On the power and risks of the percent of predicted.

Because proper interpretation of pulmonary function tests (PFTs) depends critically on the predicted normal values, the availability of new reference values, such as those from the National Health and Nutrition Examination Survey (NHANES)1 or the Health Survey for England,2 poses important challenges and evaluative opportunities for investigators/authors, respiratory care clinicians, and pulmonary-function laboratory directors. Specifically, in an era in which large population-based assessments provide new reference values that may differ from pre-existing values,1,2 adoption of new reference values may confound longitudinal interpretation of PFTs, especially if the observer pays undue attention to the percent-of-predicted value to the exclusion of considering the actual value or the normal predicted value to which the observed value is being compared. This issue has special timeliness, because the American Thoracic Society recently endorsed use of the NHANES reference values for spirometry,3 prompting all laboratories that comply with this recommendation to switch to the new NHANES reference values. Consider the recent example of a situation from our practice, which emphasizes the potential pitfalls of inattention to changes in the reference value used and of overreliance on the percent-of-predicted value. Mr X, a patient with known alpha-1 antitrypsin deficiency,4 was referred to us from a center in Chicago for consideration of specific therapy with pooled-human-plasma alpha-1 antitrypsin, which is an effective (albeit expensive) treatment.5 Available records from the referring physician included a printout of the most recent PFT results (with spirometry, volumes, and diffusing capacity of the lung for carbon monoxide [DLCO]) and handwritten notes from prior PFT sessions. There appeared to be a dramatic drop in his DLCO from normal (2 years ago) to 68% of predicted currently. This perceived change prompted concern about progressive loss of alveolar/capillary units, consistent with emphysema, and a recommendation to initiate alpha-1 antitrypsin augmentation therapy and to seek a second opinion at The Cleveland Clinic. Later retrieval and review of the actual printouts from the patient’s earlier PFT sessions showed that the previously used DLCO standards were based on those of Burrows et al6 (in which the predicted DLCO is 28 mL/min/mm Hg for this patient), but that the predicted values had more recently been changed to those of Crapo et al7 (in which this patient’s predicted DLCO is 42 mL/ min/mm Hg). Importantly, the change in the standard values used was unannounced on the PFT report printout and escaped the referring clinician’s attention, which prompted concern about the falling DLCO and the out-of-city referral. In fact, testing in our laboratory indicated that the patient’s DLCO had not changed over several years; rather, the impression of a dramatic drop in DLCO was based on undue attention to the percent-of-predicted DLCO value, which had, in fact, only appeared to decrease because they had switched to a higher reference value (ie, from the reference equations of Burrows et al6 to those of Crapo et al7). As another example of the power and risks of the percent of predicted and the potential pitfalls of switching reference equations, percent-of-predicted values are often criteria for consideration for clinical procedures and inclusion or exclusion from clinical trials. As an example, in the recent National Emphysema Treatment Trial (NETT) of lung-volume-reduction surgery (LVRS) for severe emphysema,8–10 the high-risk group (for which LVRS is now deemed contraindicated) was defined as individuals whose forced expiratory volume in the first second (FEV1) was 20% of predicted and who had a homogeneous pattern of emphysema on chest computed tomography and/or a DLCO of 20% of predicted.10 Furthermore, eligibility for inclusion in the NETT was also based on percent-of-predicted lung function, including an FEV1 45% of predicted (and 15% of predicted in those 70 years old), total lung capacity 100% of predicted, and residual volume 150% of predicted.8 Imagine the difficulties that might be posed in generalizing the results of the NETT if different lung-function reference values were used. For example, in the worst hypothetical case, patients who would be ineligible because they are in the high-risk NETT group according to the reference values used in the NETT11 could be eligible for LVRS if other reference values were used, and they therefore might be placed at risk for adverse outcomes. To underscore the impact and risks of using different reference equations on the interpretation of an individual patient’s pulmonary function, classification of lung-disease severity, and eligibility for clinical trials, Tables 1 and 2 present percent-of-predicted values for 2 “standard” patients (66-year-old white men, 173 cm tall, weight 70 kg), based on several different reference equations.6,7,11–17 As shown in Table 1, use of 5 different reference values for FEV1 and DLCO produces discordant conclusions regarding the patient’s candidacy for LVRS based on whether