Cited by 373SARS-CoV-2 and COVID-19 ResearchCOVID-19 epidemiological studiesCOVID-19 Clinical Research Studies

Evidence for increased breakthrough rates of SARS-CoV-2 variants of concern in BNT162b2-mRNA-vaccinated individuals

Open access
Kustin, Talia, Harel, Noam, Finkel, Uriah, Perchik, Shay, Harari, Sheri, Tahor, Maayan, Caspi, Itamar, Levy, Rachel +21 more·2021·Nature Medicine

Published version of a preprint

Abstract

The BNT162b2 mRNA vaccine is highly effective against SARS-CoV-2. However, apprehension exists that variants of concern (VOCs) may evade vaccine protection, due to evidence of reduced neutralization of the VOCs B.1.1.7 and B.1.351 by vaccine sera in laboratory assays. We performed a matched cohort study to examine the distribution of VOCs in infections of BNT162b2 mRNA vaccinees from Clalit Health Services (Israel) using viral genomic sequencing, and hypothesized that if vaccine effectiveness against a VOC is reduced, its proportion among breakthrough cases would be higher than in unvaccinated controls. Analyzing 813 viral genome sequences from nasopharyngeal swabs, we showed that vaccinees who tested positive at least 7 days after the second dose were disproportionally infected with B.1.351, compared with controls. Those who tested positive between 2 weeks after the first dose and 6 days after the second dose were disproportionally infected by B.1.1.7. These findings suggest reduced vaccine effectiveness against both VOCs within particular time windows. Our results emphasize the importance of rigorously tracking viral variants, and of increasing vaccination to prevent the spread of VOCs.

AI snapshot

Written from the paper’s full text.

TL;DR

In a matched cohort study of 813 SARS-CoV-2 genome sequences from Israel, BNT162b2 vaccine breakthrough infections were disproportionately caused by B.1.351 at least 7 days after the second dose and by B.1.1.7 between 14 days after the first dose and 6 days after the second dose. The authors conclude this suggests reduced vaccine effectiveness against both VOCs within particular time windows, while noting small variant sample sizes and that the design cannot measure absolute vaccine effectiveness.

Key findings

  • Among 813 sequenced samples, B.1.1.7 was predominant (712/813 sequences), while B.1.351 had an overall frequency of 1.6% (13/813).
  • In dose2 vaccinees versus matched unvaccinated controls, B.1.351 was significantly overrepresented among discordant pairs (McNemar OR 8:1; one-sided exact McNemar test, P = 0.02), whereas B.1.1.7 was not (OR 6:4; P = 0.38).
  • In dose1 vaccinees versus controls, B.1.1.7 was significantly overrepresented (McNemar OR 26:10; P = 0.006); conditional logistic regression gave an OR of 2.4 (95% CI 1.2–5.1). B.1.351 in dose1 was too sparse for statistical analysis (one infection in each category).
  • Seven of eight B.1.351 dose2 cases were isolated 7–13 days after the second dose, and the eighth was isolated exactly 14 days after the second dose; no increased breakthrough of either variant was observed 2 weeks after the second dose, though data were limited (76 cases).

Methods

  • Matched cohort study using Clalit Health Services (Israel) data. Vaccinated infected cases were matched to unvaccinated infected controls on sampling date (±3 days), age, sex, ethnic sector, and geographic location. Dose1 was defined as a positive PCR test from 14 days after the first dose to 6 days after the second dose; dose2 was defined as a positive PCR test at least 7 days after the second dose.
  • Complete viral genome sequencing was performed on RNA from nasopharyngeal swabs, yielding 813 samples from different individuals: 149 dose2-control pairs, 247 dose1-control pairs, plus additional unpaired samples. Lineage assignment used Pangolin with stringent consensus sequences, and a maximum-likelihood phylogenetic tree was reconstructed.
  • Primary analysis used a one-sided exact McNemar test on discordant pairs; a conditional logistic regression including age was used as a sensitivity analysis for the dose1 B.1.1.7 data.

Limitations

  • The authors state that the main caveat was the small sample size of WT and B.1.351 variants, driven by the high frequency of B.1.1.7 and low frequency of B.1.351 in Israel; they also caution that McNemar ORs do not necessarily represent the OR of breakthrough and that absolute numbers, especially for B.1.351, were very small.
  • The authors state that the design observes VOCs conditioned on infection, ignores non-infected individuals, and does not measure absolute infection rates, so it cannot deduce vaccine effectiveness against either variant. They also note sequencing limitations excluded very low-viral-load samples (only Ct values ≤33 were collected), the dose2 cohort may include individuals infected before boost immunity was fully established, and behavioral biases may exist because vaccinees were exempt from quarantine and testing after exposure and were likely tested mainly when symptomatic.
  • Evident from the design: as an observational matched study, it cannot fully exclude unmeasured confounding, and the variant distribution among infected individuals does not directly quantify protection against infection or disease.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41591-021-01413-7

Cites

Statistics

Nothing to recompute.

Claims the paper makes

0 of 14 cited sources checked · 0 supported · 0 flagged

  • Findingown data

    Among BNT162b2 vaccinees who tested positive at least 7 days after the second dose, the proportion of B.1.351 was significantly higher than among matched unvaccinated controls (McNemar OR 8:1, P=0.02).

    “a significantly higher proportion of B.1.351was observed in dose2 cases versus unvaccinated controls (McNemar OR of 8:1; one-sided exact McNemar test, P = 0.02).”
    Quote in the paperSubstantial claim
  • Findingown data

    B.1.351 was at an overall frequency of 1.6% in the sample of both vaccinated and non-vaccinated individuals (13/813 sequences).

    “B.1.351was at an overall frequency of 1.6% in our sample of both vaccinated and non-vaccinated individuals (13/813 sequences) (Fig. 1b), similar to previous reports of B.1.351frequency in Israel from January 2021 16 .”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [16] South African variant found in about 1% of all positive tests. MOH https://www.gov.il/en/departments/news/22022021-01 (2021).

      Not checked in the free run.

  • Findingown data

    In this study, no higher viral load was observed in B.1.1.7 infections compared to other variants.

    “We note that in this study we did not observe higher VL in B.1.1.7 infections as compared to other variants (see also ref. 28 )”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [28] Walker, A. S. et al. Increased infections, but not viral burden, with a new SARS-CoV-2 variant. Preprint at medRxiv https://doi.org/10.1101/2021.01.13.21249721 (2021). source

      Not checked in the free run.

  • Findingown data

    No evidence was found for increased breakthrough rates of B.1.1.7 a week or more after the second dose.

    “yet we find no evidence for increased breakthrough rates of B.1.1.7 a week or more after the second dose (Figs. 2 and 3).”
    Quote in the paperSubstantial claim
  • Interpretationreasoning

    Mass vaccination with two doses controls and contains the spread of the VOCs B.1.1.7 and B.1.351.

    “suggesting that while vaccine breakthrough infections at particular windows of time may be more frequent with the VOCs B.1.1.7 and B.1.351,mass vaccination with two doses controls and contains their spread.”
    Quote in the paperSubstantial claim
  • May combine two claims. 7
  • The quote only partly states this claim. 1
  • Classified as probably not a substantial claim. 22
  • May combine two claims.

    Findingcitation

    Higher viral load was noted in B.1.351 infections, while lower viral load was noted in vaccinees compared to unvaccinated individuals.

    “yet we did note higher VL in B.1.351infections, while noting lower VL in vaccinees as compared to unvaccinated individuals (Supplementary Fig. 6) 29,30 .”
    Quote in the paperNot a substantial claim0 of 2 sources checked
    • cites [29] Levine-Tiefenbrun, M. et al. Initial report of decreased SARS-CoV-2 viral load after inoculation with the BNT162b2 vaccine. Nat. Med. 27, 790–792 (2021). source

      Not checked in the free run.

    • cites [30] Petter, E. et al. Initial real world evidence for lower viral load of individuals who have been vaccinated by BNT162b2. Preprint at medRxiv https://doi.org/10.1101/2021.02.08.21251329 (2021). source

      Not checked in the free run.

  • The quote only partly states this claim.

    Findingown data

    Among BNT162b2 vaccinees who tested positive between 14 days after the first dose and 6 days after the second dose, the proportion of B.1.1.7 was significantly higher than among matched unvaccinated controls (McNemar OR 26:10, P=0.006).

    “a significantly higher rate of B.1.1.7 was observed in dose1 cases versus unvaccinated controls (McNemar OR of 26:10; one-sided exact McNemar test, P = 0.006).”
    Quote in the paperSubstantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    One dose1 control sequence, originally classified as WT (B.1.235), was located within the B.1.351 clade on the phylogenetic tree.

    “Following classification by Pangolin, we noted that one dose1 control sequence, originally classified as WT (B.1.235),was located within the B.1.351clade on the phylogenetic tree.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Findingown data

    B.1.1.7 was the predominant strain in Israel over the entire sampling period (712/813 sequences), increasing in frequency over time.

    “we noted that B.1.1.7 was the predominant strain of the virus in Israel over the entire sampling period (712/813 sequences), increasing in frequency over time (Fig. 1a).”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The pair of the misclassified dose1 control sequence was classified as B.1.1.7.

    “Its pair was classified as B.1.1.7”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Findingown data

    The dose2 B.1.351 case shown in the 14-20 days category was isolated exactly 14 days after the second dose.

    “The dose2 B.1.351case that is shown in the 14-20 days category was isolated exactly 14 days after the second dose.”
    Quote in the paperNot a substantial claim
  • May combine two claims.

    Interpretationcitation

    These results are generally aligned with in vitro neutralization assays showing a large reduction in neutralization against B.1.351 and little to no reduction against B.1.1.7 in fully vaccinated individuals.

    “These results are generally aligned with those from in vitro neutralization assays that have shown a large reduction in neutralization against B.1.351,and little to no reduction against B.1.1.7 in fully vaccinated individuals [7][8][9][10][11]17 .”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [7] Liu, Y. et al. Neutralizing activity of BNT162b2-elicited serum. N. Engl. J. Med. 384, 1466–1468 (2021). source

      Not checked in the free run.

    • cites [8] Xie, X. et al. Neutralization of SARS-CoV-2 spike 69/70 deletion E484K and N501Y variants by BNT162b2 vaccine-elicited sera. Nat. Med. 27, 620–621 (2021). source

      Not checked in the free run.

    • cites [9] Collier, D. A. et al. Sensitivity of SARS-CoV-2 B.1.1.7 to mRNA vaccine-elicited antibodies. Nature 593, 136–141 (2021). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Samples for which Pangolin labels of the strict and majority-rule consensuses did not coincide were discarded.

    “Samples for which Pangolin labels of the strict and majority-rule consensuses did not coincide were discarded.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    Regions with fluctuating ambiguity (continuous subsequence of length 20 or lower flanked by ambiguous bases) were masked as N.

    “We also noted some regions with fluctuating ambiguity: if a continuous subsequence of length 20 or lower was flanked on both ends by ambiguous bases, we masked out this entire subsequence by assigning it with N, under the assumption it represents unreliable sequencing.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: A continuous subsequence of length 20 or lower flanked by ambiguous bases represents unreliable sequencing.
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The failed parameter match was most often age, sex or municipality.

    “We found that the failed parameter match was most often age, sex or municipality.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductreasoning

    Two consensus sequences were constructed for each sample: one based on majority rule, and a stricter consensus requiring at least 80% of reads to support a given base.

    “To overcome these limitations, we constructed two consensus sequences for each sample, one based on majority rule, and a more strict consensus sequence where we required at least 80% of reads to support a given base.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    To increase dose2 matching, matching on date of PCR sampling was enforced, but four out of five matches in the remaining parameters were allowed, prioritizing municipality, then sector, then age, then additional criteria.

    “To increase dose2 matching, we enforced matching on the date of PCR sampling, but allowed for four out of five matches in the remaining parameters, while prioritizing municipality, then sector, then age, and then the additional criteria.”
    Quote in the paperNot a substantial claim
  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Each case was matched to a control using six parameters: date of sampling for PCR (±3 days), sex, age (±10 years), municipality of residence, geographical district of residence and sector.

    “Next, each case was matched to a control using six parameters: date of sampling for PCR (±3 days), sex, age (±10 years), municipality of residence, geographical district of residence and sector.”
    Quote in the paperSubstantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductreasoning

    A conditional logistic regression was used as a sensitivity analysis to include age as a possible confounder in case matching was not sufficient.

    “A conditional logistic regression was used as a sensitivity analysis to include age as a possible confounder in case that matching was not sufficient, under the assumption that it was sometimes only partially mediated through matching.”
    Quote in the paperNot a substantial claim
    • The authors state this as an assumption: Age effect was sometimes only partially mediated through matching.
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    Ten control samples served as controls for both a dose1 and a dose2 sample.

    “We note that ten control samples served as controls for both a dose1 and a dose2 sample.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S1 (asserted), S3 (load-bearing).

    Premisecitation

    B.1.1.7 is associated with lower Ct values, corresponding to higher viral loads.

    “it has been previously reported that B.1.1.7 is associated with lower Ct values, corresponding to higher VLs 3”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [3] Davies, N. G. et al. Estimated transmissibility and impact of SARS-CoV-2 lineage B.1.1.7 in England. Science 372, eabg3055 (2021). source

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Study design and conductnone given

    Only samples with Ct values of 33 or lower were collected.

    “Following matching, RNA from cases and controls was obtained from the main testing laboratories of CHS, with one major limitation: only samples with Ct values of 33 or lower were collected.”
    Quote in the paperSubstantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Premisecitation

    Amplicon dropout of amplicons 74 and 76, both covering the spike gene and some lineage-defining alterations of B.1.1.7 and B.1.351 (such as E484K and N501Y), has been previously noted.

    “We and others have previously noted amplicon dropout of amplicons 74 and 76 (ref.32 ), both of which cover the spike gene, and in particular some of the lineage-defining alterations of B.1.1.7 and B.1.351(such as E484K and N501Y).”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites ref.32

      Not checked in the free run.

  • Classified as probably not a substantial claim.

    Interpretationcitation

    Breakthrough cases observed might be due to immune evasion mediated by particular alterations present in these strains.

    “From a biological point of view, the breakthrough cases observed in this study might be due to immune evasion, mediated by particular alterations present in these strains 9,[23][24][25][26][27] .”
    Quote in the paperSubstantial claim0 of 3 sources checked
    • cites [9] Collier, D. A. et al. Sensitivity of SARS-CoV-2 B.1.1.7 to mRNA vaccine-elicited antibodies. Nature 593, 136–141 (2021). source

      Not checked in the free run.

    • cites [23] Greaney, A. J. et al. Comprehensive mapping of mutations to the SARS-CoV-2 receptor-binding domain that affect recognition by polyclonal human serum antibodies. Preprint at bioRxiv https://doi.org/10.1101/2020.12.31.425021 (2021). source

      Not checked in the free run.

    • cites [24] Weisblum, Y. et al. Escape from neutralizing antibodies by SARS-CoV-2 spike protein variants. Elife 9, e61312 (2020). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductnone given

    In both consensus sequencing assignments, sequencing coverage of at least ten reads was required.

    “In both types of consensus sequencing assignments, we required sequencing coverage of at least ten reads.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Thirty-two pairs in which one sample did not undergo successful sequencing were discarded from the paired analyses.

    “Thirty-two pairs in which one sample did not undergo successful sequencing were discarded from the paired analyses (but see Supplementary Fig. 3).”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    All sequences in the study were manually verified for phylogenetic location.

    “This is in line with recent concerns regarding misclassifications of Pangolin 37 , and led us to manually verify the phylogenetic location of all sequences in our study.”
    Quote in the paperNot a substantial claim0 of 1 sources checked
    • cites [37] Dudas, G. et al. Travel-driven emergence and spread of SARS-CoV-2 lineage B.1.620 with multiple VOC-like mutations and deletions in Europe. Preprint at medRxiv https://doi.org/10.1101/2021.05.04.21256637 (2021). source

      Not checked in the free run.

  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    The conditional logistic regression was performed only on the dose1 B.1.1.7 analysis, as not enough data were available in other categories.

    “The regression was performed only on the dose1 B.1.1.7 analysis, as not enough data were available in other categories.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductnone given

    The study was approved by the CHS institutional review board (no.0016-21-COM2) and was exempt from the requirement for informed consent; it was further approved by the Tel Aviv University ethics committee (0002706-1).

    “The study was approved by the CHS institutional review board (no.0016-21-COM2) and was exempt from the requirement for informed consent.The study was further approved by the Tel Aviv University ethics committee (0002706-1).”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing). May combine two claims.

    Study design and conductown data

    In preliminary analyses, matching often failed for dose2 samples due to their small sample size and increasing proportions of vaccinated individuals in older age categories across time.

    “In preliminary analyses, we noted that matching often failed for dose2 samples due to their small sample size, as well as due to increasing proportions of vaccinated individuals in older age categories across time, in line with the vaccine rollout policy in Israel.”
    Quote in the paperNot a substantial claim
  • Probably not a substantial claim: may not pass S3 (load-bearing).

    Study design and conductown data

    Some vaccinees (7.7%) received the second vaccine dose more than 28 days after the first vaccine, yet were still annotated as described.

    “We also note that some vaccinees (7.7%, Table 1) received the second vaccine dose more than 28 days after the first vaccine, yet were still annotated as described above.”
    Quote in the paperNot a substantial claim
26 of 26
  • The paper relies on this without support.

    In Israel, vaccinees were exempt from quarantine and testing following exposure to a positive patient.

    “vaccinees in Israel were exempt from quarantine and testing following exposure to a positive patient”
  • The authors state this as an assumption.

    A continuous subsequence of length 20 or lower flanked by ambiguous bases represents unreliable sequencing.

    “under the assumption it represents unreliable sequencing”
  • The authors state this as an assumption.

    Age effect was sometimes only partially mediated through matching.

    “under the assumption that it was sometimes only partially mediated through matching”
  • The authors state this as an assumption.

    Breakthrough is ordinal.

    “assuming ordinality of breakthrough”
4 of 4

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • The distribution of B.1.1.7 in breakthrough cases was similar to its prevalence in the community.

    “We identified the B.1.1.7 variant in 85% of cases, similar to its prevalence in the community.1,16”
  • At the time of the study, B.1.1.7 was the most widespread variant in Israel and accounted for up to 94.5% of SARS-CoV-2 isolates.

    “At the time of this study, the B.1.1.7 variant was the most widespread variant in Israel and accounted for up to 94.5% of SARS-CoV-2 isolates.1,16”
  • Reports have been published in which certain variants of concern were more prevalent in breakthrough infections.

    “In contrast, reports in which certain variants of concern were more prevalent in breakthrough infections have been published as well.8,16,26”
  • Among nine BNT162b vaccine breakthrough infections in an Israeli case-control study, eight were with the Beta variant even though the Alpha variant predominated during the study period.

    “Among nine BNT162b vaccine breakthrough infections in an Israeli case-control study, eight were with the Beta variant even though the Alpha variant predominated during the study period 226 .”

The citation lists haven’t been fetched yet

References 39 · Cited by 373

Research works printed but not linked: 1

  1. Liu, Y. et al. Neutralizing activity of BNT162b2-elicited serum—preliminary report. N. Engl. J. Med. 384, 1466–1468 (2021). · DOI · Paper page

Authors
Kustin, T., Harel, N., Finkel, U., Perchik, S., Harari, S., Tahor, M., Caspi, I., Levy, R., Leshchinsky, M., Ken Dror, S., Bergerzon, G., Gadban, H., Gadban, F., Eliassian, E., Shimron, O., Saleh, L., Ben-Zvi, H., Keren Taraday, E., Amichay, D., Ben-Dor, A., Sagas, D., Strauss, M., Shemer Avni, Y., Huppert, A., Kepten, E., Balicer, R. D., Netzer, D., Ben-Shachar, S., & Stern, A.
Affiliations
  • Tel Aviv University (IL)
  • Clalit Health Services (IL)
  • Rabin Medical Center (IL)
  • Ben-Gurion University of the Negev (IL)
  • Emek Medical Center (IL)
Corresponding author
Stern, Adi
Venue
Nature Medicine
Published
2021-06-14
Volume
27
Issue
8
Pages
1379-1384
ISSN
  • 1078-8956 (ISSN-L)
  • 1546-170X
OpenAlex
W3170724610
Cited by
373
References
39
Citations by year
2026: 6 · 2025: 11 · 2024: 17 · 2023: 36 · 2022: 145 · 2021: 156 · 2020: 1
FWCI
28.00
Type
journal article
Language
en
Open access
Locations
  • Nature Medicine (publishedVersion, cc-by)
  • medRxiv (acceptedVersion, cc-by-nc-nd)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
Funders
  • European Commission (HORIZON2020, 852223)
  • Tel Aviv University
  • H2020 European Research Council (852223)
Topics
  • Health Sciences › Medicine › Infectious Diseases › SARS-CoV-2 and COVID-19 Research
  • Physical Sciences › Mathematics › Modeling and Simulation › COVID-19 epidemiological studies
  • Health Sciences › Medicine › Infectious Diseases › COVID-19 Clinical Research Studies
Keywords
SARS-CoV-2, variants of concern, BNT162b2 mRNA vaccine, breakthrough infection, vaccine effectiveness, Alpha variant, Beta variant, nasopharyngeal swab, COVID-19 pandemic
MeSH terms
COVID-19, SARS-CoV-2, COVID-19 Vaccines, BNT162 Vaccine, Adolescent, Adult, Aged, Aged, 80 and over, Child, Child, Preschool, Female, Humans, Infant, Infant, Newborn, Male, Middle Aged, RNA, Messenger

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Claims read by deepseek-v4.1-flash under definition version 3 (extract-v3) on 2026-10-06.

Classified by jev-1.13-20260917 (gold-guideline-v0.2) on 2026-10-06.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review