Interobserver agreement regarding the Fleischner Society diagnostic criteria for usual interstitial pneumonia patterns on computed tomography

Abstract

Objective: To assess interobserver agreement among radiologists regarding the current Fleischner Society diagnostic criteria for usual interstitial pneumonia (UIP) patterns on computed tomography (CT). Materials and Methods: Using the Fleischner Society criteria for UIP CT patterns, five raters, working independently, categorized the high-resolution CT (HRCT) scans of 44 patients with interstitial lung disease who underwent lung biopsy. The raters also evaluated the presence, extent, and distribution of the most relevant imaging findings, as well as indicating their level of confidence in the most likely diagnosis and in up to three diagnostic hypotheses. Results: There was moderate to substantial interobserver agreement regarding the UIP patterns on HRCT—kappa statistic (κ) = 0.59-0.61. Interobserver agreement for the binary scores was substantial (κ = 0.77-0.79), whereas that for the presence of honeycombing was almost perfect (κ = 0.81-0.96). There was agreement regarding at least one of the three diagnostic hypotheses in only 36.4% of the cases. For the level of confidence in the most likely diagnosis, there was only slight to fair agreement (κ = 0.19-0.21). Conclusion: Interobserver agreement regarding the current Fleischner Society CT criteria for UIP was moderate to substantial among raters with varying levels of experience. There was only slight to fair agreement regarding the diagnostic hypotheses and for the level of confidence in the most likely diagnosis.

Pneumonia intersticial usual. Estamos “falando a mesma língua” e “vendo as mesmas coisas” ao analisar a tomografia computadorizada?

“…Pelas questões levantadas, é essencial que ao darmos este diagnóstico tenhamos convicção de não somente estarmos, em nível individual, de acordo com os critérios estabelecidos, mas também de estarmos coletivamente dando o mesmo diagnóstico aos mesmos pacientes de forma reprodutível, e para tal, estudos como o de Westphalen et al (15) , publicado no número anterior da Radiologia Brasileira, calculando a variabilidade interobservadores ao diagnosticar PIU são extremamente necessários. Ao demonstrar concordâncias moderadas a altas entre os observadores ao classificar o padrão de PIU pode-se ter maior segurança da precisão ao http://dx.doi.org/10.1590/0100-3984.2022.55.3e1 Editorial classificar o padrão intersticial observado, o que, em última análise, contribui para a homogeneização dos diagnósticos e, consequentemente, dos tratamentos oferecidos aos pacientes com doenças pulmonares intersticiais fibrosantes, o que tem implicações prognósticas importantes (16) .…”

“…Embora a comparação entre estudos que analisam a concordância interobservadores seja difícil pelos seus desenhos diferentes, tem-se a impressão de uma melhora constante e progressiva dos resultados na avaliação das pneumonias intersticiais (15,17,18) . A causa desta possível melhora é difícil de ser determinada, porém, possivelmente advém do melhor entendimento das doenças intersticiais ao longo do tempo, associado a publicações frequentes de novos consensos, os quais transportam estes novos conhecimentos para a prática diária de forma objetiva, direta, organizada e padronizada, fornecendo instrumentos que, aliados a experiência e treinamento, levam ao aumento da precisão diagnóstica entre os radiologistas, e podendo, finalmente, culminar com melhora na acurácia diagnóstica das doenças intersticiais pulmonares.…”

Usual interstitial pneumonia. Are we “speaking the same language” and “seeing the same things” when analyzing computed tomography scans?

“…Although comparisons across studies that analyze interobserver agreement is difficult due to their different designs, there http://dx.doi.org/10.1590/0100-3984.2022.55.3e1-en Editorial is the impression of a constant, progressive improvement in the evaluation of interstitial pneumonia (15,17,18) . The cause of such an improvement is difficult to determine.…”

“…In view of the issues raised, it is essential that when we make a diagnosis of UIP we are convinced that we are not only acting in accordance with the established criteria at an individual level but are also collectively giving the same diagnosis to the same patients in a reproducible way. Therefore, there is an urgent need for studies like the one that was conducted by Westphalen et al ( 15 ) and published in the previous issue of Radiologia Brasileira, in which interobserver variability is taken into account when a diagnosis of UIP is being made. By demonstrating moderate to high interobserver agreement when identifying the UIP pattern, one can have greater certainty of accuracy when classifying the interstitial pattern observed, which, ultimately, contributes to the homogenization of diagnoses and, consequently, of the treatments offered to patients with fibrosing interstitial lung diseases, such homogenization having major prognostic implications ( 16 ) .…”