Writing the results section

A results section reports what happened once the study was run. Its data are selected for what bears on the study's objectives, worded in the past tense and exact about what each number counts, and ordered from the subjects through the primary outcome to the secondary ones. Each result is built from an orientation, its data and a statement of what it shows.

Selecting the data

A results section is made by choosing, and a box of handmade chocolates pictures the choice: a shop holds twenty or thirty varieties and a box holds six, so each one has to be chosen with care. The data kept are those that support the primary and the key secondary objectives. Where CONSORT is followed, the demographic data and the flowchart of which patients were included, which were excluded and why already take one of the six. ^{1, 2}

Each result in the section should serve the key message. Asking of each data point, yes or no, whether it does keeps the writer from filling the section with material copied from the clinical study report, or from other articles in a review article. ³

Authors can pull the selection two ways: some may be reluctant to include data that weaken their results, and others may want to include all their data, even data outside the manuscript's focus. The section includes all the data a peer reviewer would expect to see, so it helps to think like a peer reviewer during outlining and to ask the questions a reviewer would ask before the final content is decided. ⁴

What the reporting guidelines require

Reporting guidelines such as CONSORT and the ICMJE recommendations give some guidance on what a results section contains and little or none on how to organise it. A journal's instructions for authors may add what to include, and they do not explain how to organise it either. ⁵

The CONSORT statement guides the reporting of randomised clinical trials and can be adapted to other study designs. It requires the results section to include: ^{5, 6}

• participant flow
• the dates defining recruitment and follow-up
• the demographic and clinical characteristics of each group
• for each group, the number of participants included in each analysis and whether the analysis was by the originally assigned groups
• for each primary and secondary outcome, the result for each group with the estimated effect size and its precision, and for binary outcomes both the absolute and the relative effect sizes are recommended
• the results of any other analyses performed, including subgroup and adjusted analyses, distinguishing the pre-specified from the exploratory
• adverse events

The list does not set out a logical flow, but following its order can help. ⁵

The ICMJE recommendations add principles of presentation. Results are presented in logical sequence across text, tables and figures, with the main or most important findings first. The text does not repeat all the data in the tables or figures and emphasises or summarises only the most important observations. Data are provided on all primary and secondary outcomes identified in the methods. Extra or supplementary material and technical detail can go in an appendix, where it is accessible without interrupting the flow of the text, or be published solely in the electronic version of the journal. ⁶

The order of the section

Once the results to include are decided, they are presented in a logical order. For a randomised clinical trial, a flowchart of patients through the trial, such as the one CONSORT provides, helps in making an informed analysis of the treatment course, and it also aids communication between author and writer. A case report discusses the patient background, medical history and the other details needed to understand the choice of treatment. With the aims of the study and of the manuscript in mind, a medical writer can help authors decide what is essential to their narrative while providing the objectivity an honest description of the findings needs, and may also need to advise on the statistics, such as P values, confidence intervals and odds ratios, that would support the argument. ⁴

One structure for a clinical study runs in three moves and can be adapted to any type of study. ⁶ Start with a section describing the subjects, titled Subjects or Patients. Its first sentence states how many subjects were included, when, and how they were split up or randomised. What happened to them follows, then their baseline characteristics. A complex study includes a CONSORT flow diagram, and a simple one can summarise the flow of subjects in the text. In a worked example, one paragraph gives the enrolment totals by arm, the enrolment and completion dates, the withdrawals of consent before treatment, the non-completions with their main reason, the resulting number who completed, and a short description of the population by sex, average age, ethnicity and prior corticosteroid treatment, referring to the flow diagram and the baseline table. ⁷

Follow with the primary outcome or outcomes, presented first because the goal of the study and the main focus of the section should be to address the primary objective. For a study assessing the superiority of a vaccine over the current standard vaccine, the section might simply be titled Immunogenicity. Within it the main analysis, here the superiority analysis, comes first, followed by related analyses such as antibody titres, cellular immune responses and the factors predicting immune responsiveness or its lack. ^{8, 7}

Finish with the secondary or exploratory outcomes, organised into logical chunks, each with its own title. Within each chunk the most important assessments come first, followed by secondary information or more detailed or exploratory analyses. For a vaccine study the next logical section might be safety, titled Safety or Adverse events. There the most important information might be the expected or solicited reactions, such as injection-site swelling and itching, followed by unsolicited adverse events and, finally, analyses of the risk factors for either. ⁷

Each method described in the methods section should have results, and organising the results in the same order as the methods, from most to least important, may help. ⁹

The parts of each result

The results section is built from three components, and two further considerations apply in writing it. ^{10, 11}

The first component, context re-orientation, can be as little as a subheading and a sentence. It makes the results section self-sufficient, so the reader has less need to reread the introduction and methods. ¹² For example, under the subheading Relative molecular weight determination: To determine the relative molecular weights of proteins A and B in order to ascertain their structural relation, the purified proteins were electrophoresed on polyacrylamide. ¹³

The second, data presentation, usually centres on a table or figure. So excessive repetition of the data in the text is verbose and unsophisticated to an insightful reader, who would see the repetition as a ploy to pad the article. ¹¹

The third is the result statement with a data-based preliminary interpretation, results being data-based observations, trends and comparisons. The movement of protein A corresponded to a marker with a molecular weight twice that of protein B states a result. The molecular weight of protein A was twice that of protein B, an observation consistent with protein A as a dimer of protein B adds its interpretation. ¹⁴

The first further consideration is tense. A journal article is a retrospective, so its tense is primarily past, which conventionally conveys an observation of past information rather than a current observation of the graphical data. The past tense also conveys understatement: something was, against the time-independent, overstated truism something is. ¹⁵

The second is premature inference. A preliminary interpretation in the results is controversial because it resembles the discussion. But the discussion extends the results into four kinds of paragraph, namely conclusion-support, limitation and counterargument, recommended future research, and conclusion-consequence, all built on result statements already made in the results, and none of them appears in the results. ^{10, 16}

Each component of the results should be oriented to its method description. Tables and figures convey more information than descriptive text, and the text should be used for observations, not to repeat the data. ¹⁷

Findings are described in a logical sequence, fully yet succinctly. The section should open with some explanatory prose before sending the reader to the first table, and tables and figures organise numbers in ways words cannot. Each table or figure has a legend that lets it be understood without the text, and tables and figures should not duplicate the data in the text, the numbers being presented in one place only. ¹⁸

Tense and numbers

The results section reports what happened after the study was done, and its wording has to say so. Tenses often get mixed up, and a result is sometimes worded so that it reads as a method, which is worth checking. The wording also has to be exact about who did what: 590 dermatologists completed the questionnaire is a different claim from 590 dermatologists were sent the questionnaire. Where 600 were sent the questionnaire and 590 answered it, the two numbers are expressed differently. ^{19, 20, 3}

A results section is largely writing about numbers, to describe comparisons or to highlight differences that are statistically or clinically significant. One device for clear sentences about numbers is the "4 W's": ²¹

• who: the group or subgroup
• what: the units of measurement
• when: the duration of analysis or the year
• where: the study site or the geographic population
• how much: the magnitude and direction of the effect or response