Code compression techniques might be useful to meet code size constraints in embedded systems. In the average case, the impact of code compression on the performance is double-edged: on one side, the number of accesses to memory hierarchy is reduced because several instructions are coded in a single word, and this is likely to reduce the execution time; on the other side, the decompression penalty increases the processing time of compressed instructions. Nevertheless, experimental results show that the execution time might be lowered by code compression. In this paper, our goal is to analyze the impact of code compression on the estimated Worst-Case Execution Time of critical tasks that must meet at the same time code size constraints and timing deadlines. Changes in the access patterns to the instruction cache are indeed likely to alter the accuracy of the cache analysis within the process of determining the WCET. Experimental results show that, besides reducing the code size, our code compression scheme also improves the WCET estimates in most of the cases.