1. Quality assessment of included studies.

Study	Allocation	Concealment	Study groups	Eligibility	Blinding	Attrition	Delphi list	Jadad score	Comment
Kasseroller 1998	random allocation; method of sequence generation not reported	unclear	no information on comparability of study groups regarding risk factors at baseline	eligibility criteria specified; number of eligible patients not stated	patient: yes; care provider: yes; outcome assessor: yes ‐ unblinding of treatment status after end of intervention	3 dropouts	1‐0‐0‐1‐1‐1‐1‐0‐0	1‐1‐1	Selection bias possible: allocation concealment unclear, no information on distribution of risk factors at baseline. Observer/detection bias possible: no control procedure reported whether blinding was successful, details on outcome assessment not reported Performance bias possible: concomitant treatment not sufficiently described
Mücke 2007	random allocation; method of sequence generation not reported	serially numbered, sealed, opaque envelopes	no information on comparability of study groups regarding risk factors at baseline	eligibility criteria specified; number of eligible patients not stated	patient: no; care provider: no; outcome assessor: no	dropouts / withdrawals not reported in most recent publications	‐‐	‐‐	Preliminary results of an ongoing study (Delphi List and Jadad Score will be given when final report is available) Selection bias unlikely, but possible: allocation concealment adequate, but no information on distribution of risk factors for outcome. Observer/detection bias possible: no blinding. Performance bias possible: concomitant treatment unclear
Zimmermann 2005	random allocation; method of sequence generation not reported	unclear	no information on comparability of study groups regarding risk factors at baseline	eligibility criteria specified; number of eligible patients not stated	patient: unclear; care provider: unclear; outcome assessor: unclear	no dropouts or withdrawals	1‐0‐0‐1‐0‐0‐0‐0‐0	1‐1‐1	Selection bias possible: allocation concealment unclear, information on distribution of risk factors at baseline. Observer/detection bias possible: study was reported as double‐blinded ‐ informed consent form, however, contained information on intervention status, reliability and validity of outcome assessment unclear. Performance bias possible: concomitant treatment not sufficiently described. Randomisation and blinding unclear: informed consent form for participants contains the sentence "you have been allocated to the group receiving selenium supplements", while study is reported as double‐blinded