Validation of the lower limit of normal diffusing capacity for detecting emphysema.

BACKGROUND Diffusing capacity for carbon monoxide (D(Lco)) has been regarded as reliable for detecting emphysema. The lower 5th percentile of the reference population has been used as the lower limit of normal (LLN) for D(Lco), without clinical validation. OBJECTIVES We performed this study to validate the LLN for D(Lco) and to determine the optimum cutoff LLN value for detecting emphysema. METHODS A total of 197 COPD patients and 103 healthy adult subjects were included. COPD patients with emphysema were defined as COPD patients in whom volumetric CT showed that the volume fraction of the lung at less than -950 Hounsfield units at full inspiration was more than 15%. All other COPD patients were defined as COPD patients without emphysema. All measured D(Lco) values were transformed to estimates of reference population percentiles. ROC curve analysis was used to validate and to determine the optimum cutoff percentile value as the LLN for D(Lco). RESULTS Of the 197 COPD patients, 126 were classified as having emphysema and 71 as without emphysema. On ROC curve analysis, the lower 5th percentile used as the LLN for D(Lco) had a sensitivity of 68.3% and a specificity of 98.1% to differentiate COPD patients with emphysema from healthy subjects. The lower 9th percentile was the best LLN cutoff value for detecting COPD patients with emphysema. CONCLUSION The lower 5th percentile of the reference population may not be the best LLN cutoff value for D(Lco) for detecting emphysema.