4. ROBINS‐I assessment by study: Ray 2018.

Study name:Ray 2018
Risk of bias domain	Assessments by outcome	Support for judgment	Conclusion
Bias due to confounding	All outcomesa	Quote: "multivariate analysis was performed in R version 3.3.2 (2016‐10‐31). We applied a combination of multiple imputation and propensity‐matched pairing in the comparative between‐group analysis. Propensity matching was based on a logistic regression model and yielded 65 matched pairs. Background variables used for matching were age at procedure; length of time with LUTS; baseline IPSS; IPSS QoL; IIEF; Qmax; and PVR". Judgment: although authors likely used an appropriate analysis method to control confounding factors, concerns for confounding may remain. In addition, multivariate analysis including propensity‐matched pairing was reported only for IPSS and IPSS QoL. For all other outcomes in the review, risk of bias due to confounding could be considerable.	Serious
Bias in selection of participants into the study		Judgment: selection of participants into the study was not based on participant characteristics observed after the start of the intervention and the start of follow‐up and the start of the intervention likely coincided for most participants. As inclusion criteria were not reported in detail in protocol as well as in publication, there are concerns for postintervention variables that influenced selection likely to be associated with intervention (e.g. prostate volume).	Moderate
Bias in classification of interventions		Quote: "the British Society of Interventional Radiologists and the British Association of Urological Surgeons co‐funded the online UK Register of Prostate Embolization (UK‐ROPE), which was built and hosted by Dendrite Clinical Systems Ltd". Judgment: this study was based on the ongoing authorized registry (UK‐ROPE) that predefined the interventions.	Moderate
Bias due to deviations from intended interventions		Judgment: although this study was based on the prospective enrolled registry (UK‐ROPE), no information was provided with regard to co‐intervention.	No information
Bias due to missing data	Urologic symptom scores, QoL, erectile function, ejaculatory disorders, and hospital stay	Judgment: although the proportion of participants with missing data was similar across interventions, about 2/3 participants in each group were included in the analysis.	Serious
	Major adverse events, retreatment, minor adverse events, and AUR	Judgment: all participants were included in the analysis.	Low
Bias in measurement of outcomes	Subjective outcomesb	Quote: "there was no blinding (either clinician or participant) in this single‐arm observational study". Judgment: given that study outcomes were subjective, outcome measures were likely influenced by knowledge of the intervention received.	Serious
	Objective outcomesc	Judgment: although objective outcomes are unlikely influenced by knowledge of the intervention received in outcome assessment, participants and personnel were not blinded.	Serious
Bias in selection of the reported result	All outcomesa	Judgment: protocol was published and study outcomes were well predefined and described. In addition, study author provided unreported data via email.	Low
Overall	—	Judgment: serious risk of bias in ≥ 1 domain, but not at critical risk of bias in any domain.	Serious

AUR: acute urinary retention; IIEF: International Index of Erectile Function; IPSS: International Prostate Symptom Score; LUTS: lower urinary tract symptoms; PVR: postvoid residual; Qmax: maximum flow rate; QoL: quality of life; ROBINS‐I: risk of bias tool to assess non‐randomized studies of interventions.
^aAll review outcomes reported in study: urologic symptom scores, QoL, major adverse events, retreatment, minor adverse events, erectile function, AUR, ejaculatory disorders, and hospital stay.
^bUrologic symptom scores, QoL, major adverse events, erectile function, ejaculatory disorders, and minor adverse events.
^cRetreatment, AUR, and hospital stay.