New study develops improved mortality prediction model for COPD patients

In a recently published study in EClinicalMedicineA group of researchers has developed and validated a model to predict the risk of death specifically for chronic obstructive pulmonary disease (COPD) (a progressive lung disease that causes difficulty breathing) using probabilistic graphical modeling to improve disease management strategies.

Study: Development and validation of a mortality risk prediction model for chronic obstructive pulmonary disease: a cross-sectional study using probabilistic graphical modellingPhoto credit: Jo Panuwat D/Shutterstock.com

background

COPD is one of the leading causes of death worldwide. Predictive models such as body mass index, obstruction, dyspnea, exercise capacity (BODE), age, dyspnea, obstruction (ADO) and dyspnea, obstruction, smoking status, frequency of exacerbations (DOSE) help identify high-risk COPD patients, but primarily focus on overall mortality.

Traditional models such as regression models and random survival forests are limited to associative predictions that lack causal insights. In contrast, probabilistic graphs or causal graphs can identify potential cause-effect relationships from observational data by factoring out confounding factors.

Further research is needed to refine and validate COPD-specific mortality predictors in different populations and to explore the underlying biological mechanisms for targeted interventions.

About the study

The present study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines, with all participants providing informed consent.

The discovery cohorts were drawn from the COPD Genetic Epidemiology (COPDGene) Study, which included 10,198 current and former smokers aged 45 to 80 years with a smoking history of >10 pack-years.

Data collected included demographic, spirometric, clinical, and chest CT scan data, as well as all-cause mortality and COPD-specific mortality, defined by criteria that exclude deaths due to comorbidities such as cardiovascular disease (CVD) or cancer. The final analysis focused on 8,610 participants with complete baseline and follow-up data.

For external validation, the ECLIPSE (Evaluation of COPD Longitudinally to Identify Predictive Surrogate End-points) study was used, which included 2,501 participants from the United States (US) and Europe. Complete three-year mortality data were available for the 2,312 individuals.

Directed probability graph models were constructed using the CausalCoxMGM method to identify direct predictors of mortality while controlling for confounding factors.

These models were compared with ADO, updated BODE indices, and standard machine learning approaches, with performance assessed by cross-validation using Harrell's concordance index.

Vital capacity-forced vital capacity (FVC) %predicted, age, history of pneumonia, oxygen saturation, forced expiratory volume in 1s (FEV₁)/FVC ratio, 6-minute walk The exercise capacity, dyspnea (VAPORED) risk score was developed using seven characteristics associated with COPD-specific mortality and its accuracy was validated in the ECLIPSE cohort to predict survival at one, two, and three years.

Study results

In the Phase 2 study, which included a subset of Phase 1 participants, notable changes in clinical covariates were observed. These changes included an expected five-year increase in age and a significant reduction in patients in the more severe Global Initiative for Chronic Obstructive Lung Disease (GOLD) categories.

In addition, there was a significant increase in the incidence of comorbidities such as cardiovascular disease and diabetes. The BODE index showed a significant decrease, while the ADO index increased in phase 2 participants compared to phase 1. Despite these changes, survival functions were not statistically different between the two phases, indicating consistent overall mortality in both phases.

The ECLIPSE study used for external validation differed significantly from COPDGene in that it had a higher proportion of male participants and lower ethnic diversity. The ECLIPSE cohort also included more severe COPD cases, as reflected in higher ADO, BODE and updated BODE indices, as well as a higher overall mortality rate. This difference was significant, particularly in terms of the number of deaths observed within the first three years.

The study's analysis identified characteristics that were directly associated with COPD-specific mortality. While there was considerable overlap in the variables that affect overall mortality and COPD-specific mortality, some differences also emerged.

For example, FVC percentage was strongly associated with COPD-specific mortality, while FEV₁ The predicted percentage was more relevant for overall mortality. Comorbidities such as cardiovascular disease and diabetes were only associated with overall mortality.

Graph-based prediction models developed from these findings outperformed traditional indices such as ADO and the updated BODE index in predicting both overall mortality and COPD-specific mortality. The models showed that they can classify patients into different risk groups more effectively than the BODE index.

For external validation, the VAPORED risk score, developed using seven clinical variables, was tested on the ECLIPSE cohort. The VAPORED model significantly outperformed ADO and BODE and updated BODE indices in several predictive metrics, particularly in estimating the probability of concordance.

The model predictions were well calibrated for survival probabilities of one, two and three years in the ECLIPSE study.

In addition, a web-based tool has been developed that allows users to calculate and compare mortality risk using the VAPORED score and the BODE and ADO indices. This tool is available as a Shiny app and allows clinicians and researchers to assess and visualize mortality risks using key clinical variables.

Conclusions

In summary, this study used probabilistic graph modeling to identify features directly associated with COPD-specific and overall mortality and differentiate them from simple correlates. Using clinical data, researchers developed the VAPORED risk score, which outperformed traditional indices such as ADO and BODE in predicting mortality.

In addition, the study identified unique factors such as internet access and certain biological markers associated with increased overall mortality risk. This approach demonstrated superior predictive power and opens new opportunities for targeted interventions and the potential for the development of more comprehensive mortality risk scores for COPD patients.