Statistical integrity report: Analysis of factors affecting vitamin K1 levels in preterm infants on long-term parenteral nutrition (DOI: 10.3760/cma.j.cn101451-20250709-00268)

Academic fraud report · Geng Detector

Summary

This integrity review concludes that the paper '长期肠外营养早产儿维生素K1水平的影响因素分析' by Huo Yuan, Zhang Guo, Zhao Yang, Li Wanyi, and Gao Hongxia, published in the Chinese Journal of Neonatology (Chinese & English), shows severe statistical anomalies consistent with fabricated or carelessly fabricated data. Reported as 'confirmed' (实锤) by the original Chinese reviewer. Five major issues were identified in Table 1 (p. 423): (1) a negative P-value of −0.957, mathematically impossible as P∈[0,1]; (2) multiple negative Kruskal-Wallis H statistics (e.g., −0.080, −0.043, −0.843, −1.488), which cannot be negative; (3) inconsistent use of 'H value' as a label for both two-group Mann-Whitney U and multi-group Kruskal-Wallis comparisons; (4) duplicated median and IQR values (8.4 (3.0, 11.5)) across independent subgroups; (5) a H statistic of 7.237 yielding an expected P≈0.02–0.03, but a negative P is reported to support a non-significance conclusion. Confidence is high for findings 1–4 based on basic statistical theory. Limitations: the report relies on a single table image; raw SPSS output and underlying data were not independently verified.

Verdict

🔴 Confirmed — severe statistical anomalies indicating likely data fabrication or gross negligence in the production of Table 1.

Key findings

Negative P-value of −0.957 reported for the 14-day vitamin K1 comparison across parenteral nutrition duration groups (14~21 d, 22~28 d, >28 d), contradicting the definition of probability.
Multiple negative Kruskal-Wallis H statistics (−0.080, −0.043, −0.843, −1.488) for two-group and multi-group comparisons; H is bounded ≥ 0.
Methodological inconsistency: the Methods section cites both Mann-Whitney U and Kruskal-Wallis H tests, but Table 1 uniformly labels all test statistics as 'H值', regardless of group number.
Duplicate numeric entries (8.4 (3.0, 11.5)) appearing in both the non-exclusive breastfeeding subgroup (n=48) and the full-course glucocorticoid subgroup (n=46) for the 28-day vitamin K1 measurement, suggesting copy-paste error or fabricated underlying data.
A reported H=7.237 (df=2) would correspond to a genuine P≈0.02–0.03; the paper instead reports a negative P and concludes the differences are not statistically significant.

Evidence highlights

Table 1 (p. 423): H=7.237 paired with P=−0.957 for 14-day vitamin K1 across three PN duration groups.
Table 1 (p. 423): H values of −0.080, −0.043, −0.843, and −1.488 across several subgroup comparisons labeled as Kruskal-Wallis.
Table 1 (p. 423): Identical 28-day vitamin K1 medians and IQRs (8.4 (3.0, 11.5)) for non-exclusive breastfeeding (n=48) and full-course glucocorticoid (n=46) subgroups.
Methods section states use of SPSS 26.0 with Mann-Whitney U or Kruskal-Wallis H tests; SPSS cannot output negative P-values or negative H statistics.
Discussion claims 'no statistically significant differences' in 14-day and 28-day vitamin K1 levels across PN duration groups, contradicting the magnitude of H=7.237.
DOI: 10.3760/cma.j.cn101451-20250709-00268.

Notes

All statistical claims in this report are inferred from the published Table 1 and Methods text; raw SPSS output files were not available for independent verification.
The authors are affiliated with Gansu Provincial Maternal and Child Health Hospital / Gansu Provincial Central Hospital.
The reviewer recommends contacting the authors for raw data, posting concerns on PubPeer, and reporting to the editorial office of 中华新生儿科杂志（中英文） and the authors' institutional research ethics committee.
This AI-assisted report is for academic discussion only and does not constitute a formal finding of misconduct.