A population-based analysis using 2 large genomic biobanks suggests that genetic variants that are traditionally associated with inherited retinal degenerations (IRDs) manifest clinically less often than previously assumed. In a study published in The American Journal of Human Genetics, investigators analyzed genomic and clinical data from the National Institutes of Health–supported All of Us Research Program and the UK Biobank to estimate the population-level penetrance of Mendelian variants that are thought to cause IRDs. They reported that only 9% to 28% of individuals who carry IRD-associated genotypes showed evidence of retinal disease, depending on the disease definition used.
The researchers screened 317,964 All of Us participants with linked whole-genome sequencing and electronic health record (EHR) data for pathogenic, likely pathogenic, or loss-of-function variants in IRD-associated genes. After extensive manual curation, 481 individuals were identified with what the authors defined as “definite IRD-compatible genotypes,” which represented 167 distinct variants across 33 genes. Of these participants, 77.8% carried variants in autosomal dominant genes, 15.8% had variants in X-linked genes, and 6.4% had homozygous variants in autosomal recessive genes. The most frequently represented genes included PRPH2, CRX, RHO, RPGR, BEST1, and RP1, which accounted for more than 70% of the cohort altogether.
To estimate penetrance while accounting for diagnostic variability, the authors developed 3 nested sets of ICD-9 and ICD-10 codes: a narrowly defined “IRD” set, a broader “retinopathy” set, and a highly inclusive “screening” set. They calculated disease annotation frequency (DAF) as the proportion of genetically at-risk individuals with at least one corresponding diagnostic code. The DAF ranged from 9.4% for the IRD-specific code set to 28.1% for the broadest screening set. All 3 estimates were significantly enriched compared with the prevalence of those codes in the overall All of Us population (P<.001). The findings indicate that most individuals who carried IRD-associated genotypes did not have EHR evidence of retinal degeneration.
Gene-specific analyses showed variability in annotation frequency. Variants in PRPH2, RPGR, and RHO demonstrated significant enrichment across multiple code sets, while BEST1 and RP1 variants did not show statistically significant enrichment for any retinal disease code set.
To corroborate the EHR-based findings, the researchers examined retinal imaging from the UK Biobank. Among 482 UK Biobank participants who carried variants shared with the All of Us cohort, 68 had available color fundus photography or optical coherence tomography scans. Masked grading identified definite bilateral IRD-compatible abnormalities in 16.1% of these individuals. When cases with less definitive abnormalities were included, the maximum phenotype frequency rose to 27.9%, which closely matched the upper penetrance estimates derived from EHR data. Only genes that showed enrichment in the All of Us analysis—most notably PRPH2 and RHO—were associated with imaging abnormalities in the UK Biobank cohort.
The authors evaluated whether age, sex, smoking history, socioeconomic status, diabetes, or health care utilization influenced disease annotation. Multivariable analyses found no significant association between these factors and penetrance. Restricting analyses to participants older than 60 years modestly increased disease annotation frequency but did not eliminate the gap between observed penetrance and historical assumptions of near-complete expression.
Based on their findings, the authors estimated that IRD-compatible genotypes may be present in 0.7% to 2.1% of the general population. This rate is substantially higher than the estimated prevalence of clinically diagnosed IRDs. The researchers concluded that IRD-associated variants frequently show incomplete penetrance at the population level and that additional genetic or environmental factors may be required for disease manifestation. They noted that these findings have implications for genetic testing, variant interpretation, and the design of future studies aimed at identifying modifiers of retinal disease expression.
Penetrance estimates may have been lower as a result of biobank recruitment bias, as well as UK Biobank imaging data, which is limited to the posterior pole and could have missed early manifestations of IRDs in the retinal periphery. Because of imaging as well, penetrance could have been inflated by including patients with AMD. Other limitations expanded on subject recruitment, including rare variants that were present in All of Us but that were not detected in the UK Biobank, age differences between subjects recruited to both data sets that could affect penetrance, and generalizability to autosomal recessive IRDs, compared with X-linked or autosomal-dominant IRDs.
The authors concluded that the low penetrance estimates for IRD-associated variants “suggests that IRDs result from the interaction of rare genetic variants of large effect size with the individual’s genetic or environmental background.” They continued, “This model explains both widespread incomplete penetrance and variable expressivity seen in IRDs. In turn, this suggests that genetic testing has a low PPV [positive predictive value], discouraging presymptomatic screening. The low PPV implies that the population prevalence of IRD-compatible genotypes is 0.7% to 2.1%, which is significantly higher than the currently estimated 1-in-3,450 prevalence of those affected by IRD. This represents an opportunity to discover modifying factors of penetrance and expressivity in IRDs.” RP