Abstract
Background
With the rising use of artificial sweeteners as sugar substitutes, concerns regarding their impact on cardiovascular health have emerged. Artificially sweetened beverages are the primary source of diet sweeteners, but despite approval by national food agencies, evidence of their association with cardiovascular events has not been conclusive. Our Meta-Analysis assessed the relationship between artificially sweetened beverage consumption and long-term outcomes of cardiovascular events in extended follow-up cohorts.
Methods
Medline, Embase, and Cochrane databases were systematically searched for cohort studies investigating the incidence of all-cause mortality, cardiovascular mortality, stroke, and coronary heart disease among individuals with high consumption of ASB compared to minimal or no consumption. Pooled event hazard ratios with 95% confidence intervals were calculated using a random-effects model in R software, with heterogeneity assessed via I² statistics.
Results
We included twelve prospective cohorts comprising 1,224,560 patients. Analyses were conducted on patient groups with data adjusted for co-founding, such as dietary factors and comorbidities. One or more daily dose of Artificially sweetened beverages was significantly associated with a higher risk of all-cause mortality (HR 1.14; 95% 1.03 to 1.26; p < 0.01;), Cardiovascular mortality (HR 1.29; 95% 1.1 to 1.53; p < 0.01), and stroke (HR 1.15; 95% 1.01 to 1.32; p = 0.04;).
Conclusion
In this meta-analysis, we found a significant association between high consumption of ASBs and increased incidence of ACM, CVD, and stroke, highlighting potential long-term cardiovascular implications.
Graphical abstract
Introduction
With the growing obesity epidemic , artificial sweeteners (AS) have risen as a “healthier” option when compared to added sugar in food and beverages due to their low-calorie properties and potential benefit in weight reduction ; their primary source globally are the artificially sweetened beverages (ASBs). ASBs are drinks that use artificial sweeteners instead of regular sugar or other caloric sweeteners. These synthetic substitutes offer the same sweet taste as sugar but far fewer calories. However, many concerns have arisen due to the lack of assertive evidence confirming their benefits and even lesser evidence proving their safety in the long timespan.
Previous studies found no strong correlation between using artificial sweeteners and higher cancer risk, , , while some studies found weak correlations with incident diabetes compared to the non-consuming population. However, cardiovascular risk factors favor using ASBs when compared to added sugar. , , , Thus, many studies have been done, but no conclusive findings were found, especially for the outcomes of cardiovascular events that may arise from the use of ASBs.
Many meta-analyses about AS have been made, but current literature lacks a meta-analysis that analyzes ASBs in prospective cohorts using hazard ratio, for long-term follow-ups, and the association of ASBs with cardiovascular outcomes such as stroke, myocardial infarction, and cardiovascular death. Therefore, we performed a meta-analysis with considerably more studies than previous systematic reviews and using pooled hazard ratios to assess the relationship between ASB consumption and long-term cardiovascular outcomes in long-follow-up prospective cohorts.
Methods
This systematic review with meta-analysis was registered in the International Prospective Register of Systematic Reviews (PROSPERO) under protocol number CRD42024558091. This study was designed and conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) reporting guideline.
Study Eligibility
Inclusion in this meta-analysis was restricted to studies that met all the following eligibility criteria: (1) prospective cohorts; (2) studies providing comparable data for the outcomes; (3) studies providing data of patients that drink ASBs in comparison to those who don’t or have minimum consumption; (4) studies available for review in English and full text. We excluded from this analysis studies: (1) no comparison group; (2) mixing the use of ASBs with other beverages in the same group; (3) data available in a non-comparable measure; (4) not having the outcomes being evaluated.
Search Strategy
We systematically searched PubMed (MEDLINE), Embase, and Cochrane Central Register of Controlled Trials (CENTRAL) from inception to June 2024 with the following search terms: “Artificial Sweeteners”, “Artificial Sweetener”, “Cardiovascular Risk”, “Mortality”, “Death”, “Cardiovascular Disease”, “Artificially sweetened”, “Diet drinks”, “soft drink” and “Sugar-sweetened” as according to Cochrane. In addition, the references of included studies, reviews, and meta-analyses were evaluated for additional studies. The systematic search was conducted independently by two different reviewers (L.M.B. and M.L.R.D.), and disagreements were resolved by arbitration with a third author (I.Q.C.N.).
Data Extraction and Outcomes of Interest
We extracted data for baseline information such as body mass index and duration of follow-up, sample size, beverage consumption assessments, beverage consumption types consumed, covariates evaluated, database origin, and comparable outcomes. The outcomes were (1) All-cause mortality (ACM), (2) Cardiovascular Mortality (CVM), (3) Stroke, (4) Myocardial Infarction (MI), and (5) Coronary Heart Disease (CHD). Four authors independently extracted the data (I.Q., C.F.M., A.H.T., and M.L.R.D.) following predefined search criteria and quality assessment. Disagreements were resolved by consensus and with a senior author.
Any differences related to the data extraction were resolved by rechecking the full text of the study or by discussion among authors. We collected the reported outcomes hazard ratios (HR) and 95% Confidence Intervals (CI). Each study had a multivariable model of correction for each outcome, and we synthesized the analysis using the most statistically similar and adjusted results from each study ( Table 1 ).
| Author | Year | Nation | Database origin | Mean Follow up/Follow up time | Total number (n) | Female (n) | Mean age/age range | ASB patient exposed (n) | Co-founders adjusted | Primary outcome |
|---|---|---|---|---|---|---|---|---|---|---|
| Anderson | 2020 | UK | UK Biobank (22 centers) | 7 years | 209,535 | 122,216 | 40–69 | 40,791 | Model 4: Sex, Age, ET, IC, HQ, PA, SB, BMI, SS, AI, TSI, TFI, DD, HT and TSB. | All-cause mortality |
| Gardener | 2012 | US | Northern Manhattan Study (NOMAS) | 10 years | 2,564 | 1,641 | 69 | 2,111 | Model 4: TFI, AHM, BP, BS, HDL, LDL, TGC, WC, DD, CD and PVD. | Vascular events |
| Mulle | 2019 | 10 European Countries ⁎⁎ | European Prospective Investigation into Cancer and Nutrition (EPIC) | 8 years | 451,743 | 321,081 | 50.8 | 252,357 | Model: BMI, PA, ES, SS, CP, MS, DD and TFI. | All-cause mortality |
| Naomi | 2023 | EU | Lifelines Cohort Study | 9.8 years | 118,707 | 70,764 | 45 | 66,332 | Model 3: Age, Sex, ES, AI, SS, PA, SB, BMI, DD and TEI. | All-cause mortality |
| Pacheco | 2024 | US | Nurses Health Study (NHS) / Health Professionals Follow-Up Study | 15 years | 105,148 | 65,73 | 48.80 | 79,922 | Model: Age, Race, ET, PA, SS, AI, MS, BP, TFI, HCT, BMI, HT and FH | Cardiovascular disease |
| Pase | 2017 | US | US (Framingham Heart Study Offspring) | 43 years | 3,029 | 1,587 | 62 | 1,686 | Model 3: Sex, Age, TEI, BP, AHM, CD, TC, DM, HDL, AF, LVH, HT and WC. | Stroke and dementia |
| Vyas | 2015 | US | US (Womenˋs Health Initiative Observational Study) | 3 years | 59,614 | 59,614 | 62.80 | 21,277 | Model 4: Age, Sex, SS, BMI, IC, ES, DM, BP, HCT, AI, TEI, DD, PA, HT and TC. | Cardiovascular disease |
| Zhang | 2020 | US | US (National Health and Nutrition Examination Survey) | 7.9 years | 31,402 | 15,560. | 47.08 | 22,715 | Model 3: Sex, Age, Race, ES, AI, SS, PA, HT, BMI, HCT, DM, CD, CA and TSB. | All-cause mortality |
| Konning | 2012 | US | Health Professionals Follow-up Study | 22 years | 42,883 | 0 | 40 –75 | 23,324 | Model: Age, Sex, SS, PA, AI, WV, DD, TEI, BMI, HCT, HT and DM. | Coronary heart disease |
| Malik | 2019 | UK | Nurses Health Study (NHS) / Health Professionals Follow-Up Study | 15 years | 118,363 | 88,520 | 59.85 | 84,976 | Moddel 2: Age, SS, AI, MS, HU, PA, FH, ET, HT, HCT, BMI, DD and TEI. | Cardiovascular disease mortality |
| Mossavar | 2019 | US | US Womenˋs Health Initiative Observational Study) | 11.9 years | 81,714 | 81,714 | 63,6 | 29,397 | Moddel 4: Age, Sex, ES, Race, DM, CD, HT, HCT, SS, AI, CHD, CI, AU, DD and MET. | Cardiovascular disease |
| Fung | 2009 | US | Nurses Health Study (NHS) | 24 years | 88,520. | 88,520. | 30–55 | 54,510 | Model 2: Age, SS, AI, FH, MS, PA, AU, HU, HT, HCT and DD. | Coronary heart disease |
| TOTAL | – | – | – | 10.82 years | 1,313,222 | 70,724,347 | – | 679,398 | – | – |
Data for the outcomes was divided into three groups: 2 or more daily dose intake, one or more daily dose intake, and moderate intake (Supplementary Table 1), which was a mixed intake from different analyzed intakes of each study compared to a minimum consumption group, it’s important to point out that the two or more daily intake group was a subgroup analysis of the main group reporting only studies that had two or more daily intake as their maximum intake. All studies defined minimum consumption as one or less per month or no consumption. No “low intake” group or similar was analyzed as most studies diverged in the intakes of smaller intake groups, making the comparison between studies clinically irrelevant. A dose was defined as a 12-oz serving, similar to all studies, except in European studies, where a dose is usually considered 11.2 oz.
Data Analysis
Exposure effects for binary outcomes were compared using pooled HR with 95% CI using the generic inverse variance method. Heterogeneity was examined with Cochrane’s Q test and I² statistics. Considering the expected heterogeneity in effect measures inherent to each study in this meta-analysis, the restricted maximum likelihood random effects model was applied to all outcomes. Sensitivity analysis using leave-one-out plots was made to explore heterogeneity and show the dependence of the results on each study. All statistical analyses were conducted in the R software (version 4.3.2). All the analysis was calculated, re-evaluated, and corrected by three investigators (I.Q., M.L.R.D., and V.A.).
Quality Assessment
We assessed the methodological quality of all included studies under the Cochrane Collaboration’s tool. We also evaluated the risk of bias in all studies using the Newcastle Ottawa Scale (NOS). Disagreements were resolved by consensus.
Results
Study Selection and Baseline Characteristics
The database search yielded 2,529 potential studies from three sources (EMBASE, MEDLINE, and CENTRAL) and backward snowballing from reviewed studies. After removing duplicate records and studies with exclusion criteria based on title and abstract review, 53 results remained. Those 53 studies were thoroughly examined for the inclusion criteria, and 15 were included in the quantitative review (Supplementary Figure 1). Six of these studies provided data for ACM, five for CVM, four for stroke, and four for CHD. Only one study provided data for MI; therefore, it wasn’t comparable. We used different population and cohort databases to compare outcomes and exclude potential patient overlapping ( Table 1 ). Studies from different databases had different target populations, follow-up periods, and baseline risk factors. The cohorts had no significant differences that weren’t statistically adjusted. All studies had a similar multivariable correction for co-founding factors. The total population of analyzed studies was 1,224,560 patients with a mean follow-up time of 10.82 years.
Pooled analysis of all studies
The meta-analysis showed that high consumption doses are associated with a higher hazard of cardiovascular outcomes. The pooled analysis of 1 or more daily doses intake showed that patients had a statistically significant higher hazard for ACM, CVM, and stroke. The pooled HR for ACM was (HR 1.14; 95% 1.03 to 1.26; p = 0.009; I²=79%; Fig. 1 A), CVM was (HR 1.29; 95% 1.1 to 1.53; p = 0.002; I²=63%; Fig. 1 B), and stroke was (HR 1.15; 95% 1.01 to 1.32; p = 0.036; I²=25%; Fig. 1 C).
