VUS, ClinVar stars and why a genetic result can change

Five classes for one question

A classification answers one question about one variant: does it cause a given condition? The guideline builds the answer from several kinds of evidence, among them how common the variant is in the population, what computer predictions say about its effect, laboratory studies of the protein, and whether it travels with the disease through a family^[1]. Each piece is weighted as very strong, strong, moderate or supporting, and fixed rules combine the weights into one of the five classes^[1].

Two of the words carry a rough number. The authors proposed that likely pathogenic and likely benign mean greater than 90 percent certainty, a definition they described as "common, albeit arbitrary", and they noted that most variants lack the data to put any number on their class at all^[1].

A class belongs to a variant and a condition, never to the person who carries it. ClinVar asks submitters to classify a variant in the context of its mode of inheritance, so a variant that causes a recessive disease is reported as pathogenic even though one copy of it will not cause the disease^[3]. That is why a pathogenic carrier result and a pathogenic result for your own health read so differently; our guide to carrier status covers the difference.

What uncertain means

Uncertain is a statement about the evidence, not a measured chance of disease. A VUS is the default class: the guideline assigns it when the criteria for the other classes are unmet or contradict each other^[1]. A variant seen in only a handful of people sits there for want of data, and so does one where one line of evidence points to harm and another points away from it.

Instead, the guideline asks for efforts to resolve the variant toward pathogenic or benign. For reports that carry a VUS in a gene related to the reason for testing, it recommends that laboratories suggest periodic inquiry by healthcare providers, to learn whether knowledge of the variant has changed^[1].

VUS are a routine result. In a study of 1,689,845 people who had gene panel tests at the laboratory Invitae between 2014 and 2022, 41.0 percent received at least one VUS and 31.7 percent received VUS as their only result^[5]. The more genes a panel covered, the more VUS each person received, and 86.6 percent of the VUS were missense changes, which swap one amino acid in a protein for another and which the authors single out as needing better methods of interpretation^[5].

Ancestry changes the odds of a VUS. Per gene sequenced, people outside the European White group received more of them: in the Invitae data about 46 genes had to be sequenced to find one VUS in a Pacific Islander person, against 71 in an Ashkenazi Jewish person, and White people had a lower rate than every group except the Ashkenazi Jewish and Hispanic groups^[5]. The authors call for more diverse representation in genomic databases^[5]. One mechanism follows from the rules themselves: a variant too common to cause a rare disease counts as evidence for benign^[1], and that evidence cannot be gathered for a variant whose carriers are absent from the databases.

Two laboratories followed their VUS for years

The first study followed 1.45 million people tested for hereditary cancer risk at one commercial laboratory from 2006 onwards^[4]. Of 26,670 unique VUS, 2,048 (7.7 percent) were reclassified during the study. Of those, 1,867 (91.2 percent) were downgraded to benign or likely benign, after a median of 1.17 years, and 178 (8.7 percent) were upgraded to pathogenic or likely pathogenic, after a median of 1.86 years^[4].

The second study, the Invitae data above, is larger and covers panels for hereditary disease of every kind, not cancer alone. Of 512,983 unique variants ever reported there as VUS, 37,699 (7.3 percent) were reclassified, affecting 163,977 people. About 80 percent went to benign or likely benign and the remaining 20 percent to pathogenic or likely pathogenic^[5]. Clinical evidence did the most to move a VUS up, above all a variant found to have arisen new in a child with the condition, while population data counted more often in moves down than in moves up^[5].

Each number comes from one laboratory's own patients, and 95.6 percent of the people in the cancer study were women^[4]. The times describe the VUS that did move, and say nothing about how long any particular VUS will wait: in both studies more than nine in ten unique VUS had not been reclassified when the data were collected^{[4, 5]}. The cancer study also counted every time a VUS had been reported to someone. Because the same variants recur in many people, 24.9 percent of all reported VUS results were reclassified, about three times the share of unique variants^[4].

ClinVar is an archive of submissions

ClinVar stores what others conclude. Clinical testing laboratories, research laboratories, expert panels and other groups submit classifications^[7], and ClinVar files each submission as its own record and gathers every record about one variant onto one page, under a summary classification it derives from them by fixed rules^[3]. Its documentation is direct about who decides: the classification on a submitted record is provided only by the submitter, and ClinVar staff do not curate submitted classifications^[3].

Counted from ClinVar's own file of 4 October 2026, uncertain significance is its most common classification. Of 4,588,517 variant records in the GRCh38 VCF, 2,377,011 (52 percent) carry it as their summary classification, against 358,760 (7.8 percent) that are pathogenic, likely pathogenic or both: about 6.6 uncertain records for each disease-causing one^[6]. Most of those VUS rest on one submitter's word. For 2,015,242 of them, 85 percent, the review status is a single submitter^[6].

When submitters disagree, ClinVar says so instead of averaging. Since June 2022 it reports a conflict when one submission is on the pathogenic side and another is uncertain, when one is pathogenic and another benign, or when one is uncertain and another benign. Pathogenic beside likely pathogenic is not a conflict and is shown as Pathogenic/Likely pathogenic^[3]. A classification from an expert panel or a practice guideline replaces all the others in the summary, though the other submissions stay on the record^[3].

#CHROM  POS        ID     REF  ALT  INFO, four of its fields
7       117479283  51029  T    G    GENEINFO=CFTR:1080|LOC111674463:111674463
                                    CLNSIG=Conflicting_classifications_of_pathogenicity
                                    CLNSIGCONF=Uncertain_significance_(4)|Benign_(3)|Likely_benign_(1)
                                    CLNREVSTAT=criteria_provided,_conflicting_classifications

What ClinVar's stars measure

The gold stars beside a ClinVar classification record its review status: how the classification was reached and by whom^[8]. They do not grade how severe the condition is, how likely a carrier is to develop it, or how strongly the variant acts. A three-star benign classification and a three-star pathogenic one have been reviewed to the same standard.

One star asks less than it might seem. To earn it, a submitter documents its classification terms and criteria and supplies evidence or a public contact; ClinVar does not review the details of those criteria^[9]. Expert panel status is granted through ClinGen, which reviews each panel's application and membership^[9]. By ClinVar's own count, seven in ten variant records rest on a single submitter, and one in two hundred on an expert panel or a practice guideline^[2].

A second opinion counts because laboratories applying the same guideline often disagree. When nine laboratories classified 99 variants under the 2015 rules, they agreed on 34 percent at first and on 71 percent after discussing the evidence together^[10]. Four laboratories comparing their ClinVar submissions on 6,169 shared variants found 88.3 percent concordant; reassessing 242 of the 724 disagreements, with current criteria and shared internal data, resolved 211 of them and raised overall agreement to 91.7 percent^[11].

How our report uses the stars

Aimosti does not classify variants. The report reads ClinVar's summary classification and review status for each variant it examines in your file, from the release named on the report, currently ClinVar's July 2026 release. Inside the 1,054 regions the report examines, that release holds 79,440 records classified pathogenic or likely pathogenic^[12]. Five rules decide what reaches the page.

1. Only pathogenic and likely pathogenic classifications become findings. Uncertain significance, benign, likely benign and conflicting classifications are read and then left out.
2. Two stars or more: the finding goes in the main list.
3. One star from a single submitter: the match is shown separately, under lower-confidence matches, as a lead rather than a result.
4. No stars: never reported.
5. On a sequenced genome, the wider ClinVar screen's section on dominant conditions rated actionable lists only variants with at least two stars.

More than half of those records, 42,287 of 79,440, rest on a single submitter, so a large share of what the report could match falls in the separately shown one-star group^[12]. In the carrier section, the one kind of result that does not rest on a ClinVar classification is a short list of founder variants a peer-reviewed study has shown to cause disease, each labelled as published research; our page on XXYLT1 retinal dystrophy explains one. How these grades sit within the report's three confidence tiers is set out in Confidence Tiers explained.

A genetic result is dated by its release

ClinVar changes every week. Its data are updated on Mondays, and an archived release comes out on the first Thursday of each month^[13]. A report read against one release says nothing about the next, so ours names the ClinVar release it used in its Sources & versions section.

We measured this on our own panel, comparing the two releases record by record^[14]. Between the same two releases ClinVar added 2,828 records inside those regions, 469 of them pathogenic or likely pathogenic, and withdrew 151. A record counts as reclassified when its main classification term changed. A change in stars alone is not counted, so the figure understates how much moved.

Most of the movement happens in the middle class. In the cancer-testing study only 0.7 percent of unique variants first classified pathogenic or likely pathogenic, and 0.2 percent of those first classified benign or likely benign, later moved to a different clinical category, against 7.7 percent of VUS^[4].

What re-analysis can and cannot do

The ACMG separates two kinds of second look. Variant-level re-evaluation reconsiders the classification of variants already reported; case-level reanalysis reviews every variant in an exome or genome, reported or not, and may run the data through alignment and variant calling again^[15]. Its statement calls variant classification a dynamic process, and judges routine re-evaluation of a laboratory's whole internal database likely to be impractical^[15]. Laboratories do issue amended reports: the cancer-testing study counted 59,955 issued because a variant had been reclassified^[4].

Our re-analysis is narrower than either. A report is fixed at the ClinVar release it names. With the re-analysis subscription, we keep your normalised variant file. After we install a new ClinVar release, we start a re-read of that kept file ourselves; it does not run on a fixed schedule, and because it reads the kept file it works long after your original upload has been deleted. The Genome Watch page then lists what changed. A variant that has become pathogenic or likely pathogenic appears as a new result, one that has left those classes is listed as no longer reported, and a move between pathogenic and likely pathogenic shows as reclassified. Without the subscription, the report stays dated to its release.

Re-reading has limits. It does not classify a VUS: a new classification reaches our report only once a submitter or an expert panel deposits it in ClinVar. It does not align reads or call variants again, so a variant the original file never listed stays unseen, and in a plain VCF an unread position still looks like a reference one, as our note on missing VCF records explains. And it is a separate channel from a clinical laboratory's amended report, which the guideline encourages laboratories to consider sending proactively when a variant reported as pathogenic or benign has to be reclassified^[1].