6. Studies with a major selection bias evident in treatment allocation to CB DOT or clinic DOT. This included:
a. studies in which patients were allocated to CB DOT because they were not compliant with clinic DOT or had refused to participate in clinic DOT;
b. non-randomised studies in which clinic DOT and CB DOT were delivered to patients living in different geographical locations and where these locations were not matched on population size, level of economic activity or accessibility of health services – for example, patients from rural locations or living far from a health centre allocated to one type of DOT, with patients from an urban area or living near to a health centre allocated to the other, and
c. non-randomised studies in which patient characteristics (e.g. socio-economic status) or severity of TB disease influenced the allocation of DOT type.
Poor study quality for other reasons than selection bias was not an exclusion criterion.