The primary exposure of interest was the daily intake of NO₃ via drinking water, as mg L^-1 NaNO₃. The primary outcome of analyses included mean differences in body weights between NO₃-treated and control groups, the secondary outcome was the mean changes in body weights between NO₃-treated and control groups, final vs. baseline body weight of the animals.

Multiple electronic databases, including PubMed, Scopus, and Embase, were searched for relevant published papers. The primary search was supplemented with hand screening and searching of the citation lists within the papers and then electronic searching using google scholar. Searches were performed without restrictions on years or language through 25 February 2018, and updated through May 2020. A structured search strategy using various combinations of keywords i.e. nitrate, weight, and rat, and Boolean terms were conducted to identify records in each database.

Since the effect of NO₃ on body weight was not investigated as a primary outcome, all animal studies that used NO₃ treatment and reported baseline and final body weights were included in the analyses, regardless of the primary aim of the study. The papers were initially screened using the titles and abstracts; irrelevant records, reviews whether meta-analysis, systematic and narrative, letters to editors and conference papers were excluded. Then, two investigators independently reviewed the full-text records for inclusion and exclusion criteria, and potentially relevant full-text articles were finally retrieved for data extraction. Studies were excluded from the analyses if they did not report baseline and final body weights of the animals, assessed acute effects of NO₃, single dose of NO₃ (Kuzenkov and Krushinskii, 2014[24]; Peleli et al., 2015[33]); experiments conducted in female rats were also excluded because of the low number of studies and the sex-difference in body weight gain in rats (Shi and Clegg, 2009[38]). NO₃-treatment studies that used a different rout of administration other than drinking water (e.g. NO₃-rich foods) were also excluded. Finally, studies that reported the outcome as mean ± standard deviation (SD) or standard error (SE) per each experimental group, and provided information regarding the number of animals within each group, were included in our meta-analysis. Figure 1(Fig. 1) provides details on the literature search and the screening processes.

The quality and risk of bias pertaining to the included studies were assessed independently by two investigators (S.J and S.Gh) using the Systematic Review Center for Laboratory Animal Experimentation Risk of Bias Tool (SYRCLE's RoB tool) for animal intervention studies (Hooijmans et al., 2014[16]), with some modifications. In brief, this quality assessment tool is a modified version of the Cochrane RoB tool and includes 10 questions in order to determine potential sources of bias including, selection bias (sequence generation, baseline characteristics, allocation concealment), performance bias (random housing, blinding), detection bias (random outcome assessment and blinding of outcome assessor) attrition bias (incomplete outcome data), and reporting bias (selective outcome reporting) (Hooijmans et al., 2014[16]). In the current study, the risk of reporting bias by selective outcome reporting was not considered, since body weight measurements were not always described as being part of the study protocol (Schipper et al., 2018[37]), and we had excluded studies that had no report of body weights. We also did not consider random outcome assessment, blinding of outcome assessor, and incomplete outcome data as domains of quality assessment, since our outcome of interest, that being body weights of animals, was not a primary outcome in the included studies; finally, the maximum value of the quality score of the studies would be 6. When there were disagreements between the reviewers, this was resolved through consensus-oriented discussion or by consulting a third investigator.

The following information was carefully extracted from the included studies: First author's name, date of publication, sample size of both treatment and control groups, age, strain (Wistar, Sprague Dawley), health status of the animals (healthy vs type 2 diabetes), dose of NO₃ (mg L^-1), the amount of water intake (mL), exposure period (week), mean body weight values (g), and standard deviations of the baseline and final body weights in the treatment and control groups. If a study had several measurements of body weights for a period of less than < 4 weeks, only the last reported values were considered in a separate arm. Furthermore, in studies with multiple treatment groups with various NO₃ doses, each treated group was considered as a separate arm. When data were presented only graphically, the outcome values were extracted using a simple reproducible method by Adobe Photoshop (Gheibi et al., 2019[11]).

Quantitative interpretation of data was based on the weighted mean difference and 95 % confidence interval of the final body weights of intervention and control groups. We also conducted a meta-analysis using the absolute mean differences in body weights relative to the baseline values in both NO₃-treated and control groups. Further sub-group analyses were conducted to investigate the potential sources of heterogeneity, and to evaluate whether mean differences in the body weights might be affected by the dose of NO₃, the study duration, or health status of the animals. The treatment effects were accordingly assessed in the following predefined sub-groups including study duration (<4, 4-8 and >8 weeks), dose of NO₃ (72.94 ≤ or > 72.94 mg L^-1, the median of the doses), and the health status of the animals (healthy vs. type 2 diabetes).

To identify potential moderators, which could explain the variance in body weights in response to NO₃ supplementation, random effect meta-regression analyses were also conducted. Using meta-regression, we assessed potential effects in baseline body weights and the age of the animals, dose of NO₃, study duration and overall exposure to NO₃ (calculated as doses of NO₃ × study duration) on the pooled effect sizes. A sensitivity analysis was also conducted by exclusion of the experiments [Hezel, et al. (Hezel et al., 2016[13]) and El-Wakf et al. (El-Wakf et al., 2009[6], 2015[7])] used high doses of NO₃ (620 and 401 mg L^-1 d^-1).

To assess statistical heterogeneity in the meta-analysis, the I² statistic was used according to specific categories (low=25 %, moderate=50 %, high=75 %) (Higgins et al., 2003[14]). The I², a measure of inconsistency between the study results, represents the extent of overlap of the confidence intervals (CIs) of the effect sizes and quantifies the proportion of the observed dispersion (attributed to between-study differences and not to random error) (Higgins et al., 2003[14]; Hooijmans et al., 2014[15]). Although both the fixed- and random-effects models were used to estimate the pooled mean differences in body weights in response to NO₃, findings from the random-effect models were reported because of existing significant heterogeneity among the studies.

Potential publication bias was assessed using funnel plots and Eggers's regression test asymmetry for the included studies. Statistical analyses were conducted using STATA version 11 SE (StataCorp LP, TX, USA). All tests were two-tailed, and a probability level <0.05 was considered statistically significant.

Of the 175 studies (173 yielded through search of databases + 2 yielded through supplementary hand searching), 130 studies were excluded after initial title/abstract screening, and 45 full-text documents were assessed for eligible criteria. Finally, 11 eligible studies were included in the qualitative and quantitative analyses. Details of the assessment process, exclusion criteria, and number of excluded documents are provided in Figure 1(Fig. 1).

The experiments were conducted between 2004 and 2019 and included a total of 43 intervention (n=395) and 43 control (n=395) arms. All studies reported the effect of NO₃ on body weight as a secondary outcome and 7 studies were conducted on animal models of type 2 diabetes induced by a combination of high-fat diets and low dose streptozotocin (STZ). Mean baseline body weights of the animals were 281 g (with a range of 38 to 630 g). Duration of the experiments ranged from 1.4 to 25 weeks. Mean dose of NO₃ treatment was 135 mg L^-1 (ranged from ~22 to 620 mg L^-1); eleven experiments used KNO₃ as the source of NO₃ while the rest used NaNO₃. The experimental details of each matched treatment-control group, including age, strain, exposure period, number of animals per experiment, and the dose of NO₃, are provided in Supplementary Table 1.

According to SYRCLE's RoB tool, mean quality values of the studies was 1.43; overall, there was a high risk of bias mainly due to uncertainties regarding random housing, blinding of intervention, and allocation concealment. Results of quality assessment of the studies are provided in Supplementary Table 1.

Overall, the final body weights were significantly lower in the NO₃ supplemented groups compared to controls (WMD= -16.8 g, 95 % CI= -27.38, -6.24; P=0.002) (Figure 2(Fig. 2); References in Figure 2: Ashmore et al., 2015[2]; El-Wakf et al., 2009[6], 2015[7]; Gheibi et al., 2018[10]; Hezel et al., 2016[13]; Khalifi et al., 2015[21]; Khorasani et al., 2019[22]; Norouzirad et al., 2019[31]; Oghbaei et al., 2018[32]; Roberts et al., 2015[35]; Zaki et al., 2004[40]); after exclusion of 10 experiments within the arms corresponding to high doses of NO₃ (620 and 401 mg L^-1 d^-1), overall mean differences in body weights between the groups decreased by approximately 37 % but remained statistically significant (WMD= -10.1 gr, 95 % CI= -19.0, -1.19, P=0.026).

The mean differences in body weights between NO₃ supplemented groups and controls was greatest in the long-term compared to mid-term (>8 vs. 4-8 weeks) study durations (WMD= -23.16 g, 95 % CI=-35.6, -10.7 g vs. WMD= -9.19 g, 95 % CI= -17.04, -1.35 g), and was not statistically significant in short-term intervention periods (WMD= 2.34 g, 95 % CI= -13.34, 18.02 g). Highest compared to the lowest doses of NO₃ (≤ 72.94 vs. > 72.94 mg L^-1 d^-1) was also related to a greater mean difference in body weights between the groups (WMD= -31.92 g, 95 % CI= -52.90, -10.94 g vs. WMD= -8.45 g, 95 % CI= -18.32, 1.41 g). No significant differences in body weights were observed between healthy animals and type 2 diabetic animals (Table 1(Tab. 1)). Mean body weight changes in both NO₃ supplemented groups and controls are shown in Figures 3 and 4 (References in Figure 3(Fig. 3) and 4(Fig. 4): Ashmore et al., 2015[2]; El-Wakf et al., 2009[6], 2015[7]; Gheibi et al., 2018[10]; Hezel et al., 2016[13]; Khalifi et al., 2015[21]; Khorasani et al., 2019[22]; Norouzirad et al., 2019[31]; Oghbaei et al., 2018[32]; Roberts et al., 2015[35]; Zaki et al., 2004[40]); compared to controls, NO₃ treated rats gained lower body weights during the study period (WMD= 93.69 g, 95 % CI= 77.25, 110.13 g vs. WMD= 110.87 g, 95 % CI= 91.37, 130.37 g).

Meta-regression analyses showed that the mean changes in body weights in the animals were affected by age, baseline body weights of the animals and study duration (Table 1(Tab. 1)); older animals and those with higher baseline body weights showed a lower weight gain in response to daily doses of NO₃.

The amount of water intake was similar in both NO₃-treated rats and controls (WMD= -1.14 mL, 95 % CI= -2.56, 0.29, P=0.118) (Supplementary Figure 1). The amount of water intake remained constant in the NO₃-treated rats during the study period (WMD= 1.11 mL, 95 % CI= -0.51, 2.73, P=0.18) (Supplementary Figure 2).

An overall symmetric distribution of the studies around the mean effect size observed in the funnel plot indicated a low risk for publication bias. According to results of Egger's regression test, no evidence was observed regarding publication bias among the studies. High heterogeneity values were observed in our meta-analysis models; subgroup analysis (Table 2(Tab. 2)) indicated that experiments with longer duration (> 8 weeks) and those that used higher doses of NO₃ (> 72.94 mg L^-1 d^-1) might be potential sources of heterogeneity for the outcome.