A critical point in cognitive radio spectrum sensing is the ability to detect presence and absence of primary users as fast as possible at very low SNR. In this paper, sequential power detection by cumulative sum and recursive generalized likelihood ratio test is used to detect free spectral slots of opportunity. The benefits of these change detection algorithms are the adaptive sensing window, the low processing burden and the optimality in sense of maximum likelihood. A spectrum utilization efficiency metric is proposed that put a cost on late detections as well as on false alarms that might give rise to harmful interference into the primary system. The efficiency metric is then simulated versus the size of the free slot of opportunity and for different SNR. The detectors presented are found useful for cognitive radio.