Prospective analysis of incident disease among individuals of diverse ancestries using genetic and conventional risk factors

Wei Wang, Nicholas Eriksson, Matthew K. McIntyre, Rafaela Bagur Quetglas, Bertram L. Koelsch, David A. Hinds, Stella Aslibekyan, Adam Auton, Michael V. Holmes, Suyash ShringarpuremedRxiv 2023 · DOI

This study by the 23andMe Research Team represents one of the largest and most ancestrally diverse prospective evaluations of Polygenic Risk Scores (PRS) to date. By following over a million individuals for a year, the researchers demonstrate that genetic scores can predict not just who will get a disease, but how early it might strike—and, crucially, how much a healthy lifestyle can offset that genetic risk.

TL;DR

Researchers used a massive dataset of over 8 million 23andMe participants to show that Polygenic Risk Scores (PRS) can accurately predict the one-year incidence and age of onset for 12 common diseases across diverse ancestries. The study highlights that while high genetic risk significantly accelerates disease onset, a favorable lifestyle can reduce the risk of coronary heart disease and type 2 diabetes by over 60%, effectively neutralizing a high genetic predisposition.

What Problem Does This Paper Address?

Most genomic research has historically suffered from a "European-centric bias," meaning the genetic markers used to predict disease risk often don't work as well for people of non-European descent. Furthermore, many studies are "cross-sectional"—they look at people who already have a disease (prevalent cases) rather than predicting who will develop it in the future (incident cases).

This paper tackles two major gaps:

  1. Portability: Can PRS effectively predict risk across European, Latine, and African American populations?
  2. Clinical Utility: Does a genetic score provide better information than traditional risk factors (like high blood pressure), and can lifestyle changes actually move the needle for someone with "bad" genes?

Methods

The researchers split their data into two groups: a discovery dataset (7.4 million people) to build the models and a prospective cohort (1.1 million people) to test them.

  • PRS Construction: They used GWAS (Genome-Wide Association Studies) to identify genetic variants associated with 12 diseases. They employed a "stacked pruning and thresholding" (P+T) approach, which they extended to handle multiple populations simultaneously.
  • Lifestyle Risk Score (LRS): For Coronary Heart Disease (CHD) and Type 2 Diabetes (T2D), they used an XGBoost machine learning model to combine 40 self-reported lifestyle factors (diet, smoking, exercise, etc.) into a single score.
  • Prospective Evaluation: Unlike many studies that look backward, they measured the "cumulative incidence"—the proportion of healthy people who developed a disease within exactly one year of follow-up.

Datasets

The study utilized the 23andMe research database, which is unique in its scale and diversity:

  • Discovery Dataset: ~7.4M individuals (78.5% European, 15.9% Latine, 5.6% African American).
  • Prospective Cohort: ~1.1M individuals (including 127,000 Latine and 33,000 African American participants).
  • Endpoints: 12 clinical conditions, including CHD, T2D, hypertension, anemia, blood clots, hypothyroidism, kidney stones, Crohn’s disease, glaucoma, psoriasis, and eczema.

Key Insights and Ideas

  • Genetic Risk is an Accelerator: High polygenic risk doesn't just increase the likelihood of disease; it moves the timeline. Individuals in the top 10% of genetic risk for hypertension developed the condition 16 years earlier on average than those in the bottom 10%. For T2D, the gap was nearly 10 years.
  • The "Lifestyle Shield": This is perhaps the most empowering finding. For individuals in the highest 10% of genetic risk for CHD, having a favorable lifestyle resulted in a 73% relative risk reduction. This brought their actual risk down to the population average, essentially "canceling out" their genetic disadvantage.
  • The Diversity Gap Persists: While the PRS worked well for Latine individuals (similar to Europeans), the predictive power was more modest for African Americans. This underscores the critical need for even larger and more diverse genomic datasets to ensure equitable healthcare technology.
  • Synergy with Conventional Risk: PRS isn't meant to replace a blood pressure cuff. Instead, it acts as a risk multiplier. If you have high blood pressure and a high PRS for CHD, your risk is nearly 10 times higher than someone with neither.

Key Results

  • Relative Risk (RR): Comparing the top 10% to the bottom 10% of genetic risk, the RR for developing T2D was 12.53 in Europeans and 10.92 in Latines, but lower at 4.00 for African Americans.
  • Incident Prediction: Even within a short one-year window, the PRS successfully stratified risk for all 12 endpoints.
  • Lifestyle Impact: Favorable lifestyle characteristics were associated with a 64% lower RR for T2D among those at high polygenic risk.
  • Feature Importance: For CHD risk, the most important lifestyle factors were smoking, exercise, alcohol, and salt intake. For T2D, they were exercise, alcohol, red meat, and sugar.

Limitations and Future Work

  • Self-Reporting: The study relies on participants self-reporting their medical diagnoses and lifestyle habits, which can introduce "measurement error" or bias.
  • Selection Bias: 23andMe customers may not perfectly represent the general global population (e.g., they may be more health-conscious).
  • Ancestry Sample Sizes: Despite the large total numbers, the African American cohort had fewer incident cases for certain rare conditions (like Crohn’s or blood clots), making those specific results less certain.
  • Future Directions: The authors suggest that future work should focus on integrating PRS into clinical workflows to see if "closer monitoring and earlier intervention" for high-risk individuals actually improves long-term health outcomes.

Summary updated

Abstract

Abstract Background Human genetics provides opportunities for enhancing disease prediction through polygenic risk scores (PRS). Method We used a dataset from 23andMe (6.77M European, 1.30M Latine, and 0.45M African American individuals). Using cross-sectional data for PRS construction and a prospective cohort for evaluation, we estimated PRS-associated cumulative incidences after one year of follow-up for 12 clinical endpoints. Results The cumulative incidence of disease at one year was consistently higher among individuals in the top 10% of each PRS. Relative risks (RRs) comparing top to bottom 10% of PRS differed across diseases (e.g. European RR 2.12 for eczema vs 12.53 for T2D). Estimates were similar between Europeans and Latines however were more modest for African Americans (e.g. T2D RR 10.92 for Latines vs. 4.00 for African Americans). Clinical manifestation occurred earlier among those in top vs bottom 10% of polygenic risk: 16yrs for hypertension, and 9.5yrs for T2D. Among participants at elevated conventional risk of CHD or T2D, those in the top 10% PRS had a 10-20 fold higher RR of disease incidence vs those not at conventional risk. Among individuals at high polygenic risk of CHD or T2D, favorable lifestyle characteristics associated with 64-73% lower RR of developing disease over 1-year, with cumulative incidence equivalent to the population average. Conclusion In an ancestrally-diverse cohort, individuals in the top 10% PRS had higher 1-year disease incidence and earlier age of clinical manifestation. PRS provided risk stratification beyond conventional risk factors. Lifestyle characteristics markedly lowered disease incidence among those at elevated polygenic risk.
BibTeX
@misc{wang2023prospective,
  author = {Wei Wang and Nicholas Eriksson and Matthew K. McIntyre and Rafaela Bagur Quetglas and Bertram L. Koelsch and David A. Hinds and Stella Aslibekyan and Adam Auton and Michael V. Holmes and Suyash Shringarpure},
  title = {{Prospective analysis of incident disease among individuals of diverse ancestries using genetic and conventional risk factors}},
  year = {2023},
  doi = {10.1101/2023.10.23.23297414},
  howpublished = {medRxiv},
}
Figure