We study the efficiency of two estimators of quantile difference with left truncated and right censored data. Based on deficiency, the performances of the estimators are compared. The theoretical results and numerical studies show that the smoothed estimator is more efficient.