Abstract
Genomic data are increasingly incorporated into high-throughput approaches such as the Phenome-Wide Association Study (PheWAS) to query potential effects of targeted therapies. Genetic variants, such as the interleukin-6 receptor (IL6R) genetic variant rs2228145 (Asp358Ala), have been identified with a downstream effect similar to the drug tocilizumab, which targets IL6R. These variants can be used to screen for potential protective or harmful signals across a broad range of traits in large biobanks. To date, there are limited approaches to determine whether these effects may differ across diverse populations to inform potential differential drug effects, especially in populations under-represented in clinical trials. In this study, we developed and applied an approach to detect heterogeneous associations, using the IL6R variant as an example, in African (AFR) vs. European (EUR) ancestry. We identified a total of 29 traits with a differential association between the IL6R variant, with notable differences including a lower risk of type 2 diabetes in AFR vs. EUR, and a higher white blood cell count. These findings can inform ongoing studies targeting IL6 and provide a general approach to test for heterogeneity of associations in PheWAS.
Introduction
Most clinical trials have historically been skewed toward populations of European ancestry, creating a "diversity gap" in modern medicine. This imbalance leaves a critical question unanswered: will a targeted therapy be as effective or as safe for everyone? To predict drug effects without the need for decades-long clinical trials, researchers utilize Phenome-Wide Association Studies (PheWAS) and Mendelian Randomization. By identifying natural genetic variants that mimic a drug's mechanism—such as the IL6R variant rs2228145, which reduces IL-6 signaling similarly to the arthritis drug tocilizumab—researchers can forecast clinical outcomes.
However, standard statistical methods often lack the power to detect "heterogeneity," the subtle but vital differences in how these variants behave across different genetic ancestries. This study introduces a novel statistical framework to bridge this gap, ensuring that the benefits of genomic medicine are equitably understood.
Methods
The hetFDR Framework
The researchers developed a novel statistical procedure called hetFDR (False Discovery Rate controlled heterogeneity testing). Unlike standard tests that examine ancestry groups in isolation, hetFDR operates on the assumption that if a genetic variant has a strong average effect across the entire population, it is more likely to exhibit differences between sub-groups. The procedure follows three steps:
- Mean Effect Calculation: It determines the average impact across all groups and calculates a "heterogeneity statistic" to measure the difference between groups.
- Weighting: It uses the mean effect to prioritize traits with significant overall signals in the search for ancestry-specific differences.
- FDR Control: A rigorous False Discovery Rate control (set at 10%) is applied to ensure findings are statistically robust and not merely noise.
Datasets and Cohorts
The study utilized three major biobanks to ensure diversity and validation:
- VA Million Veteran Program (MVP): The primary discovery cohort, featuring one of the world's largest populations of African ancestry (AFR) individuals (~105,000 AFR vs. ~439,000 EUR).
- UK Biobank (UKB): Used for validation (~7,500 AFR vs. ~459,000 EUR).
- Mass General Brigham (MGB) Biobank: Also used for validation (~2,900 AFR vs. ~49,000 EUR).
Results
Identification of Differential Traits
The application of hetFDR identified 29 traits with significant differential associations between AFR and EUR ancestries. The study confirmed that the IL6R variant successfully mimics IL-6 inhibitors, as evidenced by reduced C-reactive protein (CRP) and increased hemoglobin levels across groups.
Ancestry-Specific Insights
- Type 2 Diabetes (T2D): A major finding was that the IL6R variant is associated with a significantly lower risk of T2D and lower HbA1c levels specifically in individuals of African ancestry (OR 0.956). In contrast, the effect in the European group was negligible (OR 1.001).
- Hematological and Infection Signals: In the AFR group, the variant was linked to a much higher increase in absolute neutrophil counts and a modest increase in the odds of serious infection. This suggests that while IL-6 inhibitors might be more effective for certain metabolic conditions in this group, clinicians may need to monitor for infection more closely.
- Cardiovascular Consistency: Interestingly, the variant's protective effect against cardiovascular diseases, such as aortic aneurysms, remained consistent across both ancestries, suggesting these benefits are universal.
Statistical Performance
Simulation studies demonstrated that hetFDR achieved substantially higher power—up to 0.3 higher—than standard methods like the Benjamini-Hochberg procedure when detecting these critical group differences.
Discussion and Limitations
While these findings provide a roadmap for more inclusive precision medicine, several limitations remain. The MVP cohort is 91.3% male, which may not represent the general population perfectly. Furthermore, while the study focused on African and European ancestries, future research must expand to include Hispanic, Asian, and other underrepresented groups.
Ultimately, while genetic proxies are powerful tools for drug discovery and safety screening, real-world clinical trials stratified by ancestry remain the gold standard for confirming these findings. The hetFDR approach provides a scalable method to prioritize which drugs and traits require such focused investigation.
Summary updated