Olive oil consumption and risk of type 2 diabetes in US women^,,

Abstract

Background: Olive oil has been shown to improve various cardiometabolic risk factors. However, to our knowledge, the association between olive oil intake and type 2 diabetes (T2D) has never been evaluated in the US population.

Objective: We aimed to examine the association between olive oil intake and incident T2D.

Design: We followed 59,930 women aged 37–65 y from the Nurses’ Health Study (NHS) and 85,157 women aged 26–45 y from the NHS II who were free of diabetes, cardiovascular disease, and cancer at baseline. Diet was assessed by validated food-frequency questionnaires, and data were updated every 4 y. Incident cases of T2D were identified through self-report and confirmed by supplementary questionnaires.

Results: After 22 y of follow-up, we documented 5738 and 3914 incident cases of T2D in the NHS and NHS II, respectively. With the use of Cox regression models with repeated measurements of diet and multivariate adjustment for major lifestyle and dietary factors, the pooled HR (95% CI) of T2D in those who consumed >1 tablespoon (>8 g) of total olive oil per day compared with those who never consumed olive oil was 0.90 (0.82, 0.99). The corresponding HRs (95% CIs) were 0.95 (0.87, 1.04) for salad dressing olive oil and 0.85 (0.74, 0.98) for olive oil added to food or bread. We estimated that substituting olive oil (8 g/d) for stick margarine, butter, or mayonnaise was associated with 5%, 8%, and 15% lower risk of T2D, respectively, in the pooled analysis of both cohorts.

Conclusions: Our results suggest that higher olive oil intake is associated with modestly lower risk of T2D in women and that hypothetically substituting other types of fats and salad dressings (stick margarine, butter, and mayonnaise) with olive oil is inversely associated with T2D.

INTRODUCTION

Type 2 diabetes mellitus (T2D)⁹ is one of the most prevalent chronic diseases worldwide, affecting more than 382 million people, or 8.3% of all adults, in 2013 (1). In 2035, the estimated number of adults with diabetes will increase to 592 million, ∼10% of the world’s adult population (1). T2D has become a paramount public health problem given that it leads to severe complications and is considered a strong risk factor for cardiovascular disease and mortality (2, 3). Primary prevention programs, including long-term lifestyle modifications, have shown to be effective in delaying or preventing T2D (3).

Evidence suggests that the type of dietary fat rather than the total amount of fat plays a role in T2D risk (4, 5). Replacing SFAs with PUFAs or MUFAs has been shown to be beneficial in the prevention of T2D (5). Olive oil is considered a healthy source of vegetable fat (6, 7). The health effects of olive oil are traditionally attributed to its high MUFA content, mainly oleic acid (8). But olive oil, especially the extra-virgin variety of olive oil, also contains other bioactive compounds, such as lipid molecules (squalene, tocopherols) and hydrophilic compounds, such as polyphenols (6), that could exert by different mechanisms favorable effects on insulin sensitivity and T2D (9).

Recently, secondary analyses of the PREvención con DIeta MEDiterránea (PREDIMED) randomized primary-prevention nutrition trial involving elderly individuals at high cardiovascular disease risk (10) showed a 40% reduction in the incidence of T2D among participants assigned to a Mediterranean diet supplemented with extra-virgin olive oil compared with a low-fat control diet (11). This inverse association was slightly stronger in women than in men, although the observed between-sex difference was not substantial. In addition, in the same study, extra-virgin olive oil intake was inversely associated with cardiovascular disease and mortality (12). Furthermore, this trial was conducted in Spain, a Mediterranean country with a historically high average amount of customary olive oil consumption.

It is unclear whether olive oil is beneficial for diabetes prevention in healthy adults in populations outside of Mediterranean regions, namely in the United States, who have low average consumption of olive oil in comparison to Mediterranean populations. Therefore, we longitudinally examined the association between olive oil consumption and T2D in women from 2 large cohort studies with 22 y of follow-up—the Nurses’ Health Study (NHS) and NHS II—to test the hypothesis that higher olive oil consumption leads to lower risk of developing T2D. In addition, we estimated the hypothetical effect of substituting olive oil for other types of fat and salad dressing on the predicted risk of T2D.

METHODS

Study population

We used data from 2 prospective cohort studies: the NHS (age range: 37–65 y) and NHS II (age range: 26–45 y). Detailed information on these cohorts has been described elsewhere (13–16). Briefly, the NHS consists of 121,700 female nurses who were enrolled in 1976. The NHS II, established in 1989, consists of 116,671 younger female registered nurses. At enrollment and biannually, questionnaires were administered to collect and update medical and lifestyle information. Follow-up rates exceed 90% in each 2-y cycle for the 2 cohorts.

For the present analysis, we defined baseline as the first year that a question regarding olive oil intake appeared in the food-frequency questionnaire (FFQ): 1990 for the NHS and 1991 for the NHS II. We excluded women who had diagnoses of T2D, cardiovascular disease, or cancer at baseline; those who died before baseline; participants who had implausible energy intake (<500 or >3500 kcal/d) or who left >9 or >10 items blank, respectively, on the baseline FFQ; those who returned only the baseline questionnaire; and participants with missing information on olive oil variables. After exclusions, data from 59,930 NHS and 85,157 NHS II participants were available for the present analysis. The study was approved by the Human Research Committee of Brigham and Women’s Hospital and the Human Subjects Committee Review Board of the Harvard T. H. Chan School of Public Health. The completion of the self-administered questionnaire was considered to imply informed consent.

Dietary assessment

Dietary intake was measured by validated >130-item FFQs administered every 4 y. The reproducibility and validity of these FFQs have been demonstrated in detail elsewhere (15, 17–19). Participants were asked how often, on average, they consumed each food of a standard portion size over the prior year. There were 9 possible responses, ranging from “never or less than once per month” to “6 or more times per day.”

In the current study, on the basis of the distribution of responses to 2 questions related to olive oil intake, we divided olive oil intake into salad dressing olive oil and olive oil added to food or bread. The total olive oil consumption was the sum of salad dressing olive oil and olive oil added to food or bread. We categorized olive oil intake (as salad dressing olive oil, as added to food or bread, and as total olive oil intake) into 4 categories [never/almost never, ≤1/2 tablespoon/d (≤4 g/d), >1/2 to ≤1 tablespoon/d (>4 to ≤8 g/d), and >1 tablespoon/d (>8 g/d)], and we also analyzed olive oil intake as a continuous variable. We considered 1 tablespoon to be equivalent to 8 g olive oil. Total energy intake and an overall index of compliance with a healthy dietary pattern, the Alternative Healthy Eating Index (AHEI) (20), were calculated from FFQs.

Ascertainment of incident T2D

A supplementary questionnaire regarding symptoms, diagnostic tests, and hypoglycemic therapy was completed by participants who reported a diagnosis of diabetes. T2D cases were confirmed if at least one of the following was reported on the supplementary questionnaire according to the 1997 American Diabetes Association criteria (21): 1) one or more classic symptoms (excessive thirst, polyuria, weight loss, hunger) plus fasting plasma glucose concentrations ≥7.0 mmol/L or random plasma glucose concentrations ≥11.1 mmol/L, 2) at least 2 elevated plasma glucose concentrations on different occasions (fasting concentrations ≥7.0 mmol/L, random plasma glucose concentrations ≥11.1 mmol/L, and/or concentrations ≥11.1 mmol/L after ≥2-h oral-glucose-tolerance test) in the absence of symptoms, or 3) treatment with hypoglycemic medication (insulin or oral hypoglycemic agent). Only confirmed cases were included in the analysis.

The validity of the supplementary questionnaire for the diagnoses of diabetes has been previously described (22, 23). Of a random sample of 62 nurses from the NHS reporting T2D in the questionnaire, the diagnosis was confirmed in 61 (98%) after medical record review by an endocrinologist blinded to the supplementary questionnaire information (22).

Assessment of covariates

In biennial follow-up questionnaires, participants provided updated information on age, lifestyle, ethnicity, ancestry, family history of disease, medication use, and risk factors for chronic diseases, such as body weight, cigarette smoking, physical activity, menopausal status, and hormone use.

Statistical analysis

Person-years of follow-up for each participant were calculated from the return of the baseline questionnaire to the date of diagnosis of T2D, death, or end of follow-up (30 June 2012 for the NHS and 30 June 2013 for the NHS II), whichever came first.

We used Cox proportional hazards models to estimate the HRs and 95% CIs of developing T2D according to olive oil intake (salad dressing olive oil, olive oil added to food or bread, and total olive oil intake). We stratified analyses jointly by age (in mo) at baseline and calendar year to control for secular trends in consumption and age. In multivariate analyses, we further adjusted for ethnicity (white or nonwhite), ancestry (Southern European/Mediterranean, other Caucasian, African American, Hispanic, Asian, or other), smoking status (never, former, or current smoker: 1–14 cigarettes/d, 15–24 cigarettes/d, or ≥25 cigarettes/d), alcohol intake (0, 0.1–4.9, 5.0–9.9, 10.0–14.9, or ≥15.0 g/d), physical activity (<3.0, 3.0–8.9, 9.0–17.9, 18.0–26.9, or ≥27.0 metabolic equivalent task-hours/wk), family history of diabetes (yes or no), history of hypertension (yes or no), history of hypercholesterolemia (yes or no), multivitamin use (yes or no), postmenopausal status (yes or no), menopausal hormone use (no, past, or current hormone use), quintiles of AHEI, and total energy intake in kcal/d. Our final multivariate model was further adjusted for BMI (in kg/m², continuous). The covariates were updated every 2–4 y by using the most recent data for each follow-up interval.

To better represent long-term diet and to minimize within-person variation, we created cumulative averages of olive oil intake and other types of fats and salad dressing from baseline to censoring events (24). Because participants may alter dietary patterns after the diagnosis of major illness, we stopped updating dietary variables when participants reported a diagnosis of stroke, myocardial infarction, angina, or cancer, although follow-up continued until death or the end of the study period (24). Missing values during follow-up were replaced by using the carry-forward method.

We conducted prespecified subgroup analyses according to baseline age (<60 and <40 compared with ≥60 and ≥40 y in the NHS and NHS II, respectively), BMI (<25 compared with ≥25), family history of diabetes (yes compared with no), ancestry (Southern European/Mediterranean compared with other Caucasian compared with other), and above compared with below the median of each of the AHEI score, total vegetable intake, green vegetable intake, and lettuce intake. The χ² P values for the difference between strata were calculated.

We also estimated the predicted associations of hypothetical substitution of 8 g olive oil for the same amount of other types of fats and salad dressings [margarine (stick, tub/soft), butter, mayonnaise, and other vegetable oils] with T2D by including both continuous variables in the same multivariable model, which also contained nondietary covariates and total energy intake. We used the difference between regression coefficients and in variances and covariances to derive the HRs and 95% CIs of the substitution analyses. This method has been described in a previous study (25).

Tests for linear trend were conducted by assigning the median value to each category of intake and modeling this value as a continuous variable. HRs from multivariate models in each cohort were pooled by using a fixed-effect inverse-variance meta-analysis. P values for heterogeneity of study results were calculated by using the Cochran Q test and quantified with the I² statistic. All P values were 2-sided, and an α level of <0.05 was considered statistically significant. Data were analyzed with the SAS package, version 9.3 (SAS Institute).

RESULTS

During 22 y of follow-up, we documented 5738 incident T2D cases in the NHS and 3914 cases in the NHS II. The baseline characteristics of the participants according to total olive oil consumption in the 2 prospective cohorts are shown in Table 1. Compared with women who never consumed olive oil, those who had higher olive oil intake were older, tended to exercise more, had a lower BMI, and were more likely to have Southern European/Mediterranean ancestry. They also consumed more fish, whole grains, fruit and vegetables, nuts, and total energy and had a higher AHEI score. The mean intake of total olive oil in the highest category (>1 tablespoon/d) was 13.25 g/d in the NHS and 20 g/d in the NHS II, or ∼2–2.5 tablespoons/d.

TABLE 1.

Baseline characteristics according to total olive oil consumption categories in the 2 prospective cohorts¹

	Olive oil intake
	NHS (1990)				NHS II (1991)
Characteristic	Never/almost never	>0–≤1/2 tablespoon (>0–≤4 g/d)	>1/2–≤1 tablespoon (>4–≤8 g/d)	>1 tablespoon (>8 g/d)	Never/almost never	>0–≤1/2 tablespoon (>0–≤4 g/d)	>1/2–≤1 tablespoon (>4–≤8 g/d)	>1 tablespoon (>8 g/d)
Participants, n	32,166	15,321	9433	3010	50,773	23,033	9424	1927
Total olive oil, g/d	0.00 ± 0.002	1.29 ± 1.00	5.48 ± 1.28	13.25 ± 6.76	0.00 ± 0.00	1.54 ± 1.92	7.27 ± 4.17	19.99 ± 12.97
Salad dressing olive oil, g/d	0.00 ± 0.00	1.27 ± 1.76	4.77 ± 2.80	8.96 ± 6.36	0.00 ± 0.00	1.07 ± 0.87	4.23 ± 1.73	7.84 ± 5.04
Olive oil added to food or bread, g/d	0.00 ± 0.00	1.13 ± 0.92	4.15 ± 1.94	7.70 ± 5.55	0.00 ± 0.00	1.00 ± 1.38	4.68 ± 2.39	9.92 ± 6.24
Age, y	55.75 ± 7.13	55.76 ± 7.09	55.80 ± 7.07	55.87 ± 7.06	36.02 ± 4.70	36.08 ± 4.64	36.20 ± 4.59	36.31 ± 4.53
BMI, kg/m2	25.56 ± 4.82	25.25 ± 4.51	25.10 ± 4.44	25.08 ± 4.47	24.75 ± 5.40	24.11 ± 4.94	23.91 ± 4.72	24.09 ± 4.94
Physical activity, MET-h/wk	14.59 ± 20.30	16.37 ± 23.43	17.87 ± 23.20	19.00 ± 26.57	19.26 ± 25.59	22.50 ± 28.85	25.22 ± 30.68	28.34 ± 37.36
Family history of diabetes, %	28.06	27.38	26.54	27.85	34.40	32.99	31.43	32.80
Ethnicity, white, %	97.56	98.01	98.70	98.94	96.24	96.55	97.75	97.75
Southern European/Mediterranean ancestry, %	14.83	18.01	22.13	27.76	15.52	19.87	24.82	31.80
Current smoker, %	16.77	17.21	16.26	13.70	12.00	12.11	12.19	12.94
Current menopausal hormone use/current oral contraceptive use, %	38.26	39.67	41.82	38.96	10.93	11.21	10.78	8.36
Hypertension, %	26.37	26.60	26.49	26.84	6.21	5.63	4.89	5.06
High blood cholesterol, %	33.46	36.49	37.44	37.62	14.39	13.71	13.58	12.94
Total energy intake, kcal/d	1700 ± 502.80	1743 ± 511.47	1845 ± 507.14	1995 ± 544.49	1746 ± 539.2	1801 ± 546.66	1917 ± 547.46	2066 ± 553.41
Alcohol, g/d	4.26 ± 8.76	5.82 ± 9.89	7.02 ± 10.66	7.73 ± 11.32	2.44 ± 5.37	3.87 ± 6.73	4.77 ± 7.02	5.34 ± 8.03
PUFA to SFA ratio	0.57 ± 0.19	0.57 ± 0.18	0.61 ± 0.18	0.69 ± 0.24	0.51 ± 0.15	0.52 ± 0.15	0.56 ± 0.16	0.63 ± 0.21
SFA, % total energy	10.88 ± 2.56	10.56 ± 2.47	10.31 ± 2.45	10.13 ± 2.46	11.42 ± 2.43	11.00 ± 2.37	10.66 ± 2.33	10.50 ± 2.48
MUFA, % total energy	11.89 ± 2.58	11.82 ± 2.49	12.59 ± 2.41	13.92 ± 2.90	11.94 ± 2.47	11.71 ± 2.32	12.30 ± 2.31	13.48 ± 2.81
PUFA, % total energy	5.88 ± 1.64	5.75 ± 1.45	6.00 ± 1.40	6.71 ± 1.94	5.63 ± 1.41	5.54 ± 1.24	5.76 ± 1.31	6.30 ± 1.85
trans Fat, % total energy	1.57 ± 0.60	1.47 ± 0.55	1.38 ± 0.52	1.28 ± 0.52	1.71 ± 0.63	1.57 ± 0.57	1.45 ± 0.53	1.33 ± 0.53
Fish, servings/d	0.26 ± 0.19	0.30 ± 0.20	0.33 ± 0.21	0.37 ± 0.25	0.24 ± 0.24	0.29 ± 0.24	0.33 ± 0.26	0.39 ± 0.38
Poultry, servings/d	0.32 ± 0.19	0.35 ± 0.19	0.37 ± 0.19	0.40 ± 0.21	0.44 ± 0.32	0.47 ± 0.31	0.50 ± 0.31	0.53 ± 0.36
Dairy, servings/d	2.03 ± 1.09	2.09 ± 1.10	2.11 ± 1.10	2.14 ± 1.11	2.22 ± 1.44	2.30 ± 1.47	2.38 ± 1.49	2.48 ± 1.56
Red meat, servings/d	0.84 ± 0.57	0.79 ± 0.55	0.77 ± 0.55	0.72 ± 0.54	0.59 ± 0.41	0.51 ± 0.37	0.49 ± 0.37	0.45 ± 0.36
Soda, servings/d	0.81 ± 0.95	0.78 ± 0.92	0.76 ± 0.92	0.78 ± 0.99	1.57 ± 1.52	1.38 ± 1.44	1.30 ± 1.46	1.28 ± 1.50
Nuts, servings/d	0.11 ± 0.18	0.13 ± 0.18	0.13 ± 0.18	0.16 ± 0.25	0.07 ± 0.19	0.10 ± 0.21	0.11 ± 0.23	0.15 ± 0.31
Whole grains, servings/d	18.77 ± 15.64	19.24 ± 15.69	19.54 ± 14.99	19.34 ± 15.25	6.26 ± 2.97	6.62 ± 3.05	7.00 ± 3.08	7.24 ± 3.45
Fruit and vegetables, servings/d	5.13 ± 2.13	5.52 ± 2.19	6.07 ± 2.26	6.99 ± 2.56	4.63 ± 2.66	5.30 ± 2.78	6.35 ± 3.09	8.04 ± 3.98
Coffee, servings/d	3.79 ± 2.87	3.97 ± 2.74	4.08 ± 2.71	4.16 ± 2.70	1.45 ± 1.70	1.63 ± 1.64	1.78 ± 1.63	1.96 ± 1.69
Multivitamin supplement use, %	35.37	38.20	40.14	38.22	42.54	45.57	45.54	44.04
AHEI score	51.18 ± 10.71	53.59 ± 10.57	55.99 ± 10.54	58.98 ± 10.44	46.51 ± 10.57	50.31 ± 10.48	53.57 ± 10.97	57.12 ± 10.92

¹Values are standardized to the age distribution of the study population. AHEI, Alternative Healthy Eating Index; MET, metabolic equivalent task; NHS, Nurses’ Health Study.

²Mean ± SD (all such values).

In age-adjusted models, total olive oil intake was associated with a substantially lower risk of developing T2D in both studies (Table 2). In multivariable-adjusted models, these associations were attenuated, but a trend of reduced risk associated with high olive oil intake remained, especially in the NHS. In the pooled multivariable-adjusted model without BMI, the HR (95% CI) was 0.87 (0.80, 0.96) for participants consuming >1 tablespoon olive oil/d (>8 g/d) compared with those who never/almost never consumed olive oil (P-trend < 0.001). After adjustment for BMI, the association was slightly attenuated but remained statistically significant at 0.90 (0.82, 0.99; P-trend < 0.001) in the highest category of olive oil intake (>8 g/d) compared with the lowest (never/almost never). Each additional 8-g intake of olive oil was associated with a 6% reduced risk of T2D (HR: 0.94; 95% CI: 0.90, 0.99).

TABLE 2.

Risk of type 2 diabetes according to total olive oil intake¹

	Categories of total olive oil intake
	Never/almost never	>0–≤1/2 tablespoon (>0–≤4 g/d)	>1/2–≤1 tablespoon (>4–≤8 g/d)	>1 tablespoon (>8 g/d)	P-trend	Per 8-g increment in olive oil intake
NHS
Total olive oil, g/d	0.00 ± 0.002	1.29 ± 1.00	5.48 ± 1.28	13.25 ± 6.76
Cases/person-years, n	1828/343,925	2590/454,876	708/154,534	471/103,687
Age-adjusted model	1 (referent)	0.88 (0.83, 0.94)3	0.76 (0.69, 0.83)	0.68 (0.61, 0.75)	<0.001	0.81 (0.77, 0.86)
Multivariable model 1	1 (referent)	0.99 (0.93, 1.06)	0.92 (0.84, 1.01)	0.87 (0.78, 0.97)	0.002	0.92 (0.87, 0.97)
Multivariable model 2	1 (referent)	1.01 (0.95, 1.08)	0.97 (0.88, 1.06)	0.91 (0.81, 1.01)	0.038	0.94 (0.89, 0.99)
NHS II
Total olive oil intake, g/d	0.00 ± 0.00	1.54 ± 1.92	7.27 ± 4.17	19.99 ± 12.97
Cases/person-years, n	1623/582,787	1814/626,771	323/146,101	154/63,491
Age-adjusted model	1 (referent)	0.78 (0.73, 0.84)	0.64 (0.56, 0.72)	0.56 (0.48, 0.66)	<0.001	0.69 (0.62, 0.75)
Multivariable model 1	1 (referent)	0.96 (0.89, 1.03)	0.90 (0.79, 1.01)	0.87 (0.73, 1.03)	0.052	0.93 (0.85, 1.02)
Multivariable model 2	1 (referent)	0.96 (0.89, 1.03)	0.89 (0.79, 1.01)	0.87 (0.73, 1.04)	0.051	0.95 (0.87, 1.04)
Pooled
Age-adjusted model	1 (referent)	0.84 (0.80, 0.88)*	0.72 (0.67, 0.77)*	0.65 (0.59, 0.71)	<0.001	0.78 (0.75, 0.82)*
Multivariable model 1	1 (referent)	0.98 (0.94, 1.03)	0.92 (0.85, 0.99)	0.87 (0.80, 0.96)	<0.001	0.93 (0.88, 0.97)
Multivariable model 2	1 (referent)	0.99 (0.94, 1.04)	0.94 (0.88, 1.01)	0.90 (0.82, 0.99)	<0.001	0.94 (0.90, 0.99)

¹Multivariable model 1 was adjusted for age (y), ethnicity (white or nonwhite), ancestry (Southern European/Mediterranean, other Caucasian, African American, Hispanic, Asian, or other ancestry), smoking status (never, former, or current smoker: 1–14, 15–24, and ≥25 cigarettes/d), alcohol intake (0, 0.1–4.9, 5.0–9.9, 10.0–14.9, or ≥15.0 g/d), physical activity (<3.0, 3.0–8.9, 9.0–17.9, 18.0–26.9, or ≥27.00 metabolic equivalent task-hours/wk), family history of diabetes (yes or no), history of hypertension (yes or no), history of hypercholesterolemia (yes or no), multivitamin use (yes or no), postmenopausal status (yes or no) and menopausal hormone use (no, past, or current hormone use), quintiles of the Alternative Healthy Eating Index score, and total energy intake (kcal/d). Multivariable model 2 was also adjusted for BMI. *P-heterogeneity in the fixed-effect inverse-variance weighted meta-analysis pooling both cohorts was <0.05. NHS, Nurses’ Health Study.

²Mean ± SD (all such values).

³HR; 95% CI in parentheses (all such values).

Salad dressing olive oil intake was associated with lower risk of T2D in age-adjusted models (Table 3). Compared with those who never consumed salad dressing olive oil, the pooled HR (95% CI) of T2D in the age-adjusted model was 0.69 (0.63, 0.75) in those consuming >1 tablespoon salad dressing olive oil/d (P-trend <0.001). Similarly to total olive oil intake, multivariate adjustment for covariates attenuated to nonsignificance for these associations.

TABLE 3.

Risk of type 2 diabetes according to intake of salad dressing olive oil and olive oil added to food or bread¹

	Never/almost never	>0–≤1/2 tablespoon (>0–≤4 g/d)	>1/2–≤1 tablespoon (>4–≤8 g/d)	>1 tablespoon (>8 g/d)	P-trend	Per 8-g increment in olive oil intake
Categories of salad dressing olive oil intake
NHS
Salad dressing olive oil, g/d	0.00 ± 0.002	1.20 ± 0.72	5.24 ± 4.28	11.78 ± 5.46
Cases/person-years, n	2794/685,241	1498/375,914	1000/260,074	448/122,768
Age-adjusted model	1 (referent)	0.93 (0.87, 0.99)3	0.83 (0.77, 0.89)	0.69 (0.62, 0.76)	<0.001	0.79 (0.75, 0.84)
Multivariable model 1	1 (referent)	1.02 (0.96, 1.08)	1.00 (0.93, 1.08)	0.92 (0.82, 1.02)	0.134	0.95 (0.90, 1.01)
Multivariable model 2	1 (referent)	1.03 (0.97, 1.10)	1.01 (0.94, 1.09)	0.93 (0.83, 1.03)	0.212	0.96 (0.90, 1.01)
NHS II
Salad dressing olive oil, g/d	0.00 ± 0.00	1.13 ± 0.64	5.16 ± 1.35	11.77 ± 5.84
Cases/person-years, n	2371/801,265	1020/393,677	383/171,707	140/52,502
Age-adjusted model	1 (referent)	0.84 (0.78, 0.90)	0.68 (0.61, 0.75)	0.68 (0.57, 0.81)	<0.001	0.71 (0.64, 0.78)
Multivariable model 1	1 (referent)	0.99 (0.92, 1.07)	0.90 (0.80, 1.00)	0.96 (0.81, 1.15)	0.193	0.94 (0.86, 1.04)
Multivariable model 2	1 (referent)	1.07 (0.99, 1.16)	0.93 (0.83, 1.03)	1.01 (0.85, 1.21)	0.446	0.96 (0.87, 1.05)
Pooled
Age-adjusted model	1 (referent)	0.89 (0.85, 0.93)*	0.78 (0.74, 0.83)*	0.69 (0.63, 0.75)	<0.001	0.77 (0.73, 0.81)
Multivariable model 1	1 (referent)	1.01 (0.96, 1.06)	0.97 (0.91, 1.03)	0.93 (0.85, 1.02)	0.052	0.95 (0.91, 1.00)
Multivariable model 2	1 (referent)	1.05 (1.00, 1.10)	0.99 (0.93, 1.05)	0.95 (0.87, 1.04)	0.145	0.96 (0.91, 1.01)
Categories of olive oil added to food or bread intake
NHS
Olive oil added to food or bread, g/d	0.00 ± 0.00	1.27 ± 0.98	5.33 ± 1.27	11.91 ± 5.44
Cases/person-years, n	1994/366,956	2827/504,925	600/140,653	176/44,487
Age-adjusted model	1 (referent)	0.86 (0.80, 0.91)	0.71 (0.64, 0.77)	0.65 (0.55, 0.76)	<0.001	0.72 (0.66, 0.79)
Multivariable model 1	1 (referent)	0.98 (0.92, 1.04)	0.90 (0.82, 0.99)	0.83 (0.70, 0.97)	0.003	0.89 (0.82, 0.97)
Multivariable model 2	1 (referent)	0.99 (0.94, 1.06)	0.92 (0.84, 1.02)	0.87 (0.74, 1.01)	0.021	0.91 (0.84, 0.99)
NHS II
Olive oil added to food or bread, g/d	0.00 ± 0.00	1.15 ± 0.92	5.19 ± 1.22	11.90 ± 5.29
Cases/person-years, n	1603/579,772	1970/679,938	288/132,597	53/26,843
Age-adjusted model	1 (referentS	0.77 (0.72, 0.83)	0.64 (0.56, 0.72)	0.54 (0.41, 0.72)	<0.001	0.63 (0.55, 0.72)
Multivariable model 1	1 (referent)	0.96 (0.89, 1.03)	0.92 (0.81, 1.05)	0.78 (0.59, 1.03)	0.055	0.95 (0.84, 1.08)
Multivariable model 2	1 (referent)	1.00 (0.94, 1.08)	0.99 (0.87, 1.12)	0.79 (0.60, 1.05)	0.236	1.00 (0.88, 1.13)
Pooled
Age-adjusted model	1 (referent)	0.82 (0.78, 0.86)	0.68 (0.63, 0.74)	0.62 (0.54, 0.71)	<0.001	0.69 (0.65, 0.75)
Multivariable model 1	1 (referent)	0.97 (0.93, 1.02)	0.91 (0.84, 0.98)	0.82 (0.71, 0.94)	<0.001	0.91 (0.85, 0.98)
Multivariable model 2	1 (referent)	1.00 (0.96, 1.05)	0.95 (0.88, 1.03)	0.85 (0.74, 0.98)	0.010	0.94 (0.88, 1.01)

¹Multivariable model 1 was adjusted for age (y), ethnicity (white or nonwhite), ancestry (Southern European/Mediterranean, other Caucasian, African American, Hispanic, Asian, or other ancestry), smoking status (never, former, or current smoker: 1–14, 15–24, and ≥25 cigarettes/d), alcohol intake (0, 0.1–4.9, 5.0–9.9, 10.0–14.9, or ≥15.0 g/d), physical activity (<3.0, 3.0–8.9, 9.0–17.9, 18.0–26.9, or ≥27.00 metabolic equivalent task-hours/wk), family history of diabetes (yes or no), history of hypertension (yes or no), history of hypercholesterolemia (yes or no), multivitamin use (yes or no), postmenopausal status (yes or no) and menopausal hormone use (no, past, or current hormone use), quintiles of the Alternative Healthy Eating Index score, and total energy intake (kcal/d). Multivariable model 2 was also adjusted for BMI. *P-heterogeneity in the fixed-effect inverse-variance weighted meta-analysis pooling both cohorts was <0.05. NHS, Nurses’ Health Study.

²Mean ± SD (all such values).

³HR; 95% CI in parentheses (all such values).

Olive oil added to food or bread was associated with lower risk of T2D in the pooled age-adjusted and multivariable models (Table 3). In comparison with never-consumers, the pooled association in the fully adjusted model for those in the highest category of olive oil added to food or bread was 0.85 (0.74, 0.98; P-trend = 0.010).

In hypothetical substitution analyses and pooling results from both cohorts, replacing 8 g stick margarine, butter, or mayonnaise with 8 g olive oil was substantially associated with a predicted lower risk of T2D: 5% lower (HR: 0.95; 95% CI: 0.89, 1.00) when replacing stick margarine, 8% lower (HR: 0.92; 95% CI: 0.87, 0.97) when replacing butter, and 15% lower (HR: 0.85; 95% CI: 0.80, 0.90) when replacing mayonnaise (Figure 1). Replacing 8 g tub/soft margarine (HR: 0.96; 95% CI: 0.88, 1.06) or vegetable oils (HR: 0.96; 95% CI: 0.89, 1.03) with olive oil was not associated with predicted risk of diabetes. When the models were further adjusted for BMI, a potential mediator, the associations were attenuated to nonsignificance for all the predicted risk estimates except for when olive oil was replacing mayonnaise (HR: 0.92; 95% CI: 0.86, 0.97).

FIGURE 1.

Risk of type 2 diabetes associated with substitution of 8 g olive oil for equivalent amounts of other fats and salad dressings in both cohorts pooled (NHS and NHS II) (n = 145,087). HRs were adjusted for age, ethnicity, ancestry, smoking status, alcohol intake, physical activity, family history of diabetes, history of hypertension, history of hypercholesterolemia, multivitamin use, postmenopausal status and menopausal hormone use, quintiles of the Alternative Healthy Eating Index score, and total energy intake. NHS, Nurses’ Health Study.

In the prespecified subgroup analyses (Supplemental Table 1), pooled subgroup results indicated that those who consumed a healthier diet; consumed more total vegetables, green vegetables, and lettuce; and had the highest total olive oil intake had a lower risk of developing T2D than their counterparts who were high olive oil consumers but otherwise had less healthy diets. Women reporting Southern European/Mediterranean ancestry and the highest amount of olive oil intake (median intake in the highest quartile of olive oil consumption was 11.07 g/d) had a 23% (HR: 0.77; 95% CI: 0.63, 0.94) lower risk of T2D, which was a greater risk reduction than the nonsignificant risk reduction observed in the non-Mediterranean subgroups with high intake of olive oil (other Caucasian ancestry HR: 1.13; 95% CI: 0.92, 1.39; other ancestry HR: 0.93; 95% CI: 0.82, 1.05). Similarly, having Southern European/Mediterranean ancestry and a high intake of salad dressing olive oil was associated with a 20% lower risk of T2D (HR: 0.80; 95% CI: 0.66, 0.96) or high intake of olive oil added to food or bread was associated with 26% lower risk of T2D (HR: 0.74; 95% CI: 0.57, 0.96) compared with those with no Mediterranean ancestry who had high intakes.

DISCUSSION

The results of our study suggest that higher olive oil intake has favorable effects on T2D prevention, even in populations with relatively low amounts of intake compared with their Mediterranean counterparts. In these US women, followed for 22 y, we observed inverse associations between total olive oil consumption and T2D risk. Olive oil added to food or bread was inversely associated with diabetes risk, and a trend, although nonsignificant, for lower risk of T2D for higher intake of salad dressing olive oil was also observed. These associations were attenuated but remained substantial after adjusting for BMI in pooled analyses of both cohorts. In addition, we estimated the hypothetical association on diabetes risk of replacing margarine (stick and tub/soft), butter, and mayonnaise with the same amount of olive oil and observed a lower risk of diabetes with such substitutions, particularly in older women.

To our knowledge, only one other prospective cohort study has evaluated the specific association of olive oil as a unique dietary factor in incident T2D. That study, conducted in a Spanish population consuming higher amounts of this type of oil, found no association between olive oil intake and T2D (26). However, the small number of cases of incident T2D in that population may have contributed to the null finding (26). Recently, results of secondary analyses from the PREDIMED trial indicated that a Mediterranean diet supplemented with extra-virgin olive oil reduced the risk of T2D by 40% over 4 y of follow-up in a cardiometabolically at-risk and older population (11). The effect observed in PREDIMED is considerably stronger than the association observed in the present study, likely owing to a number of study-specific factors, chiefly among them that olive oil was an intervention in PREDIMED, whereas there was no intervention in the NHS cohorts. In addition, the PREDIMED population had notable at-risk baseline characteristics, had a very high intake of olive oil compared with the present study, and typically used an extra-virgin varietal. Nevertheless, our results of a 10% lower risk of T2D with higher olive oil intake lend additional support to olive oil’s potential role in diabetes prevention, even in populations outside the Mediterranean.

The etiology of T2D is multifactorial, so the ability to isolate the impact of a single food on T2D is likely to be influenced by other factors. Higher olive oil intake appeared to be more favorable in those individuals who had healthier diets or higher intake of total vegetables, green vegetables, and lettuce, whereas null associations were generally observed in those with less healthy diets. These analyses suggest that consuming olive oil in the context of a healthy diet may offer greater protection against diabetes, implying a synergistic effect on diabetes risk. Although olive oil itself is not presumed to be a panacea for an unhealthy diet, it is a key food component of healthy dietary patterns such as the Mediterranean diet, not only because of its nutritional properties but also because of the pattern’s cultural heritage.

With respect to cultural heritage, our subgroup analysis further suggested that individuals with Southern European/Mediterranean ancestry appeared to benefit more from high olive oil intake than those with other Caucasian or other ancestries. This observation may be explained because olive oil use is traditionally the primary cooking/dressing oil/fat in women with Mediterranean ancestry as opposed to, say, English or German ancestry, for whom butter or other fat is mainly used (27). Moreover, those participants with Mediterranean ancestry may have been more likely to be culturally and dietetically exposed to olive oil for a longer time, know better how to cook with olive oil, and consume it in larger amounts, even when they select it at the same frequency as their non-Mediterranean counterparts. Our observation that olive oil added to food or bread was associated with a reduced risk of T2D further adds to the idea that olive oil consumers may be manifesting a true cultural heritage, alongside consuming more vegetables and less processed meat. It is also possible that olive oil dressing is rarely based on extra-virgin olive oil, whereas olive oil added to food/bread is more likely to be extra virgin, especially among women with Mediterranean ancestry. This scenario may potentially explain why the association for olive oil dressing was weaker than that for olive oil added to food or bread and why a stronger inverse association between olive oil and T2D was observed in women of Mediterranean ancestry.

A substantial body of evidence supports that the type of fat rather than total dietary fat is a key determinant of the risk of T2D and other chronic diseases (5). The principal component of olive oil is oleic acid (6, 7), a MUFA that has been shown to decrease insulin resistance (28) and may be beneficial in managing T2D (29). High-MUFA diets improved glucose metabolism more than Western diets in a weight maintenance trial, as well as improved glycemic control and lipoprotein profile in diabetic individuals (30–32). In contrast, in our previous analyses in the NHS cohorts of predominantly healthy women, no substantial associations were found between MUFA intake and risk of T2D (33). However, the specific association of olive oil, which is rich in anti-inflammatory and antioxidant molecules, was not evaluated. Therefore, other important sources of MUFA intake in US populations, such as meats and dairy, may have confounded the association between MUFA intake and risk of T2D.

At baseline, women with higher intake of olive oil were slightly leaner than those who never consumed olive oil. Previous studies have also found no association between olive oil consumption and weight gain (34). Despite olive oil being high in fat and energy content, its high MUFAs are probably more readily metabolized in the postprandial state, induce thermogenesis, and increase total energy expenditure compared with other fats such as saturated fat (35).

Besides olive oil’s MUFA content, particularly in the extra-virgin and virgin varieties, it is an important source of vitamin E, polyphenols (mainly flavonoids), and other minor bioactive phytochemicals and lipid molecules (e.g., squalene, tocopherols) (6). Olive oil is thought to have anti-inflammatory and antiatherogenic effects (36), many of which are thought to underscore its importance in cardiovascular disease prevention. Combined, these bioactive components may reduce oxidative stress (37) and improve endothelial function (7), lipid profile (38), insulin sensitivity, and glycemic control (7), which are considered strong risk factors for T2D. A recent cross-sectional, population-based study supports these potential mechanisms, observing that individuals who consumed olive oil (compared with sunflower oil) had a 48% lower risk of obesity, a 51% lower risk of impaired glucose regulation, and a 47% and 60% lower risk of hypertriglyceridemia or low HDL cholesterol, respectively (9).

To our knowledge, the present study is the first prospective cohort study to test the associations between incident T2D and replacing salad dressing and other types of fat (stick and tub/soft margarine, butter, and mayonnaise) with olive oil. Although further studies are needed to confirm these findings, and our results on the substitutions were based on statistical predictions, it is suggested that consuming olive oil instead of other fat-based salad dressings or animal fat sources could be beneficial in the prevention of T2D. The results are in accordance with previous evidence from controlled intervention studies indicating that substituting saturated and trans fatty acids with MUFAs or PUFAs may be beneficial in terms of insulin sensitivity and consequently in reducing the risk of T2D (5).

The strengths of the present study include the large sample size, long duration of follow-up, and use of repeated measures of diet and lifestyle variables. On the other hand, several potential limitations of our study deserve attention. First, our study population consisted of female health professionals, which may limit the generalizability of our findings to other groups, particularly other ethnic groups or those from different socioeconomic strata. However, the biological effects of olive oil are likely to be the same in diverse populations. Second, although validated (17), the FFQ and self-reported diet can produce measurement error in olive oil intake. However, the use of repeated measures of diet to calculate cumulative averages reduced random measurement error caused by within-person variation and also accommodated dietary changes over time. Third, our FFQ was unable to distinguish between varietals of olive oil intake; extra-virgin varietals may have differential effects on T2D risk owing to higher polyphenol and other nutrient content. Thus, pooling all possible varietals as was done here may have obscured the plausibly stronger preventive benefits of extra-virgin olive oil. Finally, because of the observational nature of the present study, we are unable to demonstrate a causal association, and residual confounding remains a possibility, although we carefully controlled for a number of diabetes risk factors.

In conclusion, our longitudinal data in 2 large cohorts of women, after a long follow-up and using repeated measurements of diet, suggest that increased olive oil consumption was associated with a lower risk of incident T2D in women. From a public health standpoint, replacing fat-based salad dressings with olive oil, which is high in MUFAs and other antioxidant compounds, should be considered to reduce cardiometabolic risk. Further studies may be needed to evaluate the effects of high amounts of olive oil intake, as well as its specific varieties, on the risk of T2D and other cardiovascular disease risk factors.