A register-based study of ALS outcome: suggestions for the design of clinical trials