Table II.

Certainty of evidence

GRADE domain	Judgement according to outcomes of interest	Concerns about certainty domains
Methodological limitations of the studies	Functioning: Three studies had some concerns regarding ROB (22, 24, 28), and 2 had high risk of bias (7, 31), conservatively the trials were judged to have very serious methodological limitations. In the studies investigating lung function, 1 study was judged to have low ROB (30), 3 to have some concern (25, 27, 32), and 3 to have high risk of bias (26, 29, 33). Conservatively the trials were judged to have very serious methodological limitations Quality of life: Only 1 study reported this outcome (24). The study was judged to have some concerns regarding ROB. The trial was judged to have serious methodological limitations.	Very serious Serious
Indirectness	The patients, interventions and comparators in the studies all provided direct evidence to the clinical question at hand.	Not serious
Imprecision	Functioning: Five studies reported on function, with a total of 119 participants (very low) (7, 22, 24, 28, 31). Two studies reported small improvements (7, 22), and 3 with non-significant results likely because of enrolling a small number of participants, and presence of clinical heterogeneity (age, progression) (24, 28, 31). The evidence was judged to have serious imprecision. Seven studies reported in lung function parameters as outcome (25-27, 29, 30, 32, 33). One study reported improvements in lung function (32), while in 6 studies lung function remained unchanged or declined with non-significant changes (25-27, 29, 30, 33). The evidence was judged to have serious imprecision. Quality of life: The only study including this outcome reported non-significant improvement in favour of intervention (24). The evidence was judged to have serious or very serious imprecision.	Serious
Inconsistency	Functioning: The direction and magnitude of effect varied across the different trials. Overall the results showed either small or no change in functioning in favour exercise training. The evidence was judged to have serious inconsistency. In the studies who investigated change in lung function, the direction and magnitude were similar across all except 1 of the studies with no change in lung function (32). The evidence was judged to not serious inconsistency.	Serious/very serious Serious Not serious
Publication bias	Functioning: Publication bias was not strongly suspected because both negative and positive trials were published, and search for studies were comprehensive. Publication bias was not strongly suspected with respect to lung function, except in 2 studies without reported outcome data for the time-points and separate arms for the groups of intervention (26, 29). In addition to this, publication bias was not strongly suspected, because both negative and positive trials were published, and search for studies were comprehensive. Quality of life: Publication bias was not strongly suspected, because a non-significant improvement in favour intervention was reported (24).	Not suspicious Not suspicious

Grade evidence by ROB judgements was considered as; low to be no serious or serious, unclear to be equal to serious or very serious and high ROB to be very serious. If GRADE domains were judged as serious, they were downgraded by 1 point, and very serious, certainty of evidence was downgraded by 2 points.