Table 2.

Risks of bias within trials in sarcoidosis patients reporting fatigue as an outcome measure.

Author (year)	Lower et al. (2007)29	Erckens et al. (2011)30	Heij et al. (2012)13	van Rijswijk et al. (2013)31	Lower et al. (2013)32	Strookappe et al. (2015)33	Marcellis et al. (2015)34	Strookappe et al. (2015)35
Sequence generation	Low risk: Random sequence computer-generated	High risk: No randomization (not RCT)	Low risk: Computer generated randomization code	High risk: No randomization (not RCT)	Unclear: No statement on randomization procedure	High risk: No randomization (not RCT)	High risk: No randomization (not RCT)	High risk: No randomization (not RCT)
Allocation concealment	Low risk: Pharmacy-controlled allocation	High risk: No concealment, all patients receive drug	Low risk: Pharmacy-controlled allocation	High risk: No concealment, all patients received drug	Unclear: No statement on allocation procedure	High risk: No concealment, all patients on programme	High risk: No concealment, all patients on programme	High risk: No concealment, all patients on programme
Blinding of participants, personnel and outcome assessors	Low risk: Double-blind with low risk of breaking blinding	High risk: No blinding	Low risk: Only allocating pharmacist unblinded	High risk: No blinding	Low risk: Double-blind with low risk of breaking blinding	High risk: No blinding	High risk: No blinding	High risk: No blinding, including assessors of physical function pre- and post.
Incomplete outcome data	Unclear: No description of missing data points or handling of missing data	Unclear: No description of missing data points or handling of missing data	Low risk: Missing data compensated for by taking forward last value	Unclear: No description of missing data points or handling of missing data	Unclear: No description of missing data points or handling of missing data	Unclear: No description of missing data points or handling of missing data	Unclear: No description of missing data points or handling of missing data	Unclear: No description of missing data points or handling of missing data
Selective outcome reporting	Low: All outcomes reported	Unclear: No protocol available	Unclear: No protocol available	Unclear: No protocol available	Low: All outcomes reported	Unclear: No protocol available	Unclear: No protocol available	Unclear: No protocol available
Other sources of bias	No other clear causes of bias identified; crossover ensures groups balanced, patients are own controls. Small sample.	Study design (case series) limits conclusions – no comparator group to eliminate placebo effect.	Baseline imbalance in FAS and health status score (SF36) between groups – significantly lower fatigue scores in placebo arm.	Retrospective review – data collected pre- and post-intervention but high risk of bias from retrospective nature	No other clear causes of bias identified; crossover ensures groups balanced, patients are own controls. Small sample.	Not an RCT, also participants enrolling on programme would self-select as motivated people? generalizable	Not an RCT, also participants enrolling on programme would self-select as motivated people? generalizable	Patients choosing the intervention would be a self-selecting cohort; controls not randomized but refused intervention
Overall risk of bias	Low – Well designed crossover trial, though only small sample	High – Study design (case series) means no blinding, randomization or comparator.	Low – Well designed RCT but not powered to look at change in fatigue.	High – Design (retrospective case series) has no blinding, randomization or comparator.	Unclear – Issues with description of randomisation allocation and concealment mean study at risk of bias	High – Study design (case series) means no blinding, randomisation or comparator.	High – Study design (case series) means no blinding, randomisation or comparator.	High – Self-selecting intervention group, high risk of bias given control group declined intervention

RCT: randomized controlled trial.