Exacerbation history alone is unlikely to provide clinical utility for predicting future chronic obstructive pulmonary disease (COPD) exacerbations and could be associated with a risk of harm, according to data from 3 separate clinical trials.
Findings published in the journal CHEST showed that multivariable clinical prediction models were more accurate than exacerbation history alone for predicting the future risk of COPD exacerbations, however, they required setting-specific recalibration to yield higher clinical utility.
“Contemporary management of chronic obstructive pulmonary disease (COPD) relies on exacerbation history to risk-stratify patients for future exacerbations,” wrote researchers from the University of British Columbia, Vancouver, Canada. Multivariable clinical prediction tools, they added, “combine several patient characteristics to increase the accuracy of risk stratification. Unlike exacerbation history alone, they can quantify (eg, in risk %) and communicate future risk with patients to enable shared decision-making. Importantly, prediction models are flexible and can be updated to accommodate baseline risk in different settings.”
To examine how exacerbation risk prediction models, compared to exacerbation history alone, perform in different patient populations, investigators analyzed data from 3 clinical trials that enrolled 4107 patients at different levels of moderate or severe COPD exacerbation risk:
The exacerbation risks were low, medium, and high in the 3 respective trials. Researchers compared the performance of 3 risk-stratification algorithms: exacerbation history; the model that Loes C.M. Bertens, PhD, and colleagues in the Netherlands developed in 2013 (Bertens); and the latest version of the Acute COPD Exacerbation Prediction Tool (ACCEPT).
The team compared the area under the curve (AUC) and net benefit (measure of clinical utility) among the 3 risk stratification algorithms to predict exacerbations in the next 12 months. They also evaluated the effect of model recalibration on clinical utility, according to the study abstract.
Compared to exacerbation history, ACCEPT had better performance in all 3 samples (ΔAUC: 0.08, 0.07, 0.10; P≤.001). Researchers observed that Bertens had better performance than exacerbation history in SUMMIT and TORCH (ΔAUC: 0.10 and 0.05; P<.001), but not in LOTT. No algorithm was superior in clinical utility across all 3 samples.
Before recalibration, Bertens generally outperformed ACCEPT and exacerbation history in low-risk settings, while ACCEPT outperformed the other algorithms in high-risk settings. After recalibration, the risk of harm was substantially mitigated for both ACCEPT and Bertens.
“We found that risk prediction models generally had better discriminatory performance compared with exacerbation history alone. However, when considering clinical utility, no algorithm emerged as universally better than others. Critically, it was found that all three risk stratification algorithms had the risk of causing harm,” concluded the authors. They added the caveat that prediction models cannot be universally applied but should be adapted for different populations based on the overall exacerbation risk of the group being considered.
Reference: Khoa Ho J, Safari A, Adibi A, et al. Generalizability of risk stratification algorithms for exacerbations in COPD. CHEST. Published online December 9, 2022. doi:10.1016/j.chest.2022.11.041.