Table 3. Essential components to report in randomized clinical trials with respect to the analysis.

Statement about intention to treat (ITT) for trial participants with available outcome data

Claim of ITT: if individuals were analyzed in the groups to which they were randomized with details about any post-randomization exclusions *

No claim of ITT, e.g. if analysis exclusively focused on individuals who complied with the study protocol (‘per protocol’ or ‘as treated’ analysis)

Statement about the handling of missing outcome data (MOD) †

A) No MOD (complete follow-up)

B) Individuals with MOD were not considered in the analysis (complete/available case analysis)

C) Imputation with explicit description. Options include individuals with MOD were considered in the analysis:

i) assuming all experienced the outcome of interest, ‡

ii) assuming none experienced the outcome of interest, ‡

iii) assuming a worst case scenario (i.e. individuals with MOD in the experimental group experienced the outcome of interest and those in the control group did not), ‡

iv) assuming a best case scenario (i.e. individuals with MOD in the experimental group did not experienced the outcome of interest and those in the control group did), ‡

v) last observation carried forward,

vi) censored at the time lost to follow-up in a survival analysis,

vii) multiple imputation,

viii) any other imputation/modelling that needs to be specified.

D) Two or more of the options in B & C (sensitivity analysis)

^*It may be appropriate to exclude randomized patients in order to achieve efficiencies while preserving prognostic balance between groups if two conditions are met [23]: (1) allocation to treatment or control could not possibly influence whether a particular randomized individual met criteria for post-randomization exclusion, (2) the decision about post-randomization is made without possible bias (commonly achieved through review blinded to allocation).

^†There are various ways of handling missing data; we provide illustrative examples for reporting purposes.

^‡For dichotomous outcome data.