LOCF

LOCF, or last observation carried forward, is a statistical method used to handle missing data in longitudinal studies. When a data point is missing for a participant at a given time, the value from their most recent available observation is used in place of the missing value for all subsequent analyses.

Procedure: For each subject with missing data, replace each missing value at time t with the last observed value prior to t. If a participant has no prior observation, data are typically excluded or treated as missing. This approach creates a continuous data set that can be analyzed with standard techniques.

Advantages: LOCF is simple to implement, preserves the sample size, and yields an easy-to-interpret, continuous trajectory for each participant. It can facilitate intention-to-treat analyses by avoiding case-wise deletion.

Limitations: LOCF assumes no change after the last observation, which may be unrealistic and can bias results toward stability. It tends to underestimate variance and can distort time trends, especially when outcomes naturally evolve or when dropout is related to unobserved outcomes. It is also sensitive to the timing of dropout and to nonrandom missingness.

Alternatives: Modern analyses often use approaches that reflect uncertainty about missing data, such as multiple imputation (MI), mixed-model repeated measures (MMRM), or maximum likelihood methods. Complete-case analyses or available-case analyses may be used but can reduce power and introduce bias if data are not missing completely at random.

Regulatory and practical considerations: Many statisticians discourage LOCF, particularly for long gaps or informative missingness. Pre-specifying the handling of missing data in a statistical analysis plan and conducting sensitivity analyses with alternative methods are common practices in contemporary trials.