Sugar- and Artificially-Sweetened Beverages and Cancer Mortality in a Large U.S. Prospective Cohort

In the United States, sugar-sweetened beverages (SSB) are the top contributors to added sugar in the diet (1, 2), and approximately 50% of adults consume one serving or more on any given day (3). Although energy and added sugar intake from SSBs decreased from 2003/2004 to 2015/2016 (4), approximately 60% of adults still exceed the 10% limits of total energy from sugar advised in the US Dietary Guidelines (1). In contrast, artificially-sweetened beverage (ASB) consumption increased in U.S. adults from 1999/2000 to 2009/2010 from 17.8% to 21.2% for females and from 13.9% to 19.0% for males (5, 6), with more recent consumption prevalence estimates from 2009 to 2012 NHANES of 34% and 28% in women and men, respectively (7).

The evidence that sugary drinks increase the risk of weight gain, overweight, and obesity is well established (8) from both randomized controlled trials and meta-analyses of prospective studies of weight change (9). Compared with calories as part of a solid meal, the extra calories consumed in soft drinks are not typically offset by reductions from other sources. SSB consumption may therefore increase cancer risk or mortality via excess body fatness, which is strongly related to several cancer types (10). Carcinogenic mechanisms potentially related to SSBs, both through—and independent of—body weight, involve perturbations of glucose and insulin homeostasis, inflammation and oxidative stress, and location of body fat (e.g., visceral adiposity; refs. 8, 11–15). Although generally not considered causally associated with weight gain (with some exceptions; refs. 16–18, 19), ASB consumption is higher in overweight and obese individuals, among whom calories saved may be compensated with solid food intake (20). Provocative data in mice suggest that alteration of the gut microbiome with artificial sweetener consumption may lead to metabolic disturbances (21–24), although a trial of aspartame or sucralose in humans showed minimal impact on human gut microbiota composition (25). Artificial sweeteners may also influence insulin resistance (26) and inflammation (27), but most available data in humans are from small and relatively short-term studies.

A growing number of epidemiologic studies have comprehensively examined SSB and ASB consumption in relation to incidence or mortality from all cancers or obesity-related cancers combined. In three of seven (28–34) prospective cohort studies, SSBs were associated with increased cancer-related mortality (28) or cancer incidence (29, 33). Two (28, 29) reported positive associations even after controlling for BMI, suggesting a BMI-independent role of SSBs in carcinogenesis, and in one, BMI was presumed to be in the causal pathway and thus not included in the statistical models (33). Associations of ASBs with these cancer outcomes were mostly null (28–30, 32, 33). Although generally regarded as safe by regulatory agencies worldwide, whether artificial sweeteners may promote development of cancer is controversial (35, 36), highlighting the need to examine associations of ASBs and cancer outcomes.

For individual cancers, a recent meta-analysis of 27 observational studies found significant associations between SSB intake and breast and prostate cancers, and positive associations with colorectal and pancreatic cancers; for ASB consumption, the study found a suggestion of increased risk for pancreatic cancer (37). Data are sparse and inconsistent for other cancers (37).

We examined associations of SSB and ASB consumption with mortality from all cancers combined, obesity-related cancers combined, and with 20 individual cancer types, in a large, prospective cohort of U.S. men and women. Associations were examined without and with statistical adjustment for BMI, and for both sexes separately and combined.

In 1982, 1,184,284 men and women ages 28 years or older were enrolled by American Cancer Society (ACS) volunteers in all 50 states, the District of Columbia and Puerto Rico into the Cancer Prevention Study-II (CPS-II; ref. 38), a prospective cohort study designed to examine risk factors for cancer mortality. At baseline, participants completed a 4-page confidential questionnaire on medical history and demographic and lifestyle exposures. Informed consent to participate in the study was implied by return of the self-administered questionnaire. The study was conducted in accordance with recognized ethical standards (e.g., Declaration of Helsinki). All aspects of the CPS-II study have been approved by the Institutional Review Board of Emory University.

Participants were excluded from this analysis if they had a personal history of cancer at baseline other than nonmelanoma skin cancer (n = 75,708), provided no information on consumption of any beverage type (n = 28,709), reported uninterpretable or potentially erroneous (>10 drinks/day) beverage intakes for SSBs or ASBs (n = 6,583), were men older than age 90 or women older than 95 at enrollment (n = 672; ref. 39), had diabetes at baseline to limit potential biases for diabetes-related cancers (n = 45,562), or reported only prior but not current consumption of either SSBs or ASBs (n = 92,273). This left 416,313 men and 518,464 women in the analytic cohort (total n = 934,777).

The baseline questionnaire included a grid that asked: “How many cups, glasses, or drinks of these beverages do you usually drink a day, and for how many years?”, with write-in reporting by frequency and duration. “Diet soda or diet iced teas” were considered ASBs, whereas “nondiet colas” and “other nondiet soft drinks” were considered SSBs. Participants were instructed to record “1/2” if they consumed the beverage less than once a day, but at least three times a week. If participants reported a change in consumption during the 10 years prior to enrollment, they were asked to record their previous intake of each beverage type. Never drinkers of SSBs and ASBs were defined as participants who wrote zero or left blank their current and previous consumption. Former drinkers who wrote zero or left blank the current intake amount and provided a non-zero amount for previous intake, were excluded as above.

Cause of death was obtained for 99.3% of all known deaths in CPS-II. From 1982 to 1988, cause of death was ascertained by personal inquiries made by volunteers to participants, which were verified by obtaining death certificates. From 1989 through end of follow-up on December 31, 2016, vital status and cause of death were obtained via linkage to the National Death Index (NDI). For this analysis, the outcomes of interest were underlying causes of death from all cancers combined [International Classification of Disease (ICD)-9 codes 140–209; ICD-10 codes C00–C97)], obesity-related cancers combined [esophagus (C15), stomach (C16), colorectum (C18–C21), liver (C22), gallbladder (C23–C24), pancreas (C25), post-menopausal breast (C50), uterus/endometrium (C54–C55), ovary (C56), kidney (C64–C66, C68), and multiple myeloma (C88, C90)], and from 20 specific cancer types where ≥25 deaths had accrued in the highest category of consumption (ICD codes: leukemia (C91–95), non-Hodgkin lymphoma (C96), brain (C70–C72), melanoma (C43), bladder (C67), prostate (C61), small intestine (C17), lung (C33–C34), and larynx/oral cavity/pharynx (C01–C14, C32), in addition to those listed in obesity-related cancers above). Women who reported hysterectomy (n = 132,292), uterine surgery (n = 6,056), or surgically induced menopause (n = 31,803) were excluded from uterine cancer mortality analyses; ovarian cancer mortality analyses additionally excluded women who reported a history of oophorectomy (n = 10,162). Breast cancer mortality models were restricted to post-menopausal women (n = 367,978) at baseline.

Follow-up time was calculated from completion of the baseline questionnaire until date of death or December 31, 2016, whichever came first. We censored follow-up at age 90 years for men and 95 years for women to limit potential bias from person-time of older individuals who died but did not link with the NDI (39). Cox proportional hazards models computed HRs and 95% confidence intervals (CI) for associations of SSB and ASB consumption with risk of cancer-related death, using categorical [never (ref), <1 drink/day, 1 drink/day, ≥2 drinks/day] and continuous (per 1 drink/day) exposures. All Cox models were stratified on single year of age, and adjusted for sex in sex-combined models. HRs were additionally adjusted for race/ethnicity, smoking, marital status, education, red and processed meat consumption, fruit and vegetable consumption, and alcohol consumption (categories defined in Table 1 and footnotes to Table 2). Missing data, ranging from 1% to 6% for most covariates, 9% for dietary variables and estrogen use and 17.5% for alcohol, were modeled using a missing indicator variable. SSB and ASB consumption were mutually-adjusted (drinks/day). Among women, analyses of breast (postmenopausal), ovarian, and uterine cancer mortality additionally included parity, age at menarche, estrogen use, oral contraceptive use, and age at first live birth; ovarian and uterine cancer mortality analyses additionally included menopausal status. Aspirin use, multivitamin use, physical activity, and family history of cancer were not included in the final models because of observed minimal impacts on the HRs. BMI (kg/m²) was modeled as a categorical variable, with the top and bottom 0.1% extremes (<15 and ≥46.4) set to missing (<18.5; 18.5–<22.5; 22.5–<25; 25–<27.5; 27.5–<30; 30–<35; 35+; missing). Sensitivity analyses: (i) excluded the first 2 years of follow-up, (ii) restricted the analysis to never-smokers, and (iii) stratified by BMI category [normal weight (18.5–<25), overweight (25–<30), obese (≥30), adjusting for continuous BMI within categories].

The Cox proportional hazards assumption, and interactions of SSBs and ASBs with sex and BMI, were tested by including a cross product term of continuous SSB or ASB (drinks/day) with continuous time, BMI, or sex in the final model, using the Wald Test. P values < 0.05 were considered statistically significant. P_trend in categorical models was evaluated using median values in each beverage category and modeling as a continuous variable. All analyses were performed using Statistical Analysis Software (SAS) version 9.4 (SAS Institute Inc.).

The data generated in this study can be requested according to CPS-II data access policies and procedures, available at www.cancer.org.

During a median of 27.7 (IQR, 17.4–34.3) years of follow-up, 135,093 CPS-II participants died from cancer (25.9% of all deaths). Baseline characteristics of study participants according to distribution of SSB and ASB intake are provided in Table 1. Compared with nonconsumers of SSBs, daily consumers tended to be younger, male, Black or other race/ethnicity, current smokers, to have higher BMI, to not drink alcohol, to have lower educational attainment, lower fruit and vegetable and coffee intake, and higher red and processed meat intake. In contrast, those who drank ASBs regularly were more often women, White, former smokers, and fruit, vegetable, and decaffeinated coffee consumers. ASB drinkers were more likely than SSB drinkers to have overweight or obese BMI.

SSB consumption was not associated with all-cancer mortality in men, women, or in both sexes combined (Table 2). However, consumption of ≥2 servings of SSB/day, compared with never consumption, was associated with a 5% increased risk of BMI-related cancers in men and women combined (HR, 1.05; 95% CI, 1.01–1.08; P_trend = 0.057), with 1% to 2% increased risk per drink/day in men (P_trend = 0.033) and in both sexes combined (P_trend = 0.009).

After including BMI in the model (Supplementary Table S1), the effect estimates for SSB consumption and BMI-related cancers combined was null.

Of the individual cancers, SSB consumption was positively associated with risk of colorectal cancer (HR for ≥2/day vs. never = 1.09; 95% CI, 1.02–1.17; P_trend = 0.011) and kidney cancer (HR, 1.17; 95% CI, 1.03–1.34; P_trend = 0.056) in men and women combined, and non-Hodgkin lymphoma (NHL) in men (HR, 1.25; 95% CI, 1.11–1.41; P_trend < 0.001), and men and women combined (HR, 1.16; 95% CI, 1.05–1.28; P_trend = 0.001). Inverse associations of SSB consumption with lung cancer were observed. With inclusion of BMI in the model (Supplementary Table S1), the association with colorectal cancer, kidney cancer, NHL, and lung cancer remained or were slightly attenuated.

After restricting the analyses to never-smoking men and women to limit potential residual confounding from smoking history (Fig. 1; Supplementary Table S2), SSBs were no longer inversely associated with esophageal cancer (HR, 1.12; 95% CI, 0.83–1.51; P_trend = 0.814) but remained statistically significantly inversely associated with lung cancer (HR, 0.81; 95% CI, 0.70–0.94; P_trend < 0.001). Associations with colorectal cancer and kidney cancers were further from the null and higher, whereas the association with NHL was attenuated and not statistically significant. With further control for BMI (Supplementary Table S3), the association with colorectal cancer was only statistically significant for men.

In models stratified by BMI (Supplementary Table S4), significant increased risk was limited to individuals with BMI ≥30 kg/m² for all BMI-related cancers combined, cancers of the larynx/oral cavity/pharynx and kidney but the P_interaction was not statistically significant.

ASB consumption (Table 3) was not associated with all-cancers combined, but a statistically significant increased risk of BMI-related cancers was observed. When BMI was included in the model, the association with obesity-related cancers was null and the association with all cancers combined was slightly inverse (Supplementary Table S5).

In models not including BMI, ASB consumption was associated with a statistically significant 16% increased risk of pancreatic cancer (4% increase for each serving/day) in men and women combined and separately. High vs. low consumption of ASBs was also associated with increased risk of uterine and gall bladder cancers, and continuous models suggested increased risk of small intestinal and postmenopausal breast cancers (Table 3). After controlling for BMI, however, all associations became null, except for pancreatic cancer in men and both sexes combined (Supplementary Table S5). Regardless of BMI-adjustment, ASB consumption was statistically significantly associated with lower risk of death from lung cancer. However, among never-smokers (Fig. 2; Supplementary Table S6), this association became null, and a positive association emerged for liver cancer and melanoma (men only). With further control for BMI among never smokers (Supplementary Table S7), the association with liver cancer became null, but the positive association with pancreatic cancer persisted.

In BMI-stratified models, a statistical interaction with ASB consumption was observed for prostate cancer and NHL, but all confidence intervals crossed 1.0 (P_interaction < 0.03; Supplementary Table S8).

In sensitivity analyses, excluding the first 2 years of follow-up had minimal impact on the HRs. Cox proportional hazards violations were noted for a limited number of primarily smoking-related cancers, as indicated in the footnotes to Tables 2 and 3. Analyses stratified by 10-year follow-up time period for SSBs and ASBs, for men and women combined, are provided in Supplementary Figs. S1 and S2. Associations of SSBs with lung and esophageal cancers were most strongly inverse during the first 10-year follow-up period.

Among almost one million U.S. adults followed for up to 34 years, regular consumption of SSBs was not associated with death from all cancers combined but was associated with an increased risk of mortality from obesity-related cancers. Controlling for BMI eliminated the association, suggesting that SSBs may lead to increased mortality risk through excess body mass. Individual cancer mortality associated with SSB consumption included colorectal cancer, kidney and bladder cancers, and NHL; after controlling for BMI, associations with colorectal cancer, kidney cancer, and NHL persisted, suggesting BMI-independent mechanisms. Similarly, ASB consumption was positively associated with mortality risk from BMI-related cancers but was no longer associated when controlling for BMI. Because the link between ASB consumption and body weight is inconclusive, this suggests confounding by BMI. However, a positive association of ASB consumption with pancreatic cancer mortality was robust to adjustment for BMI and in analyses limited to never smokers. While confounding cannot be ruled out, this association deserves further study.

Excess body fatness is causally related to at least 13 cancer types (10), and the evidence of an association of SSBs with weight gain, overweight, and obesity is considered strong (8). The 5% increased risk of mortality from obesity-related cancers we observed with ≥2 SSBs/d vs. never, unadjusted for BMI, is consistent with findings of the Melbourne Collaborative Cohort which reported a 10% increased risk of BMI-related cancers for >1 drink/day vs. never or <1/month (33). High vs. low intake quintile of sugary beverages was associated with adiposity-related cancer risk in the Framingham cohort, but only among participants with excessive central adiposity, with an HR, 1.59; 95% CI, 1.01–2.50 (P_trend = 0.057; ref. 34). Together, these studies provide support that increased risk of BMI-related cancer incidence and mortality with SSB consumption is due at least in part to BMI being in the causal pathway. Our study did not replicate BMI-independent positive associations of SSBs with risk of overall cancer mortality (28) or incidence (29), observed in two prior cohort studies, but is consistent with null findings for SSB intake and cancer mortality from the EPIC cohort (HR, 0.92; 95% CI, 0.78–1.09 for ≥2 glasses/day vs. <1 glass/month; ref. 30).

Death from several individual cancers was also positively associated with SSB consumption in this study, particularly for colorectal cancer and kidney cancer. The risk of death from colorectal cancer was 9% higher with consumption of ≥2 drinks/day in CPS-II participants, an association that remained after controlling for BMI, and strengthened among never smokers. Although many previous studies reported positive associations between SSB and colorectal cancer risk or mortality (29, 30, 33, 34, 40, 41), most were not statistically significant. Hodge and colleagues (33) reported a 28% (95% CI, 4%–57%; P_trend = 0.12) increased risk of colorectal cancer for ≥1/day versus <1/month of sugar-sweetened soft drinks, not controlling for BMI, in a study of Australian adults. A 2010 pooled analysis of 13 cohort studies including 5,604 incident cases and that adjusted for BMI (40) found no association (HR, 0.94; 95% CI, 0.66–1.32) with colon cancer risk comparing ≥550 g/day (about 18 oz) SSB to nonconsumption. A 2021 meta-analysis of four recent cohort studies reported nonstatistically significantly increased colorectal cancer risk for high versus low intake (HR, 1.18; 95% CI, 0.99–1.41; ref. 37); most studies included BMI in the model. Hur and colleagues (42) recently reported statistically significant two-fold (RR, 2.18) and three-fold (RR, 3.41) increased risks of early onset colorectal cancer with greater SSB consumption in adulthood and adolescence, respectively, with BMI adjustment. Our findings strengthen the available evidence of an association between SSBs and colorectal cancer mortality. SSB consumption may promote visceral adiposity (13, 14), independent of body weight, which then may alter adipokine secretion and cell-signaling pathways. In addition, among patients with stage III colon cancer who participated in an adjuvant chemoprevention trial, Meyerhardt and colleagues (43) found a higher risk of recurrence and mortality associated with higher glycemic load (calculated from quality and quantity of carbohydrates), which may contribute to hyperinsulinemia, cell proliferation, and inhibit apoptosis of micrometastases (43). High glycemic foods like SSBs may also contribute to diabetes risk and inflammation (11, 12, 44), both associated with colorectal cancer risk and mortality (45, 46).

Few studies have investigated the relationships between SSBs and kidney cancer risk or mortality (33, 47–49). In a recent analysis from the EPIC cohort, Heath and colleagues (48) reported no association of SSBs with kidney cancer mortality, including 356 deaths (HR, 0.97 per 100 mL/day). In the UK Million Women Study including 588 incident cases (49), a nonsignificant positive association was observed for intake of carbonated drinks (HR, 1.18; 95% CI, 0.89–1.58), although the study was unable to distinguish sweetener type. Likewise, a 2007 pooled analysis of 13 cohorts with 1,478 cases found no association of soda with renal cell cancer risk (HR, 1.11; 95% CI, 0.89–1.38), but that study was also unable to separate sugar- and artificially-sweetened soda (47). The Melbourne Collaborative Cohort Study, the only study to omit BMI from the model, observed a nonsignificant increased risk (HR, 1.48; 95% CI, 0.87–2.53) for ≥1/day vs. <1/month, with 146 cases. This study, with 3,429 cases, is the largest and first to report a statistically significant increased risk of kidney cancer mortality associated with SSB consumption. Because the HR was not strongly affected by controlling for BMI, and strengthened among never smokers, this relationship may involve a more direct mechanism as opposed to one solely mediated by BMI, including potentially carcinogenic chemical compounds from food coloring or packaging (11, 12, 29, 50–52). Further, SSB consumption may influence cancer risk or progression through metabolic perturbations and cell-cycle disruption (8, 12, 15). Finally, positive associations of SSB consumption with NHL mortality risk, while fairly robust to BMI adjustment, became null among never smokers, suggesting that these associations were susceptible to residual confounding by tobacco.

Counter to expectations, we observed inverse associations of SSB consumption with risk of esophageal and lung cancer mortality, which may be explained by reverse causality and/or residual confounding. Individuals experiencing symptoms of esophageal cancer may avoid carbonated beverages, especially close to the time of diagnosis. Inverse associations of SSB consumption and esophageal cancer were more pronounced within the first 10 years of follow-up compared with later periods, strengthening support for this potential source of bias. In addition, the association with esophageal cancer was null among never smokers, suggesting potential residual confounding by smoking. Stronger inverse associations with lung cancer mortality were also observed during the first 10 years of follow-up. Sugar sweetened beverage consumers were generally heavier, and smokers tend to be leaner. Among those with a BMI ≥30 kg/m², the relationship of SSBs with lung cancer mortality was null, adding support to these explanations.

ASBs contribute the largest share of low-calorie sweeteners consumed worldwide (53). In contrast to the relationship of SSBs with weight gain and obesity (8), the evidence that ASB consumption leads to excess energy intake and obesity is inconclusive and weak (16, 17). In this study, ASB consumers were heavier, and ASB consumption was positively associated with risk of BMI-related cancer mortality. However, this association was null with control for BMI, suggesting potential confounding by BMI.

Regarding individual cancers, mortality from breast cancer and endometrial cancer was increased only in models that excluded BMI. In contrast, we observed a significant positive association of ASB consumption with pancreatic cancer risk which persisted with control for BMI and in models of never smokers. Several explanations may exist for this finding. Diabetes is a pronounced risk factor for pancreatic cancer (54, 55), and individuals with diabetes would be more likely to consume ASBs. We excluded individuals with a history of diabetes at baseline to limit confounding by indication, but it is possible that undiagnosed diabetes influenced these results and that many ASB drinkers were diagnosed with diabetes after baseline. Residual confounding by BMI may also play a role, as ASB drinkers were heavier than nondrinkers in this study, and may have experienced longer-term excess body fatness. Higher BMI in early adulthood has been strongly associated with pancreatic cancer risk (56), and we were unable to control for early adulthood BMI. Support for a direct influence of artificial sweeteners on metabolism and potentially cancer-related pathways was found in a study of mice which reported that saccharin, sucralose, and aspartame altered intestinal microbiota in a manner that would lead to glucose intolerance (21); however, subsequent human studies have not been confirmatory (25). Three previous cohort studies (57–59) do not report associations of ASB consumption with pancreatic cancer risk or mortality, although data from the AARP cohort suggest increased risk (fourth quintile vs. first HR, 1.35; 95% CI, 1.03–1.77; fifth vs. first quintile HR, 1.25; 95% CI, 0.94–1.66; P_trend = 0.19; ref. 59).

Strengths and limitations of this study are worth noting. The large cohort of both men and women prospectively followed for a maximum of 34 years including over 135,000 deaths from cancer enabled a comprehensive investigation of the association of these beverages with risk of 20 cancer types, including analyses stratified by sex and BMI, and limited to never smokers. However, as this was an observational study, causality cannot be inferred. We also acknowledge that measurement error was likely present due to only one measurement of diet at baseline, particularly as artificial sweeteners used in the food supply change over time (53, 60). Missing covariate data may have contributed to loss of power to detect associations, and confounded associations. By examining mortality associations, this study examined whether these beverages were associated with incidence and/or with survival from cancer, which are impossible to disentangle from mortality outcomes alone. Finally, we were unable to distinguish adenocarcinoma from squamous cell esophageal cancers, or cancers of the gastric cardia from other gastric cancers, which limited our ability to distinguish a BMI-mediated association for these cancers.

In summary, we observed positive associations of SSB consumption and BMI-related cancers overall, partly mediated by BMI, and BMI-independent associations with colorectal cancer and kidney cancer among both men and women. We also observed a robust, BMI-independent positive association of ASBs with pancreatic cancer mortality. Although SSB consumption has decreased in the United States, intake remains high, and ASB consumption levels are increasing. Continued research on the impact of both beverage types with cancer risk and mortality is warranted to determine whether these associations are causal or confounded by other lifestyle factors, and whether they are mediated through BMI.

M.A. Guinter reports personal fees from Flatiron Health, Inc. and Roche outside the submitted work. No disclosures were reported by the other authors.

M.L. McCullough: Conceptualization, supervision, methodology, writing–original draft. R.A. Hodge: Formal analysis, visualization, writing–review and editing. P.T. Campbell: Conceptualization, methodology, writing–review and editing. M.A. Guinter: Conceptualization, writing–review and editing. A.V. Patel: Methodology, writing–review and editing.

The authors express sincere appreciation to all Cancer Prevention Study II participants, and to each member of the study and biospecimen management group. We would like to thank Dr. Allison Sylvetsky for her comments on the manuscript.

The publication costs of this article were defrayed in part by the payment of publication fees. Therefore, and solely to indicate this fact, this article is hereby marked “advertisement” in accordance with 18 USC section 1734