Table 2.

Risk of bias tool.

Bias domain	Assessment question	Criteria
Sample selection bias	Were sample locations and sampling methods implementing such that sampling did not introduce systematic differences depending on the value of the exposure variable for each sample (in the case of continuous exposure data) or between the comparison groups (in the case of categorical exposure measures)?	Criteria for the judgement of “Yes” (low risk): Method for determining the sampling locations is identical and independent of exposure status (ie, sample taken from the influent and final effluent of the WWTPa) The WWTP receives only municipal, hospital, and domestic wastewater regardless of differences in treatment methods or treatment stages (primary, secondary, or tertiary) Influent wastewater does not have any form of pretreatment before being discharged in the WWTP The time between sampling at all sites is sufficiently close to render the outcomes measured at these sites comparable for the sample type in question Collection of 24-h composite samples The authors describe the frequency of sampling (daily, weekly, monthly, etc) at each site The authors describe the volume of collected samples from the influent and effluent of the WWTP Criteria for the judgement of “No” (high risk): Sampling locations are selected differently (eg, samples taken from the effluent of the grit chamber, aeration tank, and secondary clarifier) The WWTP does not receive municipal, hospital, and domestic wastewater regardless of differences in treatment methods or treatment stages (primary, secondary, or tertiary) Influent wastewaters have some form of pretreatment before being discharged in the WWTP Time between sampling at all sites is not sufficiently close Collection of grab samples Collection of grab samples The authors do not describe the frequency of sampling (daily, weekly, monthly, etc) at each site The authors do not describe the volume of collected samples Risk of bias will be considered “unclear” if there is not enough information to judge sample selection bias criteria as either “yes” or “no.” For example, if methods for determining sampling locations are not described in enough detail.
Information bias	Were outcome ascertainment methods (ie, methods for antibiotic-resistance gene, antibiotic-resistant bacteria, and antibiotic or bacterial measurements) conducted in a way that ensures the same accuracy regardless of wastewater sample type?	Criteria for the judgement of “Yes” (low risk): Identical microbiological methods are applied to all samples (ie, influent and effluent samples) for ARBb, ARGc, and antibiotic detection (eg, culture, polymerase chain reaction, genotyping, phenotypic tests, mass spectrometry, and high-performance liquid chromatography). Controlling for different laboratory factors (eg, laboratory type, technician, testing date, and instrument used) Criteria for the judgement of “No” (high risk): Application of different methods depending on the comparison group No adjustment strategy for different laboratory methods Risk of bias will be considered “unclear” if there is not enough information to judge information bias criteria as either “yes” or “no.” For example, if methods for analyses are not explained sufficiently to reach a judgement.
Confounding	Were adequate methods to control for potential confounding employed?	Criteria for the judgement of “Yes” (low risk): Restriction of the sample population (eg, samples are not collected on a rainy day and instead collected on a dry day) Samples are collected in different seasons (eg, winter and summer) Analytical confounding control (eg, stratification, regression adjustment, and test samples are stored correctly) Criteria for the judgement of “No” (high risk): The sample population is not restricted (eg, samples are collected on a rainy day) Lack of any confounding control despite being likely (eg, samples are not collected in different seasons [winter and summer] and no consideration of water salinity) Inappropriate method of confounding control (eg, test samples are not stored correctly) Controlling for confounding is correctly applied for some potential confounders, but not for all Risk of bias will be considered “unclear” if there is not enough information to judge information bias criteria as either “yes” or “no.” For example, if methods to control for confounding are mentioned but the implementation is not explained sufficiently at length to reach a judgement.

^aWWTP: wastewater treatment plant.

^bARB: antibiotic-resistant bacteria.

^cARG: antibiotic-resistance gene.