Integrity review: Prognostic value of Musashi 2 (MSI2) in cancer patients — a systematic review and meta-analysis (Frontiers in Oncology, 2022)

Academic fraud report · Geng Detector

Summary

This integrity review examines the meta-analysis by Jiang et al. (DOI: 10.3389/fonc.2022.969632) on Musashi 2 (MSI2) as a prognostic biomarker across cancer types. Verdict: no substantive evidence of academic fraud was identified. The investigation covered numerical self-consistency, statistical plausibility, and temporal/reference logic. Patient counts across 21 studies in Table 1 sum exactly to the stated 2,640 patients. Reverse-engineered Z statistics (e.g., OS overall HR 1.84, 95% CI 1.66–2.05, reported Z = 11.36; and Degree of differentiation OR 0.68, 95% CI 0.27–1.72, reported Z = 0.81) match reported values within rounding tolerance. The literature search window (through April 2022), submission date (21 June 2022), and publication date (1 December 2022) form a coherent timeline, and cited references were verifiable. One minor table-labeling irregularity (Thol F 2013 quality score '8/6') reflects dual-endpoint NOS scoring rather than fabrication. Limits: the review relied on reported numbers only; raw individual patient data were not inspected.

Verdict

No substantive fraud indicators detected. The meta-analysis appears to be an internally consistent, methodologically standard literature synthesis. No action (PubPeer post or editorial report) is recommended.

Key findings

Patient-count arithmetic validated: Summing the 21 cohort sizes listed in Table 1 (75+40+180+105+126+91+82+108+162+85+62+164+119+67+118+181+149+51+454+101+120) yields exactly 2,640, matching the abstract.
Statistical back-calculation consistent:
OS (All): HR = 1.84, 95% CI 1.66–2.05 → back-calculated Z ≈ 11.33 vs. reported 11.36 (rounding within normal Stata output range).
Degree of differentiation (Table 3): OR = 0.68 (0.27–1.72) → back-calculated Z = 0.81, identical to reported value.
Subgroup logic self-consistent: Hematologic vs. solid-tumor subgroup totals, I² heterogeneity statistics, and HR/OR values were mutually consistent across Tables 2 and 3.
Timeline coherent: Literature search cutoff April 2022; submission 21 June 2022; published 1 December 2022. No temporal "cross-time" anomalies.
Reference list plausible: Recent citations (e.g., Zhen J 2022; Kharin L 2021) fall within the search window and appear to be real publications; no fabricated references identified.
Software version acceptable: Stata 12.0, while dated, is widely used in clinical meta-analytic work and not inherently suspicious.

Evidence highlights

Table 1 cohort sum equals stated n = 2,640 exactly.
Reverse-engineered Z values from reported HR/OR and 95% CI bounds match reported Z values within rounding tolerance (Δ ≤ 0.03).
Reported submission/publication dates are internally consistent with the April 2022 search cutoff.
Table 1 labels Thol F 2013 with two quality scores ("8/6") corresponding to dual endpoints (OS and DFS), a routine dual-rating convention in NOS-based reviews.

Notes

One minor labeling irregularity was noted (Thol F 2013 quality score "8/6" for OS(M)/DFS(U)), reflecting separate NOS ratings per endpoint rather than fabrication.
This review did not access individual patient-level data underlying the included primary studies; only published aggregates were checked.
The use of Stata 12.0 is older than current best practice but is common in hospital-based meta-analyses.
Confidence: moderate-to-high for numerical consistency; lower for any conclusions requiring raw-data inspection beyond what the paper reports.