Table 2. Risk of Bias Assessment of 22 Studies Included in the Meta-analysisa.
| Source | Bias Arising From the Randomization Process or Due to Confoundingb | Bias Due to Selection Into the Study | Bias in Classification of Interventions | Bias Due to Deviations From the Intended Intervention | Bias Due to Missing Outcome Datac | Bias in Measurement of Outcomes | Bias in Selection of the Reported Results | Overall Risk |
|---|---|---|---|---|---|---|---|---|
| IDEA collaboration,1 2018 | Low | NA | NA | Some concerns | Low | Low | Low | Some concerns |
| Ito et al,20 2000 | Low | NA | NA | Some concerns | Low | Low | Low | Some concerns |
| Ji et al,32 2018 | Serious | Low | Low | Low | Low | Low | Moderate | Serious |
| Laurent et al,36 2018 | No information (moderate) | Low | Low | Low | No information (low) | Low | Moderate | Moderate |
| Cespedes Feliciano et al,26 2017 | Serious | Moderate | Low | Low | No information (low) | Low | Moderate | Serious |
| Hwang et al,31 2017 | No information (moderate) | No information (low) | Low | Low | No information (low) | Low | Moderate | Moderate |
| van Erning et al,46 2017 | Moderate | Moderate | Low | Low | No information (low) | Low | Moderate | Moderate |
| Tsai et al,45 2016 | Moderate | Serious | Low | Low | Low | Low | Moderate | Serious |
| Hassan et al,30 2015 | Serious | Moderate | Low | Low | No information (low) | Low | Moderate | Serious |
| Kumar et al,35 2015 | Moderate | Low | Low | Low | No information (low) | Low | Moderate | Moderate |
| Sgouros et al,41 2016 | Serious | Low | Low | Low | Moderated | Low | Moderate | Serious |
| Sun et al,42 2015 | Serious | Moderate | Low | Low | No information (low) | Low | Moderate | Serious |
| Kim et al,33 2014 | Serious | Low | Low | Low | No information (low) | Low | Moderate | Serious |
| Tsai et al,44 2013 | Moderate | Moderate | Low | Low | No information (low) | Low | Moderate | Moderate |
| Figer et al,29 2011 | Serious | Moderate | Low | Low | Low | Low | Moderate | Serious |
| Ahmed et al,24 2010 | Moderate | Serious | Low | Low | Low | Low | Moderate | Serious |
| Yun et al,48 2010 | Serious | Moderate | Low | Low | No information (low) | Low | Moderate | Serious |
| Chapuis et al,27 2009 | Serious | Moderate | Low | Low | Low | Low | Moderate | Serious |
| Qiu et al,39 2009 | Moderate | No information (moderate)e | Low | Low | No information (low) | Low | Moderate | Moderate |
| Kornmann et al,34 2008 | Moderate | Low | Low | Low | No information (low) | Low | Moderate | Moderate |
| Morris et al,37 2007 | No information (moderate) | Moderate | Low | Low | No information (low) | Low | Moderate | Moderate |
| Neugut et al,38 2006 | Moderate | Serious | Low | Low | No information (low) | Low | Moderate | Serious |
Abbreviations: IDEA, International Duration Evaluation of Adjuvant Therapy; NA, not applicable.
For the 2 randomized trials included in this evidence base, the columns refer to the Cochrane Collaboration’s tool for assessing the risk of bias in randomized trials. For the remaining observational studies, the columns refer to the Risk of Bias in Nonrandomized Studies of Interventions (ROBINS-I) tool.
Risk of bias arising from the randomization process is presented for randomized clinical trials, and the risk of bias due to confounding is presented for observational studies. All observational studies with the “no information” designation adjusted for variables using regression but did not specify which variables they controlled for in the analysis. In each case, we assumed that the authors adjusted for the important confounders we identified a priori.
Studies with the “no information” designation did not provide the number of patients lost to follow-up or the amount of missing data. Given the reliance on medical records, administrative databases, and cancer registries, we thought that the risk of bias due to missingness would be low.
More than 10% of patients had missing data.
The index time used to define survival time was not described.