Potential influence of observer variation in thromboprophylactic trials.