LONDON – Based on readily available clinical data, patients with chronic obstructive pulmonary disease (COPD) can be placed into five phenotypes with different characteristics and risk profiles, according to data generated by a cluster analysis and presented at the European Respiratory Society International Congress 2016.
The algorithm that places patients into these phenotypes was developed from a cluster analysis, reported Pierre-Regis Burgel, MD, PhD, professor of respiratory medicine, Université Réné Descartes, Paris. Mortality rates at 3 years for these phenotypes ranged from 2.6% to 49.5%
Dr. Pierre-Regis Burgel
“We think that this could be the basis for recognizing clinically distinct COPD phenotypes and designing better tailored management,” Dr. Burgel explained. “We also think it has potential use in routine clinical assessment.”
To create these phenotypes, data from 2,049 COPD patients enrolled in a French-Belgian collaborative cohort were evaluated with Classification And Regression Tree (CART) analyses. The five phenotypes were derived from symptom burden, respiratory function, relative age, and presence of comorbidities.
Based on these characteristics, “a set of clinical rules” to phenotype patients was developed, according to Dr. Burgel. This algorithm was then further validated with the 3,651 COPD patients enrolled in the prospective COPD Cohorts Collaborative International Assessment (3CIA) initiative.
The two initial branches in the algorithm are created by dividing patients into those with and without cardiovascular comorbidities or diabetes. In those without cardiovascular disease, the phenotypes are defined by relative symptom severity, using cut-off scores from the modified Medical Research Council (mMRC) dyspnea assessment tool and relative degrees of lung function impairment as measured with forced expiratory volume in 1 second (FEV1).
In those with cardiovascular disease or diabetes, mMRC-defined symptoms and FEV1-defined lung function impairment also create decision points in the algorithm, but age and body mass index (BMI) are additional variables that direct patients to a specific phenotype. Class 4 and 5 are reached in the absence of cardiovascular disease or diabetes only, while cardiovascular disease is a prerequisite to reach Classes 4 and 5. Class 2 is the only phenotype that can be reached through the algorithm irrespective of the presence or absence of cardiovascular disease.
Using this algorithm, each of the phenotypes was associated with similar relative hierarchy in mortality in the two cohorts, even though mortality rates for each phenotype were consistently lower in the 3CIA group.
For class 1, which was reached by patients with cardiovascular disease or diabetes, the greatest symptom burden, and the lowest lung function, the mortality rates at 3 years were 49.5% and 23.2% for the French-Belgian and 3CIA cohorts, respectively.
For class 4, which was also defined by the greatest symptom burden and the lowest lung function without cardiovascular disease or diabetes the mortality rates were 45.3% and 27.3%, respectively. The lowest mortality rates, which were 2.6% and 4.0%, respectively, were found in the class 5 phenotype, which contained patients with a low symptom burden (mMRC less than or equal to 1), relatively good lung function (FEV1 greater than or equal to 60%), and no history of cardiovascular disease or diabetes.
In classes 2 and 3, the mortality rates fell in between those of the lowest- and highest-risk phenotypes. Specifically, these 3-year mortality rates were 22.9% and 24%, respectively, in the French-Belgian cohort, and 11.1% and 14.1%, respectively, in the 3CIA cohort.
The consistency of the hierarchy of outcomes in the two cohorts provides the basis for suggesting that these phenotypes are effective for categorizing relative risk, according to Dr. Burgel. He believes that the phenotypes are clinically important, and he emphasized that the algorithm relies on clinical information that is already routinely collected and readily accessible.
“There is growing awareness that COPD phenotypes are important and are likely to be valuable in managing patients,” Dr. Burgel explained. “We feel that we have created simple rules for allocating patients that we think will be useful for research and for clinical application.”
Dr. Burgel reported financial relationships with AstraZeneca, Boehringer Ingelheim, Chiesi, GlaxoSmithKline, Novartis, Pfizer, and Vertex.