Medical observational studies · Table 1 guide
How should Table 1 be prepared for a medical paper?
Have confirmed study groups, a data dictionary and aggregate results? Use this guide to plan Table 1, check counts, units, denominators and missing data, and draft footnotes with an adaptable AI request. Formal values must come from actual analysis.
First establish which participants, time point and groups the table describes. Then fix variable definitions, units, summaries, percentage denominators and missing counts. Table 1 helps readers understand the study sample; a column of P values cannot establish group comparability or replace an analysis plan.
This guide mainly concerns medical observational designs, including cohort, case–control and cross-sectional studies, and follows STROBE item 14 and its explanation document. Cross-sectional studies may have no longitudinal 'baseline'; the title should accurately describe participant characteristics. Randomized trials require separate checks against the applicable trial reporting guideline and journal requirements.
The table below illustrates reporting structure. It contains no real patient data, executed statistical results or claims about effects. A formal table must be generated from the defined study sample and controlled analysis outputs, and checked by the researchers.
A concrete task with RigorMed
Start with a Table 1 specification you can hand to a statistician
These are suggested task inputs and outputs to request, not a new execution, a customer outcome or a guarantee of completion.
Materials to provide: Confirmed population and groups, a data dictionary, aggregate results, units and missing counts.
Requested outputs / delivery goal: A Table 1 specification and footnote draft, plus a variable-by-variable denominator and missingness check.
An example request
Using the confirmed sample and aggregate outputs, draft Table 1 column headings, variable display names, summary statistics, denominators, units and footnotes. Identify missing outputs; do not generate uncomputed means, P values or counts.
What still needs review
Formal values must come from actual controlled analysis. P values must not automatically select confounders.
What must a self-contained Table 1 explain?
Define the population and total number in each column first. The text below describes what to report; it is not a set of results to paste into a manuscript.
| Row or footnote | What to report | Evidence to check |
|---|---|---|
| Table title and group columns | Study sample, time point, group definitions and numbers in each group | Study design, eligibility criteria, grouping variable and participant flow |
| Continuous variables | Units, available counts, mean (standard deviation) or median (P25, P75) | Variable definitions, distributions and prespecified summary methods |
| Categorical variables | Category labels, counts and percentages; specify the percentage denominator | Variable coding, eligible population and valid response counts |
| Missing information | Missing counts for each variable of interest, by group when needed | Original codes, skip logic and missing-data rules |
| Table footnotes | Abbreviations, units, summary notation, denominators and applicable statistical explanations | Analysis plan and the target journal's current author instructions |
The total sample size is not automatically the denominator for every row. For example, when a variable is collected only in an eligible subgroup, explain that eligibility first. Refused, unknown, not collected and not applicable responses must not be collapsed into a single 'missing' category without checking their meaning.
Why can the same 20 people be 20.0% or 25.0%?
This is a synthetic teaching example of one ordinary, unweighted group, not research results, customer outcomes or a product run. There are 100 participants; the variable applies to all, with 80 valid records and 20 missing. Among valid records, 20 smoke and 60 do not. Fix the denominator before reporting percentages.
| Proposed row | Reporting with valid records as denominator | Recalculable basis |
|---|---|---|
| Smoking | 20 (25.0%) | 20 / 80 × 100 = 25.0% |
| Not smoking | 60 (75.0%) | 60 / 80 × 100 = 75.0% |
| Missing for this variable | 20; 20.0% of the whole group | 20 / 100 × 100 = 20.0% |
Check 20 + 60 = 80 and 80 + 20 = 100 first. The first identity requires mutually exclusive, exhaustive categories in this example; the second requires applicability to all 100 participants. Resolve inapplicable values, skip logic and duplicates against source definitions before applying these checks. In a multi-group table, check each group, not just the total column.
If the prespecified denominator is the full group of 100, the same smoking count becomes 20 (20.0%), not smoking 60 (60.0%) and missing 20 (20.0%); change the footnote accordingly. The 20.0% and 25.0% answer different questions: neither is automatically the only correct choice. Do not silently mix denominators within a category block.
Do not add the 20 missing records to the 60 not-smoking records or delete them to make percentages add to 100%. This example concerns reporting, not missingness mechanisms, imputation, group comparability or changing the formal model sample. Weighted survey percentages, matching and repeated measures need design-specific treatment; these raw-count proportions are not a substitute.
The reviewable output should include variable/group, applicable total, valid count, category counts, missing count, percentage denominator, calculation, footnote and source-summary location. Mark unavailable source counts as missing inputs. Do not reverse-engineer counts from rounded percentages. Researchers must still confirm definitions and the analysis plan.
Add to the suggested request: “Using only aggregate outputs I may lawfully supply, prepare this Table 1 denominator audit and footnotes. Retain original counts and flag non-reconciling counts, mixed denominators and source gaps. Calculate only from existing aggregates; do not invent patient records, missing values, P values or conclusions.” STROBE item 14 supports reporting characteristics and variable-specific missingness. This table and its percentages are independent teaching material, not data from that source.
Five checks from the analysis sample to the table
Fix the sample and groups
Record inclusion and exclusion criteria, the observation time, grouping variable and group counts. State whether the table describes all included participants, complete cases for a particular analysis or a prespecified subgroup. Do not delete additional records merely to make the table neater.
Verify each variable definition
Check the data dictionary and source documentation before deciding whether a variable is continuous, categorical or ordinal. Give units for age and measured quantities. A small set of ordered categories, such as disease stages, should usually be reported as category counts and percentages; numeric coding alone does not justify averaging them as a continuous quantity.
Choose the appropriate descriptive summary
The STROBE explanation document recommends means and standard deviations by study group for continuous variables, or medians and percentile ranges for asymmetric distributions. Reporting the quartile endpoints explicitly as P25 and P75 avoids confusing the interval between quartiles with the difference between its endpoints. Do not substitute a standard error or confidence interval for a standard deviation when describing sample variability. If you also report a population estimate and its uncertainty, identify it as a different statistic.
Make missingness and denominators explicit
STROBE item 14(b) calls for the number missing for each variable of interest. Report categorical counts and percentages and explain whether the denominator is the entire group or the non-missing valid observations; give available counts directly when needed. Use consistent decimal precision and rounding for percentages. Do not alter original counts to produce tidy percentages.
Trace outputs through the table and manuscript
Check group counts, category counts, valid denominators, missing counts, units and summary notation against participant flow, methods and manuscript text. ICMJE requires table titles to be understandable on their own and nonstandard abbreviations and explanations to be provided in footnotes. Journal-specific requirements still need a separate check.
Table 1 P values are not a confounder selection tool
The STROBE explanation document explicitly recommends avoiding significance tests in descriptive tables and notes that P values are not an appropriate basis for choosing confounders to adjust for. A P value above 0.05 does not establish that groups are identical, comparable or free of confounding. A smaller P value also cannot replace knowledge of the study background and the roles of variables.
If the research question genuinely requires between-group comparisons, define their purpose, estimand, method and interpretation limits in the analysis plan. Do not inspect Table 1 significance first and decide which variables to adjust for afterward. Comparing case and control characteristics in a case–control study does not, by itself, complete a confounding assessment.
Seven checks before exporting a formal table
Each item should be traceable to a definition, plan or output:
- Does the title accurately identify the study population, time point and groups?
- Do group totals agree with the frozen study sample and participant flow?
- Have variable types, category labels, codes and units been checked against the source dictionary?
- Does the notation for continuous-variable summaries match the actual statistics?
- Are available counts, missing counts and percentage denominators clear for each variable of interest?
- Have descriptive-table P values been avoided as the basis for confounder adjustment?
- Are footnotes, abbreviations, decimal precision and manuscript text consistent, and have journal requirements been checked?
Complex sampling, matching and repeated-measures designs require additional consideration of weights or correlation structures. The ordinary sample summaries on this page are not an analysis method applicable to every design. For example, distinguish NHANES raw counts from weighted population percentages; see the related weights guide for the specific method.
Frequently asked questions
Must Table 1 in a medical paper include P values?
Not necessarily. For descriptive characteristic tables in observational studies, the STROBE explanation document recommends avoiding significance tests. Establish the table's purpose and applicable design, then check journal requirements. Do not treat a column of P values as proof that two groups are comparable.
Why do categorical percentages not match the total sample size?
Check whether the variable applies only to a subgroup and whether its denominator is the whole group or the non-missing valid observations. Report category counts, missing counts and the denominator definition, and check special codes. Do not remove rows or change numbers to eliminate an apparent discrepancy.
Can the number after a mean be a standard error without saying so?
The notation must be unambiguous. A standard deviation describes sample variability; a standard error describes estimation precision. They have different meanings. Label the actual statistic explicitly. The mean (standard deviation) discussed here cannot be replaced by an unexplained mean (standard error).
Does a complete Table 1 demonstrate that an analysis plan is sound?
No. It helps readers understand the study sample but does not validate data-use authorization, study design, bias control, confounder adjustment or causal interpretation. Formal analysis and submission still require the relevant methodology, sources and accountable researchers to be checked.
Verified sources
The official sources below were checked on 2026-09-30 / RigorMed: 2026-10-13 / STROBE14: 2026-10-23. Access conditions and update dates are rechecked before publication.
- Official STROBE checklists for observational study designs
This page focuses on descriptive data in item 14. A reporting checklist is not a research-quality score.
- STROBE Explanation and Elaboration: items 14(a) and 14(b)
The original document was checked for continuous-variable summaries, variable-specific missing counts, and the limits of P values in descriptive tables and confounder selection.
- ICMJE: Preparing a Manuscript, Results and Tables
Requirements checked include absolute counts, percentages, table titles, footnotes, abbreviations and explanations of statistics.
Start with a specific task
Start with a Table 1 specification you can hand to a statistician
Bring legally usable materials and this page's example request to the workspace. Check the pre-execution material and credit notices before confirming, then review the resulting outputs.