From Categories to Continua: Rethinking Usual Interstitial Pneumonia on <scp>HRCT</scp>

Usual interstitial pneumonia (UIP) remains the most powerful imaging predictor of outcome across fibrotic interstitial lung diseases (ILDs). For more than a decade, international guidelines have relied on categorical high-resolution computed tomography (HRCT) classifications to define UIP, with the goal of standardising diagnosis and improving diagnostic confidence [1]. In practice, however, these rigid categories often struggle to capture the heterogeneity and uncertainty encountered in real-world multidisciplinary team (MDT) discussions, where many scans sit uncomfortably between guideline-defined bins [2]. In a recent publication in Respirology, Mackintosh and colleagues propose an alternative approach: treating UIP not as a categorical diagnosis, but as a probabilistic imaging phenotype whose likelihood can be estimated and meaningfully linked to prognosis [3]. Using a large, well-characterised real-world ILD cohort from the Australian IPF Registry, the authors demonstrate that the likelihood of definite UIP on HRCT is strongly associated with transplant-free survival and disease progression, outperforming conventional guideline categories. Importantly, this relationship holds not only in idiopathic pulmonary fibrosis (IPF), but also across non-IPF fibrotic lung diseases. Current ATS/ERS/JRS/ALAT guidelines require HRCT scans to be assigned to discrete categories—definite UIP, probable UIP, indeterminate, or alternative diagnosis—based on the presence or absence of specific imaging features [1]. While this framework has improved diagnostic consistency, it also imposes artificial thresholds on what is, in reality, a continuous biological process. Experienced thoracic radiologists routinely express diagnostic confidence in probabilistic terms, yet this nuance is often lost when translating impressions into guideline-defined labels [4]. Mackintosh et al. operationalise this probabilistic approach by asking expert ILD radiologists to estimate the likelihood of definite UIP rather than assigning a single categorical label. This proved highly reproducible, with excellent interobserver agreement, and demonstrated strong prognostic discrimination. As the likelihood of definite UIP increased, so too did the risk of mortality and disease progression, independent of fibrosis extent and traction bronchiectasis. Notably, interobserver agreement for applying ATS/ERS/JRS/ALAT CT criteria for UIP has previously been shown to be only moderate, even among thoracic radiologists, highlighting inherent limitations of categorical HRCT assignment [5]. One of the most compelling aspects of this study is its relevance beyond IPF. While a UIP pattern has long been recognised as a poor prognostic marker in IPF, its significance in non-IPF fibrotic ILD has been less clearly defined. Here, the authors show that increasing UIP likelihood confers worse outcomes even in patients with MDT diagnoses other than IPF, reinforcing the concept that UIP reflects shared fibrotic biology rather than a diagnosis confined to a single disease entity. As therapeutic options for fibrotic ILD expand, the clinical importance of accurate prognostic stratification is increasing. Randomised trials in progressive pulmonary fibrosis, including INBUILD and more recent phase 3 studies of the phosphodiesterase 4B inhibitor nerandomilast, have demonstrated treatment benefit across both IPF and non-IPF fibrotic lung disease [6, 7]. In this evolving landscape, imaging approaches that more precisely capture disease behaviour, such as likelihood-based assessment of UIP on HRCT, may have increasing value in identifying patients at higher risk of progression. Similarly, recent work in fibrotic hypersensitivity pneumonitis has shown that a “typical fibrotic HP” pattern on CT is associated with significantly worse transplant-free survival [8]. Together, these observations reinforce the principle that outcome prediction in fibrotic lung disease is increasingly informed by imaging phenotype and disease behaviour rather than diagnostic category labels alone. The clinical implications of a probabilistic UIP framework are potentially far-reaching. MDTs are increasingly tasked with making management decisions—such as antifibrotic therapy initiation or transplant referral—based not solely on diagnosis, but on projected disease behaviour [9]. A reproducible, prognostically meaningful estimate of UIP likelihood could enhance risk stratification and support more nuanced discussions with patients regarding prognosis. Several limitations deserve acknowledgment. The retrospective design limits causal inference, and external validation in independent cohorts is warranted. The approach also relies on subjective expert assessment, raising questions about training, calibration, and scalability. This evolution toward confidence-based, multidisciplinary diagnosis is reflected in the recent ERS/ATS update of the international classification of interstitial pneumonias, which explicitly incorporates diagnostic confidence into patient evaluation and management [10]. While prognostic discrimination is clearly improved, the extent to which this approach influences management decisions and downstream patient outcomes warrants further prospective evaluation. Beyond the pattern of fibrosis itself, other imaging features—such as mediastinal lymph node enlargement—have been shown to independently predict transplant-free survival, hospitalisation risk, and mortality in ILD cohorts [11]. Similarly, although emphysema was included in multivariable analysis, further exploration of the prognostic interaction between emphysema burden and combined pulmonary fibrosis and emphysema phenotypes may offer additional prognostic refinement [12]. This work does not invalidate existing guidelines, nor does it suggest that categorical HRCT classification should be abandoned. Rather, it highlights an opportunity to better align diagnostic frameworks with contemporary clinical practice. As fibrotic ILD management continues to evolve toward risk-based, behaviour-driven strategies alongside an expanding range of therapeutic options, imaging paradigms must evolve in parallel [13]. Future directions may include structured training in probabilistic imaging assessment, integration of quantitative and artificial intelligence–derived metrics, and prospective studies examining whether likelihood-based approaches improve decision-making and outcomes. For now, the message is a pragmatic one: UIP is not a binary entity, and acknowledging its continuum may allow more accurate prognostic assessment and support more appropriate multidisciplinary treatment decisions for individual patients. The author has nothing to report. Professor Dodd has received grant support from the Irish Lung Foundation, the St. Vincent's Hospital Group Foundation, University College Dublin and the FP7 Program of the European Commission for the randomised multicenter DISCHARGE trial (603266-2, HEALTH-2012.2.4.-2); is an Associate Editor of Radiology, and the Quarterly Journal of Medicine; is an Editorial Board member of Radiology Cardiothoracic Imaging; receives royalties as a co-author in the Amirsys Series Diagnostic Imaging—Cardiovascular and Amirsys Series Diagnostic Imaging- Emergency Radiology has received speaker fees from boehringer ingelheim.