Table 2 ROB-MEN algorithm to calculate overall risk of bias for the network estimates

Low risk

     There is no substantial contribution from evidence with suspected bias favouring one of the two treatments,

OR

     There is substantial contribution from evidence at suspected bias, but it is split more or less equally between evidence with bias favouring one of the treatments and evidence with bias favouring the other treatment

AND

     There is no evidence of small-study effects favouring one of the two treatments,

OR

     [For indirect estimates only] There is no suspected bias favouring one of the two treatments from the assessment of indirect evidence.

Some concerns

All other combinations

High risk

     There is substantial contribution from evidence with suspected bias favouring one of the two treatments

AND

     There is evidence of small-study effects favouring the same treatment,

OR

     [For indirect estimates only] There is suspected bias favouring that treatment X from the assessment of indirect evidence.