Understanding the Role of Genetic Heterogeneity in Smoking Interventions: Experimental Evidence from the Lung Health Study

Shubhashrita Basu, Jason Fletcher, Qiongshi Lu, Jiacheng Miao, Lauren Schmitz2025 · DOI

Abstract

We examine whether genome-wide summary measures of genetic risk known as polygenic indices (PGIs) provide new insights into the efficacy of the Lung Health Study (LHS)—a large, randomized controlled trial (RCT) that evaluated the effect of a smoking cessation intervention program on cessation maintenance and lung function. Results indicate that the intervention was less successful for participants with higher PGIs for smoking initiation and intensity. Given the increasing availability and affordability of genomic data, we argue that in the context of RCTs, PGIs can further our understanding of heterogeneous treatment effects and the mechanisms that may be driving them.

Introduction

Tobacco use remains the leading cause of preventable death in the United States. While smoking cessation programs are widely available, their success rates vary significantly across individuals. Twin studies have long suggested that smoking behavior is highly heritable (46–84%), yet translating this genetic component into clinical practice has proven challenging. Early attempts using "candidate gene" approaches—focusing on specific genetic markers—often suffered from low statistical power and poor replicability.

This study moves toward a precision medicine framework by utilizing genome-wide data to investigate whether polygenic indices (PGIs) can identify which individuals are most likely to benefit from behavioral and pharmacological smoking cessation treatments.

Methods

The researchers utilized data from the Lung Health Study (LHS), a multi-center RCT conducted between 1989 and 1994. The trial involved 5,887 heavy smokers with early signs of chronic obstructive pulmonary disease (COPD). The intervention combined behavior modification, nicotine gum, and, for a subset, an inhaled bronchodilator.

Polygenic Index (PGI) Construction

To measure genetic risk, the authors constructed PGIs using PRS-CS-auto, a Bayesian framework that accounts for Linkage Disequilibrium (the non-random association of alleles). This method provides a more robust measure of genetic predisposition than older techniques. The study focused on a subset of 4,145 participants of European ancestry for whom genomic data was available.

Econometric Strategy

The analysis employed an interaction model where treatment status was interacted with PGIs. Ordinary Least Squares (OLS) was used for continuous outcomes, such as cigarettes per day (CPD), while Linear Probability Models (LPM) were used for binary outcomes, such as smoking cessation. To account for population stratification, the models controlled for the first 20 principal components of the genetic data.

Key Insights

The Extensive vs. Intensive Margin

The study reveals that genetics influence smoking behavior differently across two dimensions:

  1. Extensive Margin (Quitting): Individuals with a high genetic risk for smoking initiation found it significantly harder to quit entirely.
  2. Intensive Margin (Reduction): Among those who did not quit, the intervention was particularly effective for those with the highest genetic risk for smoking intensity.

Biological Specificity

The findings were not merely a proxy for general health or psychological factors. When controlling for PGIs related to BMI or depression, the smoking-specific genetic scores remained the primary drivers of treatment heterogeneity. Furthermore, the interaction between genetics and treatment was more pronounced in the group receiving the active drug (ipratropium bromide), suggesting a potential interaction with neural reward pathways.

Results

The intervention was highly successful on average, increasing quit rates by 23 percentage points. However, genetic factors introduced significant variation:

  • Genetic Penalty: For every one standard deviation increase in the Smoking Initiation PGI, the success of the intervention in achieving cessation dropped by approximately 2.5 percentage points.
  • The Equalizer Effect: For participants who continued smoking, a one standard deviation increase in the "Cigarettes Per Day" (CPD) PGI was associated with a reduction of one additional cigarette beyond the average treatment effect. By the end of the five-year trial, treated participants with high genetic risk smoked no more than those with low genetic risk, effectively neutralizing their genetic predisposition for heavy smoking.

Discussion and Limitations

While PGIs offer a promising path toward personalized interventions, several hurdles remain. Currently, PGIs have lower predictive power in non-European ancestral groups due to the lack of diverse genome-wide association study (GWAS) data. Additionally, PGIs capture both direct biological effects and "genetic nurture" (environmental factors correlated with genetics).

The authors conclude that while "pharmacogenetic stratification" is not yet ready for routine clinical use, integrating genomic data into RCTs significantly enhances our understanding of why treatments work for some and not for others.

Summary updated

Abstract

We examine whether genome-wide summary measures of genetic risk known as polygenic indices (PGIs) provide new insights into the efficacy of the Lung Health Study (LHS)-a large, randomized controlled trial (RCT) that evaluated the effect of a smoking cessation intervention program on cessation maintenance and lung function.Results indicate that the intervention was less successful for participants with higher PGIs for smoking initiation and intensity.Given the increasing availability and affordability of genomic data, we argue that in the context of RCTs, PGIs can further our understanding of heterogeneous treatment effects and the mechanisms that may be driving them.
BibTeX
@misc{basu2025understanding,
  author = {Shubhashrita Basu and Jason Fletcher and Qiongshi Lu and Jiacheng Miao and Lauren Schmitz},
  title = {{Understanding the Role of Genetic Heterogeneity in Smoking Interventions: Experimental Evidence from the Lung Health Study}},
  year = {2025},
  doi = {10.3386/w33473},
}
Figure