Association of Sugar-Sweetened, Artificially Sweetened, and Unsweetened Coffee Consumption With All-Cause and Cause-Specific Mortality: A Large Prospective Cohort Study
Submit a Comment
Contributors must reveal any conflict of interest. Comments are moderated. Please see our information for authorsregarding comments on an Annals publication.

Abstract
Background:
Objective:
Design:
Setting:
Participants:
Measurements:
Results:
Limitation:
Conclusion:
Primary Funding Source:
Get full access to this article
View all available purchase options and get full access to this article.
Supplemental Material
- Download
- 1.63 MB
References
Comments
Sign In to Submit A CommentInformation & Authors
Information
Published In

History
Keywords
Copyright
Authors
Metrics & Citations
Metrics
Citations
If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. For an editable text file, please select Medlars format which will download as a .txt file. Simply select your manager software from the list below and click Download.
For more information or tips please see 'Downloading to a citation manager' in the Help menu.
Association of Sugar-Sweetened, Artificially Sweetened, and Unsweetened Coffee Consumption With All-Cause and Cause-Specific Mortality: A Large Prospective Cohort Study. Ann Intern Med.2022;175:909-917. [Epub 31 May 2022]. doi:10.7326/M21-2977
View More
Login Options:
Purchase
You will be redirected to acponline.org to sign-in to Annals to complete your purchase.
Access to EPUBs and PDFs for FREE Annals content requires users to be registered and logged in. A subscription is not required. You can create a free account below or from the following link. You will be redirected to acponline.org to create an account that will provide access to Annals. If you are accessing the Free Annals content via your institution's access, registration is not required.
Create your Free Account
You will be redirected to acponline.org to create an account that will provide access to Annals.
Differences in mortality risk between unsweetened coffee and sugar-sweetened coffee, and among three types of coffee
By performing a large prospective cohort study (1), Liu and colleagues concluded that moderate consumption of sugar-sweetened coffee and that of unsweetened coffee were associated with similar reductions in mortality risk. This conclusion only emphasizes the similarity in the U-shaped associations of sugar-sweetened coffee and unsweetened coffee with mortality. However, it ignores the possible difference in mortality risk between unsweetened coffee and sugar-sweetened coffee. As shown in Table 3 in Liu et al.’s article (1), in the unsweetened coffee subgroup, there were only four not statistically significant results (P ≥ 0.05): Instant & >0–1.5 Drinks/d [hazard ratio (95% confidence interval): 0.86 (0.74–1.00)], Decaffeinated & >0–1.5 Drinks/d [0.95 (0.75–1.19)], Decaffeinated & >1.5–2.5 Drinks/d [0.89 (0.71–1.11)], and Decaffeinated & >4.5 Drinks/d [0.98 (0.73–1.32)]; whereas the other 11 results were statistically significant. On the contrary, in the sugar-sweetened coffee subgroup, there were only four statistically significant results (P < 0.05): Instant & >1.5–2.5 Drinks/d [0.69 (0.56–0.86)], Instant & >2.5–3.5 Drinks/d [0.67 (0.52–0.88)], Ground & >1.5–2.5 Drinks/d [0.71 (0.54–0.94)], and Ground & >2.5–3.5 Drinks/d [0.71 (0.54–0.94)]; whereas the other 11 results were not statistically significant. It seems to suggest that unsweetened coffee has the greater possibility of having the association with reduced mortality risk compared with sugar-sweetened coffee. However, the authors did not provide a sufficient analysis on the difference in mortality risk between unsweetened coffee and sugar-sweetened coffee.
Similarly, the authors identified that three types of coffee (i.e., ground, instant, and decaffeinated coffee) had the U-shaped associations with mortality, but did not assess the differences among these three coffee types in the risk of mortality. As shown in Table 3 in the article (1), in the unsweetened coffee subgroup most of the results for instant coffee and ground coffee were statistically significant, whereas a majority of the results for decaffeinated coffee were not statistically significant. Moreover, in the sugar-sweetened coffee subgroup two of the results for instant coffee and two for ground coffee were statistically significant, whereas none of the results for decaffeinated coffee was statistically significant. The above results possibly suggest the superiority of instant coffee and ground coffee over decaffeinated coffee in the association with reduced mortality risk. It is of clinical and public health significance to verify this issue.
References
Relevant Analyses Based on Individual Participant Data Are Needed
In Liu and colleagues’ study (1), the authors evaluated the associations of artificially sweetened, sugar-sweetened, and unsweetened coffee with mortality and the associations between coffee types (including decaffeinated, ground, and instant coffee) and mortality. However, they did not assess the differences among various coffee in mortality risk. Hence, we did a further analysis using the meta-analysis method based on the data presented in Table 3 in Liu et al.’s article (1), and aimed to appraise the possible subgroup effects according to adding sugar or not (artificially sweetened, sugar-sweetened, and unsweetened), type of coffee (decaffeinated, ground, or instant) , and drinks per day, respectively.
As shown in Figure 1 (available at https://doi.org/10.6084/m9.figshare.20026568), the subgroup effect according to adding sugar or not (Figure 1A) was statistically significant (Psubgroup =0.04), and unsweetened coffee (a 23% reduction) and sugar-sweetened coffee (a 20% reduction) had the association of greater reductions in the risk of all-cause mortality compared with artificially sweetened coffee (an 11% reduction). The subgroup effect according to type of coffee (Figure 1B) was statistically significant (Psubgroup =0.03), and ground coffee (a 26% reduction) had the association of a greater reduction in mortality risk compared with instant coffee (an 18% reduction) and decaffeinated coffee (a 16% reduction). The subgroup effect according to drinks per day (Figure 1C) was statistically significant (Psubgroup <0.01), and the subgroups of >2.5-3.5 Drinks/day (a 27% reduction) and >3.5-4.5 Drinks/day (a 30% reduction) had the association of greater reductions in mortality risk compared with the other three subgroups (a reduction of 15%-18%).
The results revealed by Figure 1C are consistent with Liu et al.’s main findings: various coffee had the U-shaped association with mortality. This suggests that it is practicable to do such an analysis as we did. More importantly, Figure 1A and Figure 1B seem to reveal the possible superiority of unsweetened coffee and sugar-sweetened coffee over artificially sweetened coffee, and that of ground coffee over instant coffee and decaffeinated coffee. If the authors additionally implement relevant analyses using their individual participant data to confirm these possible superiorities, it will greatly enhance the public health significance of their article.
References
Coffee consumption, all-cause and cause-specific mortality: an epidemiological evidence
Liu et al. conducted a 7-year prospective study to evaluate the associations of coffee consumption, all-cause and cause-specific mortality (1). Compared with non-consumers, the adjusted hazard ratios of consumers with various amounts of unsweetened coffee for all-cause mortality significantly decreased. The adjusted hazard ratios of consumers with >1.5 to 3.5 drinks of sugar-sweetened for all-cause mortality also significantly decreased. With regard to the risk reduction in mortality from cancer and cardiovascular disease by coffee consumption was consistent with that with all-cause mortality. The authors recommended the moderate amounts of coffee consumption as favorable habits. I am very interesting about their epidemiological findings, although complexed causality cannot be explained by epidemiological findings. I want to present information regarding the relationship of coffee consumption with health outcomes.
There is a report on the relationship between coffee consumption and psychological wellbeing in male workers, aged 35 to 60 years (2). Psychological wellbeing was evaluated using the General Health Questionnaire (GHQ) 12-item version. Although aging, regular exercise and sleeping time were significantly associated with GHQ scale scores, coffee consumption and smoking status did not present a significant association with GHQ scales scores. I suspect that many lifestyle factors may indirectly relate to subsequent changes in mortality, and aging would deeply associated with all-cause and cause-specific mortality.
Liu et al. conducted 7-year follow-up study and the mean age of subjects was 55.6 years old. Habitual coffee consumption would be closely related to other lifestyle factors, and mechanism for U-shaped relationship may be difficult to be explained by epidemiological approach. As many factors contribute to the risk in mortality, there are also limitation to explore the causal relationship by experimental studies.
REFERENCES
Author Response to Yin et al
We thank Dr. Yin et al. (1) for their interest in our work, and we are pleased to provide responses to their comments.
Taking subgroup analyses by types of coffee in Table 3 as an example, the authors thought that we only emphasize the similarity in the U-shaped associations of sugar-sweetened coffee and unsweetened coffee with mortality and ignored the possible difference. As shown in Table3, we could see clearly that the sample-sizes for sugar-sweetened coffee subgroups were much smaller than those in unsweetened coffee subgroups. For example, the sample-sizes for unsweetened instant and ground coffee subgroups were 107 862 and 97 772, while the sample-sizes for sweetened instant and ground coffee subgroups were only 59 832 and 52 828, respectively. This may be the reason why the number of statistically significant results in sugar-sweetened coffee subgroups was less than that in unsweetened coffee subgroups. When we took sugar-sweetened coffee and unsweetened coffee subgroups as a whole, the similar U-shaped associations of sugar-sweetened coffee and unsweetened coffee with mortality were clearly observed, as shown in Table 2 and Figure.
Similarly, the authors thought that the superiority of instant coffee and ground coffee over decaffeinated coffee in the association with reduced mortality risk according to less significant results observed in decaffeinated subgroups. This may also be caused by the smaller sample-size in decaffeinated subgroup. Take unsweetened coffee as an example, the sample-sizes for instant and ground were 107 862 and 97 772, respectively, while the sample-size for decaffeinated subgroup was only 58 733. Besides this, given that previous studies (2, 3) have been done to compare associations of instant, ground, and decaffeinated coffee with the risk of mortality, the main objective of our study focused on whether adding sugar to coffee negated potential health benefits for the consumers.
In summary, relatively smaller sample-sizes in addition to few incidences of death observed due to a short follow-up time as discussed in the article caused statistically insignificant results in some subgroups. However, these results were less reliable and conclusive as 95% CIs were wild. Therefore, at this stage, according to our data, we could not draw such a conclusion that there was difference among the associations of some types of coffee as pointed by Dr. Yin et al. with the risk of mortality.
References
Author Response to Qiu et al.
Qiu et al. (1) performed meta-analyses by combing the HRs presented in Table 3 to evaluate the possible subgroup effects according to adding sugar or not (artificially sweetened, sugar-sweetened, and unsweetened), types of coffee (decaffeinated, ground, or instant), and drinks per day, respectively. We thank for their hard work and great interest in our work.
In this study, Cox models with penalized splines showed statistically significant U-shaped associations for unsweetened coffee and sugar-sweetened coffee with all-cause mortality (P< 0.001). It means that the effects of unsweetened coffee and sugar-sweetened coffee varied according to the drinks of coffee per day. This is the reason why we demonstrated the results by the drinks per day. Therefore, it is not appropriate to combine data by using meta-analyses for unsweetened coffee and sugar-sweetened coffee regardless of amounts of coffee. In addition, Qiu et al further performed subgroup effects for adding sugar or not (artificially sweetened, sugar-sweetened, and unsweetened), and types of coffee (decaffeinated, ground, or instant). It also means that they could not simply combine data according to drinks per day regardless of adding sugar or not and types of coffee.
In summary, the effects of coffee on mortality varied according to the drinks per day. It is not appropriate to combine effect-sizes regardless of amounts of coffee consumption. In addition, given that previous studies (2, 3) have been done to compare associations of instant, ground, and decaffeinated coffee with the risk of mortality, the main objective of our study only focused on whether adding sugar to coffee negated potential health benefits for the consumers.
References
Author Response to Kawada
We thank Dr. Kawada (1) for the interest in our work, and we are pleased to provide responses to their comments.
We agreed that habitual coffee consumption would be closely related to other lifestyle factors, and these lifestyle factors may have influence on the risk of mortality. Therefore, in our multivariable Cox regression model, lifestyle factors including smoking status, pack-years of smoking, physical activity level, vitamin and mineral supplement use, intake of total energy, total sugar, fresh fruit, vegetables, red meat, processed meat, alcohol, tea, milk, sugar-sweetened beverages, and artificially sweetened beverages were adjusted. Even through, lifestyle factors may possibly lead to residual confounding as we discussed this issue in the article. In addition, given the observational nature of our study, U-shaped relationship between coffee consumption and mortality need to be further confirmed by experimental study i.e., randomized controlled trial, and laboratory study on the mechanism for this phenomenon is also needed.
References
Discrepancy from the previous study in UK Biobank and potential confounding
Liu et al. analyzed the relationship between coffee consumption and mortality, stratified by the presence or absence of sugar or artificial sweeteners using data from the UK Biobank (1). They found that adequate amounts of coffee consumption with and without sugar were associated with lower all-cause mortality and cancer mortality. They are consistent with the analysis of UK Biobank data in which Loftfield et al. reported a mild reduction in mortality with coffee consumption (2). However, in another analysis in UK Biobank by Ong et al. (3), an observational study was reported in conjunction with the Mendelian randomization study, which found coffee consumption to be almost unrelated to cancer incidence and cancer mortality. The presence of discrepancy despite the analysis of the same UK Biobank dataset indicates a risk of bias and confounding. The study by Ong et al. limited the subjects to the White British cohort, whereas Liu et al. and Loftfield et al. used a dataset from various ethnicities. Even if the ethnicity variable was included in the multivariable model, it cannot be guaranteed that the model would eliminate complexed confounding, and a stratified analysis by ethnicity will be recommended.
REFERENCES
Author response to Harada et al.
We thank Dr. Harada (1) for the interest in our work, and we are pleased to provide responses to their comments.
As suggested by Dr. Harada, we performed stratified analyses by ethnicity (Table A1). We found that adequate amounts of coffee consumption with and without sugar were associated with lower all-cause mortality and cancer mortality in White ethnicity subgroup. Insignificant associations in other ethnicities subgroup may due to the limited number of sample-size. Given the observational nature of the study, the residual confounding may also exist even if multivariable were thoroughly adjusted in Cox regression model. We acknowledged this issue in the Discussion section of the article.
Mendelian randomization studies may less affected by confounding, but instrument variables were very difficult to find. As Christina C. Wee commented in her editorial for our article, it is important to recognize the limitations of such studies. As a proxy for coffee consumption, current genetic markers are relatively weak instruments leading to limited power to detect effects. Thus, as with most underpowered studies, null findings from Mendelian randomization studies do not rule out true health effects. (2). In the study by Ong et al. (3), the selected genetic markers relatively weak instruments leading to limited power to detect effects.
References:
An unappreciated link between coffee consumption and mortality – Fluid intake
In a cohort of 171, 616 adults, Liu and colleagues observed a U-shaped relationship between coffee consumption and mortality after adjusting for several co-variate lifestyle-related factors: smoking status, pack-years of smoking, physical activity level, vitamin and mineral supplement use, intake of total energy, total sugar, fresh fruit, vegetables, red meat, processed meat, alcohol, tea, milk, etc. (1). Notably, this result was consistent rather coffee consumers preferred adding sweeteners or unsweetened, decaffeinated or caffeinated, and instant or ground.
Mechanistically, Liu et al., acknowledged that the results of their study were likely not due to caffeine content alone but may function in concert with other known bioactive compounds. However, there appears to be an underappreciation for the influence of a key nutrient in all cups of coffee – water – on health. Indeed, although lifestyle factors were controlled for, given the nature of the survey used, it is difficult to tease out if self-reported daily coffee consumption is a surrogate for daily fluid intake. Stated differently, from the data presented we do not know if individuals who consumed <2-3 cups/d were chronically dehydrated/hypernatreimic, 2-3 cups/d were chronically well-hydrated, or > 3 cups/d were chronically hyperhydrated/hyponatreimic. If this were the case, then daily fluid intake or hydration status would also mirror the U-Shaped relationship observed between coffee consumption and mortality. Thus, further consideration of the impact of fluid imbalances on health is warranted.
At the cellular level, chronic fluid deficits or excess promotes unprogrammed cell death (i.e.., necrosis) via uncorrected osmotic shifts resulting in cellular shrinkage or swelling, respectively (2). Under these conditions, the likelihood of reduced tissue perfusion or hypoxia and elevated bioenergetic and oxidative stress increases, promoting less favorable health status (2). However, given that the mean age of the sample study by Liu et al. was >55 years old, low fluid intake, rather than excess, would be the more prominent concern in interpretating their data. Indeed, with aging there is an increased likelihood of experiencing fluid deficits due to age-related impaired thirst response to elevated plasma hypertonicity or intentional low fluid intake to reduce urination frequency, especially at nighttime. In fact, data from the U.S. showed a > 4-fold increase of developing a chronic disease within 3- 6 years among adults aged 51-70 with poor drinking habits (3). Hence, fluid intake should not be overlooked as an underlining mechanism or confounder in the interpretation of the results by Liu et al.
References:
Response to Dr. Yates
Reply to: An unappreciated link between coffee consumption and mortality – Fluid intake
Dan Liu1,2, MD; Qingmei Huang1, MD; Weiqi Song1, MD; Zhihao Li1, MD; Chen Mao1,3, MD
1 Department of Epidemiology, School of Public Health, Southern Medical University, Guangzhou, China
2 Department of Public Health and Preventive Medicine, School of Medicine, Jinan University, Guangzhou, China
3 Department of Laboratory Medicine, Microbiome Medicine Center, Zhujiang Hospital, Southern Medical University, Guangzhou, China
We thank Yates et al., (1) for the interest in our work, and we are pleased to provide responses to the comments.
We agreed that fluid intake should not be overlooked as an underlining mechanism or confounder in the interpretation of our results, and fluid intake may have influence on the risk of mortality. In our study, fluid intakes were considered as covariates including alcohol, milk, tea, sugar-sweetened beverage, and artificially sweetened beverage. Therefore, in our multivariable Cox regression model, fluid intakes were adjusted. Even through, fluid intakes may possibly lead to residual confounding as we discussed this issue in the article. In addition, given the observational nature of our study, U-shaped relationship between coffee consumption and mortality need to be further confirmed by experimental study i.e., randomized controlled trial, and laboratory study on the mechanism for this phenomenon is also needed.
References: