The baseline characteristics table, the Table 1 of almost every clinical paper, is the first thing a reviewer reads. It describes who was in your study, usually split by treatment or comparison group. Get it right and the rest of the paper is easier to trust. Get it wrong and a reviewer starts doubting the analysis before reaching it. The table is simple to build and easy to get wrong, and the errors are visible from the table alone.
1What belongs in Table 1
One row per baseline variable, one column per group, and an overall column for the whole sample. Put demographics first, such as age and sex, then the clinical variables that matter for your question, then anything a reader needs to judge whether the groups were comparable to begin with. Outcomes do not belong here. Table 1 describes the sample you started with, not the result you found.
2Choose the grouping variable
Pick the variable that splits the study into the groups you compare, usually the treatment arm or the exposure. The overall column summarises everyone, and the group columns let a reader see at a glance whether the arms started out similar. If your study has no groups, an overall column on its own is perfectly correct.
3Summarise each variable the right way
Continuous variables are reported as mean with standard deviation when they are roughly symmetric, and as median with interquartile range when they are skewed, which is common for biomarkers, length of stay and many laboratory values. Categorical variables are reported as counts with percentages. The single most common Table 1 error is forcing a mean onto a skewed variable. A reviewer who knows the variable spots it immediately, so let the shape of the data decide.
4Missing data and p-values
State missing data plainly, with a Missing column or a footnote, rather than quietly dropping it. On p-values, follow CONSORT: for a randomised trial, do not test baseline differences, because any difference between randomised groups is by definition due to chance. For an observational study a comparison can be informative, but a standardised difference is often more useful than a p-value.
5The mistakes reviewers catch
A mean and standard deviation on a skewed variable. Percentages that do not add up because the denominator changes with missing data. P-values in a randomised trial. Baseline rows that are really outcomes. Inconsistent decimal places from one row to the next. Each of these is visible from the table alone, which is exactly why reviewers find them so fast.
6Build your Table 1 in seconds, free
You can build a publication-ready Table 1 from your own dataset with our free Table 1 generator. Paste or upload your Excel or CSV file, choose the grouping variable, and it picks the right summary for every variable and exports to Word, PDF or reproducible R code. Everything runs in your browser, so your data is never uploaded. The judgement calls, the ones reviewers notice, are where a statistician earns their fee, and that is the part we are glad to help with.