Table 3.

Risk of bias

	Random sequence generation (selection bias)	Allocation concealment (selection bias)	Blinding (performance bias and detection bias) for all outcomes— patients?	Blinding (performance bias and detection bias) for all outcomes— providers?	Blinding (performance bias and detection bias) for all outcomes—outcome assessors?	Incomplete outcome data (attrition bias) for all outcomes— dropouts?	Intention-to-treat analysis (attrition bias) for all outcomes	Selective reporting (reporting bias)	Overall impression
Annaheim et al13	Low risk. ‘Subjects were randomised’.	Unclear risk. Not addressed.	Low risk. ‘double blind’.	Low risk. ‘researchers involved in the assessment of blood volume compartments and exercise testing were blinded’.	Low risk.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.
Birkeland et al18	Low risk. ‘randomly assigned’.	Unclear risk. Not addressed.	Low risk. ‘double blind’.	Low risk. ‘double blind’.	Low risk. ‘blinded to technicians and investigators engaged in blood sampling, sample analyses, and exercise testing’.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.
Caillaud et al24	Low risk. ‘Randomly’.	Unclear risk. Not addressed.	Unclear risk. Not addressed.	Unclear risk. Not addressed.	Unclear risk. Not addressed.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	High risk.
Connes et al20	Low risk. ‘randomly assigned’.	Unclear risk. Not addressed.	Low risk. ‘double blind’.	Low risk. ‘double blind’.	Unclear risk. Not addressed.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.
Heuberger et al6	Low risk. ‘randomly assigned’.	Low risk. ‘randomization code’ was generated by a statistician who was not involved in the execution of the study.	Low risk. ‘double blind’.	Low risk. ‘double blind’.	Unclear risk. Not addressed.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.
Ninot et al21	Low risk. ‘randomly assigned’.	Unclear risk. Not addressed.	Low risk. ‘double blind’.	Low risk. ‘double blind’.	Low risk. ‘blinded to the technicians and investigators engaged in blood sampling, sample analysis and exercise testing’.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.
Rasmussen et al22	Low risk. ‘randomly assigned’.	Unclear risk. Not addressed.	Low risk. ‘double blind’.	Low risk. ‘double blind’.	Unclear risk. Not addressed.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.
Sieljacks et al15	Low risk. ‘randomly assigned’.	Unclear risk. Not addressed.	Low risk. ‘single blind’.	Low risk. Not blinded but not likely to affect outcomes.	Unclear risk. Not addressed.	Low risk. 2 excluded, addressed and justified.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.
Thomsen et al19	Unclear risk. Not addressed.	Unclear risk. Not addressed.	High risk. ‘single blind’ for placebo group only, rHuEPO group was ‘aware’.	High risk. Providers were not blinded.	Unclear risk. Not addressed.	Low risk. No dropouts.	Low risk.	Low risk. All prespecified outcomes reported.	High risk.
Wilkerson et al23	Low risk. ‘Randomly assigned’.	Unclear risk. Not addressed.	Low risk. ‘Subjects … were blind to the group assignment’.	Unclear risk. Not addressed.	Low risk. ‘Investigators involved in the administration of the exercise tests and the subsequent data analysis were blind to the group assignment’.	Low risk. ‘One subject in the control group sustained an injury unrelated to the study and was unable to complete the post-intervention tests’.	Low risk.	Low risk. All prespecified outcomes reported.	Low risk.

rHuEPO, recombinant human erythropoietin.