Will Adding Fibrotic Score to the GAP Score Help Predict Patient Survival?

See also the article by Chahal et al in this issue.

Jonathan H. Chung, MD, is an associate professor of radiology, section chief of thoracic radiology, and vice chair for quality of the department of radiology at the University of Chicago. His primary research focus is how imaging can help patients with chronic lung diseases such as interstitial lung disease, pulmonary fibrosis, occupational lung disease, and nontuberculous mycobacterial pneumonia. Dr. Chung received the RSNA Honored Educator Award in 2013 and the Marc Tetalman Award in 2016. He is the pulmonary imaging deputy editor for Radiology: Cardiothoracic Imaging.

Introduction

Idiopathic pulmonary fibrosis (IPF) is one of the most common and is certainly the most fatal subtype of pulmonary fibrosis. IPF diagnosis is achieved through detailed history and clinical evaluation to exclude known causes of pulmonary fibrosis. Unfortunately, diagnosis is often challenging, with disease in many patients being initially misdiagnosed as other conditions, and assessment by multiple health care professionals is required before an accurate diagnosis can be made. A diagnosis of IPF portends poor prognosis with median survival between 3 and 5 years. However, the natural history of IPF varies from individual to individual. Indeed, some patients may have an acute, inexorable course and die within a year, while others may live with IPF for up to a decade. However, a sobering fact is that all cases of IPF are fatal. Until recently, there was no effective medical therapy for IPF. Now there are two medications that have been repeatedly shown to decrease the rate of worsening in patients with IPF and likely also improve survival, which has invigorated interest in IPF diagnosis and management.

The variable clinical course of patients with IPF can be challenging in regard to patient management. Reliable predictors of survival would be extremely helpful in timing lung transplantation. Judicious planning for lung transplantation is essential in IPF, given that it is the only known cure. Appropriate lung transplantation preparation is especially poignant in IPF given the high waiting list mortality associated with this condition (1). Understanding of individual patient prognosis would also guide use of therapeutic options and palliation timelines as other important aspects of IPF management. However, no single indicator has been shown to be reliable enough in isolation to be of practical use. Extent of fibrosis, CT honeycombing, traction bronchiectasis, mediastinal lymphadenopathy, older age, male sex, worse pulmonary function, greater dyspnea score, and worse 6-minute walk test result have each been associated with worst prognosis in patients with IPF (2–6).

The best validated mortality prediction model is currently the gender, age, and physiology (GAP) model which assigns an additive score based on patient sex, age, and forced vital capacity, as well as diffusion capacity for carbon monoxide (DLCO) (4,7,8). The GAP model also allows estimates of 1-, 2-, and 3-year mortality (10). Previously, total fibrosis score (combination of percentage of lung affected by reticulation and honeycombing) was shown to be an adequate substitute for DLCO in the GAP model (9). However, using fibrosis score as an adjunct to the GAP model has not been described in the medical literature.

In a retrospective analysis, Chahal et al assessed the value of combining fibrotic score with GAP score in predicting patient survival in IPF (10). Thin-section CT scans were scored for percentage involvement with reticulation and honeycombing (summated as fibrotic score) by one of two chest radiologists. Using receiver operating characteristic analysis, an optimal cutoff for fibrotic score was determined (25%) to create low and high fibrotic score groups which were then combined with low (0–3, essentially stage I) and high GAP score (>3, essentially stages II and III) groups to create four combinatorial groups. Survival in those with low GAP score but high fibrotic score was significantly worse than those with low GAP score and low fibrotic score (hazard ratio, 4.03 [95% confidence interval: 2.02, 8.07]; P <.001). Interestingly, those with low GAP score and high fibrotic score had similar survival to those with high GAP score and either low or high fibrotic score. Without imaging correlation, the survival estimates in stage I (GAP score 0–3), stage II (GAP score 4 and 5), and stage III (GAP score 6–8) are widely divergent (3-year survival of 16.3%, 42.1%, and 76.8%; respectively) (10). The potential for increased granularity in determining prognosis using fibrotic score in those with stage I fibrosis on GAP index staging is highly promising and could be readily implemented.

The results are, in of and themselves, not unexpected. The GAP score uses functional parameters to help determine patient prognosis. However, forced vital capacity and DLCO can be affected by patient effort, neuromuscular weakness, equipment calibration, respiratory technician skill, and other concomitant morbidities. The fibrotic score used in this study presents an objective representation of abnormal versus normal lung, thereby augmenting the prognostic information provided by the GAP score. Use of a general fibrotic score rather than subcharacterizing the underlying type of fibrosis (such as reticulation, honeycombing, traction bronchiectasis, and ground-glass opacity), from a practical sense, is highly attractive given the high degree of interrater variation in identifying specific findings of pulmonary fibrosis.

The results must be considered cautiously, however. Only one of two readers scored CT scans for formal analysis. One wonders if similar results would have been achieved by other readers, though this concern is moderated by the substantial agreement between readers in dichotomous determination of fibrosis score (greater than or less than 25% lung involvement). Therefore, one must be wary of using this proposed adjunct to the GAP score until it has been validated in a separate cohort and by other readers. It would have been interesting to see how quantitative assessment of total fibrosis would have performed in this setting given the absence of variation in quantitative algorithms. One of the biggest issues in interstitial lung disease has always been variability on chest CT assessment, including interreader and even intrareader variation. Using a computer-based algorithm would mitigate, if not completely eliminate, this variability.