Reader Comments

Post a new comment on this article

RE: Artificial sweeteners and cancer risk: Results from the NutriNet-Santé Population-base-cohort study.

Posted by clavecchia666 on 14 Apr 2022 at 10:59 GMT

Debras et al. (1) in a 7.7 year follow-up of 102,865 participants to the French population-based Nutri-Santé cohort reported 3358 incident cancer cases.
Dietary information was collected every 6 months on 3 non-consecutive web-based 24-hour dietary records, including artificial sweetener use, and was averaged over two years. Hence, no information is available on changes in sweetener use over time. Although allowance was made for sugar consumption in multivariate models, no corresponding Table was presented for sugar vs artificial sweetener use. It is known that sugar can be associated to excess cancer risk (2–6) .
The authors concluded that artificial sweeteners were associated with increased cancer risk with a multivariate hazard ratio (HR) of 1.13 for lower consumers and 1.14 for higher consumers. The HR for the higher consumption level were 1.15 for aspartame and 1.13 for acesulfame K. However, causal inference from observational studies on the basis of relative risks or HRs of the order of 1.1-1.2 remain open to criticism due to residual confounding and bias.
The study power, in addition, is not given by the size of the cohort, but by the number of cases. These were 744 lower consumers and 601 higher consumers for all cancers combined. Most “Total artificial sweeteners” users had consumed aspartame and acesulfame k, thus distinction of their role was not possible. The numbers were much smaller for sucralose, and hence no significant association was observed, in the absence however of any formal heterogeneity.
The numbers are also smaller for single cancers, i.e. 299 lower consumers/194 higher consumers for breast, 63/58 for prostate cancer.
More important, allowance for BMI and several other covariates in the fully adjusted model led to a substantial reduction of the HRs, despite the fact that only one term in continuous was introduced for BMI in multivariate models. Inclusion of separate terms for BMI categories plus a continuous one would be more appropriate. Despite such a likely under-adjustment, the HR of all cancers, total artificial sweeteners for lower consumers declined from 1.26, adjusted for sex and age only, to 1.14, fully adjusted. The HR for higher consumers declined from 1.19 to 1.13. Similar declines were observed for aspartame (from 1.19 to 1.13) and for acesulfame K (from 1.18 to 1.15). This indicates that residual confounding may largely or totally explain the apparent residual association between artificial sweeteners and cancer risk in the Nutri-Santé cohort.
Even more important, there is no dose-risk relationship, i.e. the HR of all cancers for total artificial sweeteners was 1.14 for lower consumers and 1.13 for higher consumers.
Finally, an analysis of non-obesity-related cancers would be of interest, to distinguish the potential role of artificial sweeteners from the well-known one of overweight and obesity, which are now the second cause of cancer after tobacco (7).

Carlo La Vecchia, Professor of Medical Statistics and Epidemiology, Department of Clinical Sciences and Community Health, Università degli Studi di Milano, Milan, Via Vanzetti 5, 20133, Milan, Italy

1. Debras C, Chazelas E, Srour B, Druesne-Pecollo N, Esseddik Y, Szabo de Edelenyi F, et al. Artificial sweeteners and cancer risk: Results from the NutriNet-Santé population-based cohort study. PLOS Med. 2022 Mar 24;19(3):e1003950. doi: 10.1371/journal.pmed.1003950.
2. La Vecchia C, Franceschi S, Bidoli E, Barbone F, Dolara P. Refined‐sugar intake and the risk of colorectal cancer in humans. Int J Cancer. 1993;55(3):386–9.
3. La Vecchia C, Bosetti C, Negri E, Franceschi S. Refined sugar intake and the risk of gastric cancer. Int J Cancer. 1998;78(1):130–1.
4. Bosetti C, Gallus S, Talamini R, Montella M, Franceschi S, Negri E, et al. Artificial Sweeteners and the Risk of Gastric, Pancreatic, and Endometrial Cancers in Italy. 2009;
5. Galeone C, Pelucchi C, Vecchia C La. Added sugar, glycemic index and load in colon cancer risk. Curr Opin Clin Nutr Metab Care. 2012 Jul;15(4):368–73.
6. Braverman-Bronstein A, Camacho-García-Formentí D, Zepeda-Tello R, Cudhea F, Singh GM, Mozaffarian D, et al. Mortality attributable to sugar sweetened beverages consumption in Mexico: an update. Int J Obes (Lond). 2020 Jun 1;44(6):1341–9.
7. Islami F, Goding Sauer A, Gapstur SM, Jemal A. Proportion of Cancer Cases Attributable to Excess Body Weight by US State, 2011-2015. JAMA Oncol. 2019 Mar 1;5(3):384–92.

Competing interests declared: Carlo La Vecchia is a member of the Scientific Advisory Panel of the International Sweeteners Association.

RE: RE: Artificial sweeteners and cancer risk: Results from the NutriNet-Santé Population-base-cohort study.

cdebras replied to clavecchia666 on 19 Apr 2022 at 11:54 GMT

Dear Pr La Vecchia, thank you for your feedback.
In order to account for changes in sweetener use over time, we conducted supplementary analyses with artificial sweetener intakes coded as time-dependent exposure variables, and results remained similar (Table G in S1 Appendix).
Regarding your concerns for sugar vs. artificial sweetener use, in order to explore the question of which, between sugar and artificial sweeteners, may be more problematic regarding cancer risk, participants were categorised into 6 classes according to their intake levels of artificial sweeteners (non-consumers, lower consumers, and higher consumers) and sugar (</≥ the French recommended limit of 100 g/day). Cancer risks were compared 2-by-2 across the 6 categories, and in particular between the categories ‘higher artificial sweetener consumption and sugar intake below the official recommended limit’ and ‘no artificial sweetener consumption and sugar intake exceeding the recommended limit’, with the latter category being the reference category. These analyses revealed increased cancer risk associated with both artificial sweetener and sugar intakes (Fig C and Table E in S1 Appendix). In particular, no difference was detected between the categories ‘higher artificial sweetener consumption and sugar intake below the official recommended limit’ and ‘no artificial sweetener consumption and sugar intake exceeding the recommended limit’ (Table F in S1 Appendix).
As already stated and discussed in our paper, indeed, causal links cannot be established by this unique study; in particular, residual confounding bias cannot be entirely ruled out, although the wide range of adjustment factors accounted for in main and sensitivity analysis models limited this risk. In addition, the limited number of cases prevented us from studying associations for other cancer sites (e.g., pancreatic, ovarian, endometrial, kidney, liver, and bladder) than the main ones presented here. As also already discussed the statistical power was sufficient to investigate overall cancer risks (3358 cases) and breast, prostate and obesity-related cancer risks (979, 403 and 2023 cases, respectively), but was too limited to study other cancer sites.
Table A in S1 Appendix displays the number and percent of participants consuming 1, 2, or 3 of the main artificial sweeteners (aspartame, acesulfame-K, and sucralose). Participants frequently co-consumed several types of artificial sweeteners, but the proportion of those who were consumers of all 3 main artificial sweeteners was low (only 7.1%). In addition, all models were mutually adjusted for artificial sweetener intake other than the one studied.
It was perfectly expected, and is usually observed in nutritional epidemiology, to have a reduction in HR between the minimally and fully adjusted models accounting for all known risk factors.
Main model was adjusted for baseline BMI (continuous variable) and weight variation as well as a weight loss diet during follow-up (model 7, sensitivity analyses). In addition, it was not included in the paper but we have tested a model with BMI as categorical variable and results remained similar (HR=1.12 [1.00-1.25], P-trend=0.05 for overall cancer for total artificial sweeteners).
We assessed linearity by comparing the model with the 3 distinct categories of sweetener intake to a model with a linear trend across these categories, using the Akaike information criterion. The comparison of the model with 3 categories of intake to the model with a linear trend across categories did not provide evidence of non-linearity (P = 0.107, 0.250, 0.348, and 0.437 for total artificial sweeteners, aspartame, acesulfame-K, and sucralose, respectively, for the overall cancer model).
Finally, analyses for non-obesity related cancers were already presented in Results A, S1 Appendix.
Best regards,
Charlotte Debras and Mathilde Touvier

No competing interests declared.