Possible UIP Pattern on High-Resolution Computed Tomography is Associated with Better Survival than Definite UIP in IPF Patients

INTRODUCTION

Idiopathic pulmonary fibrosis (IPF) is a diffuse parenchymal lung disease of unknown cause characterized by progressive loss of lung function and a poor prognosis. Median survival is 2–3 years from the time of diagnosis¹. IPF is defined histopathologically by the usual interstitial pneumonia (UIP) pattern, after exclusion of connective tissue disease, drug toxicity, and occupational/environmental exposure. Current guidelines outline clinical, high-resolution computed tomography (HRCT), and histopathologic criteria consistent with a diagnosis of IPF¹. Diagnostic guidelines categorize HRCT findings as “definite UIP” pattern, “possible UIP” pattern, or “inconsistent with UIP” pattern. Definite UIP HRCT pattern in the correct clinical setting obviates the need for a lung biopsy to diagnose IPF¹. These radiologic patterns are further described in the “High Resolution Computed Tomography Evaluation” section under Methods.

Although IPF is invariably a progressive disease associated with a poor outcome, the individual disease course is heterogeneous. Older patients and those with greater physiologic impairment at baseline have a higher mortality risk². More extensive fibrosis on HRCT is also associated with increased risk of death^2–5. Knowledge of the HRCT pattern may add relevant prognostic information. Among patients with idiopathic interstitial pneumonias (including some with IPF), those with a UIP pattern on HRCT have shorter survival compared to those with alternative HRCT patterns^6,7. Evaluation of variably-defined HRCT patterns as predictors of survival in IPF cohorts have not replicated these findings^4,8–11. We sought to assess the prognostic value of the HRCT patterns defined in the 2011 ATS/ERS/JRS/ALAT guideline¹ on IPF diagnosis, as well as individual “inconsistent with UIP” HRCT features, in patients with biopsy-proven UIP and a clinical diagnosis of IPF.

MATERIALS AND METHODS

University of Michigan patients with UIP on a surgical lung biopsy (included biopsies were dated January 2001 to February 2013), a multidisciplinary team diagnosis of IPF, and a baseline HRCT, were retrospectively identified via the electronic medical record. Documentation of multidisciplinary case review was available for all subjects. Patient characteristics, forced vital capacity (FVC) and diffusing capacity of the lung for carbon monoxide (DLCO) were documented at the time of the clinical evaluation. Transplant-free survival time is time from the HRCT to death or lung transplantation. Vital status (including date of death or lung transplantation) was confirmed via the electronic medical record and Social Security Death Master File, censored on November 17, 2014 (3 months prior to the registry search to account for reporting lag). This study was conducted in accordance with the amended Declaration of Helsinki (http://www.wma.net/en/30publications/10policies/b3/). The Institutional Review Board at the University of Michigan (HUM #00073617) approved this protocol, and need of written informed consent was waived by the IRB. Some of these results were previously presented as an abstract¹².

RESULTS

One-hundred and thirty-three patients met inclusion criteria, of whom 41 (31%) had definite UIP pattern, 77 (58%) had possible UIP, and 15 (11%) had inconsistent with UIP pattern. As shown in Table 1, mean age at the time of the HRCT was 63.6 (SD 8.1) years, 63.2% (n=84) were male, 62.1% (n=82) were ever-smokers, and the mean baseline FVC %predicted was 67.3 (SD 15.8). The mean age and baseline pulmonary function were not significantly different across HRCT pattern categories. The proportion of patients falling into each Gender-Age-Physiology (GAP) index stage¹⁹ did not differ across HRCT patterns. Sex distributions differed, with the inconsistent with UIP group having fewer males (p=0.04) compared to the other HRCT patterns, and smoking history differed with a smaller proportion of possible UIP patients having ever smoked compared to other groups (p=0.003).

DISCUSSION AND CONCLUSIONS

In a retrospectively-assembled cohort of IPF patients with histopathologically-confirmed UIP, we evaluated the prognostic significance of the HRCT patterns outlined in the 2011 ATS/ERS/JRS/ALAT guidelines for diagnosis and management of IPF, and found that patients with a possible UIP pattern have a longer lung transplant-free survival compared to those with definite UIP, independent of age, baseline lung function, smoking status, and fibrosis extent. The inconsistent with UIP pattern appears intermediate to the definite and possible UIP patterns, but with just 15 patients (or 11% of the cohort) in this category, it is difficult to draw conclusions about this group. Further, addition of the HRCT pattern variable to a multivariable model including baseline age, baseline lung function, smoking status, and fibrosis extent resulted in improvement in the c-statistic, suggesting that knowledge of this variable improves model discrimination. The guidelines-based HRCT pattern is commonly referenced in HRCT reports, making it a readily available data point providing physicians and patients useful prognostic information beyond that encompassed by clinical variables^3–5.

The extent of fibrotic HRCT features including overall fibrosis score, honeycombing, reticular pattern, and traction bronchiectasis have been associated with poor prognosis in IPF and other fibrotic idiopathic interstitial pneumonias (IIPs)^2–5. In unselected patients with fibrotic IIP, a definite UIP HRCT pattern has been associated with significantly worse survival compared to possible or inconsistent with UIP patterns^6,7. In contrast, a difference in survival by overall HRCT pattern has not been consistently observed in patients with IPF^4,9,10. Sumikawa and colleagues evaluated similarly defined HRCT patterns (“Definite UIP”, “Consistent with UIP”, and “Suggestive of Alternative Diagnosis”) in 98 IPF patients with biopsy-confirmed UIP, and found no statistically significant difference in survival across these categories, although estimated survival times were longer for “Consistent with UIP” and “Suggestive of Alternative Diagnosis,” as compared to Definite UIP⁹. Yagihashi and colleagues evaluated IPF patients enrolled in clinical therapeutics trials and found no difference in survival among those with radiologically and histopathologically concordant findings (e.g. definite/probable UIP histopathology and definite UIP HRCT) compared to those with discordant findings (definite/probable UIP histopathology and Inconsistent with UIP HRCT)¹⁰. Arcadu and colleagues recently evaluated the association of 2011 guidelines-based HRCT patterns in IPF patients, and determined that after adjustment for age, gender, and baseline FVC and DLCO, survival was not significantly different between groups, despite “consistent with UIP” pattern having worse survival compared to possible/inconsistent patterns in unadjusted analysis¹¹. There are several factors contributing to differences in our findings compared to prior reports. First, inclusion criteria differed. Arcadu and colleagues¹¹ did not require histopathologic confirmation of a UIP pattern, and in fact only 54% in the possible UIP group had a diagnostic lung biopsy. Second, assignment of HRCT pattern likely differed, despite similarly described methodology. Presence of radiologic honeycombing is reported in 100%, 39%, and 24% of consistent, possible, and inconsistent groups in the Arcadu study¹¹, compared to 95%, 4%, and 27% in our definite, possible, and inconsistent groups. If radiologic honeycombing is the defining feature of a phenotype with poor outcomes, the higher prevalence of honeycombing in Arcadu and colleague’s possible UIP group may obscure survival differences when comparing to definite UIP. Of note, the handling of lung transplantation, an important event competing with death in this population, is not specifically mentioned in several reports^6,7,9, or was handled via censoring¹¹. We utilized death or lung transplant as the primary endpoint, making direct comparison with these studies difficult.

Additionally, we elected to use a traction bronchiectasis score as a measure of fibrosis extent to adjust for baseline disease severity, along with baseline FVC and DLCO. This avoids problems with collinearity when including a combined fibrosis score^20,21 (measures presence and extent of honeycombing, reticulation, and traction bronchiectasis), and HRCT pattern (defined by presence of honeycombing) in a single prediction model, as traction bronchiectasis is not integral in the distinction between the HRCT pattern categories. Traction bronchiectasis scores have been associated with increased risk of death in IPF patients^9,22, suggesting it is an appropriate surrogate for disease extent. Although a summary score of the number of lobes having traction bronchiectasis as we use here has not been validated as a predictor of transplant-free survival time in other cohorts, it was a significant univariable predictor of transplant-free survival time in this cohort.

Of HRCT findings considered inconsistent with IPF, mosaic attenuation and air trapping were more prevalent among patients in the inconsistent with IPF group compared to the definite/possible IPF groups, as were diffuse axial and craniocaudal distributions of abnormality. Yagihashi and colleagues similarly identified mosaic attenuation/air trapping as a common reason for classification in the discordant radiologic/histopathologic group, and it was not significantly associated with increased risk of death in their report (unadjusted HR=1.194, CI 95% 0.265–4.006, p=0.79)¹⁰. Similarly, we found no association between presence of mosaic attenuation and risk of death or lung transplant (HR 1.92, CI 95% 0.87–4.23, p=0.11), but with just 10 patients (7.5% of the cohort) having mosaic attenuation, power to detect a difference in hazards is limited. Ground glass was prevalent across HRCT category groups, and in unadjusted and adjusted Cox models was associated with decreased hazard of death. The reason for this association is unclear, and is counter to the observation in several other studies where ground glass attenuation was associated with increased hazard of death²², or not associated with any difference in survival^9,20. Ground glass opacities on CT scan are not consistently associated with any underlying histopathologic finding (e.g. fibrotic or inflammatory changes), a possible explanation for some of the variability in its association with outcomes.²¹ Understanding the clinical relevance of individual features considered inconsistent with UIP/IPF in the setting of biopsy-proved UIP has important diagnostic and treatment implications and requires further study.

Strengths of our study include availability of histopathology confirmation of UIP/IPF in all patients, and vital status confirmation via two sources (electronic medical record and Social Security Death Master File) to ensure complete capture of events. There were several weaknesses. Inclusion of patients collected from a single tertiary referral center, potentially limiting the generalizability of our findings. Additionally, only 95% of patients assigned by interpreting radiologists to the definite UIP HRCT category had honeycombing (and 5% did not have honeycombing), and conversely, 4% of patients assigned to the possible UIP HRCT category did have honeycombing, indicating that assignment of HRCT pattern based on guideline definitions was not perfect. Further, we had just one expert thoracic radiologist (of a panel of four) review each HRCT, rather than use consensus to determine presence of HRCT features and overall pattern, as was done in previous studies^4,9–11. There is recognized variability in interpretation of HRCT findings, with inter-observer agreement on categorization of overall HRCT pattern as defined in the 2011 consensus guidelines reported as moderate (weighted kappa 0.48 and 0.52 for general and thoracic radiologists, respectively)²³. We are unable to report on agreement across the various readers in this study.

An additional weakness of this study is the retrospective nature of patient collection. We do not prospectively document clinical characteristics such as non-pulmonary symptoms (e.g., gastroesophageal reflux disease, joint swelling, Raynaud phenomenon) as part of our multidisciplinary team review, and measurement of autoantibodies is at the discretion of the treating pulmonologist, rendering us unable to objectively compare these important clinical characteristics across HRCT groups. Second, we do not have comprehensive, prospectively collected data on treatments rendered during the study period. It is likely that some patients received immunosuppressant drugs or antifibrotic agents within or outside of clinical trials.^24–26

In conclusion, we found that a possible UIP pattern on HRCT is associated with reduced hazard of death or lung transplant compared to definite UIP pattern, and this effect persists after adjusting for baseline age, pulmonary physiology, and extent of traction bronchiectasis. The 2011 consensus guideline-based HRCT pattern may represent a widely available radiologic measure informative of prognosis in IPF patients, and possibly driven by the presence of honeycombing in almost all patients categorized as definite UIP compared to very few in the possible UIP category. Validation of these findings in a separate cohort of patients is prudent prior to broad application in practice.

Supplementary Material