Assessment of lung function and severity grading in interstitial lung diseases (% predicted versus z-scores) and association with survival: A retrospective cohort study of 6,808 patients.

Background: Pulmonary function tests (PFTs) are essential for predicting outcomes in interstitial lung disease (ILD). In 2022, an expert panel recommended using z-scores instead of the traditional % predicted cut-off values to interpret the severity of PFT abnormalities which may lead to discordant classifications in some patients. To assess the magnitude and prognostic impact of this phenomenon we compared these two approaches in predicting all-cause mortality in a large cohort of patients with ILDs.

Methods and findings: We retrospectively analyzed data from a tertiary referral center for patients with ILDs. Absolute FEV1, FVC, TLC, and TLCO values from patients' first presentations were transformed and presented as % predicted and z-scores using the most recent global lung initiative (GLI) reference values. Results were categorized for severity according to % predicted and z-score levels. Predictors of all-cause mortality over a 14-year follow-up were determined using Kaplan-Meier survival analysis and Cox proportional hazards regression. Between January 2009 and March 2023, 6,808 patients with ILDs were evaluated at the National TB and Lung Diseases Research Institute in Warsaw, Poland. Most were diagnosed with sarcoidosis, fibrotic ILD, or non-fibrotic ILD. At their first presentation, 13.2% had airway obstruction, 23.1% had low FVC (indicative of restriction by spirometry), and 45.6% had a reduced lung transfer factor (TLCO). Reclassification of spirometric indices occurred in 26.8% of patients for FEV1 and 24.6% for FVC among those with abnormal results, with most being reassigned to a less severe categories. For TLCO, 28.1% of patients with reduced values were reclassified, with most shifting to more severe categories. During the follow-up, 1,525 (22.4%) of patients died. Both low FVC and low TLCO predicted all-cause mortality, with z-score thresholds showing stronger associations with mortality. A one-unit decrease in the FVC z-score was associated with a 10.3% increase in the risk of death, while a one-unit decrease in TLCO z-score was linked to an over 30% increase in mortality risk. Limitations of this retrospective single-center study include lack of data on cause-specific mortality, potential residual confounding, and limited generalizability to non-Caucasian or younger populations.

Conclusions: The recently recommended use of z-scores leads to significant reclassification of lung function results in patients with ILDs, largely driven by age. This approach is justified by its stronger prognostic associations. Severe TLCO impairment remains a robust predictor of mortality in ILDs.