Gravar-mail: Evaluating the incident-user design in the HIV population: incident-use vs. naive?