Descriptive statistics and Table 1
Table 1 describes the people in a study, overall and by group, so readers can judge who the results apply to and how alike the groups were.
Fit
When it matters, and when it does not
It matters when
- You report any study of patients, and readers need to know who was included.
- You compare groups, and readers need to see how alike they were at the start.
- You need a summary for each characteristic that fits it: a mean, a median or a percentage.
It matters less when
- You want to choose confounders by testing baseline differences. Choose them from a causal diagram instead. Read: Directed acyclic graphs
- The sample was drawn with survey weights or by clusters. Plain summaries describe the sample, not the population; use survey methods. Read: Complex survey analysis
- You want to know whether the exposure changed the outcome. Table 1 describes the sample you have, and it is not sized to compare groups; the planned analysis estimates the effect.
Before you start
What to check in your study
- The characteristics to describeThose that show who was studied, and those the analysis adjusts for. Leave out design terms such as calendar period, site or ward.
- What each variable isA measurement, a category or a score. Decide from what it measures, never from where it sits in a list.
- The shape of each measurementRoughly symmetric values suit the mean and standard deviation. Skewed values, such as length of stay, suit the median and interquartile range.
- The groups for the columnsassignmentArms in a trial, exposure groups in a cohort, or cases and controls. Add an overall column where it helps.
- How many values are missingGive the number missing for each variable, so every percentage has a clear denominator.
- Every row describes a participant.How to check: Design terms that the model conditions on stay in the model. A comorbidity index is a number, so summarize it as one.
- Each summary fits the shape of its variable.How to check: Plot the values first. When the mean and median sit far apart, the values are skewed.
- Every percentage has a known denominator.How to check: Show the number missing per variable. A variable missing far more often in one group is itself a finding.
- The table describes the sample the estimate applies to.How to check: If the analysis is weighted, also show the weighted sample, because that is the population the estimate describes.
Plain language
How it works
Each row is a characteristic, and each column is a group compared, with an overall column where it helps. Each variable gets the summary that fits it. Use the mean and standard deviation for symmetric measurements, and the median and interquartile range for skewed ones. Give categories as a number and percentage.
In a randomized trial, reporting guidance advises against significance tests of baseline differences. In an observational study, a P value reflects sample size as much as imbalance. A standardized difference describes the imbalance itself.
The coefficient of variation is the standard deviation divided by the mean. It compares spread between measurements on different scales, such as two laboratory tests. A standard deviation of 12 with a mean of 80 gives 0.15, or 15%. It makes sense only when zero means none, as for weight. It is meaningless for changes, log-scale values or temperature in Celsius.
| Characteristic | Group 1 (n = 250) | Group 2 (n = 250) |
|---|---|---|
| Age, mean (SD), y | 64.2 (12.1) | 63.0 (13.1) |
| Female, No. (%) | 142 (56.8) | 138 (55.2) |
| Comorbidity index, mean (SD) | 2.6 (1.7) | 2.5 (1.9) |
| Insurance type, No. (%) | 62 (27.2) | 56 (25.2) |
| Baseline severity score, mean (SD) | 5.4 (2.2) | 5.2 (2.4) |
Worked example
A worked example
- Setting
- A cohort of 1,200 adults with atrial fibrillation, 600 starting each of two anticoagulants.
- Question
- Who was in the study, and how alike were the two drug groups at the start?
- Approach
- One column per drug. Age is summarized as mean and standard deviation, creatinine as median and interquartile range, and prior stroke as a number and percentage.
- Result
- In the first drug group, mean age was 72 years with a standard deviation of 9. Median creatinine was 1.1 mg/dL, interquartile range 0.9 to 1.4. Prior stroke: 84 of 600 (14 in 100).
- What it means
- The table carries no P values. A standardized difference beside each row shows how far the groups differ, and the number missing is given for every variable.
In practice
Recent studies PubMed search · not vetted
Watch out
Common mistakes
The one to remember
A p-value column in the Table 1 of a randomized trial answers the wrong question: with random allocation any baseline difference is compatible with chance, and a significance test can say neither whether randomization worked nor whether the imbalance matters. Judge imbalance by its size and by how strongly the variable predicts the outcome. The other routine error is the standard error written where the standard deviation belongs: the standard error describes how precisely the mean is known, not how much patients vary, and it makes a varied sample look uniform.
- Choosing the summary by a variable's place in a list, or using one summary for every row.Instead: Type each variable from what it measures. Give skewed measurements as median and interquartile range.
- Summarizing a comorbidity index as a number and percentage.Instead: An index is a measurement. Give its mean and standard deviation, or its median and interquartile range if skewed.
- Reporting a coefficient of variation for a change score, a log-scale value or a temperature in Celsius.Instead: Use it only for measurements where zero means none, such as weight or a drug level. Otherwise report the standard deviation.
Inside Mallard
How Mallard uses this
Mallard's illustrative Table 1 shell types each variable from its name, never its position, and leaves out design terms such as ward or calendar time. Mallard's method list may include a Table 1 with standardized differences as an adjunct.
descriptive-summary-table-one).concept.descriptive_statistics, the id a plan cue records.Work it out
Related Learn pages
Related designs and methods
Evidence
Sources & further reading
- Who is in this study, anyway? Guidelines for a useful Table 1 (opens in a new tab) · J Clin Epidemiol · doi 10.1016/j.jclinepi.2019.06.011
- CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group randomised trials (opens in a new tab) · BMJ · doi 10.1136/bmj.c869