Associations between Avocado Intake and Lower Rates of Incident Type 2 Diabetes in US Adults with Hispanic/Latino Ancestry

Background/Purpose: Hispanic/Latinos in the US are at increased risk for type 2 diabetes (T2D). Data suggest that avocado intake is associated with better glycemic control, but whether this translates to protection from T2D has not been studied. The goal of the current analyses was to examine whether consuming avocados at baseline is associated with lower incident T2D over a six-year period, compared to not consuming avocados at baseline. Subjects/Methods: Using data from a large population of US adults with Hispanic ancestry, without known or unknown T2D at baseline (N = 6159), participants were classified as avocado consumers (N = 983) or non-consumers (N = 5176) based on the mean of two 24-hour dietary recalls. Cox proportional hazard models estimated the association of avocado consumption with incident T2D (N = 656 cases) over a six-year follow-up period, in the population as a whole, and separately in those with normoglycemia vs. prediabetes at baseline. A set of three sequential models were run: the first controlling only for sociodemographic factors (“minimally adjusted” models), the second for these and health behaviors (“fully adjusted” models), and a third for both sets of covariates and


Introduction
Over 22 million U.S. adults have a diagnosis of T2D [1], and the prevalence is ~66% higher in Americans with Hispanic ancestry (Hispanic/Latinos) when compared to non-Hispanic whites [2]. Diet remains one of the main modifiable risk factors for T2D. In addition to reducing overall energy intake and increasing whole grain consumption [3] [4], some (but not universal) evidence supports a role for limited amounts of monounsaturated fatty acids (MUFAs) in supporting glycemic control [5] [6] [7] [8] [9], although any beneficial effects may occur only in those with dysglycemia, and most strongly in those meeting criteria for T2D [10] [11] [12].
The relatively high MUFA content of avocado (~6.65 g per USDA serving) suggests this fruit may offer protection from T2D, and is thought to be one reason why data have observed a lower incidence of incident cardiovascular disease (CVD) in avocado consumers vs. non consumers [13]. However, despite two intervention studies demonstrating an effect of avocado intake on postprandial glucose control [14] [15], and data from one observational study showing avocado consumption is associated with a lower prevalence of the metabolic syndrome [16], and from our own study showing an association with better glycemic control in those with T2D (submitted for publication and available upon request), research has not examined whether the same protection has been observed for incident T2D.
To address this knowledge gap, the goal of the current analyses was to examine the association of consuming avocado intake with incident T2D, in a large, geographically diverse sample of Hispanic/Latinos.

Methods
The data included in the present study were downloaded from NHLBI Biological Specimen and Data Repository Information Coordinating Center (BioLINCC; to the data and assumed responsibility for the integrity of the analyses and findings as reported.

Participants and Procedure
HCHS/SOL is a community based prospective cohort study of US Hispanic/Latinos [17] [18]. We included all participants who consented for their data to be included in

Sampling Design
The study design for HCHS/SOL is described in detail elsewhere [17]. In brief, a stratified two-stage area probability sample of household addresses was selected in each of the four field centers. The first sampling stage randomly selected census block groups with stratification based on Hispanic/Latino concentration and proportion of high/low socio-economic status. The second sampling stage randomly selected households, with stratification, from US Postal Service registries that covered the randomly selected census block groups. Both stages oversampled certain strata to increase the likelihood that a selected address yielded a Hispanic/Latino household. After households were sampled, in-person or telephone contacts were made to screen eligible households and to roster its members. Lastly, the study oversampled the 45 -74 age group (n = 9714, 59.2%) to facilitate examination of target outcomes. As a result, participants included in HCHS/SOL cohort were selected with unequal probabilities of selection, and these probabilities were considered during data analysis to appropriately represent the target population. HCHS/SOL sampling weights are the product of a "base weight" (reciprocal of the probability of selection) and three adjustments: 1) non-response adjustments made relative to the sampling frame, 2) trimming to handle extreme values (to avoid a few weights with extreme values being overly influential in the analyses), and 3) calibration of weights to the 2010 US Census according to age, sex, and Hispanic heritage [17], and included in all analyses using inferential statistics.

Measures
The baseline in-person clinic visit included comprehensive assessment of biological (e.g., anthropometrics, blood draw, OGTT, ankle brachial pressure index, electrocardiogram), behavioral (e.g., dietary intake assessed with two 24-hour recalls, physical activity assessment by accelerometer and self-report, overnight sleep exam for apneic events, tobacco and alcohol assessed by self-report), and sociodemographic (e.g., socioeconomic status, migration history) factors [17] [18].
Incident diabetes was defined as participants who were free of diabetes at the baseline visit and were identified as having diabetes during the follow-up period.
The date of an incident diabetes event was defined as time of the first self-report

Dietary Intake
Participants completed two 24-hour dietary recalls, using a multi-pass technique, at an interval of 5 -45 days apart. The first diet recall covered the 24-hour period prior initiating a fast for the baseline clinic visit during which time the blood draw was collected although 24-hour recalls are subject to self-report biases and error inherent in all self-reported nutrition data, validity coefficients suggest these are not higher in Hispanic/Latino data, including that from HCHS/SOL, than expected in other US populations [20] [21]. Nutrient and food group values were derived in the Nutrition Data System for Research (NDSR [22]). To minimize for the known biases in self-reported 24-hour recall data, nutrient intake was predicted as specified by the NCI method [23], using single component SAS macros developed at NCI (available at: http://riskfactor.cancer.gov/diet/usualintakes/macros.html). This method estimates the within and between person components and corrects for the high intra-individual variation intrinsic to 24-h recalls given that individuals do not eat the same foods every day [23]. Avocado intake in servings/day and total energy intake (kcal/day) were derived from an average of the two 24-hour recalls (as avocado is an episodically consumed food, participants with only one 24-hour dietary recall were excluded).

Demographics
Age, sex, education level, smoking status, Hispanic/Latino heritage/nativity, and preferred language were obtained through in-person interview with trained assessors.

Acculturation
Following previous epidemiological studies that include Hispanic/Latinos living in the US [24], an acculturation score was constructed from three proxy meas- two scores were summed to obtain an acculturation score from 0 (least acculturated) to 5 (most acculturated) [25].

Physical Activity
Physical activity was measured in metabolic equivalent of task units (METs) of moderate and vigorous physical activity, as self-reported on an adapted version of the World Health Organization (WHO) Global Physical Activity Questionnaire [26], which asks participants to report the time spent in physical activity, and the type of activity conducted during an average week, across three life domains (work-related, transportation, and leisure/recreational).

Anthropometric Measures
Height and weight were measured by trained study staff. BMI was calculated as weight in kilograms (kg) divided by height in meters (m) squared (kg/m 2 ).

Analyses
All analyses were conducted using the latest version of R software (version 4.0.5) [27], and all values and parameter estimates are adjusted for the sampling strategy using the "survey" package in R [28], with the exception of sample sizes (N and %) for participant characteristics, which are presented as unweighted.

Participant Characteristics
Demographic information stratified by avocado intake (non-consumer vs. consumer) were calculated as a weighted mean (±standard error; SE) for continuous variables, or unweighted total number (N) and weighted percentage (%) for categorical or ordinal variables. Differences by avocado consumption, without controlling for covariates, were conducted using weighed t-tests for continuous variables (transformed where necessary) and chi-squared tests of difference for categorical variables. Although unweighted counts are provided for the number of participants with each measure, weighted degrees of freedom (df) from tests of difference corrected for sampling strategy are presented throughout.

Association of Avocado Intake with Incident Diabetes
The association of avocado intake with incident T2D was examined via coxproportional hazards models, corrected for the complex sampling design with sample weights, which accounted for the reduction in variance from stratified sampling and the increase in variance from having only a small number of clusters. The hazard ratio for T2D associated with being an avocado consumer vs.
non was calculated, using time at risk until the first diagnosis, or the last followup (if no T2D occurred).
The association between avocado intake and time to T2D was examined in three sets of models: in all models, avocado intake was specified as categorical variable (non-consumer vs. consumer). Model 1 specified age, sex, education level, total energy intake (kilocalories/day), and heritage, and acculturation as cova- ing the possibility that BMI could be a confounder, not a mediator, a final set of models were run which included the same covariates as the fully adjusted models with the addition of BMI ("fully adjusted + BMI" models). Hazard ratios (HR) and 95% confidence intervals (CIs) were reported for each model.

Participant Characteristics
Sociodemographic factors, health behaviors and BMI, stratified by baseline avocado consumption (consumer vs. non) are presented in Table 1. Non-consumers did not differ from consumers in age, gender education level, acculturation, physical activity, alcohol intake alcohol consumption, smoking status, nor BMI. Avocado consumers tended to have a better overall dietary quality as measured by the AHEI2010 (t = 6.5, df = 609, P < 0.001; Table 1), a lower BMI (t = 4.3, df = 609, P < 0.001; Table 1) and have a different distribution of Hispanic heritage (c 2 = 7.5, df = 5.1, P < 0.001; Table 1) compared to non-consumers.  Table 2). When stratified by baseline dysglycemia, there was a significant inverse association in those with prediabetes, who saw a reduction in risk of 31% (HR: 0.69 (0.48 -0.98), P = 0.04; Table 2, Figure   2). While an effect was seen in those with normoglycemia (20% reduction in risk), the inverse association between avocado consumption and risk of T2D

Association of Avocado Intake with Incident Diabetes
was not significant in this group (HR: 0.80 (0.41 -1.55), P = 0.50; Table 2

Discussion
Using data from HCHS/SOL, a large, population-based cohort of geographically diverse Hispanic/Latinos, the current analyses revealed, for the first time, that    [29]. Although the relatively high MUFA content of avocados drove our rationale for conducting these analyses, the nutrient profile of avocado contains multiple elements that could convey protection, such as (in a typical USDA serving): 345 mg of potassium (a nutrient necessary for glucose control) [30], 19.5 mg of magnesium (thought to improve glucose tolerance [31] and offer protection against the incidence of T2D), and 1.14 g of linoleic acid-a polyunsaturated fat (PUFA) which may improve insulin sensitivity [32]. Thus, it is not possible to tease out a single component responsible for the lower incidence of T2D, nor whether the relationship is due to the independent action of one or more micro-/macronutrient, or their synergistic effects.
Avocado was not significantly associated with lower incident T2D in those with normoglycemia, only in those with prediabetes, although the magnitude of effect in normoglycemics was similar to that in prediabetics (conveying 20% vs. 31% protection over a six-year period). Given the lower incidence of T2D in the Nonetheless, Hispanic/Latinos are at a higher risk of T2D than non-Hispanic white counterparts [37], and preventing the onset of chronic diseases in this population has been noted as a priority research area by the National Institutes of Health [38]. Our large-scale, longitudinal study provides additional support that avocados could form part of a diet aimed at supporting good glucose control, and highlights the need for future studies which research which examine whether the most efficacious dietary strategies are those tailored to an individual's sensitive overall glucose homeostasis and metabolic functioning. In this way, dietary approaches to prevent T2D may be made more efficacious.

Acknowledgements
This manuscript was prepared using HCHS/SOL research materials obtained from the NHLBI Biologic Specimen and Data Repository Coordinating Center and does not necessarily reflect the opinions or views of the HCHS/SOL nor the NHLBI.

Data Availability Statement
All data included in the present analyses are available from the NHLBI Biological Specimen and Data Repository Information Coordinating Center (BioLINCC; https://biolincc.nhlbi.nih.gov/home/).

Disclosures
This study was funded by Hass Avocado Nutrition Board. Hass Avocado Board did not have input into the study design, analysis nor interpretation of results.
Dr. Wood has received funding from The National Cattleman's Beef Association and Ionis Pharmaceuticals for studies unrelated to the current analyses.