Summary: The paper deals with the problem of missing covariate values in a regression model. An attractive approach to avoid this problem is to impute the missing covariate values rather than delete cases with missing covariates. The paper is devoted to a comparison of different imputation techniques or methods. The type of missing data pattern in the set of independent variables is the monotone data pattern. We assume that the missing data are missing at random (MAR). Imputation of missing values was achieved using a multivariate normal model. The main objective of this paper is to study how some imputation approaches compare when the missing data pattern is monotone. The techniques that were considered include the Last Observation Carried Forward (LOCF), Propensity Score (PS), Markov Chain Monte Carlo (MCMC) and Regression. In order to compare the performance of the proposed methods, we used originally complete data sets (no data are missing), and then we intentionally created missing values to achieve the intended goal. Missingness was imposed on covariate variables. The performance of these four methods is compared on three criteria: bias, efficiency and coverage. Data from a Diabetes study is used to illustrate the considered approaches.