Reading a Negative Result

Zetobit · Reading the Report

Reading the Report

Reading a Negative Result: What It Rules Out — and What It Only Failed to Find

A negative genetic test is the one result that quietly expires. Four different situations produce identical wording, the report distinguishes only some of them, and the evidence says the most common repair is not a bigger test.

Kanna Nandakumar, PhD · Zetobit, LLC

Of the three things a genetic report can say, the negative is the one that ends the conversation. A positive result opens a management discussion. A variant of uncertain significance announces its own unfinished state and demands follow-up. No reportable variant identified reads like closure — it gets filed, and it is rarely revisited.

But that sentence is not a statement about the patient. It is a statement about a search: a particular set of genomic positions, examined at a particular depth, interpreted against a particular body of knowledge, on a particular date. Each of those four qualifiers is a place where a real answer can sit undisturbed.

The base rate makes the scale of it plain. In a meta-analysis of 108 studies covering 24,631 pediatric probands with rare or undiagnosed disease, the pooled diagnostic yield of genome-wide sequencing was 34.2%.¹ That is the most comprehensive first-line test available, and it explains roughly one patient in three. The remaining two are not two patients without a genetic cause. Fewer than half of the roughly 10,000 rare Mendelian conditions catalogued in OMIM currently have any established genetic basis at all.²

Four situations, one sentence

A negative report can be produced by four materially different states of affairs. They call for different responses, and they are not distinguishable from the summary line.

1. The cause was never inside the search space. Exome capture targets the roughly 2% of the genome that codes for protein. Deep intronic variants, regulatory variants, copy-neutral inversions, complex rearrangements and tandem repeat expansions sit outside it by design, and mitochondrial and imprinting disorders generally require separate assays.³ On a targeted panel the boundary is narrower still, and it is dated: the gene list was frozen when the panel was validated, while roughly 250 new gene–disease associations per year were reported through the first decade of exome sequencing — a pace that has slowed since about 2017 as the remaining associations require larger cohorts to establish.^4,5
2. The cause was inside the search space but was not observed. Capture is uneven, and unevenly uneven between laboratories. Across ten sequencing centers, regions covered below 20× in at least 90% of samples spanned between 1.2% and 9.1% of the coding bases in a 4,656-gene clinically relevant list⁶ — an eight-fold difference in the size of the blind spot between labs running nominally the same test. GC-rich first exons and genes with close paralogues are the usual casualties. In one platform evaluation, the share of ACMG secondary-finding genes covered completely at ≥20× ranged from 4% to 75% depending on the platform.⁷
3. The variant was observed and correctly called, but was not recognizable as an answer. A variant in a gene whose disease role is published next year sits in this year's data, called perfectly, and is not reportable. One analysis of positive exome results found 23% fell in genes characterized within the preceding two years, and 7% were novel gene discoveries.⁸ The same category covers a variant classified as uncertain and, given the phenotype the laboratory was working from, not returned.
4. There is no monogenic cause to find. This is the only one of the four that is a true negative.

The report distinguishes some of these and not others. States 1 and 2 are documented — in the limitations paragraph, in the gene list and its version date, in the coverage table — but that documentation sits well past the summary line and is written descriptively rather than flagged as a caveat. Nothing on the report separates state 3 from state 4, because at the moment of issue the laboratory cannot separate them either. That distinction is not a reporting failure. It is a fact about when the report was written.

Four upstream states produce the same sentence. The first two are documented somewhere in the report; the last two are not distinguishable from each other at the time it is written, which is why a negative result is the only class of genetic finding whose meaning changes without anyone touching the sample.

The escalation instinct, and what the evidence says about it

The instinctive response to a negative report is to order something bigger. Sometimes that is right. The evidence suggests it is right less often than the instinct implies.

In the largest study of genome sequencing after nondiagnostic evaluation, 822 families with suspected rare monogenic disease and no prior diagnosis were sequenced, and a molecular diagnosis was reached in 218 of 744 families in the initial cohort — 29.3%. Of those 218 diagnoses, 61 involved variants that genuinely required genome sequencing to identify: intronic variants, small structural variants, copy-neutral inversions, complex rearrangements, tandem repeat expansions.³ Sixty-one of 218 is 28%.

Which means the other 72% of those answers were in variant classes a coding-region test could, in principle, already have seen. The genome did not find them by looking somewhere new. It found them because the case was examined again — later, with more published knowledge, and with the phenotype reviewed afresh. Escalating the assay and reanalyzing the case were bundled into one act, and the assay collected the credit.

That is the practical finding sitting inside an expensive study. For most negative reports, the limiting resource is not territory. It is knowledge and clinical detail — and only one of those two is something the ordering clinician can supply.

The two asks, and when to make them

There are two distinct requests, and they are routinely conflated. ACMG's statement on the subject separates them cleanly: variant-level reevaluation reconsiders the classification of a specific variant already identified, while case-level reanalysis re-runs the whole case against current knowledge. Laboratories are expected to hold separate policies for each, to make those policies available to the ordering provider on request, and to respond to external requests in a timely manner.⁹ The mechanism exists — but it is request-driven, which means that for most patients it does not happen unless someone asks.

Timing is empirical rather than arbitrary. In a three-year study reanalyzing exomes on a six-month cadence, 2.6% of previously nondiagnostic and negative reports became diagnostic, with the highest rate at roughly two years after the initial report.¹⁰ Over longer horizons the accumulation is much larger: in a cohort first sequenced in 2011–2012, cumulative reanalysis raised the diagnostic yield from 24.8% to 46.8%.¹¹ A positive result is durable and a VUS declares its own instability, but the negative is the only result class that decays silently — because the knowledge it was measured against keeps moving underneath it.

The change most likely to alter the outcome, though, costs nothing at all. In one evaluation of a phenotype-driven exome pipeline, ten causative variants had ranked poorly; three improved simply because the phenotype ontology had been updated in the interim, and for the remaining seven the authors noted that precise phenotype terms had never been supplied on the requisition.¹² The laboratory filters on what it was told. A child who has since developed seizures, a sibling who has since presented, a family history that has since surfaced — that is new evidence, and it is evidence only the clinician holds.

What to read on a negative report — and what each part actually tells you

What to find	Where it usually sits	What it settles
Gene list and its version date	Methods section, or an appended list	Whether the gene you are worried about was examined at all, and how old the list was on the day it ran
Coverage summary	Technical or QC section	Whether an answer could have been missed inside the target. Ask for per-gene coverage if a specific gene matters
Variant classes assessed	Limitations paragraph	Whether CNVs, repeat expansions, mtDNA and methylation were in scope, or need separate orders
Singleton or trio	Specimen and methods section	How much interpretive power was available for de novo and recessive calls
Clinical indication as received	Requisition detail on the report face	What the filtering was actually aimed at — and whether it still matches the patient
Report date	Header	How much knowledge has accumulated since, and whether a reanalysis request is due

Three questions to ask

1. Was the gene I am worried about in the target, and was it adequately covered? This is a coverage question, answerable by the laboratory from data it already holds, usually at no additional cost and without a new specimen.
2. Which mechanisms on my differential were out of scope? Repeat expansions, deep intronic variants, methylation defects and structural rearrangements each have to be either inside the assay or on a separate order. If the differential contains one and the assay did not, the negative never spoke to it.
3. Has anything changed — the phenotype, the family, or the calendar? If yes, the request is case-level reanalysis with updated clinical information, not a new specimen. If nothing has changed and the report is under a year old, waiting is a defensible clinical decision.

A negative report is an accurate account of a search. Whether it is also an answer depends on questions the report does not contain — and, more often than the escalation instinct suggests, on information that never left the clinic.

References

1. Diagnostic yield and clinical utility of genome and exome sequencing in pediatric rare and undiagnosed genetic diseases: a meta-analysis. Genetics in Medicine, 2025. sciencedirect.com
2. Rare disease gene association discovery in the 100,000 Genomes Project. Nature, 2025. nature.com
3. Genome Sequencing for Diagnosing Rare Diseases. New England Journal of Medicine, 2024. nejm.org
4. Highlighting rare disease research: rates of rare disease-gene association discovery. GENETICS / G3, 2023. ncbi.nlm.nih.gov
5. GPAD: extracting gene–disease association discovery information from OMIM, including the post-2017 decline in discovery rate. Database, 2024. ncbi.nlm.nih.gov
6. Characterizing reduced coverage regions through comparison of exome and genome sequencing data across 10 centers. Genetics in Medicine, 2018. gimjournal.org
7. Achieving high sensitivity for clinical applications using augmented exome sequencing. Genome Medicine, 2015. ncbi.nlm.nih.gov
8. Exome sequencing covers >98% of mutations identified on targeted next generation sequencing panels. BMC Medical Genomics, 2017. ncbi.nlm.nih.gov
9. Points to consider in the reevaluation and reanalysis of genomic test results: a statement of the ACMG. Genetics in Medicine, 2019. nature.com
10. Diagnostic yield of exome reanalysis over time: contribution of reevaluation type, timing, and patient phenotype. Genetics in Medicine, 2026. gimjournal.org
11. Reanalysis of Clinical Exome Sequencing Data. New England Journal of Medicine, 2019. nejm.org
12. Rapid and accurate interpretation of clinical exomes using Phenoxome: a computational phenotype-driven approach. Genome Medicine, 2019. ncbi.nlm.nih.gov

Reading the Report is Zetobit's series for the people who receive genomic output rather than produce it — clinicians, principal investigators outside genomics, and the teams who have to act on a result. Zetobit, LLC · Lexington, KY · zetobit.com