The current study was performed according to the Strengthening the Reporting of Observational Studies in Epidemiology [17] (Supplementary Table S1).
Study design and setting
This retrospective, multicentre cohort study was conducted at eight tertiary care hospitals in Japan. A secondary analysis was performed using the same dataset used in previous studies [18], which was constructed by directly accessing the electronic medical record.
Patients
The source population comprised patients aged 40 years and older who were admitted to eight hospitals from January 2016 to February 2019. We included patients who might presented with AE-IPF as assessed using an algorithm based on the International Statistical Classification of Diseases and Related Health Problems, 10th Revision (ICD-10) (Supplementary Table S2). Patients with a claim of secondary interstitial pneumonia (e.g. connective tissue disease-associated interstitial lung disease and chronic/fibrotic hypersensitivity pneumonitis) or malignancy were not included (Supplementary Table S3). Next, the electronic medical records including chest high-resolution computed tomography (HRCT) findings were reviewed to determine whether the participants met the diagnostic criteria for AE-IPF. A diagnosis of AE-IPF was made by two respiratory physicians according to the American Thoracic Society/European Respiratory Society/Japanese Respiratory Society/Latin American Thoracic Association clinical practice guidelines for IPF [19], and the diagnostic criteria of the International Working Group for AE-IPF [20], which was confirmed by two radiologists by evaluating chest HRCT findings independently. The inclusion criteria were reported previously [18]. In brief, we combined the assessment of ICD-10 codes (e.g., J84.1, J84.9) and chart reviews about HRCT findings. We excluded patients with 1) secondary interstitial pneumonia, 2) comorbid advanced-stage cancer, 3) unilateral pneumonia, pneumothorax or pulmonary embolism upon admission, 4) refusal of treatment and 5) those whose ESMCSA or PMCSA could not be assessed.
Quantitative image analyses
Data about CT scan were obtained at admission due to AE-IPF were collected. The procedure was performed with 1-5-mm-thick samples at 1-5-mm slice intervals. CT images constructed using the standard algorithm at mediastinal setting were used for the quantitative analysis of skeletal muscles. We used a single transverse slice at the lower margin of the 12th thoracic vertebra for the quantitative measurement of ESM (Fig. 1A). The superior aspect of the aortic arch was identified. Then, the first transverse image above the arch was used for the quantitative assessment of PMs (pectoralis major and minor) (Fig. 1B). Both methods were similar, as in previous studies [3,6,7,12]. The bilateral ESMs and PMs were identified and manually shaded using a predefined attenuation range of 29–150 HU. The sum of the bilateral muscle areas was adopted as the ESMCSA and PMCSA. Skeletal muscles were evaluated independently by trained respirologists (Y. I. and N. A.) who were blinded to the clinical data. The average of the two scores was obtained to collect data in each case. All images were analysed using a commercially available software SYNAPSE VINCENT version 3 (FUJIFILM Medical Systems, Tokyo, Japan). Based on a previous study, patients were classified into the high and low groups with cut-off values of 25.6 cm2 for ESMCSA and 20.4 cm2 for PMCSA [12].
Outcome
The primary outcome was the time taken for 90-day mortality.
Data collection
We collected data about demographic characteristics and clinical, laboratory and HRCT findings at the baseline. Two radiologists calculated the HRCT score using HRCT images, as reported in previous literature [21]. Based on previous studies [15,22,23] and clinical experience, the following were considered as confounding variables: age, sex, BMI, long-term oxygen therapy (LTOT) use before AE, PaO2/FiO2, and serum LDH, C-reactive protein (CRP) and serum KL-6 level.
Sample size
It is challenging to estimate short-term prognosis according to skeletal muscle mass in AE-IPF due to the lack of prior literature. Therefore, we aimed to collect as many cases as possible without calculating the sample size.
Statistical analyses
Continuous variables and categorical variables were presented as median (interquartile range) and frequency (%), respectively. Missing data were imputed via multiple imputation using chained equations. Twenty datasets were imputed, and we combined the results using the Rubin’s rules [24]. The survival curves of ESMCSA in the low and high groups were generated using the Kaplan–Meier method. We conducted the log-rank test to compare the Kaplan–Meier curves of each group. Next, we performed multivariable analysis using three Cox proportional hazard models with an adjustment for potential confounders below: model 1, age, sex and BMI; model 2, model 1 + LTOT before AE and PaO2/FiO2 and model 3, model 2 + serum LDH, CRP and KL-6 levels. We performed all statistical analyses using STATA/SE version 16.0 (Stata Corp., College Station, TX, USA).
Ethical considerations
This retrospective study was approved by the Ethics Committees of Kyoto University Graduate School and Faculty of Medicine (R2071), and the Ethics Committes of Saiseikai Kumamoto Hospital, Hyogo Prefectural Amagasaki General Medical Center, Iizuka Hospital, Kobe City Medical Center General Hospital, Kameda Medical Center, Hoshigaoka Medical Center, Okinawa Chubu Hospital, and Japanese Red Cross Medical Center. Because this study was a retrospective observational study, the need for written informed consent was waived by the Ethics Committees of Kyoto University Graduate School and Faculty of Medicine. The patients were provided with an opportunity to opt out. All the methods were performed in accordance with relevant guidelines and regulations.