Writing the Statistical Results Section

Overview

The statistical Results section reports what was observed and estimated.

It should present:

• participant numbers and analysis populations;
• missing data and exclusions;
• descriptive characteristics;
• primary outcome estimates;
• confidence intervals;
• relevant test results;
• secondary and sensitivity analyses;
• harms or adverse outcomes where relevant.

The section should follow the order of the objectives and Methods. It should not introduce new statistical procedures or explain why findings matter clinically; those tasks belong mainly in the Methods and Discussion.

ICMJE recommends presenting results in a logical sequence, giving the most important findings first, reporting absolute numbers alongside percentages and avoiding unnecessary duplication between text, tables and figures [1].

Why this matters

A technically correct analysis can still be reported poorly.

Common problems include:

• leading with p-values instead of effect estimates;
• omitting denominators;
• reporting only statistically significant outcomes;
• repeating every table entry in prose;
• presenting adjusted results without crude context;
• using causal language for observational associations;
• hiding missing data or excluded participants;
• describing a non-significant result as no effect.

A good Results section lets readers reconstruct the evidential path from participants to estimates.

When to use this framework

Use the framework for:

• journal manuscripts;
• theses;
• clinical-trial reports;
• observational studies;
• diagnostic studies;
• reports to funders or ethics committees.

The exact structure should follow the study design and relevant reporting guideline.

A logical sequence

Participant flow

Report:

• screened;
• eligible;
• enrolled or randomised;
• followed;
• analysed;
• missing primary outcome;
• excluded and reasons.

For randomised trials, a participant-flow diagram should be used with CONSORT 2025 [2].

Descriptive characteristics

Present clinically relevant baseline characteristics using descriptive summaries.

The purpose is to describe the sample and assess applicability, not to test whether randomisation succeeded.

Primary outcome

Report:

• group summaries;
• primary effect estimate;
• confidence interval;
• p-value if part of the inferential framework;
• analysis population;
• adjustment method;
• direction of effect.

Secondary outcomes

Report prespecified secondary outcomes consistently. Avoid emphasising only favourable findings.

Sensitivity and supplementary analyses

State whether conclusions changed under alternative assumptions or populations.

Harms

Benefits and harms should be reported with denominators and uncertainty where feasible.

Data example

Simulated results structure

simulated data

Result component	Value
Randomised	240
Primary outcome observed	226
Telemonitoring observed mean	132.4 mmHg (SD 11.9), n=114
Usual-care observed mean	138.1 mmHg (SD 12.3), n=112
Adjusted mean difference	−5.9 mmHg
95% CI	−8.5 to −3.3 mmHg
Blood-pressure control	82/120 versus 65/120
Risk difference	14.2 percentage points
95% CI	1.8 to 25.9 percentage points

Note: The calculations underlying these simulated adjusted estimates, confidence intervals and p-values are not shown here.

Reporting denominators

A percentage must be linked to a count and denominator.

Write:

82 of 120 participants (68.3%)

rather than:

68.3%.

If the denominator differs across outcomes, state it in every relevant row or footnote.

Missing outcomes should not be silently removed from the denominator.

Reporting numerical outcomes

A numerical result usually requires:

• observed number;
• mean and SD, or median and quartiles;
• measurement unit;
• effect estimate;
• confidence interval;
• test statistic and degrees of freedom where useful.

Avoid using mean ± SD if the plus–minus notation could be confused with an interval.

Reporting categorical outcomes

Report:

• events and total;
• percentage;
• absolute effect;
• relative effect where useful;
• confidence intervals;
• time horizon for risks.

An odds ratio should be labelled as an odds ratio and not translated into a probability ratio.

Reporting regression results

For each primary predictor, report:

• coefficient or transformed effect;
• units;
• reference category;
• confidence interval;
• p-value where relevant.

Do not report only whether a variable “remained significant”.

Worked example

R-assisted reporting

Analysis objects can be converted into clean tables rather than copied as raw console output.

primary_result <- data.frame(
  outcome = "Systolic blood pressure at 12 weeks",
  estimate = -5.9,
  ci_lower = -8.5,
  ci_upper = -3.3,
  unit = "mmHg",
  p_value = 0.000012
)

within(
  primary_result,
  {
    formatted_effect <- sprintf(
      "%.1f %s (95%% CI %.1f to %.1f)",
      estimate,
      unit,
      ci_lower,
      ci_upper
    )

    formatted_p <- ifelse(
      p_value < 0.001,
      "<0.001",
      sprintf("%.3f", p_value)
    )
  }
)

This reduces transcription errors and supports consistent rounding.

Confidence intervals and non-significant findings

A non-significant finding should be reported using its estimate and interval.

Example:

The odds ratio for serious adverse events was 0.82 (95% CI 0.31 to 2.15).

This interval is compatible with substantial benefit and substantial harm. Writing “there was no difference” would overstate certainty [3].

Rounding

Use precision appropriate to:

• measurement scale;
• sample size;
• estimate uncertainty;
• clinical interpretation.

Typical practices include:

• one decimal place for blood pressure;
• two decimal places for ratios;
• three decimal places for ordinary p-values;
• $p<0.001$ for smaller values.

Do not write p=0.000.

Tables, figures and text

Use text to emphasise the central findings.

Use tables when readers need exact values across several outcomes or predictors.

Use figures for:

• distributions;
• trends;
• confidence intervals;
• survival curves;
• calibration;
• interactions.

Do not reproduce the same complete dataset in text, table and figure.

Interpreting the Results section

Presenting the results

Suggested reporting

Use a consistent sequence: population and denominator, descriptive value, effect estimate, confidence interval and p-value where relevant. SAMPL recommends exact numerical reporting and prioritising estimates over binary significance labels [4].

Common mistakes

Knowledge check

What should lead a primary outcome result?

Answer and reasoning

The group values and primary effect estimate with its confidence interval, not the p-value alone.

Why report 82/120 as well as 68.3%?

Answer and reasoning

The count and denominator show how the percentage was constructed and how much information it represents.

How should a non-significant result be described?

Answer and reasoning

Report the estimate and interval and state that the data do not exclude the null; do not claim equality unless an equivalence design supports it.

Where should interpretation of clinical mechanisms appear?

Answer and reasoning

Primarily in the Discussion, not the Results section.

Why automate table generation from analysis objects?

Answer and reasoning

It reduces transcription, rounding and inconsistency errors and strengthens reproducibility.

Key take-home messages

References

1. International Committee of Medical Journal Editors. Recommendations for the conduct, reporting, editing, and publication of scholarly work in medical journals: preparing a manuscript for submission. Updated 2026. https://www.icmje.org/recommendations/browse/manuscript-preparation/preparing-for-submission.html
2. Hopewell S, Chan AW, Collins GS, Hróbjartsson A, Moher D, Schulz KF, et al. CONSORT 2025 statement: updated guideline for reporting randomised trials. BMJ. 2025;389:e081123. https://doi.org/10.1136/bmj-2024-081123
3. Altman DG, Bland JM. Absence of evidence is not evidence of absence. BMJ. 1995;311(7003):485. https://doi.org/10.1136/bmj.311.7003.485
4. Lang TA, Altman DG. Basic statistical reporting for articles published in biomedical journals: the SAMPL guidelines. Int J Nurs Stud. 2015;52(1):5–9. https://doi.org/10.1016/j.ijnurstu.2014.09.006