Academic Integrity Review: Efficacy, Safety, and Cost-effectiveness of Amphotericin B in Invasive Fungal Disease — A Retrospective Cohort Study

Academic fraud report · Geng Detector

Summary

This review assesses a 2026 retrospective cohort study on amphotericin B formulations (DOI: 10.12114/j.issn.1007-9572.2025.0374) for potential data fabrication and methodological misconduct. The overall verdict is highly suspicious. Key issues include: (1) severe denominator inconsistencies between Table 1 (baseline characteristics) and Table 2 (subgroup analyses), where patients counted in baseline (e.g., malignant tumor cases: 4 in ABCD group; prior renal disease: 4 in AMB-D group) disappear or are reduced in subgroup denominators without explanation; (2) mixed and conflicting statistical reporting in Table 1, where age, PLT, and CRP are presented as median (IQR) for some groups and mean ± SD for others, inconsistent with a single Kruskal-Wallis H test; (3) inappropriate post-hoc testing, where independent-sample t-tests are reported after ANOVA/Kruskal-Wallis without multiple-comparison correction; (4) careless textual errors such as the mislabeling of L-AMB as 'L-MBL' in the abstract and an arithmetic anomaly (1/3 reported as 33.0% rather than 33.3%). Confidence is high for internal numeric inconsistencies and medium for image-related concerns, as original figures were unavailable for pixel-level analysis.

Verdict

🟠 Highly suspicious. Multiple internal data inconsistencies and statistical irregularities strongly suggest data fabrication or severe sloppiness. The numerator/denominator mismatches between Table 1 and Table 2 cannot be explained by simple typographical error.

Key findings

Denominator shrinkage between tables: Baseline patient counts in Table 1 do not match subgroup denominators in Table 2. For example, Table 1 reports 4 malignancy cases in ABCD (n=31), but Table 2 shows 2/3 for the same subgroup; AMB-D prior renal disease drops from 4 to 1/1; L-AMB prior renal disease drops from 2 to 0/1.
Inconsistent statistical descriptors: Table 1 mixes median (IQR) with mean ± SD for the same variables (age, PLT, CRP) across groups, despite a single overall test (Kruskal-Wallis H, marked with superscript c) being referenced.
Inappropriate post-hoc testing: A two-group t-test (t=2.501, P=0.016) is reported for creatinine clearance without any indication of multiple-comparison correction (e.g., Bonferroni).
Copy-paste/textual errors: Abstract contains the nonsense abbreviation 'L-MBL' (likely L-AMB); Table 2 shows 1/3 reported as '33.0%' instead of 33.3%.
Image concerns (unverified): Figures 1 and 2 (Kaplan-Meier curves) all show P > 0.05 and unusually clean step patterns given small samples (~20 per group); pixel-level analysis not possible without source files.

Evidence highlights

Table 1 vs Table 2 — Malignancy: ABCD baseline = 4/31 (12.9%); Table 2 subgroup denominator = 3.
Table 1 vs Table 2 — Prior renal disease: AMB-D 4/20 → 1/1; L-AMB 2/? → 0/1.
Table 1 — Age reported as median 66.0 (56.8, 72.8) for AMB-D vs mean ± SD 58.5 ± 14.5 for L-AMB, with Kruskal-Wallis H noted in the footnote.
Page 4 — Creatinine clearance pairwise comparison reported as t=2.501, P=0.016, no multiple-comparison adjustment indicated.
Abstract — 'L-MBL group' (nonexistent compound; L-AMB is the study drug).
Table 2 — 1/3 reported as '33.0%' (correctly 33.3%).
DOI: 10.12114/j.issn.1007-9572.2025.0374

Notes

Image reuse/duplication could not be assessed; the report relied on text extraction only. Independent figure-level inspection is recommended.
Authors and journal should be asked to provide raw SPSS output, original subgroup coding, and full follow-up time records to verify the survival curves.
Findings do not constitute a definitive determination of misconduct; institutional investigation is required for a final verdict.