Table 2

The Risk of Bias In Non-randomised Studies-of Interventions (ROBIN-I) assessment

StudyRisk of bias due to confoundingRisk of bias in selection of participants into the studyRisk of bias in classification of interventionsRisk of bias due to deviations from the intended interventionsRisk of bias due to missing dataRisk of bias in measurement of the outcomeRisk of bias in selection of the reported resultOverall risk of bias
Kao et al22LowLowLowLowLowLowLowLow
Manz et al23LowLowLowLowSerious (missing ECOG status, excluded from analysis)LowModerate (multiple subgroup analysis)Serious
Hong et al24Serious (unadjusted for confounding variables)LowLowLowLowLowLowSerious
Feng et al28LowLowLowLowLowLowLowLow
Xu et al27Serious (small sample size unadjusted for user acceptability of AI and clinical experience)LowLowLowLowLowModerate (multiple subgroup analysis)Serious
Hosny et al32LowLowLowLowLowLowLowLow
McIntosh et al30LowLowLowLowLowLowLowLow
Wong et al25LowLowLowLowLowSerious (unblinded participants with two subjective outcome measures)LowSerious
  • The risk of bias for each domain of the study is colour-coded. High risk of bias: red; moderate risk of bias: orange and low risk of bias: green.

  • AI, artificial intelligence; ECOG, Eastern Cooperative Oncology Group.