New Study Validates Algorithm for Predicting Long-Term Mortality in COPD

Female patient in hospital bed looking at xray with clincian
Female patient in hospital bed looking at xray with clincian
New research assesses the prognostic value of a COPD mortality algorithm, proposed in 2017, that includes comorbidities, lung function, age, BMI, and dyspnea.

The predictive value of a proposed clinical cluster algorithm for long-term mortality in patients with chronic obstructive pulmonary disease (COPD) has been validated in 2 different patient cohorts, according to a study in COPD.

The use of cluster analysis to derive patient subgroups that are linked with clinically relevant outcomes has been the subject of previous research. In 2017, Burgel et al developed an algorithm for predicting COPD mortality that divided patients into 5 groups, each based on a cluster of characteristics relating to age, body mass index, dyspnea grade, pulmonary function measures, and comorbidities (cardiovascular disease/diabetes). In the current study, investigators sought to validate this algorithm by assessing its ability to predict outcomes in 2 prospective cohorts that followed patients for 9 to 12 years in Sweden and the Netherlands.

The 2 studies used to validate the algorithm were the PRAXIS study, a prospective COPD cohort study initiated in Sweden in 2005, and the Rotterdam study, a prospective cohort study of multiple diseases initiated in the Netherlands in 1990. The COPD mortality algorithm validation cohort included 784 participants from the PRAXIS study (mean age, 64 years; 58% female) and 735 participants from the Rotterdam study (mean age, 71.5 years; 43% female).

Investigators divided individuals from the PRAXIS and Rotterdam studies into the 5 cluster groups defined by the algorithm:

  • cluster 5, characterized by no cardiovascular comorbidities and/or diabetes, a modified Medical Research Council (mMRC) score of 0 to 1, and forced expiratory volume in 1 second (FEV1) of at least 60%;
  • cluster 4, characterized by no cardiovascular comorbidities and/or diabetes, an mMRC score of 3 to 4, and FEV1 of less than or equal to 35%;
  • cluster 3, characterized by at least 1 cardiovascular comorbidity and/or diabetes, an mMRC score of 0 to 2, and either (1) age greater than 70 years, and FEV1 of at least 50%, or (2) age less than or equal to 7 years, and BMI of at least 30 kg/m2;
  • cluster 2, characterized by either (1) at least 1 cardiovascular comorbidity and/or diabetes, an mMRC score of 0 to 2, age of less than or equal to 7 years, and BMI less than 30 kg/m2, or (2) no cardiovascular comorbidities and/or diabetes and an mMRC/FEV combination different than clusters 4 and 5; and
  • cluster 1, characterized by at least 1 cardiovascular comorbidity and/or diabetes and either (1) an mMRC score of 3 to 4, or (2) an mMRC score of 0 to 2, age greater than 70 years, and FEV1 of less than 50%.

The researchers then conducted survival analyses for all-cause mortality at 3 and 9 years for both the Rotterdam and Praxis cohorts, with an additional 12-year analysis conducted for the longer-term PRAXIS study. Meta-analyses of the Cox models for 3- and 9-year survival were also conducted.

In an analysis of the PRAXIS study alone, cluster 5 had the lowest mortality rate at the 3-, 9-, and 12-year analyses; cluster 4 had the highest 9-year mortality rate of 72%. Clusters 1 and 4 had a 12-year mortality rate of 79%. In the analysis of the Rotterdam study alone, no patients fulfilled the criteria for cluster 4. Cluster 5 had the lowest mortality rate at 9 years (5%), and cluster 1 had the highest mortality rate at 9 years (44%).

In analyzing patients from both study cohorts, investigators found that patients in cluster 5 had the best prognosis, as indicated by the hazard ratio for mortality, while  patients in clusters 1 and 4 has the worst prognosis. Meta-analyses of the age- and sex-adjusted Cox regression models showed that clusters 1 and 3 had a higher mortality risk compared with cluster 5 at 3 years, but cluster 2 did not. Clusters 1, 2, and 3 had an increased mortality risk at 9 years vs cluster 5.

“Burgel’s clinical clusters can be used to predict long-term mortality risk. Clusters 1 and 4 are associated with the poorest prognosis, cluster 5 with the best prognosis and clusters 2 and 3 with intermediate prognosis,” the study authors concluded.

Among several study limitations, lung function in the PRAXIS study was not registered at study inclusion. Additionally, the COPD definitions differed between the Rotterdam and Praxis studies; In the Rotterdam study, the definition was based primarily on the pulmonary function testing results for a routine visit to the research center and did not require relevant exposure or respiratory symptoms; in the PRAXIS study, only doctor-diagnosed patients with COPD were included. Finally, a large proportion of patients in the PRAXIS study did not have postbronchodilatory spirometry results in their medical records from 2000 to 2004.

“We propose that these clusters should be used to a larger extent in both primary and specialist care for better planning of follow-up and health care utilization,” said the researchers, who nevertheless noted that further evaluation of the algorithm is needed.

Disclosure: Some of the study authors declared affiliations with biotech, pharmaceutical, and/or device companies. Please see the original reference for a full list of authors’ disclosures.

Reference

Gagatek S, Wijnant SRA, Ställberg B, et al. Validation of clinical COPD phenotypes for prognosis of long-term mortality in Swedish and Dutch cohorts. COPD. 2022;19(1):330-338. doi:10.1080/15412555.2022.2039608