How do you handle missing data in clinical trials?

1 Understand the mechanism

The first step is to understand the mechanism or reason behind the missing data. There are three main types of missing data: missing completely at random (MCAR), missing at random (MAR), and missing not at random (MNAR). MCAR means the missingness is unrelated to any variable in the study. MAR means the missingness is related to some observed variables, but not to the outcome of interest. MNAR means the missingness is related to the outcome or some unobserved variables. The type of missing data affects the choice of methods and assumptions for handling it.

2 Choose an appropriate method

The second step is to choose an appropriate method for handling the missing data. There are two broad categories of methods: complete-case analysis (CCA) and imputation. CCA means excluding the cases with any missing values from the analysis. This is simple and valid if the data are MCAR, but can reduce the sample size and power, and introduce bias if the data are MAR or MNAR. Imputation means replacing the missing values with plausible values based on some rules or models. This can preserve the sample size and reduce bias, but can also introduce variability and uncertainty, and require strong assumptions.

3 Evaluate the sensitivity

The third step is to evaluate the sensitivity of the results to the missing data and the chosen method. This means comparing the results obtained from different methods or scenarios, and assessing how robust they are to the assumptions and uncertainties involved. For example, you can use multiple imputation to generate different sets of imputed values, and compare the estimates and confidence intervals across them. You can also use different models or parameters for imputation, and test how they affect the results.

4 Report the details

The fourth step is to report the details of the missing data and the chosen method in the study report or publication. This means describing the amount, pattern, and mechanism of the missing data, the rationale and procedure for the chosen method, and the results of the sensitivity analysis. This can help the readers and reviewers to understand and evaluate the quality and credibility of the results, and to compare them with other studies.

5 Prevent the occurrence

The fifth step is to prevent the occurrence of missing data as much as possible in the design and conduct of the study. This means planning and implementing strategies to minimize the dropout, loss to follow-up, non-compliance, or measurement error that can cause missing data. For example, you can use incentives, reminders, feedback, or adaptive designs to encourage and facilitate the participation and retention of the subjects. You can also use quality control, standardization, or validation procedures to ensure the accuracy and completeness of the data collection and entry.

6 Consult an expert

The sixth step is to consult an expert in statistics or biostatistics if you have any doubts or questions about how to handle missing data in clinical trials. Missing data is a complex and nuanced topic that requires careful consideration and application of appropriate methods and assumptions. An expert can help you to choose the best method for your study, to perform the analysis correctly and efficiently, and to interpret and report the results accurately and transparently.

7 Here’s what else to consider

This is a space to share examples, stories, or insights that don’t fit into any of the previous sections. What else would you like to add?