*3.3. Quality Assessment*

The detailed results of quality assessment are presented in Tables S3 and S4. First, the quality of the eighteen included animal studies was assessed through the ARRIVE guidelines. As a result, the included animal articles adequately provide an accurate title and abstract, a structured and thorough introduction, an ethical statement only for mammalian studies, and an adequate study design except for two studies [44,45] which are unclear. None of them justified the sample size, and consequently, the use of a too small number of animals may lead to a lack of experimental statistical significance given the use of too many animals may lead to unnecessary wastage of resources and ethical issues. Only one study did not clearly describe statistical methods [42]. Baseline characteristics (body weight, age, and gender) before treatment are reported in five of the total of studies [8,40,48–50]. For twelve studies [35–37,39,41–47,51], body weight was not specified, and for three studies [38,41,42], age was not reported. All studies adequately reported and interpreted their results in terms of numbers analyzed, outcomes, adverse events, interpretation, and generalizability.

Secondly, the risk of bias of the included animal studies was assessed using SYRCLE's tool. In regards to sequence generation, in twelve out of eighteen studies, the allocation sequence was randomly generated and applied. However, in eleven out of 12 studies, the investigators did not describe the sequence generation process such as the use of a random number table or a computer random number generation. Only in the study of Zhang et al. [50], mice were randomly allocated into the control group and the TiO2 NPs' group using a web-based randomization service. For all studies, it is not clear how animals were allocated to different groups. In addition, for all studies, all groups had similar characteristics at baseline. Regarding allocation concealment, the concealment was not clear for all studies. Indeed, no studies have explicated the concealed procedure when the investigators have allocated the animals to different groups. Moreover, all included studies have a high risk of performance bias. Indeed, the animals did not randomly house during the experiment and it is not clear whether the investigators did not blind from knowledge which intervention each animal received during the experiment. Additionally, overall, it is not specified whether the investigators did not select animals at random for outcome assessment. However, the outcome assessment methods are the same in both groups for all studies. In regards to attrition and reporting bias, the risk is low for all studies since the outcome data reported in each study was completed for each outcome. All primary outcomes have been reported. Finally, the studies did not report other problems that could result in a high risk of bias. As a conclusion, according to SYRCLE's risk of bias tool, the quality of each study is debatable due to an inadequate or unclear randomization of allocation, housing and outcome assessment, and a lack of blinding. However, the studied population has similar characteristics at baseline making the sample homogenous and avoiding confounding bias. Moreover, in regards to the reporting of outcomes (complete outcome data reporting, adequate outcome reporting), the risk of bias is low.
