Sugar Sweetened Beverages Intake and Risk of Non-Communicable Chronic Diseases in Longitudinal Studies: A Systematic Review and Meta-Analysis with 1.5 Million Individuals

Leonardo Pozza Santos (  leonardo_pozza@yahoo.com.br ) Universidade Federal de Pelotas https://orcid.org/0000-0002-3993-3786 Denise Petrucci Gigante Federal University of Pelotas: Universidade Federal de Pelotas Felipe Mendes Delpino Federal University of Pelotas: Universidade Federal de Pelotas Ana Paula Maciel Federal University of Pelotas: Universidade Federal de Pelotas Renata Moraes Bielemann Federal University of Pelotas: Universidade Federal de Pelotas


Introduction
Non-communicable chronic diseases (NCD) are the leading cause of deaths worldwide [1]- [3], and behavioral risk factors, like unhealthy eating habits, are the main determinants for the development of NCD [4], [5]. Despite the adherence of healthy dietary patterns have recently improved, mainly in high-income countries [6], the consumption of unhealthy food items is still high [6]- [8]. In this context, sugarsweetened beverages (SSB) are among the unhealthy food items most associated with adverse outcomes, especially with NCD [9]- [12].
SSB are beverages sweetened with different forms of added sugar, including carbonated and non-carbonated soft drinks, fruits and sports drinks, and they are characterized by low nutritional quality [13].
In 2010, almost 200 thousand deaths around the world were attributable to SSB consumption, being type 2 diabetes the leading cause [14]. Singh et al. [14] also estimated that the majority of mortality attributable to SSB intake occurred in low-and middle-income countries (LMIC), while high-income countries accounted for a quarter of the total deaths. Since then, the consumption of SSB is apparently decreasing, although the prevalence remains high, mainly in speci c sub-groups such as men and younger individuals [15]- [19].
Association between SSB intake and health-related outcomes is well reported in literature. Several investigations have shown that high SSB intake increases the risk of obesity-related diseases [9], [10], [12], [20]. However, the effects of SSB consumption on NCD, like type 2 diabetes, appear to be only partially explained by obesity status, being part of this effect independent of adiposity [11]. Despite the scienti c literature regarding the association between SSB consumption and NCD has increased in the last years, the magnitude of the association is quite heterogeneous among studies.
As SSB intake is an important and suitable target for environmental interventions [20], understanding the strength, consistency and biological gradient of the association between SSB consumption and health-related outcomes is important. Based on these evidence, comprehensive approaches to reduce SSB consumption, such as public health interventions and taxation, may be better planned.
Therefore, we performed a systematic review and meta-analysis of longitudinal studies with adults to assess the association between SSB intake and NCD, namely type 2 diabetes, obesity, coronary heart disease (CHD) and stroke.

Search strategy
This study is a systematic review and meta-analysis performed according to the PRISMA statement [21]. We searched studies that observed the association between SSB intake and type 2 diabetes, obesity, CHD or stroke in adults published until December 2020. We considered SSB as any drink with added sugar, such as soft drinks, arti cial fruit juices, and sports drinks.
The articles' search was carried out using the Pubmed, Lilacs, Web of Science, Cochrane, Embase, and Scopus databases. The terms used on the articles' search were based on the MeSH terms along with other non-MeSH terms to complement the searching, as following: "soft drinks", "processed juice", "sugar-sweetened drinks", "sugar-sweetened beverage", "fruit juices", "arti cially-sweetened soda", "arti cially sweetened beverages", "sugar-sweetened soft drinks", "diet soda", "fruit-avored drinks", "sports and energy drinks", "bottled fruit juices" with the Boolean operator "OR" between the terms, and "AND" "obesity", "coronary disease", "diabetes", "cerebral vascular accident", "cerebrovascular accident", "stroke" and "cerebrovascular disease", also utilizing "OR" between the terms. This systematic review was registered and approved on the International Prospective Register of Systematic Reviews (PROSPERO) (CRD420212342060).

Inclusion criteria
We included longitudinal studies (cohort) carried out with adults (≥ 20 years old) not belonging to speci c groups (individuals with an initial diagnosis of any morbidity or institutionalized, for example) and investigating the association between the intake of one or more types of SSB with at least one of the four outcomes of interest (type 2 diabetes, obesity, CHD and stroke).

Exclusion criteria
Systematic and non-systematic reviews, cross-sectional and intervention studies, and studies with animals or in vitro were excluded. We also excluded those investigations that reported data from dietary patterns without showing results exclusively for sugar sweetened beverages.

Study selection
Two independent reviewers (APM and FMD) conducted the articles' selection, and, in case of disagreements, the judgment was left to a third one (DPG). After excluding duplicates, the selection started by reading the titles, followed by abstracts, according to the inclusion and exclusion criteria. Finally, the eligible articles were selected from a full reading of the manuscripts. We also evaluated the reference lists of all included articles to identify relevant studies that might not have been initially located.

Meta-analysis
In the meta-analysis, we included studies that evaluated the association between SSB intake and risk of type 2 diabetes, obesity, CHD and/or stroke, and reported data as Odds Ratio (OR), Hazard Ratio (HR) or Relative Risk (RR). We pooled the RR across studies through the DerSimonian and Laird random-effects model [22]. We considered HR values as RR since their results are very similar [23]. ORs were converted into RRs using the following equation: RR=OR / [(1-P o ) + (P o X OR)], in which Po represents the incidence of the outcome of interest in the nonexposed group [24]. We used the xed-effects model to calculate article-speci c RR for studies that assessed values strati ed by sex or different subgroups. In case of missing information for OR, HR or RR in the manuscripts, we contacted authors in two different occasions, with a delay from 30 days between each contact.
For studies assessing the isolated risk of different types of SSB, we included results for the most consumed beverage in the sample under analysis. For all studies, the cut-off point assumed was one or more servings of SSB per day. We considered low/moderate intake one or less portion of SSB per day and the intake of two or more portions of SSB per day was classi ed as high intake. The HR, OR and RR with the greatest adjustments for confounding were considered as the nal effect.
We used the Higgins I 2 test to estimate statistical heterogeneity among studies [25]. This test varies from 0-100% and the higher the values the higher heterogeneity. We used the cut-off points proposed by Higgins et al. [26] to classify heterogeneity, considering moderate heterogeneity those values of I 2 between 50% and 74.9% and high heterogeneity when I 2 values were ≥75%. Analyses were performed through R language utilizing the Rstudio program and miniMeta package [27].

Quality of evidence and risk of publication bias
The methodological quality of the studies was assessed by two independent reviewers (APM and FMD), using the Newcastle-Ottawa Scale (NOS), which assesses longitudinal studies quality based on sample selection criteria, comparability between groups, and evaluation of exposure or outcome [28]. This scale includes eight items, and each of them receives a star when the study is classi ed as high quality, except for the comparability item, which can receive two stars when results were adjusted by sex and other factors. Thus, the total NOS scale score varied from 0 to 9. We classi ed studies which scored less than ve points in the NOS with poor quality, those studies with ve or six points as medium quality, while articles which scored seven or more points were classi ed as high-quality studies.
Risk of publication bias was assessed using the funnel plot, and the funnel plot asymmetry was tested using the Egger's and Begg's test [31].

Results
Page 4/26 Figure 1 summarizes the studies selection process. A total of 18755 titles and abstracts were screened after duplicates have been removed. The screening of titles and abstracts retrieved 98 articles for further consideration. After exclusion of 69 manuscripts which have not met the inclusion criteria, our systematic review included ndings from 27 longitudinal studies evaluating the association between SSB intake and type 2 diabetes, obesity, CHD and stroke ( Figure 1).
Notwithstanding the 27 included studies have been conducted in 12 different countries, more than a half of them came from United States. In addition, almost all studies were conducted in high-income countries, with only three of them from middle-income countries (China, Thailand, and Mexico). Almost half of the included studies enrolled between 10,000 and 50,000 individuals in the sample. In terms of follow-up period, almost 20% of the selected papers followed their sample for more than 20 years. Finally, type 2 diabetes was the most frequent outcome in the included studies (n=15). Stroke was the outcome in ve, while obesity was the outcome in four studies and CHD in three of them (Table 1). Table 1 Characteristics of the 27 papers selected by systematic review on the longitudinal association between sugar sweetened beverages intake and noncommunicable chronic diseases in adults. The scores obtained in the NOS varied from 5 to 9. The majority of articles were classi ed as high-quality studies (21 studies; 77.8%), while six were classi ed as medium quality. None of the included articles scored less than ve points in the NOS. Five studies classi ed as medium quality did not clearly state about the representativeness of the exposed cohort, ve studies did not report on the assessment of outcome and four medium quality studies did not state well the adequacy of follow-up cohorts (Supplementary table 1).

Association between SSB intake and type 2 diabetes
Fifteen investigations assessed the association between SSB intake and diabetes (Table 2). In 13 studies, diabetes was the only outcome [9], [32]- [43], while one study assessed diabetes and weight gain as the main outcomes [44] and another one assessed diabetes and stroke [45]. Results were considerably consistent in showing that SSB intake has been associated with higher risk of type 2 diabetes, since only one study did not nd any statistically signi cant association [32]. Schultze et al. [44], Palmer et al. [33], de Koning et al. [35] and Hirahatake et al. [41] showed that the consumption of more than one portion of SSB per day increased the risk of type 2 diabetes in United States, after controlling for several confounders, including dietary intake and body mass index. In addition, study conducted by the InterAct consortium from eight European countries also revealed positive effects of SSB intake on the risk of type 2 diabetes after adjustment for demographic and lifestyle characteristics [9]. Studies which analyzed SSB intake in grams, milliliter or ounce instead portion also found positive associations with type 2 diabetes.
Four different studies showed an increased risk of type 2 diabetes as increased the consumption of SSB, independent of sociodemographic and nutritional characteristics, with effect measures varying from 1.18 RR to 1.34 HR [37], [38], [40], [45]. One of these studies also adjusted the analyses for waist circumference [38], suggesting that the effects of SSB intake on risk of type 2 diabetes appear to be independent of central adiposity.
The positive effects of SSB intake on the risk of type 2 diabetes was also observed in studies conducted in middle-income countries.
Stern et al. [42], analyzing data from almost 80,000 women from the Mexican Teachers' Cohort, found that the consumption of one portion of sweetened soft drinks per day increased the hazard of type 2 diabetes in 1.27 times (HR = 1.27; 95C.I. 1.16 -1.38), after adjusting for several confounders, including dietary intake and body mass index. In 40,000 Chinese Singaporeans, SSB intake increased the risk of type 2 diabetes, independent of lifestyle and dietary confounders [43]. The authors also found that 5-year weight gain was an effect modi er of the association between SSB intake and type 2 diabetes. Finally, data from a cohort study from Thailand [39], observed that SSB intake increased the risk of type 2 diabetes more than two times, independent of demographic and nutritional characteristics. However, positive results were found for women only.

Association between SSB intake and obesity
The effect of SSB intake on obesity status was investigated by four studies with three of them presenting positive associations [46]- [48] and one with null results [49] (Table 3). In the Framingham Heart Study [46], the consumption of one or more soft drinks per day increased the odds of obesity by 31% (OR = 1.31; 95%C.I. 1.02; 1.68), while the consumption of two or more soft drinks per day increased the odds by 50% (OR = 1.50; 95%C.I. 1.06; 2.11). Data from the Korean National Health and Examination Survey [47] showed that men and women who have consumed one or more portion of SSB per day presented higher odds of obesity when compared to those who consumed less than one portion per day. Both studies adjusted analyses for several confounders, but only the rst one adjusted for dietary characteristics. Garduno-Alanis et al. [48], analyzing data from the Health, Alcohol and Psychosocial factors in Eastern Europe, found that SSB intake increased the risk of obesity, independent of several confounders, including dietary characteristics as well as medical history of CVD, cancer and diabetes. Interesting to notice that results were only signi cant in Poland and Russia, but not in Czech Republic.

Association between SSB intake and coronary heart disease
Three longitudinal studies included in our systematic review hypothesized that SSB intake is a risk factor for CHD (Table 3). One study included only women in the sample [50], one study enrolled only men [51], and the other one included both men and women in the sample [52]. The investigation that analyzed data for men and women from the Harvard Pooling Project observed that the intake of more than one portion of SSB per day increased the hazard of CHD (HR = 1.08; 95%C.I. 1.02; 1.14). Strati cation by sex revealed that results were signi cant for men but not for women [52].
In addition, the consumption of two or more portions of SSB per day increased almost 30% the risk of CHD in 90 thousand women from the Nurses' Health Study [50]. The authors adjusted the analyses for several confounders, including lifestyle and nutritional characteristics, and health status. De Koning et al. [51], in turn, found that more than one portion of SSB increased by almost 20% the risk of CHD (RR = 1.19; 95%C.I. 1.11 -1.28) in more than 40 thousand men from the Health Professionals Follow-up Study. The results were also adjusted for several confounders.

Association between SSB intake and stroke
Five studies investigated the association between SSB intake and stroke [53]- [57] (Table 3). Larsson et al. [54], analyzing data from the Cohort of Swedish, showed that the consumption of two or more portions of SSB was associated with higher risk of stroke (RR = 1.19; 95% C.I. 1.04; 1.36). Nevertheless, when they strati ed results by sex, association was signi cant for men only (RR = 1.22; 95% C.I. 1.02; 1.36). The authors adjusted the results for several confounders including dietary characteristics and waist circumference, suggesting that the effect of SSB on stroke may be independent of central adiposity.
Bernstein et al. [57] and Pacheco et al. [56], on the other hand, found signi cant results for women only. The rst study observed that the consumption of one or more portion of SSB increased the risk of stroke only in women (RR = 1.14; 95% C.I. 1.02; 1.27), after 28 years of follow-up [57]. Similarly, one or more portion of SSB per day (more than 355g) increased the hazard of stroke in more than 100,000 women, after adjustment for several confounders, including dietary characteristics and body mass index [56].
Nevertheless, two longitudinal studies did not observed effects of SSB intake on the risk of stroke after controlling for potential confounders in 40,000 men and women from Japan [53] or in almost 3000 men and women from US [55], after 18 and 10 years of follow-up, respectively.

Meta analysis
From the 27 studies selected in the systematic review, 26 were included in our meta-analysis as one investigation did not present OR, HR or RR in results. Figure 2 shows that the grouped effect of SSB on obesity and type 2 diabetes was positive. When obesity was the main outcome analyzed, we observed a positive pooled effect of SSB intake on the risk of obesity in the four studies included in the metaanalysis, with low heterogeneity among them (I 2 = 36%; p-value = 0.20). Results comparing low/moderate intake vs. no intake, and high intake vs. no intake (available for only two studies) showed that intake of SSB increased the risk of obesity by, on average, 17% (RR = 1.17; 95%C.I. 1.10 -1.25).
We also observed a positive pooled effect of low/moderate intake of SSB on the risk of type 2 diabetes in adults In Figure 3, we have the meta-analysis results for the association between SSB intake and CHD as well as stroke.

Discussion
Our systematic review and meta-analysis, which included longitudinal studies with adults, showed considerably consistent results on the association between SSB intake and NCDs. According to the prospective studies included in our review, arti cially sugar-sweetened beverages increased the risk of type 2 diabetes, obesity, CHD and stroke, with increased risks of low/moderate SSB intake ranging from 10% for stroke to 20% for type 2 diabetes.
Type 2 diabetes was the outcome presenting the strongest association with SSB intake according to our systematic review and metaanalysis. In fact, this is not the rst systematic review showing prospective association between SSB intake and type 2 diabetes.
Imamura et al. [11] have found that higher consumption of SSB increased the incidence of type 2 diabetes by 18%, with adiposity attenuating this effect (RR = 1.13; 95% C.I. 1.06 to 1.21). The authors also estimated that, over 10 years, two million people developed type 2 diabetes in US and almost 100 thousand in UK due to consumption of SSB [11]. Greenwood et al. [59] have found similar results: SSB intake increased the risk of type 2 diabetes by 20% (RR 1.20; 95%C.I. 1.12; 1.29), with the risk being attenuated after adjustment for BMI, indicating that adiposity is involved in the causal pathway. In our review, most studies which included type 2 diabetes as the outcome have adjusted results for BMI or waist circumference, and there were not differences among effect measures of studies adjusting or not adjusting analyses by measures of adiposity.
Results for the other outcomes (obesity, CHD and stroke) were less evident in our review, both due to the smaller number of selected studies as well as due to the smaller magnitude of the associations. In consonance with our ndings, previous published reviews have already demonstrated that SSB intake is positively associated with weight gain and cardiovascular diseases [12], [13], [60]. Malik et al. [60] have found that effects of SSB intake on weight gain is observed not solely in adults, but even in children. The authors showed that an increase of one-serving in SSB intake per day incremented weight gain in 0.12 Kg in adults and BMI in 0.05 Kg/m 2 in children [60]. In addition, two systematic reviews and meta-analyses of prospective studies have suggested that SSB consumption increases the risk of SSB intake may increase the risk of NCDs by different causal pathways. The rst one is through weight gain and excess adiposity caused by excessive daily energy intake [13]. But weight gain and adiposity are not the only causal pathway involved in this relationship, as most studies included in our review have adjusted the results for anthropometric measures, like BMI and waist circumference, and, even so, results remained signi cant. SSB intake also increases the risk of NCD via high glycemic load, which, in turn, leads to insulin resistance, in ammation and dyslipidemia [13], [63]- [65]. This pathway may help to better understand why the type 2diabetes was the outcome presenting the strongest association with SSB intake.
Another interesting result evidenced by our study is the lack of investigations on the longitudinal associations between SSB intake and NCD from LMIC. None of the selected studies were conducted in low-income countries and only three in middle-income countries, which is troublesome since LMIC are the settings with the highest burden of deaths attributable to SSB consumption [14]. Estimates for global consumption of SSB with data from 2010 indicated that Latin America and Caribbean, a region comprised by LMIC, presented the highest rates of SSB intake [8]. Moreover, despite there is current evidence of decreasing in SSB intake, trends information on SSB consumption are mostly available for high-income countries with few published data from LMIC.
There is an increasing concern on the standardization of food classi cation to assess dietary intake in epidemiological studies [66], [67].
Despite the concerning about that, standardized and harmonized protocols to assess SSB intake have not been developed so far. It becomes more evident when we look on the way SSB intake was assessed and categorized in the included studies. While some investigations evaluated only the frequency of consumption, not considering the portion size, other classi ed consumption in grams, milliliters or ounce.
But even those studies taking portion size into account, the cut-off to de ne high consumption also varied. In addition, the reviewed studies included different types of SSB, since solely soft drinks up to beverages with sweetened milk, tea, or coffee. It is important to de ne a standard approach to assess SSB intake in epidemiological studies to make comparisons easier as well as to objectively de ne a cut-off of SSB intake that indicates health risks at both individual and populational levels.
Finally, we should also consider that more than a half out of the 27 investigations included in our review followed-up the enrolled individuals for less than 10 years. Despite scarce evidence on how long takes for environmental factors, such as dietary intake, induce NCDs [68], [69], studies following individuals for short periods may not have enough time to address the cumulative exposure of SSB intake associated with the onset or the clinical detection of the diseases analyzed. Considering the long latency period of NCD [70], longitudinal studies should have to consider following the sample for longer periods, which is a challenge in the research practice.
Our study has limitations that are important to be mentioned. The rst limitation is the possibility of residual confounding, which is common in observational studies [71]. Even though the included studies have adjusted for several confounders like socioeconomic, dietary and body composition characteristics, errors in measuring these characteristics can exist resulting in residual confounding in our meta-analysis. In addition, gray literature, such as academic thesis and conference papers, is more likely to present null results and have not been included here, which can increase the risk of publication bias [72].
As strengths of our systematic review, we can indicate the inclusion of only longitudinal studies, which decreases the risk of reverse causality in the associations investigated. In addition, our systematic review focused in more than one NCD and nd similar results for type 2 diabetes, obesity, CHD and stroke, despite differences in the number of included studies for each outcome. This aspect of our study helps to increase the consistency in the association between SSB intake and NCDs.
In conclusion, longitudinal studies with adults demonstrated that consumption of SSB intake appears to increase the risk of type 2 diabetes, obesity, CHD and stroke. The evidence was stronger for type 2 diabetes, and more studies including obesity, CHD and stroke as the main outcome are needed. In addition, longitudinal studies about the association between SSB intake and NCD in LMIC are crucial to better understand this association in these settings. Political programs and actions are important to be developed to reduce the SSB intake and, by consequence, decrease the worldwide incidence of NCD.

Declarations Funding information
This study was supported by Brazilian National Research Council (CNPq) (grant number 442801/2019-0).

Con ict of interest
The authors declare they do not have any con ict of interest.
Availability of data and material Not applicable.
Author's contribution LPS: Data analysis, drafting, critical revision and editing of the manuscript. RMB: Proposed the idea, supervision of the project and critical revision of the manuscript. FMD and APM: literature search, data analysis and critical revision of the manuscript. DPG: supervision of the project, literature search and critical revision of the manuscript.

Consent for publication
All authors have read and approved the nal version of the manuscript.

Figure 2
Pooled effect for the association between sugar sweetened beverages intake and obesity and type 2 diabetes among adults.