ABSTRACT
Conclusions remain controversial between the consumption of sugar and artificially sweetened beverages (SSBs and ASBs) and mortality.
We systematically searched the PubMed, Embase, Cochrane Library and Web of Science databases from their inception date to 1st January 2020, prospective cohort studies researching the mortality risk and SSBs or ASBs consumption were included. Random effects meta-analyses and dose–response analyses were performed to measure the association. Subgroup analyses and sensitivity analyses were further performed to explore the source of heterogeneity. Publication bias was assessed by Funnel plots and Egger’s regression test.
Across all 15 cohorts, 1211 470 participants were included. High SSB consumption was associated with a higher risk of all-cause mortality (hazard ratio [HR], 1.12; 95% confidence interval [CI], 1.06–1.19, P < 0.001; and cardiovascular disease [CVD] mortality [HR 1.20, 95% CI, 1.05–1.38, P < 0.001]), and high ASBs consumption showed similar result (HR 1.12, 95% CI, 1.04–1.21, P = 0.001 for all-cause mortality and HR 1.23, 95% CI, 1.00–1.50, P = 0.049 for CVD mortality), both showed a linear dose–response relationship.
High consumption of both ASBs and SSBs showed significant associations with a higher risk of CVD mortality and all-cause mortality. This information may provide ideas for decreasing the global burden of diseases by reducing sweetened beverage intake.
Background
A previous study has reported that highly energy dense and sweet products account for a larger portion of our diet.1 Sugar-sweetened beverages (SSBs), in particular, were considered to play a key role.2,3 SSBs are defined as beverages with added caloric sweeteners, such as high-fructose corn syrup, sucrose or fruit juice concentrates. Excessive intake of SSBs has been associated with a higher health risk of weight gain,4 type 2 diabetes,5–7 metabolic syndromes8,9 and cardiovascular diseases (CVDs),10,11 in which high glycemic load and obesity have been considered to play an important role.12 However, previous meta-analyses13,14 mentioned a non-significant association between SSBs consumption and all-cause mortality. Moreover, a recent prospective study reported that there is no significant association between SSB consumption and cardiovascular mortality15 in the elderly population, in contrast to a previous study.16
In contrast to SSBs, artificially sweetened beverages (ASBs) refer to all types of low-calorie or diet beverages using artificial sweetener, such as aspartame, saccharin and xylitol, which provides a sweet taste but with lower caloric intake.17 ASBs are thought to potentially reduce excess energy intake18 and may be an ideal alternative for sweet-tasting beverage lovers. Several prospective studies have suggested that ASB consumption is not associated with several kinds of cancer.19,20 Nonetheless, concerns have been raised throughout their use. As early as 1997, a case-control study21 reported a higher risk of childhood brain tumors associated with aspartame consumption. Results from recent cohort studies have also indicated that ASBs, similar to SSBs, are related to a higher risk of obesity, diabetes,22–24 biliary tract cancer25 and insulin resistance.26 Animal experiments have also indicated that excessive ASB consumption may increase the incidence of kidney dysfunction.27,28
Though the health risk of ASB and SSB consumption has been continuously indicated by animal experiments and population-based studies,7,28–30 the significance of the findings has been inconsistent, and the association of ASBs and SSBs with all-cause mortality13,14 and specific diseases remains controversial. To assess the association between the habitual consumption of sweetened beverages and all-cause mortality and the difference between specific diseases, a systematic review and meta-analysis of prospective cohort studies were conducted to obtain a comprehensive and reliable result.
Methods
The review was registered in the PROSPERO (international prospective register of systematic reviews; https://www.crd.york.ac.uk/PROSPERO; ID: CRD 42019140581) and written based on the preferred reporting items for systematic reviews and meta-analyses statement.31 Patients or the public were not involved in the design, conduct, reporting or dissemination of this research.
Search strategy
We searched four main databases (PubMed, Embase, Cochrane Library and Web of Science) with keywords including ‘sweetened beverages’, ‘SSB’, ‘ASB’, ‘soft drinks’, ‘carbonated beverages’, ‘energy drinks’ and ‘mortality’, combined with ‘Title/Abstract’ and ‘MeSH/Emtree’ from their inception date to 1st January 2020. The language was limited to ‘English’.
Selection of articles
Original studies were considered eligible if they met the following criteria: (i) prospective population-based studies of adults; (ii) the type of beverage was consistent with the definition of ASBs or SSBs (as provided earlier); soft drinks were classified as SSBs unless the author indicated artificially sweeteners were used, while studies merely about milk, juice, tea and alcohol were excluded; (iii) the studies included more than one beverage consumption group and a reference group; (iv) the primary endpoint was all-cause mortality, and the secondary endpoint was cause-specific mortality. Studies could be included if at least one outcome was reported and (v) the adjusted hazard ratios (HR), odd ratios (OR), or relative risk (RR) and 95% confidence interval (CI) were presented or necessary original data could be extracted; studies with no available full texts, animal studies, conference abstracts, reviews, case reports or abstracts were excluded.
Two investigators (LHY and YH) independently screened all titles and abstracts to select the studies that met the inclusion criteria. When opinions were controversial, a third reviewer (LHY) would resolve this controversy and reach a consensus with the group.
Data extraction and quality assessment
The basic information for the original articles included the following: name of the cohort and the first author, year of publication, location, average age, follow-up time, type of beverage, size of sample, group assignment and adjusted covariates. Additionally, if the participants of original articles were divided into three or more groups, the adjusted HRs or RRs between the different ASB or SSB consumption groups and reference group were extracted separately. For studies using calorie intake as a grouping criterion, we converted the boundary level through standard calories or beverage energy as much as possible.32 When several endpoints of CVD or cancer mortality were reported, estimates of the endpoints with the most coverage were extracted. If a study did not report the HR or RR, we extracted the original data to calculate the pooled effect under fix effect model. If multiple articles were based on the same cohort, only the study with the longest follow-up time, the largest number of cases, the most adjusted covariates, and/or the maximum information was included in data synthesis. When several endpoints of CVD mortality or cancer mortality were reported, the endpoints with the most coverage were used (hierarchy: CVD mortality, mortality from specific CVD endpoints; cancer mortality, mortality due to specific types of cancer).
The Newcastle-Ottawa Scale (NOS)33 was used to assess the risk of bias of the included cohort studies. Articles that received seven or more points were considered high quality. Covariates were divided into eight parts: age and sex, race and socioeconomic status, physical activities, smoking and drinking, body mass index, diet factors, clinical features and other types of beverages. Subgroup analysis was also performed to explore the source of bias.
Data synthesis
Forest plot was performed to synthesize HRs and 95% CI across studies. We extracted estimates between the highest consumption group and endpoints, and an inverse-variant weighed random effect model method was applied in the synthesis process. The potential heterogeneity was presented by the I2 statistics; when I2 was 0–25%, 26–50%, 51–75% or >75%, the pooling outcome was considered to represent non-significant, low, moderate and high heterogeneity,34 respectively. Subgroup analysis was performed to explore the source of heterogeneity, with the potential source of heterogeneity including the definition of exposure, grouping criteria, study sites, number of participants, sex ratio, follow-up period, quality of study, baseline age and adjustment of clinical features. The publication bias was evaluated by Egger’s linearity regression test and funnel plot. Sensitivity analysis was further conducted by sequential stepwise omitting of one study to evaluate the influence of a single study on the overall effect estimate. The data synthesis was performed by STATA 14.0 (College Station, Texas 77845, USA, Serial number: 401406267051), and a P value < 0.05 was considered to be statistically significant except where otherwise specified.
As the highest categories vary between studies, we further performed dose–response analysis. The dose–response meta-analysis was performed using the method described by Greenland and Longnecker35 and Orsini.36 For studies containing at least three consumption groups, the number of cases and total cases or person-year data in each group and the natural log of corresponding estimates were used, and dose was defined as the median value of each group. For the infinite interval, the difference between the median and the bound of adjacent interval was considered to be consistent with the adjacent interval. Doses were all converted to milliliters, and 250 ml/day of both SSBs and ASBs was considered a measurement unit, as described by the largest and most up-to-date study.
Results
From the above four databases, 2859 studies were retrieved. Two reviewers (LHY and YH) screened the titles and abstracts and chose 84 full-text articles for the assessment. Ultimately, 14 studies reporting 15 cohorts were included,15,16,29,32,37–46 14 of which were considered high quality according to the NOS (Supplementary Table S1). The screening process and search strategy in PubMed are presented in Fig. 1.
Flow diagram of including studies.
In total, 1211 470 adult participants were included in this study, with 137 310 death cases; the sample size of a study42 was <1000. Ten cohorts29,32,37,38,40,42–45 reported association between SSBs and mortality outcome, 5 cohorts32,37,39,41 reported an association of ASBs and 2 cohorts15,16 only reported an association between SSB consumption and CVD mortality. Most studies used structured food-frequency questionnaires to evaluate beverages intakes (two repeated and seven validated), one cohort used a 1-year recall of food42 (baseline only), one used a face-to-face questionnaire41 (baseline only) and one used a modified 7-day diet history method45 (validated). Covariates such as age, sex, behavior, morbidity and diet, were adjusted differently in original articles by a Cox regression model or matched cohort.42 (As shown in Supplementary Table S2).
SSB consumption and mortality
The pooled multivariable adjusted HRs comparing the highest SSB consumption with the reference group was 1.12 (95% CI, 1.06–1.19, P < 0.001; Fig. 2) for all-cause mortality, 1.20 (95% CI, 1.05–1.38, P < 0.001; Fig. 2) for CVD mortality, 0.96 (95% CI, 0.84–1.10, P = 0.564; Fig. 2) for cancer mortality and 1.22 (95% CI, 1.01–1.47, P < 0.001; Fig. 2) for other cause of mortality. However, high heterogeneity was observed between studies (I2 = 74.3% for all-cause mortality, I2 = 76.1% for CVD mortality, I2 = 86.4% for cancer mortality, I2 = 87.0% for other cause of mortality). Subgroup analysis showed that studies of soft drinks contributed to the heterogeneity and showed a non-significant association with all-cause mortality. High-quality studies of CVD mortality showed non-significant between-study heterogeneity, and studies with a long follow-up period (>20 years) showed a higher risk of CVD mortality. Among cohorts reporting cancer mortality, the follow-up period and definition of beverages contributed to the high heterogeneity, while the studies with a larger sample size (>10 000 in total) showed lower heterogeneity (Supplementary Table S3). For other causes of mortality, different definitions of the endpoint were the main source of heterogeneity. Sensitivity analysis showed that omitting each study did not alter the significance of all results (Supplementary Fig. S1). In addition, the funnel plot showed a visually symmetric result (Supplementary Fig. S2), and the Egger’s test suggested that there is no significant publication bias in the included studies in terms of the highest SSB consumption and mortality (P = 0.643 for all-cause mortality, 0.907 for CVD mortality and 0.205 for cancer mortality, Supplementary Fig. S3).
Forest plot of the association between high consumption of SSBs and mortality.
Based on the dose–response analysis, we observed a linear relationship between SSB consumption and all-cause and cause-specific mortality. Each 250 ml/day serving of SSB was associated with a significantly higher risk of all-cause mortality (HR 1.05, 95% CI, 1.03–1.08, P for non-linearity = 0.8943) and CVD mortality (HR 1.13, 95% CI, 1.06–1.20, P for non-linearity = 0.5678). Moreover, consistent with the synthesized results of the highest SSB consumption group, there was a non-significant relationship between SSB consumption and cancer mortality (HR 0.96, 95% CI, 0.89–1.05, P for non-linearity = 0.3309) and other causes of mortality (HR 1.06, 95% CI, 0.84–1.32, P for non-linearity = 0.1407; Fig. 4).
ASB consumption and mortality
Across the four included studies reporting an association between ASB and risk of mortality in 5 cohorts, 5 cohorts reported all-cause mortality, 3 cohorts reported CVD and cancer mortality and 1 cohort reported other cause of mortality. The pooled multivariable adjusted HRs comparing the highest ASB consumption group with the reference group was 1.12 (95% CI, 1.04–1.21, P = 0.008; Fig. 3) for all-cause mortality, 1.23 (95% CI, 1.001–1.50, P = 0.049; Fig. 3) for CVD mortality, 1.04 (95% CI, 0.97–1.12, P = 0.224; Fig. 3) for cancer mortality and 0.91 (95% CI, 0.51–1.63, P = 0.752; Fig. 3) for other cause of mortality. The between-study heterogeneity was high for all-cause mortality (I2 = 79.3%, P = 0.001) and CVD mortality (I2 = 82.5%, P = 0.003). Considering the limited studies reporting cause-specific mortality, subgroup analysis, sensitivity analysis and publication bias were only performed for the primary endpoint. In the subgroup analysis, one of the hypothesized factors had a no significant effect on the heterogeneity among studies reporting all-cause mortality (Supplementary Table S4). A sensitivity analysis showed that omitting each study did not alter the significance of the result (Supplementary Fig. S4), the funnel plot was visually symmetric and Egger’s test suggested a non-significant publication bias (P = 0.966, Supplementary Fig. S4).
Forest plot of the association between high consumption of ASBs and mortality.
The dose–response analysis performed for the association between ASB consumption and mortality indicated a linear relationship (as shown in Fig. 4). Similar to SSB, every 250 ml/day of ASB was increased with a higher risk of all-cause mortality (HR 1.04, 95% CI, 1.01–1.07, P for non-linearity = 0.1015) and CVD mortality (HR 1.067, 95% CI, 1.001–1.136, P for non-linearity = 0.0683). However, no association was observed between ASB consumption and cancer mortality (HR 1.01, 95% CI, 0.991–1.02, P for non-linearity = 0.6585).
Dose–response relationship between SSB and ASB consumption and mortality.
Discussion
Main finding of this study
This meta-analysis demonstrated that high consumption of both SSB and ASB is associated with a 12% higher risk of all-cause mortality compared with the reference group, in which CVD mortality might be the main cause of the higher risk of death. Both kinds of beverages showed a linear dose–response relationship with all-cause mortality. These findings may help inform the control of excessive SSB or ASB consumption.
What is already known on this topic?
Recent studies in cancer patients have suggested that both ASBs18 and SSBs43,47 may not increase all-cause mortality. Narain et al.13 performed a meta-analysis of SSB consumption and risk of mortality, which showed that SSB consumption was related to a 13% higher risk of stroke and a 22% higher risk of myocardial infarction, but not related to all-cause mortality, which is partly in contrast to the present study. And a recently published dose–response analysis performed by Pei Qin et al.48 reported similar result to this study, while we further summarized studies of cause-specific mortality and found the CVD mortality may be the main contributor of mortality. Conversely, De Koning et al.11 indicated that intake of ASBs, unlike SSBs, does not increase the risk of coronary heart disease. The opposite conclusions are possibly due to the differentiation of ASBs and SSBs because of the low calories amounts contained in artificial sweeteners,49 considering the higher risk of weight gain with SSBs50, which was elucidated to be closely related to a higher risk of death51. Similarly, Schwingshackl et al.14 suggested that there was a nonlinear relationship between SSBs intake and risk of mortality, however, the result showed a 7% higher risk with increasing intake of SSBs up to 250 ml/day after nonlinear analysis,14 which also provided supported our conclusions. More interestingly, several studies22,37,50 reported that ASB consumption showed a slightly lower risk of mortality than the same amount of SSB consumption, whereas a large cohort study32 reported that only the highest ASB consumption group showed this result, and the author pointed out that ASBs had a higher threshold of the nonlinear J-shaped association with risk of mortality than SSBs, which is similar to our results.
What this study adds?
This meta-analysis compared the risk of death from high consumption of ASB and SSB, suggesting that the health risk of ASB is similar to that of SSB, which is mainly related to CVD. Based on the subgroup analysis, the heterogeneity may derive from the different definitions of soft drinks included in studies of high SSB consumption. The scope of a soft drink is broader than SSBs, including caffeinated drinks, juice and tea drinks, among others. Juice drinks have shown a slightly lower risk of disease,52 and caffeinated beverages may have a potentially protective mechanism against mortality,53–55 which may result in heterogeneity. Importantly, studies with a follow-up period of > 20 years presented lower heterogeneity and higher CVD mortality, which indicated that more high-quality, long follow-up period studies could enhance the evidence. For studies of ASB, none of the hypothesized factors significantly altered the heterogeneity between studies. Nevertheless, the influence analysis in this study suggested that the results of all analyzed groups were robust, and the results of the funnel plot and Egger’s test revealed no publication bias, indicating that the heterogeneity was not methodological. The results remain meaningful after adjustment of multiple covariates, which indicates that sweetened beverages may be an independent risk factor for death with the exception of explicit factors such as adiposity. According to included cohorts, both ASB and SSB consumption showed a relationship with cancer mortality and other cause of mortality, this part of the results may be biased, because some articles specified related diseases or cancer.32,40,41 Previous studies suggested that SSB consumption might increase the incidence of prostate cancer and breast cancer,56,57 while SSB might improve the survival rate of colon cancer patients.47 Obviously, these results are limited and more in-depth research is needed to provide more accurate nutrition advice, this is conducive to further research.
Limitations of this study
However, there are also some limitations of this study. First, the grouping varied between studies, and the subgroup analysis indicated that the different groupings might have contributed to the heterogeneity between studies to some degree. Second, different questionnaire designations and ignorance of the deflection of habits during follow-up periods may have resulted in over- or underestimation of drinking and vague beverage types thus failing to guarantee stable intake during the subsequent period, which may thereby increase the instability of the results. Third, although the systematic search was conducted for relevant articles, these studies mainly focused on several nations; the data from populous nations such as China and India were insufficient, potentially leading to a reduction in universality of the results. Additionally, the definitions of endpoints and follow-up periods varied in the included studies, resulting in an inability to determine the affected period of the outcome. Fourth, the findings from ASB consumption were more prone to bias due to confounding for its limited number of studies, and the reverse causation might also influence the effect size in low consumption group, as reported by Malik et al.37 and more original studies are needed. Finally, covariates were adjusted more or less in all included studies, but varied greatly, which may have contributed to the inconsistences among studies.
In conclusion, high consumption of both ASBs and SSBs showed significant associations with a higher risk of CVD mortality and all-cause mortality, with a dose–response relationship. This information may provide ideas for decreasing the global burden of diseases by reducing sweetened beverage intake.
List of abbreviations
ASBs, sweetened beverages; BMI, body mass index; CI, confidence interval; EPIC, European Prospective Investigation into Cancer and Nutrition; HPFS, Health Professionals Follow-up Study; HR, hazard ratio; NHS, Nurses’ Health Study; NOS, Newcastle-Ottawa Scale; RR, relative ratio; REGARDS, the Reasons for Geographic and Racial Differences in Stroke study; SSBs, sugar-sweetened beverages; SCHS, Singapore Chinese Health Study; TE, total energy; TLWCS, The Leisure World Cohort Study, USA, the United States of America; VITAL, Vitamins and Lifestyle Study; WHI-OS, Women’s Health Initiative Observational Study.
Authors’ contributions
T.W.S., H.Y.L. and H.Y.L. conceived of the study. HYL and HYL equally contributed to this study. XFD, XJZ, RFZ and HY contributed to data interpretation and the GRADE assessment. XJZ, HYL and HY contributed to the study protocol and wrote the article. YMM, QCK, ZSL and TWS revised the article.
Acknowledgements
All the authors contributed substantially to the work presented in this article. The authors declare that they have no competing interests and all data generated or analyzed during this study are included in this published article [and its supplementary information files]. We would like to thank Chengyang Wang, Bo Yuan and Yan Yan for their help with this study. We would also express our thanks to the West China Medical College of Sichuan University for providing us STATA14.0.
Funding
This study was supported by the United Fund of National Natural Science Foundation of China (grant no. U2004110); Leading Talents Fund in Science and Technology Innovation in Henan Province (grant no. 194200510017); Science and Technology people-benefit project of Zhengzhou (grant no. 2019KJHM0001); the special fund for young and middle-aged medical research of China International Medical Exchange Foundation (grant no. Z-2018-35); the integrated thinking research foundation of the China foundation for International Medical Exchange (grant no. Z-2016-23-2001) and the study of mechanism of Gabexate Mesilate in the treatment of sepsis and septic shock (grant no. 2019-hx-45).
Conflict of interest
The corresponding author, on behalf of all authors of this manuscript, declare that: we have no financial and personal relationships with other people or organizations that can inappropriately influence our work, there is no professional or other personal interest of any nature or kind in any product, service and/or company that could be construed as influencing the position presented in, or the review of, the manuscript entitled, ``Association between intake of sweetened beverages with all-cause and cause-specific mortality: a systematic review and meta-analysis''.
Hongyi Li, Medical Master
Huoyan Liang, Medical Doctor
Han Yang, Medical Master
Xiaojuan Zhang, Medical Doctor
Xianfei Ding, Medical Doctor
Ruifang Zhang, Medical Doctor
Yimin Mao, Medical Doctor
Zhangsuo Liu, Medical Doctor
Quancheng Kan, Medical Doctor
Tongwen Sun, Medical Doctor
References
Author notes
Hongyi Li and Huoyan Liang contributed equally to this article.
