We have recently shown that anxiety and depression are common comorbidities for people with interstitial lung disease (ILD). In a cross-sectional single-centre study, the prevalence of anxiety was 31% and the prevalence of depression was 23% [1]. Anxiety and depression were not related to physiological parameters; however, dyspnoea and number of comorbidities were important contributors. The aims of this study were to determine the frequency of prolonged anxiety and depression among sufferers of idiopathic pulmonary fibrosis (IPF), and factors contributing to their persistence.Using the Australian IPF Registry, we obtained data from all individuals who had completed the self-reported Hospital Anxiety and Depression Scale (HADS) at baseline and at 12 months of follow-up [2,3]. HADS scores were classified according to standard criteria where a score >10 indicates a probable case of clinically significant anxiety or depression and a score of 8-10 indicates borderline probability. We defined prolonged anxiety and depression as being present when subjects had HADS scores ⩾8 at baseline and at 12 months follow-up. We defined worsening risk of anxiety and depression as moving from "no case" to "borderline" or "case", or moving from "borderline" to "case", at 12 months.Analysis included demographic and physiologic data, and patient-reported medical comorbidities. Breathlessness was measured using the UCSD shortness of breath questionnaire (UCSDSOBQ) which, like the HADS scale, explores symptoms over the preceding week [4]. Cough severity was rated using an unvalidated visual analogue scale (VAS), measured between 0 and 100 mm, with anchors of "no cough" and "worst cough imaginable" The persistence of anxiety and depression over 12 months of follow-up was evaluated using the Fisher's exact test. Univariate relationships were tested using Pearson's Chi-squared test for categorical variables and one-way analysis of variance for continuous variables. Variables with a relationship to prolonged anxiety or depression ( p<0.1) were entered into a logistic regression model. Model fit and the proportion of participants correctly classified by the model were evaluated using the Chi-squared test.