Do changes in effort-reward imbalance at work contribute to an explanation of the social gradient in angina?