This paper deals with the construction and optimisation of accurate condition based maintenance policies for cumulative deteriorating systems. In this context, the system condition behaviour can be influenced by different environmental factors which contribute to an increase or decrease the degradation rate. The observed condition can deviate from the expected condition if the degradation model does not embrace these environmental factors. Moreover, if more information is available on the environment variations, the maintenance decision framework should take advantage of this new information and update the decision. The question is how shall we model the decision framework for this? We propose to model the effect of the random operating environment on the system behaviour with a randomization of the Gamma process-degradation parameters. A new decision rule is introduced to update the maintenance decision in case of a significant deviation from the expected condition. The performance of the introduction of this new decision rule is discussed according different contexts: the level of knowledge on the real stress data and the restriction of a potential updating in the policy. A mathematical framework and optimization procedures are presented and numerical experiments are conducted to highlight the benefits of the different models.