Manuscript editing

What manuscript editing is, and is not

A manuscript has to do two things: report the research accurately and be easy to read. Editing serves both. It checks that the paper says what the study did and found, completely and consistently, and it improves the way that is said, so that editors, reviewers and readers can follow it. Good editing makes a paper clearer and more likely to be judged on its merits, and it removes avoidable reasons for criticism.

Editing is not authorship and not a rewrite of the science. An editor does not decide what the research means, change results or conclusions, or add claims that the data do not support. Questions that touch content, such as a number that does not match between the abstract and a table, or a conclusion that goes further than the results, are raised as queries for the authors to resolve. The authors review every change and remain responsible for the paper, as described in the research integrity and authorship statement. A paper that needs drafting from scratch is a different task, covered under the optional manuscript preparation described in the meta-analysis service.

Levels of editing

What the service includes

Editing an evidence-synthesis paper

Systematic reviews and meta-analyses have a recognizable structure, and editors and reviewers look for specific content in each part. PRISMA 2020 sets out what the title, abstract, introduction, methods, results and discussion should contain, and a manuscript that omits an item can be criticized for it. The editing follows the guideline.

Title

Identifies the paper as a systematic review, a meta-analysis or both, and says what question it addresses.

Abstract

Structured, accurate and consistent with the full text, giving the objectives, data sources, eligibility criteria, risk-of-bias methods, synthesis methods, results, limitations, conclusions and registration. PRISMA 2020 has a separate checklist for abstracts.

Introduction

Explains the rationale in the context of existing knowledge and states the objectives or questions precisely.

Methods

Gives the eligibility criteria, information sources and full search strategies, the selection and data collection processes, the data items, the risk-of-bias and effect measures, the synthesis methods and the planned subgroup and sensitivity analyses, and the protocol and registration.

Results

Presents the study selection with a flow diagram, characteristics of the included studies, risk-of-bias findings, the results of the syntheses with their uncertainty, and the assessment of reporting bias and certainty.

Discussion

Interprets the results in context, states the limitations of the evidence and of the review, and draws conclusions that follow from the evidence.

The completed checklist is checked against the manuscript item by item. See the guide to reporting a meta-analysis.

Consistency of numbers, terms and references

Inconsistency is one of the commonest problems in manuscripts and one of the most damaging to a reader's confidence. The number of included studies differs between the abstract and the flow diagram, a confidence interval in the text does not match the forest plot, an abbreviation is defined twice or never, a term is used in two senses, a table reports a different sample size from the text. Such errors arise in the course of drafting and revision, and a check that deliberately traces each number to its source catches them.

References receive the same attention. Each is checked against the source for the authors, title, journal, year, volume and pages, each in-text citation is matched to an entry and the reverse, and the list is formatted to the journal's style. Citations that cannot be verified are flagged and are not silently kept. Fabricated or inaccurate references are an integrity problem in their own right, and automated writing tools have made them a practical risk.

Language and style

Scientific writing should be as plain as the content allows. Language editing aims at sentences that say what they mean with the fewest words that do it, in a consistent, restrained register. It removes ambiguity, repairs grammar and punctuation, places the important information where readers expect it, replaces vague or inflated phrasing with precise statements, and makes terminology uniform. For authors writing in a second language, it also corrects usage that is understandable but non-idiomatic. The editing respects the author's voice and the conventions of the field, and it does not make a text ornate. Changes are tracked, and an author can reject any of them.

Numbers and units follow the journal's style: how to report confidence intervals, how many decimals, how to write p-values, and how to present effect measures with their scales so that readers know the direction of effect. Abbreviations are kept to those that help the reader, and each is defined at first use.

Tables, figures and supplementary material

Tables and figures should be understood without reading the text. Each has a title that says what it shows, labeled axes and columns with units, legends that define abbreviations and symbols, and an order that matches the text. In evidence-synthesis papers the forest plot, the flow diagram, the risk-of-bias figure and the summary of findings table are scrutinized, and each should follow convention. The review checks that they are legible at the journal's size, that the numbers match the text, and that supplementary files, such as full search strategies and excluded studies, are referenced and complete.

Common problems in evidence-synthesis manuscripts

Certain problems recur in review manuscripts, and an edit looks for them deliberately. The title does not say that the paper is a systematic review or meta-analysis, which makes it harder to find and to index. The abstract states results or conclusions that are not in the main text, or omits the registration number. The objective is vague, so readers cannot tell what question the review answers. The methods describe the search without giving the strategies or the dates, or do not say how many reviewers screened and extracted. The flow diagram numbers do not add up. Heterogeneity is reported with a statistic but not interpreted. The results are described as significant or not, without the size of the effect and its interval. The discussion restates results without considering the certainty of the evidence or the limitations of the review. And the conclusions are broader than the evidence supports, for instance recommending practice on the basis of low-certainty findings. Each of these can be corrected without altering a single result, and each, left alone, is a ready target for a reviewer.

Preparing your manuscript for editing

A few things make editing faster and better. Send the most complete version, with the tables, figures and any supplementary files, and say which version is current if several authors have contributed. Give the target journal and its instructions, or the list of journals you may try. Include the reporting checklist if you have completed one, and the protocol or registration. Say if parts of the text were produced with the help of translation or automated tools. Mark any sections you already know are weak or any comments from co-authors or reviewers you want addressed. And name the decisions that only you can make, such as which results to emphasize, so that queries can be aimed at them. The more complete the draft, the more of the effort can go to clarity and consistency and the less to filling gaps.

How the editing works

Deliverables

Optional extension: from analysis to full paper

Limitations and boundaries

Editing cannot rescue a study that is flawed in design or analysis, and it cannot guarantee that a paper will be accepted. It does not verify the underlying data, and the authors remain responsible for their accuracy. An editor can flag a conclusion that seems to go beyond the results, but whether to change it is the authors' decision. Editing for language does not replace the judgment of a subject expert about the science. Authors should also tell us if they have used automated writing or translation tools on the text, so that the checks of facts and references can be made with that in mind, and they should follow the journal's policy on such tools. Our own position is in the AI-use policy. This service provides research and publication support and does not provide clinical advice.

How long does it take?

The time depends on the length and condition of the manuscript, the level of editing, the number of tables and figures, whether the reference list needs checking in full, and how quickly the authors answer queries. A clean, well-organized draft takes less time than one that has been revised many times by several authors. Tell us the deadline at the start, so that the level of editing can be chosen to fit it.

Frequently asked questions