Already retracted

Table 2's column totals (41 and 19) do not match the entries beneath them (35 and 25), and its reported Slug-MMP-9 p-value does not reproduce from either reading. Across 15 recomputed tests, nine did not reconcile and five change significance at p<.05.

The mediation analysis reports an indirect effect of b = 0.0325 with a 95% CI of 0.054-0.0596 - a point estimate outside its own interval, which is impossible as printed. Raised at high severity, quoted from the table, and corroborated across reviewers.

The declared data links do not resolve, statistical reporting is threshold-only with no effect sizes, and the liver-function table prints 2.00 ± 0.006 next to a 2.00 ± 0.00 vehicle - implausibly tight for the scale.

UV-exposed treated cells are reported at 31.19% telomere shortening against 14.22% in untreated controls - the opposite of the protection claimed. The TERT ELISA measures enzyme concentration, not telomere length. One reference resolves to a retracted paper.

Every confidence interval in Table 2 has its lower limit above its upper limit. Alongside it: percentages that do not match their counts (524 of 798 printed as 64%), a cycle total given as both 4,738 and 4,750, and a "significant" Pearl-Index improvement with no test reported.

The safety section states "No adverse events were noticed" in a trial reporting 16 deaths and 61 heart-failure events. Table 2's incidence rates do not recompute from its own events and person-years, Table 1 prints an impossible 134%, and the composite hazard ratio is 0.39 in the abstract, 0.40 in the results.

The abstract says ten case-control studies; the results say thirteen. Egger's test returns p < 0.001 for publication bias and the conclusions are not tempered for it, and the claim that fetuin-A reflects disease progression goes beyond the association with disease presence that was tested.

Rated the vaccine implication overstated for an in-vitro result, with a truncated data-availability statement, no cell-line authentication, threshold-only p-values, and a comet assay that describes ~2,000 cells and quantifies n > 200.

Rated the headline 5:1 risk-benefit claim unsupported by the paper's own evidence, grounded on the passage where a ~100-fold VAERS underreporting multiplier and a 94% death-misattribution correction are both called conservative and neither is validated.

On a design whose entire claim is causal: no demographic detail for the study populations, no exact p-values or confidence intervals in the text, no adjustment for multiple testing across many outcomes, and a data-availability statement with no code.

Rated the promotion/prescription conclusion as over-extending the paper's own evidence, and failed design reporting: no a priori power analysis, no pre-specified inclusion criteria, no outlier handling. The 11 statistics it could recompute all held.

Flagged the tortured phrase "convolutional neural organization" - a paraphrasing-tool fingerprint - together with a dead data URL, unversioned software and no conflict-of-interest statement.

Table 9's own values rank A1 above A3 while the table declares A3 the best college, and Table 10's total-gap figures do not follow the formula the paper states for them. Both findings are quoted from the tables.

Flagged the quality appraisal applying a case-control scale to cohort studies, and participant-level variables (age, sex, health status) missing across the pooled studies. Rated 4.25 of 5 even so: the rating deducts for demonstrated problems, and one appraisal flaw does not move it far.