Predicting familial hypercholesterolemia (FH) in future generations will likely involve a combination of tools that don’t yet work together: genomic sequencing at birth or in childhood, artificial intelligence scanning medical records for missed cases, polygenic risk scores that capture genetic risk beyond single mutations, and digital platforms that make family screening faster and more reliable. Right now, over 90% of the estimated 30 million people with FH worldwide remain undiagnosed, and only a minority of those who are diagnosed receive adequate treatment. Closing that gap requires fundamentally changing how and when we look for the condition.
Why So Many Cases Go Undetected Now
FH is one of the most common inherited conditions, affecting roughly 1 in 250 people. It causes dangerously high LDL cholesterol from birth, dramatically increasing the risk of early heart disease. Yet the vast majority of affected individuals have no idea they carry it. A study of Dutch general practitioners found that only about 20% correctly identified how common FH is, and just 56% understood its inheritance pattern. These are the doctors most likely to encounter FH patients first, and their limited awareness creates a bottleneck before any screening even begins.
The current approach relies heavily on opportunistic detection: a doctor notices high cholesterol during a routine blood test, connects it to a family history of early heart attacks, and refers the patient for further evaluation. This catch-as-catch-can system misses the majority of cases, particularly in people without an obvious family history or those who don’t see a doctor regularly. Up to one-third of known FH patients receive no treatment at all.
Polygenic Risk Scores for the “Missing” 60%
Traditional genetic testing for FH looks for a single mutation in one of three well-known genes. But only about 40% of people who meet the clinical criteria for FH actually have one of these mutations. The remaining 60% have high cholesterol that looks and acts like FH, yet no identifiable single-gene cause. For many of these individuals, the culprit appears to be polygenic: dozens or hundreds of common genetic variants that each nudge LDL cholesterol slightly higher, and in unlucky combinations push it past the diagnostic threshold.
Polygenic risk scores (PRS) assign a number based on how many of these cholesterol-raising variants a person carries. In a Brazilian study of FH patients, those without a detectable single-gene mutation had the highest average polygenic risk scores, significantly above both the mutation-positive FH group and healthy controls. This suggests that PRS can help explain why someone has severe high cholesterol even when standard genetic testing comes back negative. In future generations, combining traditional mutation testing with a polygenic risk score could identify a much larger share of people at risk, catching cases that current methods miss entirely.
The limitation is that polygenic scores work best in the populations they were developed in. Scores built from European datasets perform less reliably in people of African, South Asian, or East Asian descent. Expanding these tools to be globally useful is one of the major challenges ahead.
AI Scanning Medical Records for Hidden Cases
One of the most promising near-term strategies involves machine learning algorithms that comb through electronic health records (EHRs) to flag patients who likely have FH but were never formally diagnosed. These systems look at patterns across years of lab results, medication histories, family health data, and clinical notes to identify people whose profiles match FH.
Several models are already being tested. FIND FH uses a random forest algorithm, while FAMCAT relies on logistic regression. Across studies, machine learning models consistently outperform simpler rule-based approaches. One ensemble ML model achieved an area under the curve (a measure of diagnostic accuracy) of 0.79, compared to 0.67 for a traditional scoring system. The best-performing algorithms in the literature reach scores as high as 0.95. FAMCAT2, tuned for high sensitivity, correctly identified about 95% of FH cases, though its specificity was lower at 31%, meaning it also flagged many people who turned out not to have FH. That tradeoff is acceptable for a screening tool, since the goal is to avoid missing anyone, and follow-up testing can sort out the false positives.
The real power of these tools is that they work passively, running in the background on data that already exists. No extra blood draw, no special appointment. A health system could theoretically screen its entire patient population overnight and generate a list of people who should be evaluated further.
Digital Tools for Family Cascade Screening
Once someone is diagnosed with FH, their close relatives each have a 50% chance of carrying the same mutation. Cascade screening, testing family members outward from a known case, is the most efficient way to find new diagnoses. In practice, it rarely works well. The traditional method involves mailing a templated letter to relatives explaining the diagnosis and encouraging them to get tested. There’s no way to know if the letter arrives, if the relative reads it, or if they follow through.
Digital platforms are being developed to fix this. FH Family Share, for example, is a web-based tool that lets a diagnosed patient send information about their condition directly to at-risk relatives through email or messaging, with built-in educational content and links to testing resources. Genetic counselors involved in its development noted that having a structured digital template made patients more comfortable initiating what can be an awkward conversation. These tools also allow tracking, so healthcare teams can see whether relatives have been contacted and whether they’ve sought testing.
Scaling this approach with automated reminders, integration with health system portals, and multilingual support could dramatically increase the yield of cascade screening in the coming decade.
Childhood and Newborn Screening
Catching FH before damage accumulates is the ultimate goal. Several countries already recommend cholesterol screening around age 9 to 11, but uptake is inconsistent. A 2025 cost-effectiveness study published in JAMA modeled what would happen if the U.S. adopted universal sequential screening, a cholesterol test followed by genetic testing for those above a certain threshold, at either age 10 or age 18.
The results were mixed. Screening a hypothetical cohort of 4.2 million 10-year-olds could prevent between 1,385 and 1,820 cardiovascular events over their lifetimes, depending on the cholesterol threshold used to trigger genetic testing. Screening at age 18 could prevent 1,154 to 1,448 events. Both strategies were effective in absolute terms, but neither met the standard threshold for cost-effectiveness ($100,000 per quality-adjusted life year gained). The most favorable scenario, screening at 18 with a high LDL threshold of 190 mg/dL or above, still came in at roughly $290,000 per quality-adjusted life year. The authors noted that cost-effectiveness could improve substantially if screening also led to better lifestyle interventions and lifelong cholesterol monitoring for people found to have non-FH high cholesterol.
The more ambitious vision involves screening at birth. Whole genome sequencing of newborns could theoretically identify FH-causing mutations on the first day of life. A study of healthcare professionals in Queensland, Australia, explored the feasibility and found significant enthusiasm alongside serious concerns. On the practical side, existing newborn screening infrastructure would need major expansion in laboratory capacity, trained personnel, and formal follow-up pathways. Pilot programs would need to come before any wider rollout.
The ethical questions are equally complex. FH typically doesn’t require medication until age 8 to 10, raising the question of whether it’s appropriate to screen asymptomatic newborns for a condition that won’t be treated for nearly a decade. Healthcare professionals flagged concerns about parental anxiety, the emotional weight of managing a chronic diagnosis from infancy, potential impacts on parent-infant bonding, genetic discrimination (particularly around insurance), and the possibility that genetic results could reveal non-paternity. Any newborn screening program would need robust informed consent processes that don’t currently exist in most newborn screening frameworks.
Epigenetic Markers and Beyond
Genetic variants tell you what risk someone was born with. Epigenetic markers, chemical modifications to DNA that change how genes are expressed, can reveal how that risk is actually playing out in a person’s body over time. DNA methylation patterns are emerging as a powerful layer of cardiovascular risk prediction that could eventually be applied to FH.
Researchers have already built prediction models combining methylation data with standard clinical variables. One framework called HFmeRisk, developed for heart failure prediction, used 25 specific methylation sites along with five clinical measures (age, diuretic use, BMI, albuminuria, and serum creatinine) to achieve an accuracy score of 0.90, significantly outperforming models built on clinical data alone (0.78) or methylation data alone (0.65). Notably, DNA methylation also outperformed RNA-based and microRNA-based models for predicting cardiovascular risk, suggesting it captures something about disease trajectory that other molecular markers miss.
For FH specifically, epigenetic profiling could help answer a question that genetics alone cannot: among people with the same FH mutation, why do some develop heart disease in their 30s while others remain relatively healthy into their 60s? Methylation patterns influenced by diet, exercise, stress, and other environmental factors may help explain this variability and allow for more personalized risk predictions in future generations.
Putting the Pieces Together
No single tool will solve FH prediction on its own. The likely future involves layered screening. AI algorithms will scan health records to identify adults who should have been diagnosed years ago. Universal childhood cholesterol checks, paired with genetic testing when results are concerning, will catch new cases early. Polygenic risk scores will identify the large group of people whose high cholesterol has a genetic basis even without a classic FH mutation. Digital cascade platforms will ensure that when one family member is found, their relatives are quickly and systematically reached. And epigenetic profiling may eventually allow clinicians to distinguish between people who carry a genetic risk and people in whom that risk is actively progressing toward heart disease.
The technical components for most of these approaches already exist in some form. The harder challenge is integration: building systems where a flag in a medical record automatically triggers a genetic referral, where a positive genetic test automatically launches digital family notification, and where all of this data feeds back into risk models that get smarter over time. That infrastructure doesn’t exist yet, but the pieces are falling into place faster than most people realize.