Centrilobular emphysema combined with pulmonary fibrosis results in improved survival: a response

Please see related letter http://fibrogenesis.com/content/4/1/17 webciteWe read with great interest the somewhat thought-provoking article by Todd et al. [1], which addressed the prognosis of patients with combined pulmonary fibrosis and emphysema (CPFE). The authors found that survival in CPFE was better than that of patients with idiopathic pulmonary fibrosis (IPF) without emphysema or with minimally extensive emphysema ('lone fibrosis'), in striking contrast to previous results [2-4]. From these counterintuitive results, it is argued that tobacco-smoking-induced pulmonary inflammation may be protective against the progression of fibrosis. However, in reaching this conclusion, we feel the authors have failed to sufficiently consider a number of potential biases that are evident from the stated data.Firstly, amongst biopsied patients, fibrotic non-specific interstitial pneumonia (NSIP) was more prevalent in association with CPFE (6 of 20, 30%) than in the remaining groups (7 of 53, 13%). If these proportions are extrapolated to the whole cohort, including non-biopsied cases, the prevalence of fibrotic NSIP was significantly higher in the CPFE cases. Furthermore, this imbalance may have been amplified by the inclusion of non-biopsied cases: it is well recognized that NSIP when combined with emphysema may mimic honeycombing and can, therefore, be confused with genuine IPF [5]. A bias towards NSIP would also explain the unexpectedly higher female rate in the CPFE group.Secondly, the way in which transplantation was handled in the survival analysis is very open to question. It has been usual to censor at date of transplantation, and not to handle a transplant event as a death. The authors point out that the rate in transplantation did not differ between groups. However, as at least a third of 'deaths' (n = 34) were transplant episodes, it is likely that this way of handling transplant events as deaths distorted the Kaplan-Meier curves.Thirdly, potential lead-time bias