This review aimed to find, assess, and synthesize all intervention and observational studies that quantified avocado intake in humans and associated health outcomes. A systematic literature search was conducted through PubMed, Web of Science, Scopus, and CENTRAL on October 17th, 2023, with no date restrictions, and again on April 29th, 2024, date restricted to the time between searches. Searches included terms related to the intake of avocado and MUFA and are provided in full detail in the Supplementary methods. The electronic search was supplemented with backward citation screening of all in-scope articles. Backward citation screening was performed using Scite (Henderson, NV, USA)¹ or Spidercite (Systematic Review Accelerator, Bond University, Gold Coast, Australia)² to export all reference lists of in-scope articles, where possible. Exported reference lists were subsequently uploaded to Covidence (Melbourne, Australia)³ and underwent single-review screening (title/abstract and full-text). Where full reference lists could not be exported and imported into Covidence, reference lists were manually screened for inclusion.

The 2025 USDA NESR protocol for dietary patterns and growth, body composition, and risk of obesity (21) was modified to suit an evidence map approach and greater inclusivity of diverse populations (Table 1). Modifications related to an evidence map approach (and not a systematic review) included expansion of criteria by: including cross-sectional and case–control studies and excluding studies on allergenicity and methods validation; removing restrictions on publication dates, population health status, study duration, language, and country; expanding outcomes of interest; and requiring exposure to avocado pulp (not an extract or oil thereof) to be quantified, statistically analyzed with relation to a health outcome, and isolatable in multi-component interventions. Modifications related to inclusivity were inclusion of studies on populations of any health status and those published in any language or country regardless of their Human Development Index, given numerous studies have been published in countries where English is not the primary language (especially Mexico).

Dual screening at the title/abstract and full-text levels was conducted by three reviewers in Covidence⁴. Discrepancies were resolved by discussion between reviewers or by a senior reviewer. Two reviewers independently extracted all data from each study using pre-defined sheets in Microsoft Excel, and conflicts were resolved through discussion or by a senior reviewer. No automation or AI-based tools were used for classification/categorization during screening or data extraction. Variables were extracted at either the per-study level (e.g., bibliographic information, study design, sample size, country of origin, funding source) or on a per-group level (subject characteristics, avocado and nutrient intakes, and all health outcomes measured). A complete list of all variables extracted is available in Supplementary Tables S1, S2. Data were extracted from both the main text, tables, figures, and all relevant Supplementary material. Only values (e.g., means and variation) related to avocado intake, BMI, age, sex, and blood lipids were extracted, to enable characterizing the participants studied. Where possible, data were extracted as described by the authors and re-classified or transformed for data analysis and visualization.

Data cleaning, transformation, descriptive analysis, and visualization were performed manually in Excel and Tableau Cloud (Tableau Server Version 2024.1.0). To avoid double-counting secondary analyses of the same study, each publication was classified according to whether it represented the “primary” report and if subsequent publications were conducted on unique populations. References to the count of primary reports indicate whether a given variable was assessed in the first or any subsequent publication from the same primary report unless specified otherwise. For example, the clinical trial NCT01235832 published by Wang et al., in 2015 and 2020 (22, 23) reported the same population but different outcomes. This clinical trial was treated as a single “primary report” rather than two. Conversely, the Persea americana for Total Health Study (NCT02740439) published initially by Edwards et al., 2020 (8) is detailed in three other publications (24–26), each of which differed slightly in the subject populations analyzed. Each publication’s sample size, subject characteristics, and outcomes were treated independently, whereas variables common to all four publications (e.g., study design, country of study) were treated as a single report. For simplicity, nearly all data reported in this manuscript were treated as a single report when multiple publications from the same data source (e.g., a survey, prospective cohort, clinical trial) exist, unless specified otherwise.

Similarly, multi-categorical information, such as a study on participants in both the United States (U.S.) and Canada (27) was stored in a single cell (e.g., U.S.; Canada) and then split into multiple rows. Thus, the percentage of primary reports that describe a given multi-categorical variable (e.g., funding sources, weight status, age of participants) may add to over 100%, as the categories are not mutually exclusive. Group and sub-group information (e.g., avocado intake in treatments A and B in males and females) were stored on multiple rows. All counts presented here were de-duplicated but not in the raw database. Unit conversions for avocado intake (e.g., from % energy, frequency, or other form into g/d) are listed within the database itself and in Supplementary Tables S3, S4. Protocols, Supplementary material, interactive visualizations, and raw data are freely available at https://github.com/Traverse-Science/Avocado-Evidence-Map, and archived at https://doi.org/10.5281/zenodo.12824839.

The earliest in-scope research on avocados began in 1985 (Figure 2), though, to our knowledge, the first study published on avocado intake and health was by Grant et al., 1960 (34); however, that study was out-of-scope given it lacked a control group. A steep rise in publications per year began in 2017, with 29/58 (50%) articles published between 2019–2023. The top five most commonly reported funding organizations (studies could have multiple funding sources), in order, were the Hass Avocado Board (which includes the Avocado Nutrition Center), the National Heart, Lung, and Blood Institute (NHLBI), NIH, National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK), and the USDA. In total, 58% of articles reported funding from commodity boards, 55% from research institutes, 16% from universities, and 7% from other sources (e.g., hospitals or companies). The majority of primary reports were conducted on participants in the Americas (U.S., Mexico, Canada, Jamaica, Chile, and Venezuela) (Figure 2). No studies were conducted in Europe and only one was conducted in Asia (in Indonesia). A small number of studies were conducted in South Africa, Uganda, and Australia. The geographical trend matches the funding source, as there were no funding sources from European/Asian commodity boards, and the only research institute in Asia (Korean International Cooperation Agency) funded a study in South Africa (57). The majority of primary reports were interventions (28/45, 62%) of crossover or parallel-arm design, with the remainder observational trials, typically a prospective cohort (8/45, 18%) or cross-sectional/survey study (10/45, 22%).

The intended population was usually described as “healthy” (no diagnosis of a medical disease/disorder) or described in vague terms such that it was assumed to be about the general population (Table 3). Only seven studies were intentionally designed to study participants with elevated blood lipids or those diagnosed with or at risk for type 2 diabetes (T2D). Regardless of the intent, the majority of studies were conducted in adults, fairly balanced in sex with a skew toward females, predominantly in those with overweight/obesity, and those with elevated blood cholesterol (Table 3; Figure 3). Interventions most frequently included individuals of middle (~35–50 yrs) or college ages (~20–35 yrs), and observational studies primarily described participants ages 40 yrs. or older (Figure 3). Intervention studies exhibited a bi-modal-like distribution with participant BMI clustered in ranges of 23–26 or 28+ kg/m², whereas observational studies reported ranges between these two peaks, from 24–29 kg/m² (Figure 3).

Despite the relatively small number of primary reports (22, 23, 58–61) that explicitly recruited participants for their elevated blood lipids (18%, Table 3), all interventions that reported blood lipid concentrations (17/28 primary reports) detailed elevated blood lipid concentrations by at least one metric (HDL, LDL, total cholesterol, or triacylglycerols) at some point during the study (Table 3). Participants were most likely to be rated as having elevated lipids, or conversely least likely to be rated as healthy, according to HDL and LDL concentrations. These characteristics were infrequently reported among observational studies.

There were four different designs of dietary control among intervention trials: (1) Constant macronutrient intake between groups (on a % energy basis) with enrichment of MUFA from avocado, (2) Higher vs. lower carbohydrate/fat intake (on a % energy basis) with avocado as a primary fat source, (3) Free feeding of avocado with no control over macronutrient intake or energy consumption, and (4) Studies not concerned with macronutrient intake (Table 4). Habitual intake studies represent the most translatable designs as they allow for potential dietary compensation; however, it can be difficult to isolate whether the resulting changes can be attributed to the intake of avocado itself independent of dietary compensation.

The most common comparators included the consumption of some avocado to no avocado (Table 5). However, observational trials more frequently included dose–response evaluations. Among the observational trials, most primary reports (44%) assessed the intake of whole, raw avocado, but not avocado from all food sources. In such cases, participants who consumed guacamole but not whole avocado would be considered “non-consumers” of avocado. Only 2/17 observational primary reports (62, 63) estimated intake from all food sources of avocado. This is likely why the most common intake of avocados was reported between 0–10 g/d (7% of an avocado/d, assuming 1 avocado = 150 g) among observational trials (Figure 4). Among interventions, intakes of avocado spanned between 30–490 g/d (or per meal, in the case of acute studies) but was most commonly 60–70, 150–160, 210–220, or 300–310 g/d (Figure 4). Such amounts are roughly equivalent to 0.5, 1, 1.5, and 2 whole avocados/d (assuming the average avocado weighs 150 g). The difference in exposure to avocado varies by upwards of 5-10x between observational and intervention trials.

Among the interventions, the intended energy balance of participants was either not relevant (29%, in acute studies), not explicitly controlled (32%, among habitual intake or ad libitum designs), or either a weight maintenance (25%) or weight loss (14%) design (Table 5). Sixty-four percent of interventions controlled intake such that comparisons between groups were isocaloric, 29% allowed energy intake to deviate (purposefully or due to dietary compensation), and a small proportion (18%) assessed avocado intake in a hypercaloric design. Such hypercaloric designs were more common in bioavailability or acute intake studies comparing avocado intake to the lack thereof or to different foods (53, 64–68).

Observational trials used either a food frequency questionnaire (FFQ) or food/diet record, and 88% of observational trials described the energy intake of participants compared to 65% of interventions. Intervention trials usually controlled nutrient/food intake through a prescribed diet, did not report nutrient intake at all (14%), or did so using a FFQ, 24-h recall, or 3-7d food record (7%) (Table 5).

Reporting nutrient intake was usually completed (Table 6), with 75% or fewer primary reports of interventions reporting energy or macronutrient contributions to energy. Although 88% of observational trials reported energy intake, ≤ 47% reported macronutrient contributions to energy (Table 6). Despite the objective of many studies being to investigate cardiovascular-related events and biomarkers (Figure 5), only 5/28 and 7/28 interventions reported cholesterol or fiber intake, respectively. Vitamins, minerals, and carotenoid consumption were similarly reported infrequently, however observational reports described vitamin and mineral intake more frequently than interventions (Table 6).

Among intervention trials, lipids, nutrient bioavailability, CVD risk factors (e.g., blood pressure/flow/structure/function or biomarkers, excluding anthropometrics and lipids), anthropometrics, and glycemia were the most common outcomes measured (Figure 5). Nutrient bioavailability was measured with the greatest heterogeneity (34 variables were measured across 11 studies) compared to anthropometrics or glycemia, where four variables were measured across 10 trials. Outcomes whose number of variables measured exceeded the number of primary reports indicate high heterogeneity and where greater replication would be valuable. Trials lasting longer than one month tended to have larger sample sizes than those of shorter duration. Fourteen of 28 primary interventions (50%) were conducted within one month or shorter, 11 conducted for 1–3 months (39%), and three for 3–6 months (11%).

Among observational trials, anthropometry, CVD incidence, and glycemia were the most commonly measured. Most measures were analyzed in a cross-sectional design (11/17 primary observational reports), whether that be a national survey or otherwise. CVD incidence was not measured among any cross-sectional studies and was the outcome analyzed in the greatest number of cohort studies (47, 49, 69). Follow-up lengths for all outcomes were usually 10+ years, with only one cohort study (69) reporting a shorter follow-up period. Outside of cross-sectional studies only 1–2 studies have been conducted on any given outcome, representing a general lack of longitudinal data across all topics.

Observational studies typically measured a distinct population compared with interventions (either younger or older than interventions, with a typical or overweight but not obese BMI) and reported much lower intakes of avocado than interventions (Figure 4). Figure 5 indicates both the heterogeneity in outcomes measured between the two designs as well as the general dearth of observational studies. Ultimately, the observational data and interventions were conducted on separate populations, exposure levels, and endpoints, and may be ill-suited for substantiating the other’s findings. This gap represents a promising opportunity for future epidemiological studies. We recommend a review to identify existing observational data that measured avocado intake for further secondary analysis. Doing so would bolster the available epidemiological evidence across a wide range of exposures, populations, and endpoints, thus filling some gaps between intervention and observational trials. This may be achievable by specifically identifying observational data in Europe. Despite no published research on the topic from participants in Europe, in 2023 Europe accounted for the second highest global import of avocado (0.76 billion kg), 7x higher than the next largest import region (Canada at 0.11 billion kg), and 60% that of the U.S. (1.26 billion kg) (75). Thus, the high import of avocado into Europe makes the region a novel research population and potential source for secondary data on avocado consumption and health. Replicating observational studies conducted in the US in other countries in subjects of diverse race/ethnicity, socioeconomic position, gender identity, and health disparities would also satisfy methodological recommendations to help evaluate diet-health relationships stated in the Scientific Report of the 2025 Dietary Guidelines Advisory Committee (76).

The ability to make causal inferences about the impact of avocado intake and health is limited by a few characteristics of the body of literature, most notably the ability to account for dose–response effects and dietary compensation from avocado consumption.

First, one of 28 interventions (77) study evaluated a dose–response effect in a controlled trial. While doing so was more common among observational trials, intervention studies assessed high intakes of avocado (typically >50 g/d), whereas observational studies reported low intakes of avocado (typically 0–40 g/d), an underestimate that is likely related to only 2/17 observational primary reports calculating intake from all food sources of avocado (62, 63) (Figure 4). The difference in avocado consumption between interventions and observational trials complicates the comparison of findings from one study design to the other.

More importantly, there is limited understanding of dietary compensation from avocado intake. A third of trials provided avocado in a free-feeding context, and numerous trials either did not measure, analyze, or report nutrient intakes (Tables 5, 6). Such characteristics limit the ability to experimentally isolate or statistically evaluate the effect of the background diet. To overcome such limitations, we recommend the following. First, future interventions should consider using a dose–response design to identify if increasing avocado intake has a causal impact on health. Second, estimation of exposure to avocado in observational studies should estimate intake from all possible food sources (e.g., raw avocado pulp, guacamole, sushi, etc.). Estimation of avocado intake from food frequency questionnaires may be more accurate than from 24-h recalls or 3-7d food diaries in populations that consume avocado infrequently. Third, all studies (including those of acute nature) should measure and report energy, macronutrient, and micronutrient intake at the start and end of the trial or follow-up period. Doing so would enable quantification of potential dietary compensation and the ability to control for the background dietary pattern. Lastly, alongside macronutrient and MUFA intake, fiber should be considered a critical nutrient to control for. Fiber has a known, beneficial impact on cardiovascular health, and in the United States there are approved health claims related to fiber intake and coronary heart disease (21 CFR 101.77, 21 CFR 101.81). Yet, existing study designs of interventions were primarily designed to control energy, MUFA, and/or macronutrient intake (Table 4). Fiber intake could be used among interventions with a free-feeding design or observational trials as an additional measure to statistically control for.