The main physical features of the cascade damage is described and their impacts on the surviving defect fraction and the damage accumulation are considered. It is shown that the essential features of the cascade damage are fundamentally different from the basic assumptions of the conventional mean-field approach. The production bias approach, on the other hand, is based on the physical features of the cascade damage and is therefore considered to be more appropriate for describing the damage accumulation under cascade damage conditions. However, production bias can not produce high a swelling rate when the primary cluster density is high. Therefore, a sustained high swelling rate driven by production bias must involve the annihilation of primary clusters at sinks. A number of experimental observations which are unexplainable in terms of the conventional dislocation bias for monointerstitials is considered. It is found that the production bias and cluster annihilation are necessary to explain these observations, with, in many cases, the explicit consideration of the annihilation of the primary interstitial clusters.