Can the EpiSmokEr Classifier Help to Predict Smoking-Induced Lung Damage?

Use of the Epigenetic Smoking Status Estimator (EpiSmokEr) — a machine learning–derived classifier — appears to capture relevant smoking-related biologic effects by predicting smoking phenotypes that are strongly associated with lung function and all-cause mortality, as well as identifying former smokers who have an increased risk for incident airflow limitation and death. These are among study findings recently published in Respiratory Medicine.

EpiSmokEr detects individuals’ exposure to smoking by measuring changes in their DNA, predicting smoking phenotypes based on DNA methylation levels. It is well known that tobacco smoke alters DNA methylation at thousands of CpG [cytosine–guanine] sites in nucleated blood cells, with some of these sites localized to genes that are association with inflammation and chronic obstructive pulmonary disease (COPD). Recognizing that smoking-induced epigenetic alterations may represent a useful biomarker of smoking-linked damage, the researchers sought to assess the associations of phenotypes predicted by EpiSmokEr vs self-reported smoking phenotypes with lung function and all-cause mortality in a cohort of older adults.

The investigators conducted an analysis of the prospective, longitudinal, population-based US Veteran Affairs Normative Aging Study, in which DNA methylation measurements were obtained between 1999 and 2012, with follow-up reported through 2016. All participants also completed questionnaires related to lifestyle factors.

The analysis included 784 men who provided whole blood for measurements of DNA methylation. The mean participant age was 72.6±6.8 years. Among the study participants, spirometry measurements were performed every 3 to 5 years. Airflow limitation was defined as forced expiratory volume in 1 second/forced vital capacity of less than 0.7.

The 784 participants in the study included 144 classified as “predicted ever smokers” and 640 classified as “predicted never smokers”; all contributed a total of 5414 person-years of follow-up. Among these individuals, the EpiSmokEr-predicted smoking phenotypes matched the self-reported phenotypes among 97% of never smokers, 14% of former smokers, and 71% of current smokers. Further, the EpiSmokEr classified 79% of self-reported former smokers as never smokers and 7% of former smokers as current smokers.

Despite these findings, the EpiSmokEr-predicted former smoking phenotype was more strongly associated with incident airflow limitation (hazard ratio [HR], 3.15; 95% CI, 1.50-6.59; P =.002) and mortality (HR, 2.11; 95% CI, 1.56- 2.85; P <.001), compared with the self-reported former smoking phenotype (airflow limitation: HR, 2.21; 95% CI, 1.13-4.33; P =.02 and mortality: HR, 1.09; 95% CI, 0.87-1.38; P =.46).

The risk for airflow limitation and death did not differ between self-reported never smokers and former smokers who were classified as being never smokers. In fact, the discriminative accuracy of EpiSmokEr-predicted phenotypes for mortality and incident airflow limitation was improved compared with the use of self-reported phenotypes.

The current study has several limitations. The cohort comprised older white men, which limits the generalizability of the findings. Additional studies that include diverse demographic groups will help to validate the utility of the EpiSmokEr in other patient populations. Further, more than one-third of the eligible participants did not have follow-up spirometry measurements.

The researchers concluded that the “EpiSmokEr classifier may be a useful surrogate of smoking-induced lung damage and morbidity and may help identify former smokers most at risk of developing adverse smoking-related health effects.”

Reference  

Eckhardt CM, Wu H, Prada D, et al. Predicting risk of lung function impairment and all-cause mortality using a DNA methylation-based classifier of tobacco smoke exposure. Respir Med. Published online June 2, 2022. doi:10.1016/j.rmed.2022.106896