The relationship of the long-term temperature and precipitation trends, dominated by climate change, to the skill of seasonal to interannual forecasts has been noted recently (Livezey and Timofeyeva, BAMS 2008). However, the the extent to which the skill of seasonal and interannual forecasts is due to trends or due to independent climate modes has not been rigidly tested. Missing a priori information on the expected trends, calculating a predictable climate signal from analysis of past trends is non-trivial. We attempt to separate the trends and interannual variability due to other major modes to examine the information content of the trend component of long-term variability. The component of trends attributable to major modes of climate variability versus the residual trend are compared. We determine the maximum possible skill due to perfect trend hindcasts, the skill attained from trends predicted from prior information alone, and finally, the predictability of past trends. The skill of trend forecasts is compared to the skill of seasonal forecasts which combine multiple sources of information. Lastly, we examine the stationarity of the recent climate state and estimate the sampling error in estimates of this state.