Searching for medical information online is becoming more common and the number of consumer questions increases every day. In order to support consumer search, we need to develop Question Answering tools. However, Consumer health questions may contain subjective and borderline information making them unuseful to find relevant answers or to train efficient QA tools. The objective of Consumer health question summarization is to simplify long and complex questions by extracting only relevant information without changing their intent. The problem is that the size of datasets used for training medical summarization models contain fewer questions compared to the datasets used for open domain question summarization. In fact, medical consumer health question summarization dataset has to be annotated with experts which is a time-consuming process. In this paper, we propose to use text augmentation in order to extend questions with additional information obtained with generative models (Pegasus and T5). We fine-tune BART and T5 using en expanded version of MeqSum dataset. Experiments show that our approach, which combines transformers with data augmentation improves performance over summarization baseline and state of the art methods.