Effort estimation is key in any software development organization. Over the past two decades there has been considerable activity in the area of effort prediction with approaches being typified as being algorithmic in nature such as COCOMO. However, many change request databases contain effort data for finished change requests but they do not contain data that would allow using these methods. We propose a statistical regression approach that uses cumulative effort data to predict future effort in an evolving legacy system. This method provides up to six months of effort prediction for six months in advance. It is then expanded to perform re-estimation of the prediction model to consider change-points in order to get more accurate predictions of effort.