Table 4 Source, nature and solution to encountered loss of information

Baseline data not adjusted for

Collected data not used, bias [9, 10]

Adjust for baseline in a regression model

Dichotomisation of the main continuous outcome

Loss of power, information [15, 16]

Analyse the continuous outcome as primary analysis, dichotomised outcome presented as secondary

Aggregated longitudinal data

Loss of the longitudinal nature of the data, loss of the treatment dynamic over time

Use a linear mixed model possibly with time as a covariate.

Cross sectional analysis at each time-point

Multiple testing requiring a correction, therefore loss of power

Use a linear mixed model with categorised time with time-group interactions.

No effect size provided

No information on the magnitude of the effect in term of outcome measure

Use a regression model (e.g. mixed model) and present the regression parameter for group effect with confidence interval