Introduction and objectives Spirometry is frequently carried out as part of workplace-based respiratory surveillance programmes for the detection of both obstructive and restrictive lung diseases. However, the performance of spirometry to detect restrictive lung diseases is generally poor and especially so if the prevalence of the disease in the tested population is low such as in many working populations.
Our aim was to increase the specificity and the positive predictive value (PPV) of current spirometry-based algorithms to diagnose restrictive lung diseases in the occupational health setting to reduce false positives and so the number of unnecessary and expensive referrals for lung volume measurements in hospital.
Methods We re-analysed two prospective studies of 259 and 265 tertiary care hospital consecutive patients, respectively used to derive and validate the current standard spirometry-based algorithm (FVC <85% predicted and FEV1/FVC >55%) to diagnose restrictive lung diseases (Glady CA, et al. Chest 2003). We used true lung restrictive cases (TLC <LLN predicted) as a gold standard in 2 × 2 contingency tables to estimate sensitivity, specificity, positive and negative predictive values for each potential diagnostic cut-off. Predicted values for spirometry parameters were calculated by using both Crapo and Hankinson equations. Because our target population is active workers we tested the performance of each diagnostic algorithm among subjects under 65 years old and with a simulated prevalence of restrictive disease of 10% and 1%. In addition, we compared the performance of our best diagnostic algorithm to the ones previously reported by using receiver operating characteristic (ROC) curves.
Results Our best diagnostic algorithm (FVC <70% predicted and FEV1/FVC≥0.7) had a higher specificity (96% using Hankinson prediction equation) and PPV (80% and 27% for a disease prevalence of 10% and 1%, respectively) compared to previous algorithms. For example, compared to Glady’s algorithm, among 184 people tested, ours produced only 6 (3%) false positives vs. 64 (34%), and correctly classified 91% subjects vs. 65%, corresponding to an area under the ROC curve of 0.83 vs. 0.77. The results were confirmed in the validation dataset.
Conclusions Our proposed spirometry-based algorithm accurately excludes pulmonary restriction and reduces unnecessary lung volume testing in occupational health clinical setting.