How to Write the Results Section of a Medical Manuscript: What Editors Expect in 2026

There is a persistent belief among clinical researchers that the results section is the easiest part of a medical manuscript to write. You collected the data, you ran the analyses, and now you just need to report what you found. In practice, peer reviewers disagree. Methodologists who review for major clinical journals often say the results section is the part they read most carefully, because it is the section most susceptible to presentation choices that quietly distort what the study actually showed. Outcome switching, selective p-value presentation, undisclosed post-hoc analyses, and inconsistencies between the text and the tables are all common enough that experienced reviewers actively look for them.

The context has sharpened in recent years. The CONSORT 2025 update, published simultaneously in Nature Medicine, The Lancet, PLOS Medicine, and the BMJ in April 2025, made transparency in results reporting one of its central concerns. Journals that adopted the updated checklist now expect authors to specify, in the results section itself, whether the analysis plan was pre-registered and whether any deviation from that plan occurred. That is a significant change from simply being asked to list registered outcomes in a disclosure line. The bar for what counts as a complete results section has moved, and many authors are still writing to the old standard.

The Fundamental Rule: Report, Do Not Interpret

The cleanest way to think about the results section is that it exists to answer one question: what did you observe? Not what it means, not why it happened, not how it compares to previous studies. That belongs in the discussion. The results section is a record of what your study produced, written clearly enough that a reader could assess your findings independently before they reach your interpretation of them.

In practice, the line is not always sharp. It is acceptable to note in the results section that a difference was statistically significant or that a confidence interval excluded the null. It is not appropriate to say in the results section that this finding confirms the clinical relevance of your intervention, or that it is consistent with the hypothesis you stated in the introduction. Those sentences belong in the discussion, and when they appear in the results, they usually signal that the author is compensating for data that does not stand well on its own.

Editors at journals like JAMA, the BMJ, and the New England Journal of Medicine are trained to catch this early. Interpretive language in the results section is one of the signals that a reviewer may be primed to read the rest of the manuscript skeptically. The cleaner the results text, the more your data speaks for itself, which is exactly what you want.

Start Where Your Methods Left Off: Participant Flow First

For most clinical study designs, the results section should open with a description of the study population before any outcome data appear. For randomized trials, the CONSORT 2025 checklist expects a flow diagram showing the number of participants who were assessed for eligibility, randomized, received allocated interventions, completed follow-up, and were analyzed. The text should summarize this flow briefly and flag any meaningful deviations between what was allocated and what was analyzed.

Baseline characteristics belong here as well. These are typically presented in a table rather than prose, but the text should note whether groups were comparable on key variables and whether any imbalances existed that might affect interpretation. For observational studies following the STROBE framework, this section establishes the cohort or case population and describes losses to follow-up or missing data before any analysis results are reported.

One common error is burying poor retention rates inside a table footnote without flagging them in the text. If your study lost more than ten percent of participants to follow-up, or if the analyzed population differs meaningfully from the enrolled population, that should be visible and clear in the opening results paragraphs. Reviewers who find it only in a footnote often assume the author was hoping they would not notice.

Outcome Order Matters More Than Authors Realize

The order in which you report outcomes in the results section should match the order in which you declared them in the methods section, which should in turn match the order in which they were pre-registered. Primary outcome first, then secondary outcomes, then exploratory or post-hoc analyses. This is not merely a stylistic preference. It is a transparency requirement that reviewers now check against the methods text and, increasingly, against the trial registration record.

Selective outcome reporting, where authors lead with the outcome that produced the most favorable result rather than the outcome they specified as primary, remains one of the most documented forms of publication bias. The now-famous 2008 analysis by Chan et al., and subsequent replications, showed that statistically significant outcomes were substantially more likely to be fully reported than non-significant ones from the same trials. CONSORT 2025 addressed this directly by adding a requirement that authors state explicitly in the results whether the primary outcome changed after trial initiation and provide justification if it did.

For observational studies, the comparison point is the analysis plan described in the methods section rather than a registration record, though pre-registered observational studies are increasingly common. The logic is the same: the sequence of results presentation should be predictable from the methods you described, not arranged to tell the most favorable story.

Statistical Reporting: What the Current Standards Require

A 2025 audit of 100 clinical medicine papers found that 65 percent lacked adequate descriptions of how statistical tests were applied, and 64 percent failed to report effect sizes. Those numbers are striking not because the requirements are new but because they remain so widely ignored. The SAMPL guidelines (Statistical Analyses and Methods in the Published Literature), developed by the authors of the original CONSORT statement, have been available since 2015 and are referenced in the author guidelines of most major medical journals. The problem is not that authors do not know these requirements exist. It is that results sections are often drafted under time pressure, and statistical completeness is the first casualty.

The minimum for any clinical results section is: report the test used, report the test statistic and degrees of freedom, report the exact p-value (not just p < 0.05), report the effect size, and report a confidence interval. For continuous outcomes, that typically means reporting means or medians with appropriate measures of spread (standard deviation or interquartile range), not just the change from baseline. For binary outcomes, absolute risk differences and numbers needed to treat are generally more clinically interpretable than odds ratios alone, and at least some journals now explicitly request both.

The shift away from binary significance thresholds deserves particular attention. Since 2019, the American Statistical Association has actively discouraged the dichotomization of results into significant and non-significant based on whether a p-value crosses 0.05. The Lancet family of journals has updated its statistical reporting guidance accordingly, and several other clinical journals now request that authors frame results in terms of the magnitude of the observed effect and its precision rather than in terms of whether statistical significance was achieved. If your results section says things like "the primary outcome did not reach statistical significance" without reporting what the effect was, a reviewer at any rigorous journal will flag it.

Tables and Figures: Choosing the Right Format

The results section of a clinical manuscript rarely works as pure prose. Most studies generate too many numbers to read comfortably in paragraph form, and tables and figures exist to carry the data load while the text guides readers to the findings that matter most. Getting the relationship between text and display items right is a practical skill that takes deliberate attention.

The default rule is: use a table when readers need to look up specific values or compare many numbers precisely, and use a figure when you want to show a pattern, trend, or relationship that prose alone cannot convey. A forest plot showing effect estimates across subgroups communicates something that a table of odds ratios would obscure. A Kaplan-Meier survival curve shows time-to-event data in a way no table of median survival times can fully replace. But a simple comparison of means across three groups rarely benefits from being displayed as a bar chart rather than a two-row table.

Whatever format you choose, the text should not duplicate the table or figure in full. The most common structural error in results sections is reciting every number that appears in a table, row by row, in the paragraph above it. The purpose of the text is to draw the reader's attention to the most important findings and to interpret which results deserve emphasis, not to transcribe the table in sentence form. If you find yourself writing "Group A had a mean score of 42.3 (SD 8.1), Group B had a mean score of 44.7 (SD 7.9), and Group C had a mean score of 39.6 (SD 9.2)" and those numbers also appear in Table 2, eliminate one or the other.

Each table and figure should be independently interpretable. Titles should be informative enough that a reader can understand what is being shown without reading the surrounding text. Footnotes should define all abbreviations and explain any unusual data presentation choices. The BMJ, Nature Medicine, and most journals in the Lancet family now check this during editorial screening, not just peer review, because display items that require the text to decode them slow down reviewers and suggest careless preparation.

Reporting Negative and Null Results Honestly

A substantial proportion of clinical trials do not show a statistically significant benefit for the primary intervention, and observational studies frequently find no association where researchers expected one. How authors handle these results in the text is one of the places where editorial judgment and research integrity intersect most directly.

The temptation when primary results are null is to lead the results section with a secondary or exploratory finding that was positive, to report the null primary result in a subordinate clause, or to frame negative findings with language that implies they might have reached significance in a larger sample. All three of these moves are recognized by experienced reviewers. They do not rescue a manuscript from a null finding. They instead raise doubts about whether the authors are presenting the study they designed or the study they wish they had designed.

A null result reported cleanly and completely is a legitimate scientific contribution. Journals including PLOS ONE, PLOS Medicine, Trials, and the BMJ have editorial policies that explicitly welcome null findings when the study was adequately powered and appropriately conducted. The Registered Reports format, now offered by more than 300 journals including Nature and BMJ Open, accepts in-principle commitments to publish before the data are collected, specifically to remove the publication pressure that drives outcome switching and selective reporting.

When your primary result is null, report it first, state the effect size and confidence interval, and let the confidence interval do the work. A confidence interval that excludes a clinically meaningful effect is informative. A confidence interval that is wide and includes effects ranging from clinically harmful to substantially beneficial tells reviewers and readers something different, and that distinction belongs in the results, not masked by a p-value summary.

Subgroup Analyses: A High-Risk Area

Subgroup analyses are probably the single most abused component of clinical trial results sections. A trial that shows no significant effect on its primary outcome may show a significant effect in a subgroup defined by age, sex, baseline severity, or genotype, and that subgroup finding may be true, or it may be a chance result produced by multiple testing across many possible subgroups. The results section cannot always resolve that ambiguity, but it can and should make the ambiguity visible.

The CONSORT 2025 checklist explicitly requires authors to distinguish pre-specified subgroup analyses from post-hoc subgroup analyses, and to report the statistical test for interaction rather than simply the within-subgroup p-values. A subgroup analysis that shows a significant effect within the subgroup but a non-significant test for interaction does not provide reliable evidence that the effect is genuinely different between subgroups. Many published papers still omit the interaction test entirely, and reviewers at rigorous journals now routinely request it.

The honest approach is to label each subgroup analysis clearly as pre-specified or exploratory, report the interaction test, and resist the temptation to headline a significant subgroup finding when the overall trial result was null. If a subgroup finding is the genuine scientific story of your study, say so in the discussion, with appropriate caveats, rather than burying it in the results section under a table heading that makes it look like a primary analysis.

Consistency Between Text, Tables, and Registration

Inconsistencies between different parts of the manuscript are common, and the results section is where they surface most often. A number that appears as 43.2% in the text and 43.5% in Table 1, an odds ratio cited as 1.8 in a paragraph and 1.78 in the forest plot, a p-value rounded differently in two places: none of these errors are scientifically meaningful individually, but in aggregate they undermine the impression that the manuscript was prepared carefully. Reviewers at journals with strict statistical review, including the New England Journal of Medicine and JAMA, sometimes reject manuscripts at revision stage because too many small inconsistencies surfaced after close reading.

The most consequential inconsistencies are those between the results section and the trial registration record. If your registered primary outcome was 30-day all-cause mortality and your results section reports 30-day cardiovascular mortality as the primary outcome, reviewers will notice, and they will ask for an explanation. If your registration listed four secondary outcomes and your results section reports nine, the additional five need to be clearly labeled as post-hoc, along with a note about why they were added. This is not pedantry. It is the transparency requirement that prevents selective reporting from polluting the clinical literature.

A practical consistency check that many experienced authors use: before final submission, have one co-author, who was not responsible for drafting the results section, read it against the methods section, the registration record, and the tables simultaneously, looking specifically for any discrepancy in numbers, outcome labels, or analysis descriptions. This is the kind of error that is invisible to the person who wrote the section and obvious to someone reading it cold.

Handling Adverse Events and Safety Data

For intervention trials, safety and adverse event data belong in the results section and are not optional. The CONSORT 2025 checklist added explicit items on harms reporting that had been underspecified in earlier versions, and CONSORT-Harms, a dedicated extension, provides detailed guidance for studies where harm data is a central concern. Many authors treat the adverse events table as an afterthought placed after the efficacy results, but reviewers at clinical journals now read the harms section carefully, partly because post-publication safety signals are common and partly because regulators sometimes request the underlying data for marketed interventions.

At minimum, the results section of a trial should report all-cause serious adverse events, any adverse events that led to study discontinuation, and any adverse events with an incidence above a pre-specified threshold (commonly five percent). If certain adverse event categories are more clinically relevant than others for the intervention being studied, those should be highlighted in the text rather than left to the reader to find in a large safety table. When harms data show something unexpected, report it as clearly in the results as you report the efficacy data.

A Pre-Submission Checklist for the Results Section

Most of the errors described in this guide are preventable with a deliberate review pass before submission. The checklist below is not exhaustive, but it covers the categories of problems that most commonly trigger reviewer comments and major revision requests at clinical journals in 2026.

Working through this checklist with a co-author who was not the primary writer is the most reliable way to catch the inconsistencies and omissions that survive multiple rounds of self-editing. The reviewer reading your manuscript cold will notice them. Give yourself the same advantage.

What Strong Results Sections Have in Common

The results sections that reviewers find genuinely satisfying are usually short for the amount of data they contain. They move in a logical order that is entirely predictable from the methods section. Every number in the text can be traced to a table or figure. Every analysis is clearly labeled. The effect sizes are large enough or precise enough to be informative regardless of where the p-value falls. And nothing in the results section requires the reader to trust that the authors are presenting the study honestly rather than strategically, because the structure itself makes strategic omission obvious.

That last point is the one that has become most salient in 2026. As journals have invested more heavily in statistical review, pre-registration checking, and data availability requirements, the expectations around results reporting have converged toward a single principle: the results section should be something that a reader, using only publicly available information, could independently verify. Not completely, but enough to detect the most consequential problems. When authors write with that standard in mind, the results section tends to become cleaner, shorter, and more convincing regardless of what the data actually show.

The goal is not a perfect study. The goal is a complete and honest account of the study you ran, written so that the data can speak for themselves. That is all a results section needs to do, and it is harder than it sounds.

Further Reading