Obesity and Sex-Related Associations With Differential Effects of Sucralose vs Sucrose on Appetite and Reward Processing

Key Points Question Are adiposity and sex associated with neural, metabolic, and behavioral responses to consumption of nonnutritive sweeteners (NNSs) vs nutritive sugar? Findings In this randomized crossover trial, both obesity and female sex were associated with differential neural food cue responsivity in reward processing areas following ingestion of sucralose (an NNS) compared with sucrose (nutritive sugar). Meaning These findings suggest that female individuals and those with obesity have greater neural reward responses to NNS vs nutritive sugar consumption, highlighting the need to consider individual biological factors that might influence the efficacy of NNS.


Detailed Study Overview
All the visits were conducted in the morning after a 12h overnight fast. During the screening visit, we assessed eligibility for participation in the study and demographic information and anthropometric measurements were obtained. Height was measured to the nearest 0.1cm using a stadiometer and weight to the nearest 0.1kg using a bioelectrical impedance analysis scale (Model no. SC-331S, TANITA Corporation of American, Inc.), and BMI was calculated as weight in kilograms divided by height in meters squared. Participants were split into three BMI status groups (healthy weight: 18.5-24.9 kg/m 2 , overweight: 25-29.9 kg/m 2 , and obesity: ≥ 30 kg/m 2 ) based on Centers for Disease Control and Prevention (CDC) criteria 2 . Measurement of body weight was repeated at the beginning of each MRI study session. Participants with a weight change of ≥5 lbs between study sessions were disqualified to reduce variability in brain responses due to changes in body weight. Participants ingested 300mL drinks containing either sucrose (75g), sucralose (sweetness matched), or a water control, mixed with 0.45g of nonsweetened zero calorie cherry flavoring (Kraft Foods Kool-Aid® Unsweetened Cherry Drink Mix) for palatability. The sucrose and sucralose drinks were individually sweetness matched during the initial screening visit using a blinded task where participants selected the concentration of sucralose (1.5 mM, 2 mM, or 3 mM) that best matched a 25% weight per volume concentration of sucrose. The order of the drinks was randomized using a computer-generated sequence.
Participants and experimenters were blind to the drink provided during the study sessions. Of the 76 participants in the trial who received at least one drink allocation, 2 participants received neither the sucrose nor sucralose drink allocations (due to study drop-out) and therefore were excluded from this analysis, while 74 participants received at least one of the primary drink condition allocations (i.e., sucrose and/or sucralose) and were included. Of these 74 subjects, 72 participants received both primary drink allocations, while 2 participants received only one primary drink allocation prior to withdrawing from the study (see Figure 2 for additional details).
MRI scan included a T1 structural scan (for anatomical registration) followed by participants exiting the scanner to consume the test drink within two minutes in order to reduce variability in timing of drink effects. After consuming the drink, participants re-entered the scanner and underwent a food cue task (described below) beginning at approximately 20min post-drink ingestion. Arterial spin labelling (ASL) sequences were also collected as a part of the larger study but are not included in the current analysis. During each study visit, blood samples were collected at baseline (0min), 10min, 35min, and 120min post-drink to measure plasma glucose, insulin, GLP-1, acyl-ghrelin, PYY, and leptin concentrations. The study ended with a food buffet (125min post-drink, described below). Female subjects underwent study visits during the follicular phase of their menstrual cycles to reduce cycle related variability in hunger and food cravings 3,4 .

MRI Imaging Analysis Details
To analyze fMRI data, we used several tools from the Oxford University Centre for Functional MRI of the Brain Software Library (FMRIB). MRI data were processed using the fMRI Expert Analysis Tool (FEAT) version 6.00. Eight functional volumes (8TRs) acquired at the beginning of each MRI session were discarded to account for magnetic saturation effects. fMRI data were preprocessed using motion correction, high-pass filtering (100s), and spatial smoothing with a Gaussian kernel of full width at half-maximum=5mm. Functional data were first mapped to each participant's anatomical image and then registered into standard space [Montreal Neurological Institute (MNI)] using affine transformation with FMRIB's Linear Image Registration Tool to the avg152 T1 MNI template. Food and non-food blocks were added to the general linear model (GLM) after convolution with a canonical hemodynamic response function. Temporal derivatives and temporal filtering were added, increasing statistical sensitivity. Motion confounds were generated using the tool "fsl_motion_outliers" to be used as no-interest regressors in the GLM.
For each participant, visual cue contrast maps were created on the first-level analysis. All ROI were bilateral and anatomically defined using the Harvard-Oxford atlas found in FSL (which provides probability mapping of 21 subcortical brain structures), except the hypothalamus, which is not included in the atlas and was defined bilaterally as a 2-mm spherical ROI surrounding peak glucose-responsive voxels identified previously 7 . Percent BOLD signal change was extracted from each ROI and cue contrast for each participant to identify differences in relative brain activation to food cues vs non-food cues using FSL's FEATquery.

Ad-libitum Buffet Meal
The buffet meal consisted of 32 pre-measured food and drink items. Total energy available from the buffet meal was 4650 kcal (for a full list of foods at the buffet and the calorie content of each food cue see eTable 2). Caloric value per gram or fluid ounce of each item was calculated using the Nutritional Data System for Research (NDSR) software (described below). Participants were given 20 minutes to eat any quantity they desired and instructed not to leave the room with any items. After the participant exited, each buffet item was re-weighed. Total caloric intake during the buffet meal was calculated using the difference between the pre-meal and post-meal weight for each buffet item.
To give an index of the degree of compensation for the 300kcal sucrose preload during the ad libitum buffet meal, we calculated COMPX scores using the following equation: COMPX = ((calories consumed after water conditioncalories consumed after sucrose condition) / (300)) x 100 10 . A score of 100% would indicate that the caloric intake consumed in the buffet meal was exactly 300kcal more in the water control condition to compensate for the caloric difference between sucrose (300kcal) and water (0kcal) preloads, whereas over 100% represents overcompensation for the preload calories (i.e., consuming too much after the water preload and/or too little after the sucrose preload), and 1-99% indicates some degree of compensation (i.e., consuming more after the water preload and/or less after the sucrose preload but not enough to fully compensate for the difference in preload calories) 10 .

Habitual NNS Dietary Intake Assessment and Analysis
Diet was assessed using the multipass 24-hour dietary recall, which is a validated method that provides detailed information on food and beverages consumed over the previous 24-hour period 11,12 . Each dietary interview was administered by a trained staff member, wherein volunteers were asked to recall all food and drinks items (including meals and snacks) that they ingested during the previous 24-hours. In order to account for potential daily variations in dietary intake, 24-hour recalls were captured on both weekdays and weekend days. After the dietary recalls were obtained, the data was analyzed using the Nutritional Data System for Research (NDSR) software v.2018, developed by the Nutrition Coordinating Center, University of Minnesota, Minneapolis, MN, USA 13 . Dietary recalls were also assessed for plausibility and quality using the method described by Jones et al. 14 , and using this method, 359 dietary recalls were included in the analysis (an average of 5 per participant) and 5 recalls were excluded. The variables outputted from this software that were used in the analysis were the average daily intake of acesulfame potassium (aceK), sucralose, saccharin, aspartame, or any combination thereof, in milligrams (mg). To represent habitual NNS dietary intake, we used each subjects' mean values across all recalls obtained. Participants were classified as either NNS user if average daily intake of NNS was above 0 or NNS non-user if average daily NNS intake was equal to 0.

Statistical Analysis Details
We found AUC values to be right-skewed; for statistical analyses, AUC values for glucose, insulin, GLP-1, acyl-ghrelin, PYY, and leptin were cubic root transformed to better meet the assumptions of normality. In addition, the residuals for the buffet meal models were not normally distributed, and for statistical analyses, total caloric was square root transformed. Paired t-tests and analysis of variance were used to compare NNS dietary intake between sex and BMI status © 2021 Yunker AG et al. JAMA Network Open. groups, respectively. Baseline levels of plasma glucose, insulin, and GLP-1 between the sucralose, sucrose, and water drink condition visits were compared using analysis of variance.

Whole Cohort after Sucrose vs Water (control) and Sucralose vs Water (control)
There were no significant BOLD signal differences in any ROI in response to any food cue contrasts after either the sucrose vs water or sucralose vs water drink comparisons, adjusted for covariates (age, sex, BMI status, and NNS user status) and multiple ROI and visual cue contrasts (eTables 4,5).

Effects of BMI Status after Sucrose vs Water (control)
We observed BMI status x drink interactions in the MFC in response to high-calorie food vs nonfood (p=0.02) and savory food vs non-food (p=0.04), adjusted for covariates (age, sex, and NNS user status), multiple ROIs, and visual cue contrast comparisons. The remaining associations did not meet the threshold of significance (eTable 7).
As a post-hoc analysis, we stratified results by BMI status (and adjusted for age, sex, and NNS user status as covariates) in order to understand the directionality of the interactions.
However, in data stratified by BMI status, the associations did not meet the threshold of significance (eTable 8).

Effects of BMI Status after Sucralose vs Water (control)
None of the BMI status x drink interactions met the threshold of statistical significance (eTable 9).

Effects of Sex after Sucrose vs Water (control)
There were sex x drink interactions in response to food vs non-food cues in the MFC (p=0.03), hippocampus (p=0.03), and OFC (p=0.03), in response to high-calorie vs low-calorie food cues in the dorsal striatum (p=0.04), MFC (p=0.04), insula (p=0.04), and OFC (p=0.04), in response to sweet vs non-food cues in the MFC (p=0.02) and OFC (p=0.04), and after low-calorie vs non-food in the MFC (p=0.01), OFC (p=0.01), hippocampus (p=0.02), dorsal striatum (p=0.03), and insula (p=0.03), adjusted for covariates (age, BMI status, and NNS user status) and multiple ROIs and visual cue contrast comparisons. The remaining associations did not meet the threshold of significance (eTable 12).

Effects of Sex after Sucralose vs Water (control)
There were sex x drink interactions in response to food vs non-food cues in the MFC (p=0.03), hippocampus (p=0.03), and OFC (p=0.03), in response to high-calorie vs low-calorie food cues in the dorsal striatum (p=0.04), MFC (p=0.04), insula (p=0.04), and OFC (p=0.04), in response to sweet vs non-food cues in the MFC (p=0.02) and OFC (p=0.04), and after low-calorie vs non-food in the MFC (p=0.01), OFC (p=0.01), hippocampus (p=0.02), dorsal striatum (p=0.03), and insula (p=0.03), adjusted for covariates (age, BMI status, and NNS user status) and multiple ROIs and visual cue contrast comparisons. The remaining associations did not meet the threshold of significance (eTable 14).

Whole Cohort
There was a main effect of drink on in-scanner cue-induced ratings of hunger after viewing food As a post-hoc analysis, we compared the difference of LSmeans ± SE between each drink comparison (i.e., sucralose vs sucrose, sucralose vs water, sucrose vs water) in order to understand the directionality of each significant main effect of drink on in-scanner cue-induced appetite ratings (shown in eTable 16). We observed a main effect of drink on in-scanner cueinduced ratings of hunger and wanting, but not liking, across most food cues whereby hunger ratings after viewing food (β=0.25, 95% CI: 0.03-0.47, p=0.03), high-calorie (β=0.23, 95% CI: 0.01-0.45, p=0.04), low-calorie (β=0.28, 95% CI: 0.08-0.48, p=0.01), sweet (β=0.26, 95% CI: 0.02-0.50, p=0.03), and non-food (β=0.23, 95% CI: 0.03-0.43, p=0.03) cues were higher following sucralose vs sucrose (eTable 16). In secondary analyses, hunger and wanting ratings were also consistently higher after water vs sucrose conditions, and liking ratings after viewing savory food cues were higher following water compared to the sucralose and sucrose conditions (see eTable 16 for details).

Effects of BMI Status
We observed BMI status x drink interactions on in-scanner cue-induced ratings of hunger after As a post-hoc analysis, we compared the difference of LSmeans ± SE between each drink comparison (i.e. sucralose vs sucrose, sucralose vs water, sucrose vs water) stratified by BMI status in order to understand the directionality of each significant BMI status x drink interaction on in-scanner cue-induced hunger ratings (shown in eTable 17). In analyses stratified by BMI group, individuals with overweight (β=0.52, 95% CI: 0.13-0.91, p=0.04), but not with healthy weight (β=0.36, 95% CI: 0.01-0.71, p=0.14) or obesity (β=-0.12, 95% CI: -0.57-0.33, p=0.88), had greater hunger after viewing sweet food cues following sucralose vs sucrose conditions. Secondary analyses showed that individuals with healthy weight and overweight, but not with obesity, reported greater hunger after water vs sucrose conditions across most food cue types, and participants with healthy weight also reported greater hunger to non-food cues following water vs sucralose (see eTable 17 for details). There were no significant BMI status by drink interactions for in-scanner cue-induced ratings of liking or wanting (see eAppendix 2 for details).

Effects of Sex
We observed sex x drink interactions on in-scanner cue-induced ratings of hunger after viewing As a post-hoc analysis, we compared the difference of LSmeans ± SE between each drink comparison (i.e. sucralose vs sucrose, sucralose vs water, sucrose vs water) stratified by sex in order to understand the directionality of each significant sex x drink interaction on in-scanner cueinduced hunger, liking, and wanting ratings (shown in eTable 18). Sex stratified analyses showed no differences in liking or wanting ratings after sweet food cues following sucralose vs sucrose among either females (β=0.08, 95% CI: -0.14-0.30, p=0.77 and β=0.28, 95% CI: 0.01-0.55, p=0.16, respectively) or males (β=-0.26, 95% CI: -0.51 to -0.01, p=0.16 and β=-0.15, 95% CI: -0.46-0.16, p=0.67, respectively), but secondary analyses showed that females, but not males, reported less wanting for sweet foods after sucrose vs water (see eTable 18 for details).