4. Summary of evidence quality.

	Foreit 1993	Hardy 1998	Nacar 2003	Abdel‐Tawab 2008	Lee 2011	Sebastian 2012
NOS criteria for cohort studies
Exposed cohort representativeness	‐‐‐	‐‐‐	‐‐‐	‐‐‐	‐‐‐	‐‐‐
Nonexposed cohort selection	✸	‐‐‐	✸	‐‐‐	✸	‐‐‐
Exposure ascertainment: method used	✸	‐‐‐	✸	✸	✸	✸
Comparability of groups: design or analysis	‐‐‐	✸✸	‐‐‐	✸	‐‐‐	✸✸
Outcome assessment: method used	‐‐‐	‐‐‐	‐‐‐	‐‐‐	‐‐‐	‐‐‐
Follow‐up length	✸	✸	✸	✸	✸	✸
Follow‐up adequacy	‐‐‐	‐‐‐	✸	✸	‐‐‐	‐‐‐
Intervention fidelity (>= 4 criteria)	‐‐‐	‐‐‐	‐‐‐	✸	‐‐‐	✸
Quality of evidence1,2	Poor	Very low	Very low	Low	Poor	Low

¹Evidence was initially considered moderate quality and then downgraded for 1) no stars for comparability (not controlling for confounding), 2) not meeting >= 4 of remaining 6 NOS criteria, and 3) not having intervention fidelity information for >= 4 categories. Quality grades were moderate, low, very low, or poor.
 ²We did not use criterion for 'outcome of interest not present at study start' (Assessment of risk of bias in included studies).