Limitations of LOCF method for handling missing data

MISSING data are an inevitable aspect of clinical trials and present a significant challenge for researchers.

- K. Sadasivan Pillai & Durairaj Sathishkumar

The EMA defines missing data as “data that would be meaningful for the analysis of a given estimand but were not collected.” Inadequate handling of missing information can lead to biased estimates, reduced precision, and compromised validity of treatment effects, underscoring the importance of robust statistical methods and careful trial design.

The choices made for managing incomplete datasets directly impact the validity, reliability, and interpretability of clinical trial results. The Last Observation Carried Forward (LOCF) method, once widely used, has come under intense scrutiny by statisticians and regulatory bodies for its numerous shortcomings and the risks it poses to unbiased estimation in longitudinal studies. This article provides an analysis of the disadvantages, statistical implications, and regulatory context regarding LOCF in clinical trials.

Disadvantages of LOCF

Conceptual flaw: Imputing values that may not exist

LOCF imputes the last recorded value for all subsequent missing time points. This approach can create scenarios where the imputed value does not reflect the biological reality of patient progression. For instance, in psychiatric trials, a participant may withdraw due to intolerable side effects or a worsening condition.

Under LOCF, the patient's final observed score is carried forward as if their condition remained unchanged thereafter, disregarding any true deterioration or improvement that occurred post-discontinuation. This assumption can distort the estimated treatment effect, obscure genuine patterns of response, and ultimately lead to biased or misleading conclusions regarding the efficacy and safety of an intervention.

Questionable conservatism and direction of bias