The Beverage Quality Index and Type 2 Diabetes Risk in Women: A Prospective Analysis of the Mexican Teachers’ Cohort

doi:10.21203/rs.3.rs-4713212/v1

Download PDF

Article

The Beverage Quality Index and Type 2 Diabetes Risk in Women: A Prospective Analysis of the Mexican Teachers’ Cohort

https://doi.org/10.21203/rs.3.rs-4713212/v1

This work is licensed under a CC BY 4.0 License

You are reading this latest preprint version

BACKGROUND

Individual beverages have shown divergent associations with type 2 diabetes. Whether overall beverage quality affects diabetes risk is unknown. Therefore, we estimated the association of a previously developed Beverage Quality Index (BQI) with incidence of diabetes in Mexican women.

METHODS

We included 77 484 female participants from the Mexican Teachers’ Cohort without diabetes at baseline (2006–2008). At baseline, diet was assessed using a 140-item food-frequency questionnaire. The BQI included 7 components (coffee, milk, juices, sugar-sweetened beverages [SSBs], alcohol, sugar added to beverages, and energy from beverages), with a total theoretical score ranging from 0 to 70. A higher score represents a healthier beverage intake pattern. Data on diabetes incidence were available through 2018 from self-reports or cross-linkage with administrative data. We used multivariable Cox proportional-hazard models adjusted for potential confounders.

RESULTS

Participants' mean (SD) baseline age was 45.9 (7.2) years, and BQI score was 37.3 (8.6), ranging from 9.8 to 69.3. During a median follow-up of 7.6 years, 4521 participants developed diabetes. After multivariable adjustment, when comparing extreme categories (≥55 vs. <25), a higher BQI was suggestively associated with lower diabetes incidence (HR: 0.87; 95% CI: 0.71, 1.06), but the estimation was imprecise. However, restricted cubic spline analysis showed no association between the BQI and diabetes incidence (p-nonlinearity = 0.20).

CONCLUSIONS

In a cohort of Mexican women, the BQI for overall beverage quality showed no consistent association with diabetes incidence. Further research on beverage quality indices for Mexican populations, including those with high SSBs intake, is warranted.

Health sciences/Risk factors

Health sciences/Diseases/Endocrine system and metabolic diseases/Diabetes

Health sciences/Health care/Public health/Epidemiology

beverage quality index

beverage quality

beverage patterns

type 2 diabetes incidence

ESMaestras

Mexican women

Type 2 diabetes (T2D) is a major public health problem in Mexico [1]. In 2022, the prevalence of diabetes in Mexican adults was 18.3%, with 12.6% diagnosed and 5.8 undiagnosed [2]. This prevalence is higher than the global average (10.5%) [3], and is also higher in women compared to men (20.1% vs. 16.3%) (2). Projections indicate that diabetes prevalence is expected to increase at a higher rate in middle-income countries like Mexico (21%) compared to high- and low-income countries (~ 12%) by 2045 (3). A concerning trend is the increasing onset of diabetes at a younger age, with a notably faster rise in the prevalence among young adults (aged 15–49 years) compared to those aged 50 or older over the last three decades [4]. Therefore, implementing policies to reduce modifiable risk factors for diabetes prevention, such as poor diet, is of utmost importance in Mexico [5].

Beverage intake is an important modifiable determinant of diabetes. Higher consumption of sugar-sweetened beverages (SSBs) is considered a risk factor for T2D [6], while consumption of (unsweetened) filtered coffee and tea has been associated with a reduced risk [7, 8]. Beverage patterns among Mexican adults are characterized by high intakes of beverages with added sugars [9]. In women, a significant proportion of total energy intake comes from beverages such as sodas, sweetened coffee and tea, and aguas frescas, a typical Mexican drink made with water, fruit and added sugar [10]. Among Mexican adults, SSBs are the main source of added sugar [9]. Notably, 19% of cardiometabolic disease mortality, including diabetes-related deaths, has been attributed to SSBs intake [11]. Therefore, identifying the optimal beverage pattern for diabetes prevention in Mexico is a public health priority.

Indices for adherence to beverage recommendations could serve as a metric to describe beverage patterns, enabling the translation from research to health policy and consumer education [12]. Prior studies on beverage indices showed that higher scores on the Healthy Beverage Index (HBI) and Healthy Beverage Score (HBS) were linked to lower cardiometabolic risk factors, less kidney function decline, and lower mortality risk in US adults [13, 14]. In Dutch patients with history of myocardial infarction, higher adherence to the Beverage Quality Index (BQI) was associated to lower risk of cardiovascular disease (CVD) recurrence and a potential decrease of diabetes incidence [15]. The BQI, developed in The Netherlands, is a 8-component adherence score to assess beverage quality. All components of the BQI were chosen for their established influence on health and most of them were informed by evidence-based beverage recommendations of the 2015 Dutch dietary guidelines.

Whether overall beverage patterns affect diabetes risk in Mexican women is not known. Therefore, we examined whether the BQI is associated with diabetes incidence in a large cohort of Mexican women.

Study design and population

The Mexican Teachers’ Cohort (MTC) is an ongoing prospective study comprising 115 307 female teachers aged 25 years or older who reside in a culturally, geographically, and economically diverse 12-state area in Mexico, with baseline measures collected between 2006 and 2008 [16]. Extensive information on socio-demographics, reproductive history, diet, lifestyle, and clinical conditions was obtained through a self-administered questionnaire. Follow-up is carried out every 3–4 years to ascertain disease diagnosis and to update risk factor profiles. Response to the baseline questionnaire was considered informed consent. The study was approved by the Research, Ethics, and Biosecurity Committee at the National Institute of Public Health of Mexico (INSP project number 1221).

Of the 115,307 participants recruited at baseline, exclusions were made for the following reasons: self-reported diabetes at baseline (n = 6172); implausible energy intakes (< 500 or > 3500 kcal/day), and FFQs with more than 70 missing items or the staple food section missing (n = 19 111); self-reported cancer, myocardial infarction or stroke at baseline and pregnant women (n = 3319) due to possible changes in diet after diagnosis and/or pregnancy; or no follow-up data available (n = 9220). The sample for analysis included 77 484 participants (Supplemental Fig. 1).

Dietary assessment

Dietary intake data were collected at baseline with a 140-item semi quantitative food-frequency questionnaire (FFQ), which was an extended version of a previously validated 116-item FFQ for Mexican women. Participants were asked how often, on average, they consumed a specified unit or portion size of each item over the previous year. Frequencies of consumption (10 categories) ranged from never to ≥ 6 per day. The FFQ showed good relative validity for total energy, carbohydrate, and total fat intakes compared with four 4-day 24-h recalls (Pearson r = 0.52, 0.57, and 0.63, respectively) [17].

Beverages were classified into five groups: 1) coffee (instant with and without caffeine), 2) milk (skimmed milk, semi-skimmed milk, full-fat milk, yogurt drink, and atole with milk), 3) fruit juices (fresh orange, mandarin and grapefruit juice), 4) SSBs (cola, other carbonated or non-carbonated soft-drinks, aguas frescas, sweetened probiotic fermented milk drink, sweetened soy milk, and atole with water), and 5) alcoholic beverages (beer, wine, brandy, whiskey, rum, tequila, mezcal, pulque, and aguardiente). Aguas frescas are traditional Mexican beverages prepared with water, fruit, and sugar (e.g., lemon water), but they can also be prepared with flowers (e.g., hibiscus water), or grains (e.g., horchata water). Atole is a traditional hot Mexican beverage prepared with water or milk, corn or rice flour and sugar. When some but not all items from a beverage group were missing, those values were imputed to zero. For example, if participants reported consuming skimmed milk but full fat/semi skimmed milk, yogurt drinks or atole with milk were missing then those were imputed to zero. Added sugar to foods and beverages at home was assessed with one item in the FFQ. As most Mexicans do not habitually add sugar to foods, we assumed that all reported added sugar was added to beverages like coffee, tea, aguas frescas, and atoles. Total alcohol intake (g/day) was calculated as the sum of ethanol from all alcoholic beverages. Energy (kcal/day) was additionally calculated through linkage with the USA Department of Agriculture food-composition database [18], and with a database used in the ENSANUT.

Beverage Quality Index (BQI)

The Beverage Quality Index (BQI) is a measure of beverage quality based on the 2015 Dutch dietary guidelines and emerging scientific evidence [15, 19]. It consists of component scores for intakes of filtered coffee, black and green tea, milk, juices, SSBs, alcohol, added sugar to beverages, and total energy from beverages (Table 1). Water and herbal teas were not included in the BQI because they are not part of the Dutch dietary guidelines. We were unable to evaluate black and green tea intake because mostly herbal teas are consumed in Mexico (personal communication). We excluded artificially sweetened beverages from the BQI, because their impact on chronic disease risk is not yet clear [20, 21]. Scores were based on adherence to the relevant component. Each component was scored from 0 to 10. A score of 10 represents total adherence, whereas a score of 0 represents non-adherence. Intermediate intakes were scored proportionately from > 0 to < 10. The BQI has a theoretical range from 0 to 70 points. Coffee and milk were considered optimum components, with maximum scores for intakes within the recommended range (Table 1). Juices, SSBs, alcohol, added sugar to beverages (coffee, tea, aguas frescas and atole), and total energy from beverages were considered moderation components, with maximum scores for limited consumption.

Table 1

Components and scoring system for the beverage quality index (BQI)
Component	Description in guidelines	Minimum score (0)	Maximum score (10)	Proportional scores between 0 and 10*, †
Optimum components
Coffee	Drink 2–4 cups of filtered coffee daily‡	0 or ≥ 900 mL/day	300 to 600 mL/day	> 0 to < 300 mL/day or > 600 to < 900 mL/day
Milk	Consume 1 to 2 portions of dairy daily	0 or ≥ 450 mL/day	180 to 270 mL/day	> 0 to < 180 mL/day or > 270 to < 450 mL/day
Moderation components
Juices	Limit consumption of sugar-containing beverages	≥ 250 mL/day	0 mL/day	> 0 to < 250 mL/day
Sugar sweetened beverages (SSBs)	Limit consumption of SSBs	≥ 250 mL/day	0 mL/day	> 0 to < 250 mL/day
Alcohol	Drink a maximum of 1 glass daily	F: >20 g/day M: >30 g/day	≤ 10 g/day	F: >10 to ≤ 20 g/day M: >10 to ≤ 30 g/day
Added sugar to coffee and tea	Limit addition of sugar to beverages	≥ 25 g/day	0 g/day	> 0 to < 25 g/day
Total energy from beverages	Limit amount of energy from beverages	≥15% of TEI	<10% of TEI	> 10 to < 15% of TEI
Abbreviations: F, female; M, male; TEI, total energy intake.
Components and scores were based on beverage recommendations in the 2015 Dutch dietary guidelines and emerging scientific evidence for cardiometabolic outcomes. The tea component was excluded because data on black and green tea intake were not available. The total BQI score as used in this study therefore includes 7 components and has a theoretical range from 0–70. Example of assignation of proportional scores for optimum components, e.g. milk: for intakes > 0 to < 180 mL/day, proportional scores were = (reported milk intake in mL/day / 180) 10. For intakes > 270 to < 450, proportional scores were = 10 – [(reported milk intake in mL/day – 270 / 450 − 270) * 10]. †Example of assignation of proportional scores for moderation components, e.g. SSBs: for intakes between the cut-off and threshold value, proportional scores were = 10 – [(actual intake mL/day / 250) * 10]. ‡One standard cup of coffee was considered as 150 mL.

Assessment of diabetes

Incident diabetes cases were identified through self-report in any of the follow-up questionnaires or via linkages with administrative data. Self-reported cases were defined if participants answered “yes” to two out of the following three questions: having a medical diagnosis of diabetes, use of antidiabetic medication, or date of diabetes diagnosis. The questionnaire did not specifically ask about T2D, but given the age at diagnosis, we assumed that reported cases were T2D. The validity of self-reported cases was evaluated in a subsample of 1222 women who reported having diabetes on follow-up questionnaires, and who responded to a supplementary questionnaire with extensive questions about diagnosis, pharmacological treatment (hypoglycaemic medication or insulin), and/or diabetes-related complications. Based on an algorithm developed for the MTC, 84% (95% CI: 81–86%) of women who self-reported being diabetic, confirmed diabetes diagnosis or treatment in the supplementary questionnaire [22]. Cross-linkage with clinical databases from a healthcare provider to which 78% of participants are insured (Institute for Social Security and Services for State Workers, ISSSTE) and mortality registries, were also used to identify additional diabetes cases, including incident diabetes cases as the underlying cause of death and six contributing causes of death using the E11 code from the International Classification of Diseases 10th Revision [23]. Detailed information about the two-step cross linkage procedure was previously described [22]. Follow-up for incident diabetes was carried up until December 31, 2018.

Assessment of covariables

Socioeconomic and sociodemographic information, lifestyle habits, dietary habits, anthropometric measures, and family history of diabetes were self-reported during baseline questionnaires. To account for income and health disparities between different regions in Mexico, we categorized participants based on their region of residence (north, central, Mexico City and the metropolitan area, and south). We used internet access at home (yes/no) and health insurance used for serious conditions (public, private or other) as proxies for socio-economic position [24]. Smoking status was reported as never, past, or current. Physical activity was categorized in tertiles of total metabolic equivalents of task (METs) per week of multiple-choice frequencies of activity. The correlation between the MTC questionnaire and the International Physical Activity Questionnaire was 0.55 [25]. Menopausal status was defined using an algorithm that considered last menstruation, hot flushes, hormonal treatments, hysterectomy, and oophorectomy, and it was categorized as premenopausal or postmenopausal. The following food groups were estimated from the FFQ: fruits, vegetables, red meat, processed meat, fish, and legumes (servings/day). Self-reported height (cm) and weight (kg), were used to calculate BMI (kg/m²). The reproducibility and validity of self-reported anthropometric measurements was evaluated in a subsample of 3413 participants. Standardized technician measurements were highly correlated with self-reported ones: weight (r = 0.93), and height (r = 0.84) [26].

Statistical analysis

Mean (SD) or median (IQR) for skewed continuous data, and proportions for baseline characteristics and dietary intakes were examined by BQI score categories. We used Cox proportional hazards models to estimate hazard ratios (HR) and 95% confidence intervals (CI) for the association between categories of the BQI and diabetes incidence, using the lowest category as the reference. The following prespecified categories were used:<25, ≥ 25- <35, ≥ 35- <45, ≥ 45- <55 and ≥ 55. Categories were selected based on diabetes incident rates. Person-years were computed from the date of response to the baseline questionnaire until the date of incident diabetes diagnosis, death, or response to their last questionnaire, whichever came first. For clinical or administrative databases, the end of follow-up was December 31, 2018. When the date of diabetes diagnosis was missing, we imputed it to the midpoint between the dates of the last diabetes-free questionnaire and the first diabetes-positive questionnaire. A directed acyclic graph (DAG) was used to identify potential confounders of the BQI-diabetes association, informed by previous literature. The final set of covariables was included in Supplemental Fig. 2. Model 1 was adjusted for age (years). Model 2 was additionally adjusted for area of residence (north, central, Mexico City and metropolitan area and south), socioeconomic position (internet access yes/no and health insurance used in case of emergency: public, private, or other), physical activity (tertiles of METs/week), smoking status (never, past, current and unknown), family history of diabetes (yes/no), and menopausal status (premenopausal, post-menopausal and unknown). Model 3 was additionally adjusted for energy intake (kcal/day from solid foods only) and intake of the following food groups in servings/day: fruits, vegetables, red meat, processed meat, fish, and legumes. While the DAG suggests that energy intake could potentially mediate the association between the BQI and diabetes incidence, it is important to note that energy intake was measured concurrently with the exposure. Thus, energy intake cannot be considered a mediator but rather a confounder in the association. Therefore, we decided to further adjust our multivariable model 3 for energy intake to account for its likely role as a confounder, and to minimize measurement error in dietary intake estimations [27].

Restricted cubic splines were used to examine the dose-response relationship between the BQI and diabetes incidence, using model 3. A BQI score of 25, was set as the reference and knots were placed at the 5th, 35th, 65th and 95th percentiles. The Wald chi-square test was carried out for testing nonlinearity of the association [28].

We performed sensitivity analyses to test the robustness of our findings. First, we repeated the analyses using a different definition of diabetes, including only women who reported a diagnosis plus medical treatment for diabetes. Using this definition, 87.0% (95% CI: 84.8, 89.0) of participants classified as having diabetes confirmed the diagnosis and treatment in the supplementary questionnaire. Second, we excluded participants who developed diabetes during the first three years of follow-up to address possible reverse causation, although this could introduce selection bias [29]. Third, we used a different categorization of the BQI: <20, ≥ 20-<30, ≥ 30-<40, ≥ 40-<50, ≥ 50-<60, and ≥ 60. Fourth, for understanding of whether associations were driven by individual components of the BQI, we recalculated scores by removing 1 of the 7 components at a time, and adjusted for the excluded component in model 3. Fifth, we excluded energy from beverages from the BQI because of overlap with other components, e.g. SSBs. Sixth, we used an alternative BQI, moving coffee, atole with milk and yogurt drink to the SSBs component, assuming they were sweetened. BMI was not included in the main analysis because it may act as a confounder, mediator, or effect modifier in the diet-diabetes associations. Nevertheless, we additionally adjusted model 3 for baseline BMI (categories: normal weight, overweight, obesity and unknown). Eight, we examined interaction terms of the BQI with age (years), BMI (continuous and categories), and socioeconomic position (categories). Finally, we conducted prespecified subgroup analyses by: age (below and above the median), BMI status (< 25 and ≥ 25kg/m²), and socioeconomic position, using health insurance for serious conditions as a proxy (public and private).

We used missing indicator variables (category unknown) to handle partially missing confounder information for BMI, smoking, and menopausal status (from 3.0 to 8.9%). SAS software version 9.4 was used to calculate the BQI scores, and R version 4.0.2 was used to perform all analyses. Two-sided p-values < 0.05 were considered statistically significant, also for interaction terms.

Compared with those in lower BQI categories, participants in the highest BQI category (≥ 55) were more likely to be from the northern region of the country, older, postmenopausal, current smokers, to have obesity, greater access to internet, and family history of diabetes (Table 2). Participants lost-to follow-up (n = 9244), were more likely to be from the southern region of the country, older and less likely to have access to internet compared to respondents (Supplemental Table 1). BQI scores ranged from 9.8 to 69.3, with a mean (SD) of 37.3 (8.6). The most consumed beverage was SSBs [median(IQR): 315 (169–579) mL/day], while alcoholic drinks were the least consumed [0.46 (0.0-1.42) glasses/day] (Table 2). These consumption patterns are reflected in the component scores, with alcohol having the highest score, and SSBs and energy from beverages having the lowest scores (Supplemental Table 2). Dietary intakes were generally similar across BQI categories, except for total energy and fruit intakes, with participants in the lowest BQI category (< 25) showing higher intakes compared to participants in higher BQI categories (Table 2).

Table 2

Baseline characteristics of 77 484 participants of the Mexican Teachers’ Cohort and average dietary intake by categories of the BQI
		BQI score, categories
	Total (n = 77 484)	< 25 (n = 5014)	≥ 25-<35 (n = 26 789)	≥ 35-<45 (n = 31 630)	≥ 45-<55 (n = 11 688)	≥ 55 (n = 2363)
Age, years	45.9 (7.2)	41.7 (7.3)	41.7 (7.2)	42.0 (7.2)	42.8 (7.2)	44.0 (7.1)
BMI, kg/m²	27.2 (4.7)	26.7 (4.4)	27.0 (4.6)	27.2 (4.5)	27.4 (4.6)	27.7 (4.7)
BMI, categories, %
Normal weight	32.4	35.7	34.1	32.0	30.4	28.7
Overweight	37.2	37.9	37.4	38.5	38.7	39.2
Obesity	21.0	17.8	20.1	20.8	22.0	23.7
Unknown	9.4	8.7	8.4	8.7	8.9	8.4
Mexican Regions, %*
North	12.8	10.3	14.9	19.1	25.3	34.7
Central	18.2	22.9	19.3	17.1	13.9	12.7
Mexico City and State of Mexico	15.3	25.9	23.2	23.2	23.8	25.2
South	53.7	40.9	42.6	40.6	37.0	27.4
Insurance - serious condition, %
Social Security	70.2	72.3	70.5	70.8	69.0	67.5
Private	20.3	19.1	19.0	17.6	17.7	17.6
Other	9.5	8.6	10.5	11.6	13.3	14.9
Internet access at home, %	39.5	49.0	47.8	48.4	51.9	61.9
Family history of diabetes, %	44.4	45.2	45.4	45.9	48.1	51.6
Menopausal status, %
Premenopausal	73.9	78.1	78.3	77.7	74.6	71.9
Postmenopausal	15.9	13.3	13.2	13.3	15.7	18.0
Unknown	10.2	8.7	8.5	9.0	9.7	10.1
Physical Activity, METs/week†
Tertile 1, %	33.4	29.7	31.9	31.9	34.3	33.1
Tertile 2, %	33.0	32.3	33.7	34.8	33.6	32.9
Tertile 3, %	33.6	38.1	34.4	33.2	32.0	33.9
Smoking status, %
Current smoker	10.5	7.4	8.5	9.3	10.3	14.0
Past smoker	13.4	11.5	11.1	11.5	12.6	16.4
Never smoker	72.4	78.4	77.5	76.1	74.1	66.5
Unknown	3.7	2.7	2.9	3.2	3.1	3.1
Dietary intake
Total energy, kcal/day	1817(623)	2195 (640)	1901 (627)	1748 (591)	1676 (608)	1691 (571)
Fish, s/day	0.5 (0.3–0.8)	0.6 (0.3–0.9)	0.5 (0.3–0.8)	0.5 (0.3–0.8)	0.5 (0.3–0.8)	0.5 (0.3–0.9)
Fruit, s/day	3.1 (1.8–4.9)	4.6 (2.9-7.0)	3.3 (1.9–5.2)	2.8 (1.7–4.5)	2.8 (1.6–4.5)	2.9 (1.6–4.5)
Vegetables, s/day	2.2 (1.3–3.5)	2.7 (1.6–4.2)	2.2 (1.3–3.6)	2.0 (1.2–3.4)	2.1 (1.2–3.5)	2.3 (1.3–3.9)
Legumes, s/day	0.5 (0.2–0.8)	0.5 (0.2–0.9)	0.5 (0.2–0.8)	0.5 (0.2–0.8)	0.5 (0.2–0.8)	0.4 (0.2–0.8)
Whole grains, s/day	2.1 (1.1–3.2)	2.4 (1.2–3.2)	2.0 (1.1–3.1)	2.0 (1.1–3.1)	2.5 (1.2–3.2)	2.6 (1.2–3.6)
Processed meat, s/day	0.5 (0.3–0.9)	0.5 (0.3–0.9)	0.5 (0.3–0.9)	0.5 (0.3–0.9)	0.5 (0.3–0.9)	0.6 (0.3–0.9)
Red meat, s/day	1.2( 0.7–1.9)	1.2 (0.8-2.0)	1.2 (0.8–1.9)	1.2 (0.7–1.9)	1.1 (0.7–1.9)	1.2 (0.7-2.0)
Coffee, mL/day	103 (16–240)	20 (8–69)	40 (8-189)	111 (20–240)	189 (34–291)	240 (189–600)
Milk, mL/day	179 (74–282)	447 (45–634)	169 (59–409)	189 (103–259)	143 (59–240)	155 (103–237)
Juices, mL/day	36 (8-107)	196 (107–250)	36 (21–107)	21 (8-107)	21 (8–36)	8 (8–36)
SSBs, mL/day	315 (169–579)	464 (306–772)	401 (260–694)	301 (173–516)	122 (62–228)	52 (30–88)
Alcohol, glasses/day	0.46 (0-1.42)	0.46 (0-1.42)	0.46 (0-1.42)	0.46 (0-1.42)	0.46 (0-1.57)	0.46 (0-1.57)
Added sugar to beverages, g/day	8.40 (4–13)	13 (8–21)	8 (4–13)	8 (4–8)	4 (0–8)	0 (0–4)
Energy from beverages, kcal/day	362 (207)	594 (243)	442 (208)	331 (163)	206 (113)	149 (62)
BQI score‡	37.3 (8.6)	21.7 (2.7)	30.7 (2.8)	39.5 (2.8)	48.9 (2.7)	58.6 (2.9)
Values are mean (SD) for normally distributed continuous data, median (IQR) for skewed continuous data or % for categorical data. Proportion of missing data was 8.6% for BMI, 8.9% for menopausal status, and 3.0% for smoking.^*North: Baja California, Durango, Nuevo León, Sonora; Central: Guanajuato, Hidalgo, Jalisco; South: Chiapas, Veracruz, Yucatán. †Tertiles based on the total study population. ‡BQI theoretical score range from 0 to70. Abbreviations: BQI, beverage quality index; METs, metabolic equivalents of task; s, serving; SSBs, sugar-sweetened beverages.

After a median follow-up time of 7.6 years (568 346 person-years), 4521 participants developed diabetes. After adjusting for confounders (model 3), when comparing extreme BQI categories (≥ 55 vs. <25), a higher BQI was associated with a lower incidence of diabetes (HR: 0.87; 95% CI: 0.71, 1.06), although there was uncertainty in the estimation (Table 3). In continuous analysis (P for nonlinearity = 0.20), using restricted cubic splines, HRs were close to 1 and non-significant across the entire range of BQI values. (Fig. 1).

Table 3

Risk of incident diabetes by categories of the BQI among 77 484 participants of the Mexican Teachers’ Cohort
	BQI score, categories
mean (SD)	< 25 21.7(2.7)	≥ 25-<35 30.7 (2.8)	≥ 35-<45 39.5 (2.8)	≥ 45-<55 48.9 (2.7)	≥ 55 58.6(2.9)
n/events	4715/299	25 323/1466	29 755/1875	10 950/738	2220/143
Person-years	37,582	197,696	231,020	84,806	17,242
Model 1	Ref.	0.96 (0.84, 1.08)	1.05(0.93, 1.19)	1.11 (0.97, 1.27)	1.01 (0.82, 1.23)
Model 2	Ref.	0.94 (0.83, 1.07)	1.01 (0.89, 1.14)	1.03 (0.90, 1.18)	0.88 (0.72, 1.08)
Model 3	Ref.	0.93 (0.82, 1.06)	0.99 (0.88, 1.12)	1.01 (0.88, 1.16)	0.87 (0.71, 1.06)
Estimates are HR (95% CI) from Cox proportional hazards models. Model 1 was adjusted for age. Model 2 was additionally adjusted for area of residence, socioeconomic position, physical activity, smoking, family history of diabetes, and menopausal status. Model 3 was additionally adjusted for food groups intake (fruits, vegetables, red meat, processed meat, fish, and legumes), and energy intake from foods. Abbreviations: BQI, beverage quality index; CI, confidence intervals; HR, hazard rations; SD, standard deviation.

In sensitivity analyses, using a stricter definition of incident diabetes, showed a similar association (HR model 3, BQI ≥ 55 vs. <25: 0.87; 95% CI: 0.70, 1.09) (Supplemental Table 3). Findings did not differ when excluding participants who developed diabetes within the first 3 years of follow-up (Supplemental Table 4), using an alternative categorization of the BQI (Supplemental Table 5), excluding individual BQI components one at a time (Supplemental Table 6), excluding energy from beverages from the BQI (Supplemental Table 7), moving sweetened coffee and dairy beverages to the SSBs component (Supplemental Table 8), or further adjusting for BMI (Supplemental Table 9).

BQI-diabetes associations were consistent among subgroups of age and socioeconomic position (all P _interaction >0.05), but not among subgroups of BMI (P _interaction= 0.016) (Supplemental Tables 10–15). Among participants with overweight and obesity, a higher BQI (≥ 55 vs. <25) was associated with lower diabetes incidence (HR: 0.77; 95% CI: 0.61, 0.96) in multivariable analysis (Supplemental Table 13).

We applied the BQI to assess overall beverage intake quality in a large cohort of Mexican women. There was suggestive evidence for a lower diabetes incidence among women in the upper category of BQI (indicating higher beverage quality), but findings were not statistically significant, and could not be confirmed in continuous analysis.

Research on overall beverage quality assessed with a priori developed beverage scores and cardiometabolic outcomes is limited. Two previous studies showed that higher scores on the HBI and the HBS were associated to lower cardiometabolic risk factors, less kidney function decline, and a lower risk of mortality in US populations [13, 14]. However, two other studies that used the HBI in European population-based cohorts did not find associations with cardiometabolic outcomes, including diabetes [30, 31]. In a previous study, we found that a higher BQI (T3 vs. T1), was associated with lower risk of recurrent cardiovascular disease (CVD), lower risk of CVD mortality and with a potential decrease of diabetes incidence in Dutch patients with history of myocardial infraction from the Alpha Omega Cohort [15]. Similarly, in our population of Mexican women with different beverage patterns, a potential decrease in diabetes incidence was observed.

The BQI was developed based on the current Dutch dietary guidelines [19, 32]. Based on our findings, we conclude that the BQI may not optimally reflect guidelines or intake patterns in Mexico. Previously, Rivera et.al, adapted beverage guidelines developed for the US population to the Mexican context [33]. These guidelines prioritize water, skimmed and semi-skimmed milk consumption, and in third place, unsweetened coffee and tea, and promote the limitation of full-fat milk intake. Unlike the Dutch dietary guidelines, which do not provide recommendations for milk with varying fat content due to insufficient evidence yet to make distinctions between high- and low-fat dairy (milk) [19], the Mexican context, characterized by a relatively high proportion of the population with overweight or obesity, may warrant differentiated scoring based on milk fat content.

Existing literature lends support to health effects of the individual BQI components. The unhealthy beverage components of the BQI, like SSBs, have been consistently associated with higher risk of diabetes [6]. A previous study within the MTC showed that soda consumption is associated with a higher rate of diabetes in a magnitude similar to that reported in other populations (40), and with a higher risk of other cardiometabolic diseases, such as hypertension [34]. On the other hand, coffee and tea, and possibly dairy beverages (yogurt) have been shown to be largely protective for cardiometabolic health [7, 8, 35]. However, in Mexico, healthy beverages such as coffee, tea and dairy beverages tend to be drivers of added sugar intake. Most participants (83%) added sugar to their coffee and tea, and the Mexican beverage pattern is also characterized by added sugar in other popular beverages such as aguas frescas and atoles on top of high SSBs intake. Besides, the most consumed type of coffee in Mexico is instant coffee, rather than the filtered coffee consumed in the Netherlands, for which associations with cardiometabolic outcomes are not fully clear [36, 37]. Teas in Mexico tend to be (sweetened) herbal teas, in contrast to (mostly unsweetened) black tea which is popular in the Netherlands. These factors may counteract a potential beneficial effect of dairy beverages, tea, and coffee intake. And therefore, make this score less suitable to estimate beverage quality in this population.

Moreover, the unhealthy beverage pattern of the participants from the MTC, characterized by high consumption of beverages with added sugar and thus a considerable proportion of energy coming from beverages (20%), was relatively homogeneous. Almost all participants consumed SSBs (99.5%), including those in the highest BQI category. This relative homogeneity in beverage intake patterns affected the scoring, placing most participants in the central categories of the BQI, which hampered comparisons.

Results from sensitivity analyses were in line with main findings, showing a potential lower risk of developing diabetes at higher BQI scores. Also, excluding individual components from the BQI did not greatly change the estimates. However, when excluding the component energy from beverages, the HR decreased the most, although the uncertainty around the estimate was high. We scored calories from beverages separately from other beverage components in the BQI. This approach was taken because cohort studies often adjust for total energy intake, not allowing for ad libitum intake (and overconsumption) of caloric beverages. Therefore, the role of (excessive) energy intake from beverages on chronic disease risk may be underestimated. Healthy, non-caloric alternatives (e.g. water, unsweetened coffee and tea) are available for sweetened beverages, and there is no need to quench thirst with caloric beverages, providing additional rationale for this penalty.

Upon further subgroup analysis, among participants with overweight and obesity, those with the healthiest beverage intake patterns exhibited lower diabetes incidence, with a significant interaction observed between BQI and BMI. BMI could act as a confounder, mediator or effect modifier in the BQI-diabetes association. Therefore, these findings need to be interpreted with caution because BMI could also be a collider (Supplemental Fig. 1), and conditioning on a collider could result in overadjustment bias [38].

Strengths of this study include its large sample size and prospective design with long-term follow-up, access to clinical data for diabetes case identification, detailed information on risk factors for diabetes enabling extensive control for confounding, and the conducted sensitivity analyses. Some limitations of this study should be acknowledged. Self-reported dietary intakes are prone to non-differential misclassification. Aguas frescas and atoles together with sweetened probiotic fermented milk drinks were categorized as SSBs due to their added sugar content, however, their role in cardiometabolic risk is unknow. These beverages may have potential beneficial effects (probiotics, fruits) that could counterbalance the negative effects of added sugar, which may have introduced misclassification biasing the HRs towards the null. Future studies need to clarify the role of these popular beverages on cardiometabolic health. In the FFQ, participants were asked about their added sugar intake in foods and beverages. While we could not differentiate the specific amounts for each, we attributed all reported added sugar intake to beverages such as coffee, tea, aguas frescas, and atoles. This assumption is based on the common practice in Mexico of sweetening these beverages, whereas sugar is generally not added to foods. Approximately 11% of participants did not respond to follow-up questionnaires and were considered lost to follow-up, potentially resulting in selection bias. Those lost to follow-up were slightly older and less likely to have access to internet (proxy for socioeconomic status) at baseline compared to participants included in the study. However, no other significant differences were observed. Finally, the BQI was not specifically created for the Mexican population, thus it may not be the most suitable scoring system to evaluate beverage quality in populations that consume significant amounts of beverages with added sugar.

In conclusion, in a cohort of Mexican women, the BQI for overall beverage quality showed no consistent association with T2D. Future studies should explore and develop composite beverage scores that better reflect population-specific beverage intake patterns and identify optimal beverage intake patterns for cardiometabolic health.

BQI

Beverage quality index

CVD

cardiovascular disease

DAG

directed acyclic graph

ENSANUT (acronym in Spanish)

Mexican National Survey of Health and Nutrition

FFQ

food-frequency questionnaire

HBI

healthy beverage index

HBS

healthy beverage score

hazard ratio

IQR

interquartile range

METs

metabolic equivalents of task

MTC

Mexican Teachers’ Cohort

standard deviation

SSBs

sugar-sweetened beverages.

Acknowledgements

We thank Adrian Cortes and Antonio Garcia-Anaya for assistance with data management. We also thank ISSSTE (Social Security and Services Institute for Employees of the State) and Victor Sastre from Mexico’s Ministry of Education for their technical and administrative support. In particular, we would like to thank Dr. Vesta Louise, Richardson López Collada from ISSSTE for providing the MTC access to the Comprehensive Management of Diabetes by Stages (MIDE) program and the Observatory of Diabetes and Chronic Diseases (ODEC). We are extremely grateful to the Mexican Teachers’ Cohort participants, as, without their participation, this study would not have been possible.

Author Contributions

M.G.J.C., A.M., N.K., M.G. and D.S. conceptualized and designed the study. A.M. created the diabetes dataset to conduct the research. M.G.J.C. and A.M. performed the analysis with supervision of D.S. M.G.J.C. wrote the first draft of the manuscript, and all authors edited, reviewed, contributed to the interpretation of results, and approved the final version of the manuscript. D.S. is the guarantor of this work and, as such, had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Competing Interests

No potential conflicts of interest relevant to this article were reported.

Data Availability Statement

The datasets generated during and/or analysed in the current study are available from the corresponding author upon reasonable request.

Funding

Financial support for the Mexican Teachers’ Cohort was obtained from the American Institute for Cancer Research (grant number 05B047) and the Consejo Nacional de Ciencia y Tecnología (CONACyT) (grant number S0008-2009- 1: 000000000115312). MGJC was supported by a scholarship from the CONACYT (grant number 2018-000009-01EXTF-00323).

Bello-Chavolla OY, Rojas-Martinez R, Aguilar-Salinas CA, Hernández-Avila M (2017) Epidemiology of diabetes mellitus in Mexico. Nutr Rev 75:4–12
Basto-Abreu A, López-Olmedo N, Rojas-Martínez R, Aguilar-Salinas CA, Moreno-Banda GL, Carnalla M et al (2023) Prevalencia de prediabetes y diabetes in México: Ensanut 2022. Salud Publica Mex 65:s163–s168
Sun H, Saeedi P, Karuranga S, Pinkepank M, Ogurtsova K, Duncan BB et al (2022) IDF Diabetes Atlas: Global, regional and country-level diabetes prevalence estimates for 2021 and projections for 2045. Diabetes Res Clin Pract 183:109119
Seiglie JA, Franco RR, Wirtz VJ, Meigs JB, Mendoza MA, Miranda JJ et al (2021) Regional and state-level patterns of type 2 diabetes prevalence in Mexico over the last three decades. Diabetes Res Clin Pract 177:108927
Meza R, Barrientos-Gutierrez T, Rojas-Martinez R, Reynoso-Noverón N, Palacio-Mejia LS, Lazcano-Ponce E et al (2015) Burden of type 2 diabetes in Mexico: Past, current and future prevalence and incidence rates. Prev Med 81:445–450
Malik VS, Hu FB (2022) The role of sugar-sweetened beverages in the global epidemics of obesity and chronic diseases. Nat Rev Endocrinol 18:205–218
Yang J, Mao QX, Xu HX, Ma X, Zeng CY (2014) Tea consumption and risk of type 2 diabetes mellitus: A systematic review and meta-analysis update. BMJ Open 4(7):e005632
Carlstro M, Larsson SC (2018) Coffee consumption and reduced risk of developing type 2 diabetes: A systematic review with meta-analysis. Nutr Revl 76:395–417
Sánchez-Pimienta TG, Batis C, Lutter CK, Rivera JA (2016) Sugar-sweetened beverages are the main sources of added sugar intake in the Mexican population. J Nutr 146:1888S–1896S
Stern D, Piernas C, Barquera S, Rivera JA, Popkin BM (2014) Caloric beverages were major sources of energy among children and adults in Mexico, 1999–2012. J Nutr 144:949–956
Braverman-Bronstein A, Camacho-García-Formentí D, Zepeda-Tello R, Cudhea F, Singh GM, Mozaffarian D et al (2020) Mortality attributable to sugar sweetened beverages consumption in Mexico: an update. Int J Obes 44:1341–1349
Neuhouser ML (2019) The importance of healthy dietary patterns in chronic disease prevention. Nutr Res 70:3–6
Duffey KJ, Davy BM (2015) The Healthy Beverage Index Is Associated with Reduced Cardiometabolic Risk in US Adults: A Preliminary Analysis. J Acad Nutr Diet 115:1682–1689e2
Hu EA, Anderson CAM, Crews DC, Mills KT, He J, Shou H et al (2020) A Healthy Beverage Score and Risk of Chronic Kidney Disease Progression, Incident Cardiovascular Disease, and All-Cause Mortality in the Chronic Renal Insufficiency Cohort. Curr Dev Nutr 4:nzaa088
Jacobo Cejudo MG, van Laarhoven R, Voortman T, Geleijnse JM (2024) The Beverage Quality Index and Risk of Cardiometabolic Outcomes after a Myocardial Infarction: A Prospective Analysis in the Alpha Omega Cohort. Nutr Metab Cardiovasc Dis S0939–4753(24):00197–00192
Lajous M, Ortiz-Panozo E, Monge A, Santoyo-Vistrain R, García-Anaya A, Yunes-Díaz E et al (2017) Cohort profile: The Mexican Teachers’ Cohort (MTC). Int J Epidemiol 46(2):e10
Hernández-Avila M, Romieu I, Parra S, Hernández-Avila J, Madrigal H, Willett W (1998) Validity and reproducibility of a food frequency questionnaire to assess dietary intake of women living in Mexico City. Salud Publica Mex 40:133–140
Department of Agriculture US (2009) ARS USDA National Nutrient Database for Standard Reference, Release 22. Methods and Application of Food Composition Laboratory Home Page. Accessed 13 April 2023. http://www.ars.usda.gov/nea/bhnrc/mafcl
Health Council of the Netherlands. Dutch dietary guidelines 2015 (2015) https://www.healthcouncil.nl/documents/advisory-reports/2015/11/04/dutch-dietary-guidelines-2015
Qin P, Li Q, Zhao Y, Chen Q, Sun X, Liu Y et al (2020) Sugar and artificially sweetened beverages and risk of obesity, type 2 diabetes mellitus, hypertension, and all-cause mortality: a dose–response meta-analysis of prospective cohort studies. Eur J Epidemiol 35:655–671
Drouin-Chartier JP, Zheng Y, Li Y, Malik V, Pan A, Bhupathiraju SN et al (2019) Changes in consumption of sugary beverages and artificially sweetened beverages and subsequent risk of type 2 diabetes: Results from three large prospective U.S. Cohorts of women and men. Diabetes Care 42:2181–2189
López GE, Batis C, González C, Chávez M, Cortés-Valencia A, López-Ridaura R et al (2022) EAT-Lancet Healthy Reference Diet score and diabetes incidence in a cohort of Mexican women. Eur J Clin Nutr 77(3):348–355
World Health Organization (2004) ;1:332–45. http://www.who.int/classifications/icd/icdonlineversions/en/
Hirko KA, Lajous M, Ortiz-Panozo E, Lopez-Ridaura R, Christine PJ, Lê-Scherban F et al Socioeconomic position and markers of adiposity among female teachers in Mexico. J Epidemiol Community Health. 2017:jech–2017
Medina C, Monge A, Romero M, López-Ridaura R, Barquera S, Romieu I et al (2021) Reliability and validity of the Mexican teachers’ physical activity questionnaire (MTPAQ) in a subsample of female Mexican teachers. BMC Sports Sci Med Rehabil 13(1):143
Ortiz-Panozo E, Yunes-Díaz E, Lajous M, Romieu I, Monge A, López-Ridaura R (2017) Validity of self-reported anthropometry in adult Mexican women. Salud Publica Mex 59:266–275
Kipnis V, Subar AF, Midthune D, Freedman LS, Ballard-Barbash R, Troiano RP et al (2003) Structure of dietary measurement error: Results of the OPEN biomarker study. Am J Epidemiol 158:14–21
Desquilbet L, Mariotti F (2010) Dose-response analyses using restricted cubic spline functions in public health research. Stat Med 29:1037–1057
Hernán MA, Hernández-Díaz S, Robins JM (2004) A structural approach to selection bias. Epidemiology 15:615–625
Wagner S, Merkling T, Girerd N, Bozec E, Berghe L, Van Den, Hoge A et al (2022) Quality of Beverage Intake and Cardiometabolic and Kidney Outcomes: Insights From the STANISLAS Cohort. Front Nutr 8:738803
Jacobo Cejudo MG, Ochoa-Rosales C, Ahmadizar F, Kavousi M, Geleijnse JM, Voortman T (2023) The healthy beverage index is not associated with insulin resistance, prediabetes and type 2 diabetes risk in the Rotterdam Study. Eur J Nutr 62:3021–3031
Kromhout D, Spaaij CJK, De Goede J, Weggemans RM, Brug J, Geleijnse JM et al (2016) The 2015 Dutch food-based dietary guidelines. Eur J Clin Nutr 70:869–878
Rivera JA, Muñoz-Hernández O, Rosas-Peralta M, Aguilar-Salinas CA, Popkin BM, Willett WC (2008) Consumo de bebidas para una vida saludable: recomendaciones para la población mexicana. Salud Publica Mex 50:173–195
Monge A, Silva Canella D, López-Olmedo N, Lajous M, Cortés-Valencia A, Stern D (2021) Ultraprocessed beverages and processed meats increase the incidence of hypertension in Mexican women. Br J Nutr 126(4):600–611
Salas-Salvadó J, Guasch-Ferré M, Díaz-López A, Babio N (2017) Yogurt and Diabetes: Overview of Recent Observational Studies. J Nutr 147:1452S–61S
Chieng D, Canovas R, Segan L, Sugumar H, Voskoboinik A, Prabhu S et al (2022) The impact of coffee subtypes on incident cardiovascular disease, arrhythmias, and mortality: long-term outcomes from the UK Biobank. Eur J Prev Cardiol 29(17):2240–2249
Alperet DJ, Rebello SA, Khoo EY, Tay Z, Seah SS, Tai B et al (2020) The effect of coffee consumption on insulin sensitivity and other biological risk factors for type 2 diabetes: a randomized placebo-controlled trial. Am J Clin Nutr 111(2):448–458
Cole SR, Platt RW, Schisterman EF, Chu H, Westreich D, Richardson D, Poole C (2010) Illustrating bias due to conditioning on a collider. Int J Epidemiol 39:417–420

There is NO conflict of interest to disclose

OnlineonlySupplementalMaterial.pdf

Download PDF

Review #1 received at journal
21 Aug, 2024
Reviewer #1 agreed at journal
31 Jul, 2024
Reviewers invited by journal
10 Jul, 2024
Submission checks completed at journal
10 Jul, 2024
Editor assigned by journal
09 Jul, 2024
First submitted to journal
09 Jul, 2024

You are reading this latest preprint version

The Beverage Quality Index and Type 2 Diabetes Risk in Women: A Prospective Analysis of the Mexican Teachers’ Cohort

Status:

Version 1

Abstract

Figures

INTRODUCTION

SUBJECTS AND METHODS

Study design and population

Dietary assessment

Beverage Quality Index (BQI)

Assessment of diabetes

Assessment of covariables

Statistical analysis

RESULTS

DISCUSSION

Abbreviations

Declarations

References

Additional Declarations

Supplementary Files

Status:

Version 1