A results section reports what the study found, in the order the methods promised, without interpretation. Start with participant flow, then baseline characteristics, the primary outcome, secondary outcomes and harms. Give every percentage its denominator, every comparison an effect estimate with a confidence interval and exact p-value, and leave the meaning for the Discussion.
Reviewers check the Results against the Methods line by line. This guide works at sentence level, using one invented trial. For the structure of the whole paper, see how to structure a paper with IMRaD.
Order the results section to mirror your methods
A reader should find each item the Methods promised in the same place in the Results. Use the same outcome names, time points and order, with subheadings that repeat the Methods wording. If the Methods lists three secondary outcomes, the Results reports three, including any that did not differ.
- Participant flow. How many were screened, excluded (with reasons), enrolled or randomised, lost to follow-up and analysed. For a randomised trial this matches the flow diagram in the CONSORT guideline on the EQUATOR Network.
- Baseline characteristics. Usually in Table 1, by group.
- Primary outcome. The one outcome your sample size was calculated for, reported first and in full.
- Secondary outcomes. In the order the Methods listed them.
- Harms and adverse events. By group, with the grading system named.
- Additional analyses. Subgroup and sensitivity analyses, each labelled as pre-specified or post hoc.
Give every percentage its denominator
A percentage without a count cannot be checked. "45% were women" could describe 9 of 20 people or 540 of 1,200. Write the count, the denominator and the percentage together: 54/120 (45.0%). The denominator is the number actually assessed for that item, not the number enrolled, so it can change from row to row.
- State the unit: patients, procedures, lesions or events.
- Where a denominator shrinks, say why in the same sentence or a table footnote.
- Use one rounding rule throughout.
Report the estimate, its interval and the exact p-value together
Every comparison needs three things: the size of the difference, its 95% confidence interval and the exact p-value, with the test named. A bare "p < 0.05" tells the reader neither how large the difference was nor how precisely it was measured. Choosing effect measures is covered in reporting p-values, confidence intervals and effect sizes.
For binary outcomes, give the absolute difference as well as the relative one. Write "p < 0.001" only for very small values, never "p = 0.000". When an interval includes no difference, report the estimate and its interval plainly. Do not describe a fixed result as a "trend", as if it were still moving. How to frame that finding honestly is covered in reporting negative and null results.
Keep interpretation out of the results section
Interpretation is any statement about why a finding occurred, whether it matters, or what it means for practice. It belongs in the Discussion.
These words often signal interpretation:
- "surprisingly", "importantly", "interestingly", "as expected"
- "demonstrates", "proves", "confirms", "suggests"
- "effective", "safe", "beneficial", "superior"
- "trend towards", "marginally significant", "failed to reach significance"
Replace each with the observation itself: the two values, the difference and its interval.
Decide what goes in the text, a table or a figure
Each number should appear once, in its best home. The text carries the headline result and points to the display item, without walking through a table row by row.
| Content | Best home | What the text does |
|---|---|---|
| Primary outcome, one or two numbers | Text, and a table if there are several time points | States the full result with its interval |
| Baseline characteristics | Table | Points to the table; notes any imbalance in a variable the Methods named |
| Change over time, a distribution or a pattern | Figure | Describes the pattern and gives the key numbers |
| Participant flow | Flow diagram | Gives the numbers randomised and analysed |
| Adverse events by grade | Table | Gives serious events and any deaths |
Never show the same data as both a table and a figure. For formatting, see preparing figures and tables for publication.
Account for missing data and harms in every group
Missing data is any planned measurement you do not have. State how many values are missing for the primary outcome in each group, and the reason where known. Say which analysis the headline result comes from, and report any sensitivity analysis that handled the gaps differently.
Report adverse events by group with the denominator of participants who received the intervention, and name the system used, such as Clavien–Dindo for surgical complications or CTCAE (Common Terminology Criteria for Adverse Events), with its version. "No complications occurred" needs an observation window and a count of patients followed, as set out in grading and reporting surgical complications.
Weak-to-strong sentence bank for an invented trial
The trial below is invented to illustrate the structure; its numbers are not real data. Adults having elective abdominal surgery were randomised to a preoperative education session or usual care. The primary outcome was unplanned readmission within 30 days. Of 312 people screened, 72 were excluded (58 ineligible, 14 declined), and 240 were randomised, 120 to each group. Surgery was cancelled after randomisation, before any intervention, for 2 education and 1 usual-care participant. Spread is the standard deviation (SD).
| Part | Weak sentence | Strong sentence |
|---|---|---|
| Participant flow | Most eligible patients took part and few were lost. | Of 312 patients screened, 72 were excluded (58 did not meet the criteria; 14 declined). The remaining 240 were randomised, 120 to each group (Figure 1). |
| Baseline | The groups were well matched (Table 1). | Baseline characteristics are shown in Table 1. Mean age was 61.4 years (SD 11.2) in the education group and 62.0 years (SD 10.8) in the usual-care group; 54/120 (45.0%) and 58/120 (48.3%) were women. |
| Primary outcome | Readmissions showed a trend towards reduction with education (p = 0.06). | Unplanned readmission within 30 days occurred in 9/116 (7.8%) participants in the education group and 18/114 (15.8%) in the usual-care group (risk ratio 0.49, 95% CI 0.23 to 1.05; risk difference -8.0 percentage points, 95% CI -16.3 to 0.2; p = 0.059, uncorrected chi-squared test). |
| Secondary outcome | Education significantly improved pain (p < 0.05). | Mean pain score on day 3 was 3.2 (SD 1.5) in the education group (n = 116) and 3.9 (SD 1.7) in the usual-care group (n = 114); mean difference -0.7 points (95% CI -1.1 to -0.3; p = 0.001, Welch's t-test). |
| Missing data | A few patients were lost to follow-up and were excluded. | Primary outcome data were missing for 10 of 240 participants (4.2%): 4 in the education group (2 surgery cancelled, 2 lost to follow-up) and 6 in the usual-care group (1 cancelled, 5 lost). The main analysis used participants with complete data. |
| Harms | The intervention was safe, with no serious complications of note. | Among participants who had surgery, adverse events occurred in 21/118 (17.8%) in the education group and 24/119 (20.2%) in the usual-care group. Clavien–Dindo grade III or higher complications occurred in 3/118 (2.5%) and 4/119 (3.4%) respectively (Table 3). |
Two checks on the example. The primary result's interval includes a risk ratio of 1, so the strong sentence claims no reduction. The harms denominators (118 and 119) are participants who had surgery with their allocated intervention, not the 120 randomised to each group.
Copy the templates below and replace each slot.
Flow: Of [n screened] patients screened, [n excluded] were excluded ([n] [reason]; [n] [reason]). The remaining [n] were [enrolled / randomised], [n] to [group A] and [n] to [group B] (Figure [x]).
Baseline: Baseline characteristics are shown in Table [x]. [Variable] was [value] ([spread]) in [group A] and [value] ([spread]) in [group B].
Binary outcome: [Outcome] occurred in [n/N] ([%]) participants in [group A] and [n/N] ([%]) in [group B] ([effect measure] [estimate], 95% CI [lower] to [upper]; p = [exact value], [test]).
Continuous outcome: Mean [outcome] was [mean] (SD [sd]) in [group A] (n = [n]) and [mean] (SD [sd]) in [group B] (n = [n]); mean difference [estimate] (95% CI [lower] to [upper]; p = [exact value], [test]).
Missing data: [Outcome] data were missing for [n] of [N] participants ([%]): [n] in [group A] and [n] in [group B], because [reason].
Harms: [Event type], graded by [system and version], occurred in [n/N] ([%]) in [group A] and [n/N] ([%]) in [group B] (Table [x]).
Check the results section against the Methods before you submit
Tick each outcome, time point and analysis the Methods promised, and confirm the Results reports it in the same order. Recalculate a few percentages from their counts, and confirm each interval agrees with its p-value.
At Directive Publications, the author guidelines ask you to match your manuscript to the relevant EQUATOR checklist, and its results items make a useful final pass. If you spot an error in any article we have published, tell us how to report a problem.