Table 2. Quality of included studies.

Study	Quality	Design type	Blinded pre/post	Attrition (%)
Aldred 2004 [25]	Good	RCT	Yes	12%
Dawson 2010 [27]	Good	RCT	Yes	10%
Green 2010 [31]	Good	RCT	Yes	4%
Howlin 2007 [35]	Low	RCT	No	1%
Solomon 2014 [45]	Good	RCT	Yes	12%
Akshoomoff 2004 [23]	Good	CS	Yes	9%
Anderson 2014 [11]	Low	CS	No	47%
Ben Itzchak 2009 [17]	Uncertain	CS	No	Not reported
Brian 2008 [26]	Uncertain	CS	Yes	Not reported
Chawarska 2009 [18]	Uncertain	CS	No	Not reported
Chawarska 2014 [19]	Uncertain	CS	Not reported	Not reported
Dereu 2012 [28]	Good	CS	Not reported	5%
Di Renzo 2015 [21]	Uncertain	CS	Not reported	Not reported
Freitag 2012 [29]	Uncertain	CS	No	Not reported
Gammer 2015 [22]	Good	CS	Not reported	2%
Gotham 2012 [30]	Uncertain	CS	Not reported	Not reported
Guthrie 2013 [32]	Uncertain	CS	No	Not reported
Gutstein 2007 [33]	Low	CS	Not reported	25%
Hobson 2015 [34]	Uncertain	CS	Yes, partly	Not reported
Kleinman 2008 [12]	Low	CS	No	Not reported
Klintwall 2015 [36]	Good	CS	Not reported	12%
Landa 2013 [13]	Good	CS	22	2%
Lerna 2014 [24]	Low	CS	Yes	22%
Lord 2006 [37]	Good	CS	Not reported	10%
Lord 2012 [38]	Uncertain	CS	No	Not reported
Louwerse 2015 [39]	Low	CS	Yes	26%
Macari 2012 [40]	Uncertain	CS	Yes	Not reported
Messinger 2015 [15]	Good	CS	Not reported	14%
Mosconi 2009 [41]	Low	CS	Not reported	49%
Ozonoff 2014 [20]	Uncertain	CS	Not reported	Not reported
Ray-Subramanian 2012 [42]	Uncertain	CS	Not reported	Not reported
Richler 2010 [43]	Uncertain	CS	No info	Not reported
Shumway 2012 [44]	Low	CS	Indep assessments	56%
Szatmari 2015 [46]	Low	CS	Not reported	30%
Thurm 2015 [47]	Uncertain	CS	Not reported	Not reported
Turner 2007 [48]	Good	CS	Not reported	17%
van Daalen 2009 [16]	Uncertain	CS	Not reported	Not reported
Venker 2014 [49]	Good	CS	Not reported	9%
Vivanti 2014 [50]	Good	CS	No	2%
Zachor 2007 [51]	Uncertain	CS	Not reported	Not reported

Studies are listed by design type. CS–prospective cohort study; RCT–randomised controlled trial.