Risk of bias in included studies
We included 19 trials that used appropriate methods of randomization
(e.g. computer-generated table of random numbers, draw
of lots, flip of coins), three trials that used alternate group assignment
as allocation method, and eight trials that reported using
randomization but failed to state the randomization method.
Sixteen trials used allocation concealment whereas six trials did
not. For the remainder of the trials, use of allocation concealment
was not mentioned.
Only four trials reported blinding of the outcome assessors for
objective measures. For two trials, the use of blinding was unclear.
The other trials did not use blinding. However, it is important
to point out that blinding of outcome assessors is not possible in
the case of subjective measurement tools (e.g. STAI (Spielberger
1983)) unless the participants are blinded to the intervention.
Blinding of the participants is often not feasible in music therapy
and music medicine studies. This may introduce possible bias.
The dropout rate was small for most trials, namely between 0 and
16.6%. Four trials reported dropout rates ofmore than 20%(Beck
1989; Burns 2008; Hanser 2006; Montserrat Gimeno 2008).
For eight trials, it was unclear whether there were any participant
withdrawals. Most trials reported reasons for dropout. Detailed
information on dropout rate and reasons is included in the
Characteristics of included studies table.
As a result, only one trial (Nguyen 2010) was at low risk of bias
for objective outcomes as it satisfied all criteria used to assess risk
of bias. Twenty-seven trials were at high risk of bias. One trial
(Binns-Turner 2008)was atmoderate risk of bias.Themain reason
for receiving a high risk of bias rating was the lack of blinding. As
pointed out above, blinding is often impossible in music therapy
and music medicine studies that use subjective outcomes, unless
in studies where the music intervention is compared to another
treatment intervention (e.g. progressive muscle relaxation). This
is especially true for music therapy studies that use active music
making. Therefore, it appears impossible for these types of studies
to receive a low or even moderate risk of bias even if all other
risk factors (e.g. randomization, allocation concealment, etc.) have
been adequately addressed.
It is worth noting that the Chinese trials were particularly problematic
in terms of providing sufficient information regarding risk
of bias. It is unclear, however, if this was due to incomplete translations
or lack of detail in the original trial reports.
Risk of bias is detailed for each trial in the risk of bias tables
included with the Characteristics of included studies table and the
Risk ofBias Summary (Figure 1). In addition, an overall assessment
of risk of bias can be viewed in Figure 2.