Probable usual interstitial pneumonia pattern on chest CT: is it sufficient for a diagnosis of idiopathic pulmonary fibrosis?

Study design and population

This study was approved by the institutional review board of Tosei General Hospital (Seto, Aichi, Japan). Informed consent was not required as the data were collected retrospectively and anonymously analysed.

A retrospective review of consecutive patients with ILD evaluated at Tosei General Hospital from January 2008 to March 2013 was performed. Patients included were those who underwent an initial workup for a suspected diagnosis of IIP and whose CT scans at initial workup were available. Patients excluded were those with an indeterminate for UIP or an alternative diagnosis pattern on CT [2], a diagnosis of connective tissue disease (CTD), chronic hypersensitivity pneumonitis (CHP) or other identifiable causes of ILD, those with concurrently associated malignant diseases, those lost to follow up within 1 year of the initial workup for reasons other than death and those who did not undergo a pulmonary function test (PFT) at the initial workup. All patients with clinically suspected IIP prior to SLB were included in the analysis, regardless of final diagnosis.

All patients with a UIP pattern or a probable UIP pattern on CT were included in the statistical analyses and we compared survival time and time to first AE between the two CT patterns. The same analyses were also performed in patients with a diagnosis of IPF and between IPF and non-IPF diagnoses in patients with a probable UIP pattern.

Data collection

Patient characteristics and test results were collected retrospectively from clinical charts. Initial workup data including PFTs, arterial blood gas analyses and CT scans conducted within 1 month were collected. Forced vital capacity (FVC) and diffusing capacity of the lung for carbon monoxide (D_LCO) were expressed as percentages of the predicted values [9, 10]. The final follow-up date was December 31, 2016. The date of initial evaluation, date of last follow-up with survival status and date of diagnosis of the first AE were also recorded.

Definition of an AE

An AE was defined as a clinical event meeting all the following criteria [11]: 1) an acute respiratory deterioration characterised by evidence of new widespread alveolar abnormality; 2) acute worsening or development of dyspnoea within 1  month; 3) CT scan of the chest with new bilateral ground-glass opacity superimposed on a background pattern consistent with UIP; and 4) deterioration not fully explained by cardiac failure, fluid overload or infection.

Radiological and histopathological evaluation

All CT scans were independently reviewed by two experienced thoracic radiologists (TJ and HS) and an experienced thoracic physician (YK). To evaluate inter-observer variability, scans from 100 patients chosen at random from the study population were reviewed by each reviewer and Fleiss' κ-value was calculated. Published criteria were used to categorise CT scans as definite, probable or indeterminate for UIP pattern, or as findings of an alternative diagnosis [2]. Furthermore, each reviewer was blinded to the clinical and histopathological information. Given the κ-value (0.66, 95% CI 0.52–0.79), radiological interpretation of eligible cases was performed by one of the three reviewers.

A subset of patients had undergone SLB and all lung biopsy specimens were anonymised for histopathological analysis, and reviewed by two experienced pulmonary pathologists (JF and TT) who were blinded to clinical and radiological information and consulted each other about histopathological decisions. Published criteria were used to categorise histopathological findings [2].

Final diagnoses of ILDs

As the institute where the study was conducted was a local public hospital as well as a tertiary referral hospital, almost all patients were seen at least once every 3 months in the first year after initial workup, and every 3–6 months in the following years. All information regarding emergency department visits and/or hospitalisation was also captured. Under these circumstances, all available follow-up information was reviewed when making the final clinical diagnoses.

When histopathological information was available, the final diagnoses were established through multidisciplinary discussion (MDD) by clinicians, radiologists and pathologists with reference to the published criteria for IPF [2] and the classification for IIPs [12]. During the MDD, the clinical context and course, chest imaging and histopathology (if available) were reviewed in order to provide a final diagnosis. A diagnosis of “unclassifiable ILD” was made for patients with a non-IPF final diagnosis and histopathological features suggesting an alternative diagnosis (cellular inflammatory infiltration apart from areas of honeycombing, prominent lymphoid hyperplasia including secondary germinal centres, or a distinctly bronchiolocentric distribution including extensive peribronchiolar metaplasia). A diagnosis of CHP was made when histopathological features and clinical course were compatible, even if a history of exposure to specific antigens was lacking.

If histopathology was not available, the diagnosis of IPF was made when the clinical context was that of an IIP and the CT showed a UIP pattern. The patients with a probable UIP pattern on CT but no histopathology were provisionally classified as IPF when the likelihood of the diagnosis was believed to be >70% [12], while being classified as unclassifiable ILD when the clinical context was that of an IIP and clinical course did not suggest a specific diagnosis [12–14]. Findings suggestive of CTD, CHP and other identifiable causes or associations were actively sought out at the time of initial workup and during follow-up. We validated the clinical unclassifiable ILD diagnoses using a second approach (supplementary material).

Statistical analysis

Continuous variables were presented as medians with interquartile range (IQR) as they showed a non-normal distribution. Categorical variables were summarised by number of patients and percentage. To assess the differences in variables between subgroups, the Mann–Whitney U-test was used for continuous variables and the Pearson Chi-squared test was used for categorical variables. Survival time and time to first AE were measured from the date of initial workup until the date of death or the first diagnosis of AE, respectively. The Kaplan–Meier method was applied to show unadjusted survival curves, and mortality rate and AE rate were shown per 100 person-years.

Cox proportional hazards models were used to evaluate the associations between CT pattern or IPF diagnosis and survival time. The prognostic impacts were adjusted by demographic information generally associated with survival time: age, gender, baseline FVC [15], baseline D_LCO [16–18] and use of anti-fibrotics [4, 19]. Competing risk analyses were used to evaluate the associations between CT pattern or IPF diagnosis and time to first AE, in order to account for competing risks such as death from disease progression without AE. The predictive values for time to first AE were adjusted by variables that have been reported to be associated with AE: baseline FVC [20–23], baseline D_LCO [21, 24] and use of anti-fibrotics. Multiple imputation analyses were performed to estimate survival time and time to first AE under the assumption of missing data. We carried out some imputations using all potential predictor variables in each imputation model. The overall proportion of patients with missing data was 2%, in all of whom D_LCO could not be assessed at baseline due to excessively low pulmonary function.

The adjusted annual rate of change in FVC was analysed with the use of a random coefficient regression model (with random slopes and intercepts) that included sex, age and height as covariates. The impact of CT pattern and IPF diagnosis were determined using estimated slopes for each study group on the basis of the time-by-treatment interaction term from the mixed model. Missing data were assumed to be missing at random and were not imputed.

All statistical tests were two sided and p-values of less than 0.05 were considered statistically significant. Statistical analyses were carried out using SPSS version 25.0 (SPSS Inc., Chicago, IL, USA) and Stata version 13.1 (StataCorp, College Station, TX, USA).