For practical reasons, it was possible to recruit only 59 (52%) of the 114 eligible symptomatic workers (fig. 1), potentially introducing non response bias. However, it was considered that the potential for non response bias was minimal for the following reasons