The incidental diagnosis cascade

When newborn screening shifts from biochemical assays for specific conditions to whole genome sequencing as a comprehensive screen, the results stop being binary. The biochemical assay returns a value that is either above or below a cutoff. The genome returns a sequence. The sequence contains millions of variants. Some are pathogenic for the conditions on the screening panel. Many more are uncertain. Some predict adult-onset conditions decades away. Some are pharmacogenomic variants that will become relevant the first time the child is prescribed a metabolized drug.

The infrastructure question is what to do with all of it.

The categories of incidental finding

A whole genome sequence at birth typically yields, in any large screened population, several distinct categories of finding beyond the panel target.

Carrier status for autosomal recessive conditions appears in nearly every sequenced genome. Most people carry pathogenic variants in two to four recessive disease genes. The finding is not clinically significant for the individual but is reproductively significant for the family.

Predisposition variants for adult-onset conditions are present in a smaller proportion of genomes and include BRCA1/BRCA2 variants for breast and ovarian cancer, Lynch syndrome variants for colorectal cancer, hypertrophic cardiomyopathy variants, long QT syndrome variants, hereditary hemochromatosis variants, and the rest of the ACMG secondary findings list. The American College of Medical Genetics maintains a list of genes for which actionable findings should be reported when sequencing is done for any clinical reason. The list expands as new actionable conditions are characterized.

Pharmacogenomic variants affect how individuals metabolize specific drugs. CYP2D6 variants affect codeine, tamoxifen, and several psychiatric medications. CYP2C19 variants affect clopidogrel and several proton pump inhibitors. TPMT and NUDT15 variants affect thiopurine drugs used in leukemia and inflammatory bowel disease. HLA-B*57:01 affects abacavir hypersensitivity. The list of clinically validated pharmacogenomic variants is growing and is incorporated into clinical practice unevenly.

Variants of uncertain significance, VUS, appear in genes associated with rare disease but with insufficient evidence to classify as pathogenic or benign. The classification can change over time as more cases accumulate. A VUS in 2026 may reclassify as pathogenic in 2031 if subsequent case reports establish the association.

The longitudinal value of the genome

The genome sequenced at birth has the same nucleotide sequence at age 30 that it had at week one. The clinical interpretation of that sequence at age 30 is substantially richer than the interpretation at week one because the field has accumulated thirty years of additional case data, gene-disease associations, and pharmacogenomic evidence.

The variant of uncertain significance reported at birth becomes a variant with a known classification at some later point. The pharmacogenomic variant that mattered for one drug at birth becomes relevant for several additional drugs as new gene-drug associations are characterized. The carrier status that was demographically interesting at birth becomes reproductively relevant when the individual approaches family planning.

The infrastructure that supports this longitudinal value capture has to do three things. It has to store the raw sequence data in a form that can be reanalyzed. It has to maintain a relationship with the patient through which updated interpretations can be returned. It has to manage the tension between the patient's right to access incidental findings and the patient's right not to know.

The right not to know

Not every incidental finding from a newborn genome should be returned. A pathogenic BRCA1 variant in a newborn has reproductive implications for the family but no actionable clinical implication for the newborn for at least two decades. Whether to return it at birth, at puberty, at age 18, or at the patient's request is an ethical question that the field has not resolved.

The current practice in genomic newborn screening pilots tends toward returning only conditions on the agreed panel, with adult-onset conditions deferred to a later age and pharmacogenomic findings returned only when clinically actionable for a specific prescription. The panel of returned conditions can expand over time as the patient grows and as the clinical context shifts.

The infrastructure decision is whether the system supports staged return of information across the patient's lifetime. The genome that is interpreted only once and reported in full at birth puts the entire interpretive burden on the family in the first weeks of life. The genome that is stored, with the patient's consent, and interpreted progressively as clinically relevant questions arise over decades respects the right not to know while preserving the option to know later.

The data trust as the right structure

A newborn whose genome is sequenced and returned to the family in a one-time clinical report has received a service. A newborn whose genome is sequenced, stored under fiduciary governance with the family's consent, and made available for staged return of information across the lifetime has received a relationship.

The data trust structure is the legal and operational form that supports the relationship. The trust holds the data. The patient or the patient's parents are the beneficiaries. The trustee has fiduciary duties to the beneficiary, which include returning information when actionable and respecting choices about what information to return. The trust persists across institutional changes in healthcare providers, insurance carriers, and laboratories.

The cascade of incidental findings that genomic newborn screening produces is not a problem to be avoided. It is a feature whose value compounds over decades, conditional on the infrastructure that holds the data being designed for that timescale.