, which showed a better clinical response in overweight and obese PsA patients (23).

In our study, we observed a robust positive association between more complex disease management and certain types of psoriasis, namely palmoplantar psoriasis and nail involvement, coherently with published data, showcasing how skin psoriasis, and in particular palmoplantar psoriasis, detrimentally affects patients’ quality of life (24). Similarly, cutaneous involvement in terms of extension of the skin disease was found to be associated with a difficult-to-treat condition in psoriatic arthritis patients in the work by Vassilakis et al. (14), although with no mention of the phenotype of skin psoriasis. Furthermore, our study confirms that psoriatic nail involvement imposes a higher burden of joint inflammatory disease (25) and that PsA patients who manifest both skin and joint disease have worse clinical outcomes than those with only musculoskeletal-related symptoms (8, 26).

These data underscore the deep groove between PsA and RA as clinical entities, ultimately advocating for a future PsA-D2T definition that embraces the multifaceted nature of PsA, overcoming the comfort of the D2T-RA shadow (15, 27).

Another distinguishing feature of our model compared to the previously proposed D2T definition is the criterion concerning the number of mechanisms of action a patient must fail to meet. Not all treatments for PsA are equally effective across different disease domains, meaning patients with multi-domain involvement may face a reduced therapeutic arsenal, potentially falling short of the D2T multi-refractoriness threshold (27).

In line with this, our exploratory study aimed to depart from the framework of the D2T-RA definition, shifting the critical element of our proposed model. Instead of centering a possible D2T definition solely on refractoriness to multiple b/tsDMARDs, we redefined refractoriness from a simple cumulative count of failed drugs to an algebraic construct. In this model, a higher ratio between observed and expected cycles through different therapies reflects greater management complexity, regardless of the number of mechanisms of action previously attempted.

This approach accounts for a future where PsA patients may receive therapies with multiple mechanisms of action. The crux of our definition lies in the relative urgency with which clinicians must switch or swap therapies. Consequently, even patients treated over extended periods with multiple drugs sharing a single mechanism of action (such as the previously limited options for TNFα inhibitors) can still be considered difficult to treat. From our perspective, this shift represents a necessary and pivotal evolution, redefining the core of the D2T criteria in a subtle yet meaningful way.

From a methodological point of view, it is important to note that we used the 75th percentile of the therapy-failure rate distribution to calculate the number of expected b/tsDMARDS. Adjusting this threshold would affect the sensitivity and specificity of the index. However, while the index is threshold dependent, the regression coefficients—expressed in terms of relative treatment difficulty—remain robust and reliably indicate varying degrees of treatment complexity.

Another noteworthy aspect of our index and definition is that a patient could qualify as D2T even in the absence of current signs and symptoms of active disease. This highlights the model’s focus on the overall therapeutic journey and management challenges, rather than solely on active disease manifestations.

Accordingly, a patient with a history of treatment failures in greater numbers than expected should be considered as D2T, even if, at the time of assessment, the disease is apparently in good control since the efficacy of the current therapy does not erase the non-response to previous therapies and the difficulty in identifying a treatment capable of producing an effective response. Moreover, a patient could experience sufficient control over his/her musculoskeletal manifestations while simultaneously suffering from poorly controlled psoriasis, inflammatory bowel disease, or uveitis, extra-articular manifestations often under-reported in many uni-dimensional indexes (28). Similarly, the achievement of a minimal disease activity state could still potentially leave room for up to two unhampered disease domains. Such a discrepancy between the actual disease domain activity and the composite indices score underscores a profound distance between RA and PsA.

Indeed, the centrality of the rapidity of therapy switching rather than the number of previously failed mechanisms is remarked upon also by the fact that in our cohort, the majority of difficult-to-manage patients do not reach the threshold number of failed mechanisms of action set by the definition proposed (10). At the same time, many of them meet the new definition we propose with our index. Even though our analysis did not focus on the reason for therapy failures, the need for a more rapid sequential switching of b/tsDMARDs could most likely be due to a succession of primary non-responses attributable to a biological mismatch between the drug target and the disease target (29), which theoretically identifies a more difficult-to-manage pattern than the same number of therapeutic switches observed over a longer time frame. Ultimately, in line with the path marked out in the D2T-RA field (30), Lubrano et al. provided different definitions for patients failing therapies due to persistent residual inflammation (“refractory-to-treatment PsA”) and for those with suboptimal disease control due to the overall management of the inflammatory and non-inflammatory components of the disease (“D2T-PsA”) (31). Although the importance of exploring the presence of residual inflammation in complex diseases is of paramount importance in the treatment of complex cases, our model does not offer discrimination between difficult-to-treat patients with or without persistent inflammatory disease, thus refraining from any inappropriate distinction between difficulty-to-treat and refractoriness-to-treat in our cohort.

Although our study expands the understanding of the D2T-PsA concept, several limitations must be acknowledged. First, the relatively low sample size from a monocentric cohort and the study’s cross-sectional design partially limit the generalizability of the results. In addition, potential patient selection bias should be considered, as the majority of the PsA patients at our center are “articular-predominant,” with less severe extra-articular manifestations. Therefore, independent longitudinal studies with broader populations are warranted to validate our findings and better identify treatment modifiers, encompassing the multifaceted nature of psoriatic syndrome.

Moreover, while we employed a statistical model to identify patients with higher treatment difficulty, this approach is currently feasible only in research settings. In standard clinical practice, such a model would likely be too cumbersome to implement. Given the exploratory nature of our study, more studies are needed to confirm the prognostic value of the factors identified in this study as potentially associated with treatment difficulty. Future studies could refine the statistical model into a practical scoring algorithm for D2T PsA, making it more accessible for routine clinical use.

Nevertheless, it is important to emphasize the exploratory nature of our study, which aims to explore the prognostic factors best suited for identifying truly difficult-to-manage PsA patients. To achieve this, we adapted and reinterpreted the D2T-RA criteria, focusing on factors uniquely related to the psoriatic disease spectrum. Although our study does not provide sufficient evidence to establish a definitive D2T-PsA definition, it offers valuable insights suggesting that the D2T-PsA criteria should diverge from the D2T-RA framework, as we believe is necessary.