Table 3

Discrepancies in components of results reporting for outcomes reported in both preprints and journal publications (N=67 studies; 258 outcomes)

Type of discrepancyNumber (%) of studies with at least one discrepancy between the preprint and journal publication (n=67)Number (%) of outcomes across all studies that were discrepant between the preprint and journal publication (n=258)Descriptive examples*
Outcome measurement6 (9%)8 (3%)
  • Journal publication contains more detail on how outcome was measured compared with preprint (n=3)

  • Journal publication reports an additional or different measurement than the one used for the same outcome in the preprint (eg, preprint reports four adverse events, journal publication reports 12) (n=4)

Units of measurement3 (4%)3 (1%)
  • For example, journal publication reports events, total and percentage for mortality, preprint reports only percentage; median (IQR) reported in journal publication, mean (SD) in preprint

Timepoint assessment was made10 (15%)24 (9%)
  • Journal publication reports outcomes measured over a longer timepoint than preprint (n=13)

  • Journal publication reports additional interim time points compared with preprint (n=3)

Numerical values reported24 (36%)52 (20%)
  • Differences in number of events or measurement values reported (n=17)

  • Differences in numbers of participants or denominators (n=5)

  • More adverse events reported in journal publication than preprint (n=4)

Finding of statistical significance11 (16%)16 (6%)
  • Different p-value reported with no change in significance (n=3)

  • Different p-value reported with change in significance; significant result reported in journal publication (n=1)

  • In multivariate models, journal publication and preprint report different variables as being statistically significant (n=2)

Statistical tests performed17 (25%)31 (12%)
  • Journal publication contains additional statistical analysis compared with preprint (n=7)

  • Journal publication uses different statistical adjustments compared with preprint (n=7)

  • Journal publication and preprint use different statistical tests for same data (n=3)

Subgroup analyses conducted14 (21%)24 (9%)
  • Journal publication includes subgroup analysis not included in preprint (n=6)

  • Journal publication finds statistically significant interaction for subgroup, preprint does not (n=1)

Identifying the outcome as a primary or secondary outcome1 (1%)3 (1%)
  • For example, preprint identifies the primary endpoint as safety; journal publication adds the secondary endpoint of exploration of efficacy

  • *Ns do not add to number of reported discrepancies as some studies could have more than one discrepancy and not all discrepancies have been included as examples.