Salt and obesity: a systematic review and meta-analysis of observational studies.

BACKGROUND
Existing research has investigated the association between sodium intake and obesity. We aim to conduct a systematic review and meta-analysis of reported evidence regarding the association between sodium intake and obesity.


METHODS
Multiple electronic databases (PubMed, Scopus, and Google Scholar) were searched for observational studies published until August 2016. A systematic literature review identified 11 cohort and 21 cross-sectional studies.


RESULT
Among the 32 studies identified in the systematic literature search, only 18 cross-sectional reports had sufficient data to be included in the meta-analysis. Higher sodium consumption was associated with greater BMI (weighed mean difference (WMD) = 1.24 kg/m2, 95%CI: 0.80, 1.67; I2 = 98.4%; p < .0001), and higher sodium intake was associated with 4.75 cm (95%CI: 3.25, 6.25; 90.8%; p <.0001) greater waist circumference (WC).


CONCLUSION
The present meta-analysis suggests that sodium consumption was associated with greater BMI and WC.


Introduction
Obesity has become a leading global health concern over the past three decades (Wang & Lobstein 2006). Lifestyle factors including unhealthy dietary intake and sedentary lifestyles have been associated with increased risk of obesity (Fonseca-Alaniz et al. 2007). From the time of hunters and gatherers to the year 2012, considerable increases in sodium intake (from <1.0 g/d to 9-12 g/d) have been reported (US Department of Health and Human Services 2001; Brown et al. 2009). In the modern day, 80% of total sodium intake comes from processed and restaurant food (Brown et al. 2009). Several epidemiological studies have suggested that increased sodium intake might be associated with the rises in obesity prevalence in recent decades (WHO 2014;Commission 2015). Both obesity and high salt consumption have been linked to greater risk of a variety of chronic diseases such as hypertension, diabetes, cardiovascular events, certain cancers and other health conditions (US Department of Health and Human Services 2001;WHO 2002WHO , 2014. Several explanations may directly link salt consumption and obesity. One hypothesis is that consumption of higher amounts of sugar-sweetened beverages follow salt consumption He & MacGregor 2009). Specifically, higher sodium intake drives a thirst response and promotes fluid intake (Campbell 2004;Stachenfeld 2008), which might result in sugar-sweetened beverage consumption . Furthermore, salty foods are often high in fat and energy (Azadbakht & Esmaillzadeh 2008;Rouhani et al. 2012) and are more palatable, encouraging individuals to consume greater quantities of these foods. Therefore, the direct link between sodium and obesity may be mediated by increased energy intake. However, there is evidence supporting an association of salt intake and obesity that is independent of energy intake (Ellison et al. 1980;Libuda et al. 2012;Song et al. 2013;. Although such an association has not been investigated in humans, animal models indicate that a high-sodium diet increases adiposity hypertrophy as well as plasma leptin concentrations that might be attributable to the lipogenic capacity of white adipose tissue (Fonseca-Alaniz et al. 2007). To date, a few research works have studied the association of 24hour urinary sodium (24-UNa) and obesity Grimes et al. 2013b;Song et al. 2013;Yi & Kansagra 2014;Ma et al. 2015). In a representative sample of adults from New York City, after normalizing the sodium-potassium ratio as a proxy adjustment for energy, higher sodium intake was related to greater odds of obesity (Yi & Kansagra 2014). Another cross-sectional study from the UK National Diet and Nutrition Survey showed that higher salt intake was significantly associated with greater body fat mass in both children and adults, after controlling for energy intake . In that study, a 1 g/d increase in salt intake increased the risk of obesity by 28% in children and 26% in adults . Consistent with this, in two other studies, body mass index (BMI) increased proportional to incremental salt intake Grimes et al. 2013b). Nevertheless, the association between sodium and body fat has not been comprehensively examined . A recent cross-sectional study among Korean children ) and a multiethnic cohort  showed direct links between sodium and body fat determined by dual-energy X-ray absorptiometry (DEXA).
Given the widespread use of sodium in commercial foods and the magnitude of their contribution to sodium intake, finding a link between salt and obesity could inform health policies with regard to the quantity of salt content permitted in processed and restaurant foods, and thereby improve cardiovascular health. Moreover, in the existing literature there have been considerable methodological differences in sodium intake assessment methods, study populations and design, which may lead to heterogeneity between studies. Therefore, in the present systematic review, we performed a meta-analysis of emerging literature to elucidate the association between salt intake and obesity.

Search strategy
We searched MEDLINE, Google Scholar and Scopus databases for relevant studies published until August 2016. Additional papers were obtained by hand searching of references from relevant original and review articles. Keywords from Medical Subject Headings (MeSH) were selected in the search strategy. The search terms included were equivalents of sodium and salt intake or excretion in the combination with terms for anthropometric measures, including weight, waist circumference (WC), BMI, overweight, obesity, fat mass and adiposity. No restrictions were applied regarding the time of publication, the study design or language. The present meta-analysis is based on PRISMA guidelines.

Inclusion and exclusion criteria
Two reviewers independently checked inclusion and exclusion criteria by reading the titles, abstracts, and then the full text of the articles, if necessary. For two potentially relevant studies with missing data, we contacted the research groups and asked them to provide the necessary information, but only one of them responded. Studies were included if they (i) were observational in design and (ii) reported means and standard deviation (SD) or standard error (SE) or 95% confidence intervals (CI) for anthropometric indices (weight, BMI, WC, waist-to-hip ratio) in low and high categories of salt/sodium intake or urinary sodium. Reviews, editorials, non-human studies, and letters without sufficient data were excluded.

Data extraction
The following information was extracted for each study: first author's name, publication year, sample size, age of subjects, study design and adjusted confounders. Moreover, means and their corresponding SD or SE or 95%CI of obesity-related indices were extracted from the first and the last categories of salt/ sodium. Some studies reported data stratified by gender Murata et al. 2010;Pfister et al. 2014). In this case, several effect sizes were extracted from one paper.

Statistical analysis
Reported SEs were converted to SDs. Since most of the studies had reported mean ± SD/SE of BMI and WC, and not many included other obesity indices (e.g. fat mass, weight, and WHR) or odds ratios for obesity risk, the present meta-analysis was conducted on only means of BMI and WC. STATA 11.0 (Stata Corp., College Station, TX) software was used to run the meta-analysis. The pooled effect size was calculated using a random effects model. Heterogeneity and between-subgroup heterogeneity were evaluated using the I 2 statistic and a fixed-effect model, respectively. The contribution of each study on the overall effect was estimated using sensitivity analyses. To detect the publication bias, Begg's and Egger's adjusted rank correlation test was run.

Appraisal of the quality of studies
The quality of the studies included in the meta-analysis was assessed using the Newcastle-Ottawa Scale modified for cross-sectional studies. A quality score was determined on the basis of three major components: (i) selection of study groups (0-5 points), (ii) adequacy of adjustment for confounding (0-2 points) and (iii) ascertainment of the outcome of interest (0-3 points). High-quality studies were defined as those that scored with at least seven stars on the Newcastle-Ottawa scale. Medium quality received scores of five to six stars.
One study reported sodium density as independent variable.   . Of those that attained four, not all used validated measurement tools for ascertainment of the exposure, and the study with three stars was not truly representative of the average in the target population and also did not use a validated measurement tool for exposure ascertainment. In regard to comparability criteria, the score for all studies was zero. Regarding outcome criteria, 12 studies achieved the maximum of three stars ( Higher sodium consumption was associated with greater BMI (weighed mean difference (WMD) ¼ 1.24 kg/m 2 , 95%CI: 0.80, 1.67; I 2 ¼ 98.4%; p < .0001). Heterogeneity could not be eliminated by subgroups based on sodium assessment tools (i.e. salt/ sodium intake, urinary sodium, serum sodium, salty snacks), though heterogeneity between subgroups was statistically significant (p < .0001) ( Figure S1). Incorporation of salty snacks with salt/sodium intake led to a 0.81 kg/m 2 increment in BMI, but heterogeneity was high (I 2 : 94.3%, p < .0001) ( Figure  2). In subgroups by gender (male, female and both sexes), direct associations between sodium/salt and BMI were also observed ( Figure 3). There was no evidence of publication bias (Begg's test, p ¼ .080 and Egger's test, p ¼ .629).
Individuals who had higher sodium intakes in comparison with those who consumed lower amounts of sodium had 4.75 cm (95%CI: 3.25, 6.25) greater WC; however, there was significant heterogeneity (I 2 ¼ 90.8%; p < .0001). After subgroup analyses based on gender, a significant increment in WC was observed in females and both sexes, though this association was not significant in males. This subgroup analyses eliminated heterogeneity in females, but not in both sexes (I 2 ¼ 90.7%; p < .0001) and males (I 2 ¼ 73.4%; p ¼ .053) (Figure 4). Subgroup analyses based on the method of sodium intake assessment showed that both urinary and dietary sodium were positively associated with WC (WMD for dietary sodium: 6.52 cm (95%CI: 5.07, 7.96) and WMD for urinary sodium: 3.63 cm (95%CI: 2.18, 5.08) ( Figure  5). Subgroup analyses by sodium assessment tool eliminated heterogeneity for dietary instruments (I 2 ¼ 43.4%; p ¼ .171), but not in the urinary sodium group (I 2 ¼ 87.5%; p < .0001) ( Figure 5). No evidence for publication bias was observed (Begg's test, p ¼ .711 and Egger's test p ¼ .113).

Discussion
The results of this meta-analysis show that high salt/ sodium intake was significantly associated with greater BMI and WC. Since heterogeneity was high and there were subgroup differences based on sex and sodium assessment tools, further studies are needed to explore associations between salt and anthropometric measures. To our knowledge, this is the first comprehensive systematic review and meta-analysis to evaluate the relation between sodium/salt intake and obesity. This meta-analysis of observational studies provides a more reliable conclusion than individual assessment of particular studies. Cohort studies are more indicative of causal relations because of their prospective nature and provide stronger evidence compared to cross-sectional studies. According to the ranking of evidence by study design, cohort studies are superior to cross-sectional designs, where inverse causation bias may occur. To date, only few cohort studies have directly assessed the relationship between salt intake and obesity Larsen et al. 2013;, but data from these studies could not be used for our meta-analysis due to the variability of reported outcomes, their statistical analysis, or because they reported sodium density as an exposure. Overall, both cross-sectional and cohort studies have reported direct correlations between salt intake and obesity.
In the present meta-analysis, high salt/sodium intake was positively associated with greater BMI and WC. Similar to these results, several studies have shown significant correlations between sodium intake and obesity. Yoon et al. indicated that high sodium intake might be a significant risk factor for weight gain independent of calorie intake . A longitudinal study by Libuda et al. reported that BMI and percent body fat in 3-18-year-old children were directly associated with urinary sodium . Larsen et al. showed that high sodium intake was associated with subsequent gain in body fat and loss of fat-free mass, while it was not associated with changes in BW or WC .
At the same time, a few studies have demonstrated no significant relation or even an inverse relation between sodium and BMI Jablonski et al. 2009;Visser et al. 2009;Murata et al. 2010;Baudrand et al.2014;Zhao et al. 2014;Ponzo et al. 2015). These discrepancies, as well as the substantial heterogeneity in the results of this metaanalysis, are probably related to sodium assessment method (diet or urinary sodium), the accuracy of assessment method or differences in populations (age, sex, and ethnicity) or study design type. However, due   to the limited number of studies in each subgroup, we could not determine the source of heterogeneity. This was particularly a problem regarding the determination of whether study design was a specific source of heterogeneity; as only cross-sectional effect sizes could be extracted from cohort studies (baseline or end values), it was not possible to conduct analysis on prospective effect sizes.     Figure 5. Forest plot derived from random-effects models depicting the association of salt intake with WC by sodium assessment tools.
In the assessment of study quality, most studies included in this meta-analysis were graded seven points or above. Regarding study selection, most studies used an acceptable method for selection, but confounding factors were not controlled in any study. This matter is relevant since individuals with higher salt intake are also less likely to have healthy lifestyles, e.g. they are likely to have lower physical activity levels, make less healthy food choices, and have poorer eating behaviors . Therefore, these factors may influence the association between salt and obesity. However, most studies received all possible stars for the outcome criteria.
Several reasons may explain the direct link between salt and obesity: (i) higher sodium intake drives the thirst response and promotes fluid intake (Campbell 2004;Stachenfeld 2008), which might be compensated with drinking sugar-sweetened beverages ; (ii) salty foods are often high in fat and energy; (iii) salty foods are more palatable and encourage individuals to consume greater quantities of these foods.
The current meta-analysis has some strengths. First, various subgroup analyses were performed to identify sources of heterogeneity. Second, a comprehensive systematic literature search was conducted to find all relevant articles. However, some limitations should also be considered when interpreting the results. First, cross-sectional studies cannot be used to infer causality. Second, we could not determine the dose-response association between salt intake and weight gain. Third, failure to accurately assess sodium intake in some studies (e.g. that used a food intake questionnaire instead of the gold standard of urine sodium) may influence the relation between salt and obesity indices. Dietary recall and weighted diet records often underestimate intake due to under-reporting. FFQs are more useful than dietary recall and dietary records since they assess intake over a longer period; however, accurate quantification of daily food intake is difficult. Although 24-hour urine collection is known as a gold standard and the most accurate method for assessment of sodium intake, research is limited by under-collection of data using this method (McLean 2014). Fourth, BMI has limitations for estimating the amount of muscle mass and fat. Fifth, we reviewed some studies in which the associations between salt and anthropometric measures were reported as secondary outcomes. Finally, the most important limitation of using these secondary data was insufficient control for confounders.
In conclusion, the results of this meta-analysis suggest that high salt/sodium intake is directly correlated with BMI and WC. However, several gaps still remain that warrant further investigation. Future studies are needed to (1) investigate the sources of heterogeneity inherent in this association, (2) assess other anthropometric measures, especially fat mass, (3) conduct prospective studies that specifically examine the association of salt intake and obesity while taking into account the role of relevant confounders and (4) use more precise methods to measure obesity indices, such as fat mass as well as sodium intake.