What Previum returns on a published dentistry paper: every critique pinned to its sentence, p values recomputed, figures that do not add up and references checked against eight databases.
Sample report · dentistry · real, already published paper
A paper published in a peer-reviewed journal, run through the Basic report exactly as it is, nothing edited: the manuscript with every critique pinned to the sentence that triggers it, and the code checks underneath.
The paper has already been peer reviewed and published. It is reproduced here under its Creative Commons Attribution 4.0 license (the full attribution, authors included, is at the linked DOI; they are deliberately not named here) to show what Previum returns on a real text from your field. The objections are the ones one more reviewer could have raised; that a published paper receives them says how demanding the tool is, not how weak the paper is.
Effectiveness of mobile oral health intervention on orthodontic patients’ oral hygiene and oral health literacy: a randomised controlled clinical trial. BMC Oral Health, 2025. DOI: 10.1186/s12903-025-07423-y. License: CC BY 4.0.
The manuscript presents a timely RCT of a scalable mHealth intervention in orthodontics, with appropriate randomisation and use of validated instruments. Two critical reporting and analytic weaknesses must be resolved before the work can be considered for publication: garbled IQR values that render the primary outcome table uninterpretable, and sole reliance on LOCF without sensitivity analyses. Once these are corrected and the sample-size justification is clarified for the three-month endpoint, the study will be substantially stronger. The current version therefore requires major revisions.
Quoted sentence: «Missing data were handled using the last observation carried forward»
Why it matters: LOCF can bias estimates toward the null or inflate effects depending on the pattern of dropout; without sensitivity checks the reported between-group differences at T2 cannot be considered robust.
What would fix it: Perform and report at least one additional analysis (e.g., multiple imputation or complete-case) and state the assumed missing-data mechanism; add a sentence in the limitations section acknowledging this approach.
Quoted sentence: «1.83 (IQR = 1.50–2.00.50.00)»
Why it matters: Such reporting errors prevent accurate replication and risk misinterpretation of the magnitude and precision of the primary outcome effect; a reviewer will flag this as a fundamental flaw in results presentation.
What would fix it: Correct all IQR entries to consistent two-decimal format matching the means and SDs already reported (e.g., 1.83 (1.50–2.00)), verify every cell against the raw dataset, and re-upload a clean Table 2.
In total the reviewer found 7 major and 9 minor issues; the Basic report shows the two that decide.
With the same engine the tool uses, on 11 September 2026: the full text of the paper, tables and references included, went into the Basic report with no added instructions. The checks below are arithmetic on the text and queries to eight bibliographic databases, no model involved. Not a line of the output has been edited.
Upload your manuscript (Word or PDF) and in a few minutes you have this same report on your text. Free, with your account. The manuscript is not stored.