Table 4. Risk of bias.

Included studies	Prospective design	Concurrent control	Assignment criteria reported	Control for confounding factors*	No other risk of bias factors†	Risk of bias‡
Non-randomized controlled studies
Carli 1999/2004 [36], [107]	YES	no	no	no	no	high
Hosoi 2007 [37]	no§	YES	no	no	no	high
Klingebiel 2008 [38]	YES	YES	no	no||	no	high

– information not reported in the publication or not applicable.

*Control for confounding factors; no: no adjusted analysis.

^†No other risk of bias factors; no: selection of patients unclear; except Gluckman 1979: no: 5 patients with failed first-line IST followed by second-line HSCT were analyzed in both treatment groups.

^‡Risk of bias: LOW required concurrent control group (YES), control for confounding factors (YES), and no other risk of bias factors (YES).

^§Hosoi 2007: questionnaire sent to hospitals.

^||Klingebiel 2008: The heading of table III of the paper indicates that RMS and RMS-like patients (n = 74+14 = 88) were assessed in the multivariate analysis. According to the text, patients of interest with RMS only (n = 74) were analyzed. The author confirmed the former statement that RMS and RMS-like patients (n = 88 patients) were analyzed (personal communication).