Background: Loss to follow-up or drop-out is often a major concern when analysing data from primary studies, but also when extracting data from these for use in systematic reviews. The policy of the Cochrane Collaboration is that data used in a Cochrane review should, if possible, be based on the intention-to-treat principle. But different approaches exist, as to how the participants lost to follow-up should be included in the analyses. Objectives: To quantify the effects of different imputation methods for continuous outcomes on results from primary studies and results from meta-analysis compared to available-cases-analysis. Methods: In a simulation-study the effect of using different (simple) approaches to imputation of missing data from continuous outcomes compared to available-cases-analysis will be examined. The different approaches examined are: 1) keep mean and standard deviation but inflate N to ITT-population (quite a wrong approach, included for illustrative purposes); 2) use the same mean and standard deviation for non-completers as seen among completers in the control group; re-compute mean and (pooled) standard deviation; 3) assume a worst case mean for the non-completers; re-compute mean and (pooled) standard deviation; 4) assume a best case mean for the non-completers; re-compute mean and (pooled) standard deviation. The simulation study will be based on 10,000 replications of five studies with sample size of 100 (two studies), 250 (two studies) and 500. Each replication of five studies will also be the basis for a meta-analysis. The data will be analysed using regression models allowing for dependent observations, adjusting for drop-out rates (overall and difference between groups). Results: The effect of the different approaches on effect-sizes and p-values will be presented for primary studies and meta-analyses.