While we were screening our heavy smokers for lung cancer with chest CTs, the Dutch and Belgians have been screening their own (in the NELSON trial, which will report results in 2015). They double-dipped their imaging data here to ask the question, how good is chest CT at identifying undiagnosed chronic obstructive pulmonary disease?
1,140 men (former/current heavy smokers, age ~62) had pulmonary function test data available, in addition to their chest CT results. (Not sure why all were men, but the trial overall only has 16% women enrolled. Maybe that’s why they gave it a boy’s name.)
They defined emphysema as voxels on CT that had Hounsfield units less than -950 (air). Air-trapping was also measured — they acquired both inspiratory and expiratory scans in the study.
I found the title of this article (“Identification of Chronic Obstructive Pulmonary Disease in Lung Cancer Screening Computed Tomographic Scans”) to be a bit overstated. They did not identify COPD with chest CT alone — they used CT data as one element of a model that also included body mass index, smoking status, and pack-years corrected for “overoptimism” (on the patient’s part). Plugging all these together into a logistic regression model and finding the ideal cutoff point for area-under-the-curve, they achieved the following accuracy at identifying people with FEV1/FEV < 0.70:
- Sensitivity: 63%
- Specificity: 88%
- Positive predictive value: 76%
- Negative predictive value: 79%
Very interesting. But is it useful? Not in current clinical practice, IMHO. Expiratory imaging is not routine, and using the model would require a slide rule and a few minutes per patient. If the model they used could be built into an online calculator and the image analysis made automatic/standardized, it could maybe be a valuable tool for “add-on” COPD screening for heavy smokers already getting CTs for other reasons (like lung cancer screening). The title suggests CT images alone can make the diagnosis, but as their supplemental tables show, if you remove the clinical data or the expiratory images, the predictive ability of their model (which is only OK to begin with, since higher sensitivity is what you would prefer) goes down substantially.
More to the point, why would we try to identify people with undiagnosed COPD? Almost by definition, people with undiagnosed COPD have mild or moderate disease that’s not bothering them much. That was supported by the fact that many who were identified as having airflow limitation were asymptomatic in this study. Would earlier diagnosis help them quit smoking or improve their quality of life through early inititation of bronchodilator treatment? These questions have already been asked, and kind of answered, in the context of screening with spirometry for heavy smokers. The USPSTF reviewed the mostly shoddy evidence surrounding this issue in 2008, and issued its recommendation against spirometric screening in the Annals of Internal Medicine and on the AHRQ website:
“In conclusion, screening for COPD using spirometry is likely to identify a predominance of patients with mild to moderate airflow obstruction who would not experience additional health benefits if labeled as having COPD,” the reviewers conclude. “A few individuals with severe airflow obstruction (FEV1 [forced expiratory volume in 1 second] < 50% of predicted) might benefit from pharmacologic treatments that reduce exacerbations. Hundreds of patients would need to have screening spirometry to identify 1 person with COPD whose incremental health benefit over clinical diagnosis would probably be limited to the avoidance of a first exacerbation.”
Mets OM et al. Identification of Chronic Obstructive Pulmonary Disease in Lung Cancer Screening Computed Tomographic Scans. JAMA 2011;306(16):1775-1781.