Abstract
Background & aims:
Insulin and insulin-like growth factor (IGF)-1 signaling is a proposed mechanism linking dietary protein and major chronic diseases. However, it is unclear whether animal and plant proteins are associated with biomarkers of insulin and IGF axis.
Methods:
We analyzed a total of 14,709 participants from Nurses’ Health Study and Health Professionals Follow-up Study who had provided a blood sample. Detailed dietary information was assessed using validated food frequency questionnaires. We assessed C-peptide, insulin, IGF-1, and IGF binding proteins (BP). Multivariable-adjusted linear regressions were used to examine associations of animal and plant protein intake with biomarkers after adjusting for confounders.
Results:
The medians (5th-95th percentiles) of animal and plant protein intake (% of total energy) were 13% (8–19%) and 5% (4–7%), respectively. Compared to participants in the lowest quintile, those in the highest quintile of animal protein had 4.8% (95% CI: 1.9, 7.9; P-trend<.001) higher concentration of IGF-1 and −7.2% (95% CI: −14.8, 1.1; P for trend=0.03) and −11.8% (95% CI: −20.6, −1.9; P-trend<.001) lower concentration of IGFBP-1 and IGFBP-2, respectively, after adjustment for major lifestyle factors and diet quality. In contrast, no association was observed between animal protein intake and C-peptide, insulin and IGFBP-3. The associations were restricted to participants with at least one unhealthy lifestyle risk factor (i.e., overweight/obese, physical inactivity, smoking, and heavy alcohol intake). Plant protein tended to be strongly associated with numerous biomarkers in age-adjusted analyses but these became largely attenuated or non-significant in multivariable adjustment. Plant protein intake remained positively associated with IGF-1 (P-trend=0.002) and possibly IGFBP-1 (P-trend=0.02) after multivariable adjustment. Substitution of plant protein with animal protein sources was associated with lower IGFBP-1. In additional analysis, IGF-1 and IGFBPs were estimated to mediate approximately 5–20% of the association between animal protein and type 2 diabetes.
Conclusions:
Higher animal protein intake was associated with higher IGF-1 and lower IGFBP-1 and IGFBP-2, whereas higher plant protein intake was associated with higher IGF-1 and IGFBP-1.
Keywords: animal protein, plant protein, biomarker, insulin, insulin-like growth factor, substitution
INTRODUCTION
Dietary protein is a major macronutrient that accounts for a large proportion of total calorie intake. Identifying the optimal macronutrient composition of a diet for prevention of diseases and longevity has been a critical topic in nutrition research.[1, 2] Recently, a high protein and low carbohydrate diet has gained its popularity due to its potential beneficial effects on weight loss and maintenance of muscle.[3] Beyond the amount of protein intake, mounting evidence suggests that the food sources of protein intake may associate differently with diseases. Recent findings from epidemiological studies showed that diets high in animal protein and low in plant protein are associated with increased risk of major chronic diseases and mortality.[4–8] More specifically, we recently published consistent results from two large US prospective cohorts (Nurses’ Health Study (NHS) and Health Professionals Follow-up Study (HPFS)).[4, 5, 7, 8] In these studies, when comparing extreme quintiles, higher intake of animal protein was associated with a 13% increased risk of type 2 diabetes while higher intake of plant protein was associated with a 9% reduced risk of type 2 diabetes.[5] We also found evidence that animal sources of protein such as red meat may increase risk of cardiovascular diseases.[4, 7] Similar associations were found for mortality outcomes showing that high animal protein intake was positively associated with cardiovascular mortality whereas high plant protein intake was inversely associated with all-cause mortality and cardiovascular mortality.[8] Interestingly, these associations were confined to participants with at least one lifestyle risk factor and substitution of animal protein with plant protein was associated with lower mortality.
Insulin and insulin-like growth factor (IGF)-1 signaling pathway has been postulated as an important biological mechanism linking dietary protein intake with chronic diseases such as cancer and type 2 diabetes.[9, 10] Some studies with small sample sizes showed that animal sources of protein increased insulin resistance and IGF-1, while plant sources of protein improved insulin sensitivity and metabolic factors.[11–14] However, it is still inconclusive whether different sources of protein are associated with various biomarkers of insulin and IGF axis.
We therefore used two large prospective cohorts (NHS and HPFS) to examine the association of animal and plant protein with biomarkers of insulin, IGF-1 and IGF binding proteins (BPs). We also examined whether the associations differ by major lifestyle factors and conducted isocaloric substitution analysis for various food sources of protein in relation to biomarkers. Lastly, we performed an additional analysis examining the mediating association of biomarkers of insulin and IGF axis in the relationship between protein and type 2 diabetes.
METHODS
Study population
The Nurses’ Health Study (NHS) was initiated in 1976 when 121,700 female nurses aged 30 to 55 years were enrolled. The Health Professionals Follow-up Study (HPFS) was initiated in 1986 when 51,529 male health professionals aged 40–75 years were enrolled. Participants completed questionnaires at enrollment and every 2 years thereafter to collect the information on demographic, lifestyle, and medical factors. Response rate of each questionnaire cycle exceeded 90% for both cohorts.
Blood samples were collected between 1989 and 1990 from 32,826 women (NHS) and between 1993 and 1995 from 18,159 men (HPFS) who were free of major diseases such as cardiovascular disease, cancer, and diabetes. Each participant received a blood kit and returned it to the lab via overnight courier. Details of the procedures for blood collection, handling, and storage have been described previously.[15] As previously reported,[16] the characteristics of participants were generally similar between those in the full cohort and biomarker sub-cohort (Supplementary Table 1). For the current study, we included participants who provided a blood sample (C-peptide, insulin, IGF-1, IGFBP-1, IGFBP-2 and IGFBP-3) and were previously selected for nested case-control analyses of type 2 diabetes, ischemic heart disease, myeloma, colon polyps, colon cancer, pancreatic cancer, breast cancer, ovarian cancer and prostate cancer. A total of 10,333 women (NHS) and 4,376 men (HPFS) who had dietary and biomarker information were included in the final analysis. The study protocol was approved by the institutional review boards of the Brigham and Women’s Hospital and Harvard T.H. Chan School of Public Health.
Dietary assessment
Dietary information was assessed using a validated semi-quantitative food frequency questionnaire (FFQ) including over 130 food items. Participants were asked how often, on average, they consumed a standardized portion size of each food over the past year. The nutrient and energy intakes were calculated by multiplying the consumption frequency of each food item by its nutrient and energy contents and then summing across all food items. The values were collected from the US Department of Agriculture food composition database.[17] Animal and plant protein intake were primary exposures of interest. Animal and plant protein intake was derived from reported food items by summing the products between intake frequency (servings/d) and its animal and plant protein content (g/serving) using the food composition database, and then expressed as a percentage of total energy ((protein intake (g/d) × 4 (kcal/g)) ÷ total daily calorie intake). Major sources of animal protein included processed and unprocessed red meat, poultry, fish, eggs and dairy products. Major plant protein sources included bread, cereals, pasta, beans, nuts and legumes. Validity and reliability of FFQ has been previous described.[18–22] The Spearman correlation coefficients of FFQ and 7-day dietary records were 0.56 for animal protein and 0.66 for plant protein.[8, 22] The correlation of energy-adjusted protein intake (percentage of total calorie from protein) assessed by FFQ and its recovery biomarker (urinary nitrogen) was 0.46.[21]
Covariate assessment
From the biennial questionnaires, we collected detailed information on covariates such as age, race, height, weight, smoking status, medication, and history of disease. Body mass index (BMI) was calculated by weight in kilogram divided by height in meter squared. Chronic disease score was calculated by summing the number of prevalent diseases/conditions including hypercholesterolemia, high blood pressure, diabetes, heart disease, cancer, and rheumatoid/other arthritis.[23, 24]
Biomarker assessment
In both cohorts, we measured plasma levels of C-peptide, insulin, and biomarkers of IGF axis including IGF-1, IGFBP-1, IGFBP-2, and IGFBP3. Plasma levels of C-peptide, insulin, IGF-1, and IGFBPs were determined in the laboratory of Dr. Michael Pollak (Jewish General Hospital and McGill University) by ELISA with reagents from Diagnostic Systems Laboratory (Webster, TX). The mean intraassay coefficients of variation for C-peptide, insulin, IGF-1, IGFBPs from the blinded quality control samples were less than 15%.
Statistical analysis
We pooled the data from NHS and HPFS after testing for heterogeneity (P>0.05), and used generalized linear models to examine the association of animal and plant protein intake with biomarkers of insulin and IGF axis. All biomarkers were recalibrated using the methods previously developed by Rosner et al.[25] to adjust for potential variation in sample handling and laboratory drift across batches.
Animal and plant protein intake were used as percentages of total calorie intake, and categorized into quintiles for the analysis. We used a nutrient density model with adjustment for total energy intake[26]. Thus, the coefficient for animal and plant protein indicates the substitution effect of an equal amount of energy from protein for carbohydrate or fat. We conducted three models: Age and sex-adjusted model (Model 1) and two multivariable adjusted models including prespecified variable based on the literature. Model 2 included age, sex, race, fasting status, smoking, NSAIDS, number of chronic diseases, case-control status, physical activity, BMI and total calorie intake. This model examines the overall association of animal and plant sources of protein on biomarkers with insulin and IGF axis. Model 3 further adjusted for overall diet quality using the AHEI score to more specifically examine the association of animal and plant protein per se on biomarkers of insulin and IGF axis after controlling for diet quality. This model attempts to isolate statistically whether associations are due primarily to the protein content and composition, or to other dietary factors inherent in a diet in protein. Of note, animal and plant protein intake were mutually adjusted in multivariable models. Test for linear trend was assessed by including continuous variable of animal and plant protein intake in the above models. In a secondary analysis, we also examined that association of animal protein from meats (processed and unprocessed red meat, poultry, and fish) and non-meats (eggs and dairy products) sources, and animal-to-plant protein intake ratio in relation to biomarkers, separately.
We conducted stratified analysis by age, individual lifestyle factors (BMI, physical activity, smoking, and alcohol intake) and combined lifestyle factors (healthy vs. unhealthy lifestyle groups). Based on prior literature and recommendations,[8] healthy lifestyle was defined as never smoking or ever smoking less than 5 pack-year, never or moderate alcohol drinking (<14 g/d in women and <28 g/d in men), BMI of 18.5 to less than 25 kg/m2, and physical activity of at least 150 min/week at moderate level or at least 75 min/week at vigorous level (equivalent to ≥7.5 MET h/week). Test for interaction was performed by including the interaction term of each animal or plant protein intake (continuous) and stratification variable (binary) in the models.
Moreover, substitution analysis was further performed to examine the change in biomarkers by substituting 3% energy from plant protein with various animal protein sources (processed red meat, unprocessed red meat, poultry, fish, egg, and dairy). We also examined different types of fish (fatty and non-fatty fish) and dairy product (milk, yogurt, cheese, and others), separately. For the substitution analysis, we included both plant protein and animal protein sources and other macro nutrients (saturated, polyunsaturated, monounsaturated and trans fats and carbohydrate) as continuous variables in the same multivariable adjusted regression models (partition model approach).[27] The differences in β coefficients, variances, and covariance between plant protein and animal protein sources were used to estimate the substitution effect of animal protein sources for plant protein.[4, 16]
Lastly, we conducted an additional analysis to examine the mediating association of biomarkers of insulin and IGF axis in the relationship between animal and plant protein and type 2 diabetes. Cox regression was used to estimate the hazard ratios (HR) and 95% CIs of type 2 diabetes per 1 standard deviation (SD) increase after adjustment for age, race, case-control status, alcohol, smoking, family history of diabetes, physical activity, total calorie and menopausal status and hormone use for women, BMI and AHEI. We then further adjusted for the biomarkers to see the changes in the association between animal and plant protein and type 2 diabetes. We also estimated the proportion of the associations mediated by the biomarkers using a macro developed by Donna Spiegelman and colleagues (Mediate SAS; Harvard School of Public Health; available at http://www.hsph.harvard.edu/donna-spiegelman/software/mediate/).[28, 29] This macro computes the magnitude of mediation, 95% CI and P value for significance.[30]
For all statistical analyses, two-sided P value<0.005 was considered to be statistically significant as recommended by Benjamin et al.[31], whereas P values between 0.05 and 0.005 were considered suggestive of an association. All data analyses were performed using SAS 9.4 version
RESULTS
Age and sex-standardized characteristics of participants according to quintiles of animal and plant protein intake are described in Table 1 and Supplementary Tables 2–3. The medians (5–95th percentiles and ranges) of animal and plant protein (% of total energy) were 13% (8–19% and 0–40%) and 5% (4–7% and 1–28%), respectively. Participants with higher animal protein intake had higher BMI and lower percentage of current smokers. They also had higher animal protein and lower plant protein, carbohydrate, and alcohol intake. However, participants with higher plant protein intake had healthier lifestyle behaviors and healthier diets.
Table 1.
Age and sex standardized characteristics of participants according to quintiles of animal and plant protein intake
| Type of protein, % of total energy (N=14,709) | ||||||
|---|---|---|---|---|---|---|
| Animal protein | Plant protein | |||||
| Characteristics | Q1 ≤10.4 |
Q3 >12.2 to 13.9 |
Q5 >16.0 |
Q1 ≤4.3 |
Q3 >4.9 to 5.4 |
Q5 >6.1 |
| Age, yrs | 58.3 (8.0) | 57.3 (8.1) | 56.9 (8.1) | 56.5 (8.3) | 57.3 (8.2) | 58.4 (7.6) |
| BMI, kg/m2 | 24.9 (4.0) | 25.8 (4.4) | 26.5 (4.4) | 26.2 (4.6) | 25.7 (4.4) | 25.1 (4.0) |
| Physical activity, MET-hour/wk | 21.6 (25.9) | 21.8 (25.3) | 21.5 (24.8) | 18.6 (23.5) | 21.5 (24.7) | 23.6 (25.0) |
| Male, % | 28.1 | 30.2 | 30.7 | 31.3 | 29.6 | 28.8 |
| white, % | 98.5 | 98.8 | 98.8 | 98.9 | 98.5 | 98.2 |
| Smoking, % | ||||||
| Never | 47.4 | 45.7 | 46.4 | 42.5 | 47.3 | 48.4 |
| Past | 41.7 | 43.8 | 44.0 | 42.6 | 43.5 | 44.8 |
| Current | 11.0 | 10.5 | 9.7 | 14.8 | 9.1 | 6.8 |
| NSAID use, % | 28.8 | 29.7 | 31.2 | 31.0 | 31.2 | 28.5 |
| Number of chronic diseases | ||||||
| 0 | 44.8 | 43.9 | 42.4 | 46.3 | 43.3 | 40.3 |
| 1 | 32.5 | 32.4 | 32.2 | 31.4 | 32.3 | 32.9 |
| 2 | 16.3 | 16.6 | 16.8 | 15.9 | 17.0 | 18.4 |
| 3+ | 6.4 | 7.1 | 8.6 | 6.4 | 7.3 | 8.4 |
| Dietary intake | ||||||
| Total protein, g/d | 69.5 | 85.9 | 96.5 | 89.1 | 84.3 | 80.2 |
| Animal protein, g/d | 41.6 | 61.4 | 76.8 | 70.8 | 60.5 | 49.4 |
| Plant protein, g/d | 28.0 | 24.5 | 19.7 | 18.3 | 23.8 | 30.7 |
| Total fat, g/d | 62.5 | 65.9 | 60.5 | 72.6 | 64.0 | 54.6 |
| Carbohydrate, g/d | 264.7 | 237.2 | 191.2 | 217.9 | 232.8 | 245.8 |
| Total calorie, kcal/d | 1905 | 1893 | 1700 | 1923 | 1851 | 1770 |
| Fiber intake, g/d | 21.5 | 19.7 | 18.5 | 14.5 | 19.5 | 25.7 |
| Whole grain, g/d | 27.0 | 23.0 | 21.7 | 12.9 | 22.7 | 37.2 |
| Fruit, serving/wk | 13.7 | 12.3 | 10.8 | 9.7 | 12.3 | 14.3 |
| Vegetable, serving/wk | 22.6 | 22.9 | 22.9 | 17.9 | 22.2 | 28.9 |
| Legume, serving/wk | 2.1 | 1.9 | 1.5 | 1.2 | 1.7 | 2.8 |
| Nut, serving/wk | 1.8 | 1.1 | 0.7 | 0.7 | 1.1 | 1.9 |
| Bean, serving/wk | 1.0 | 0.8 | 0.6 | 0.5 | 0.7 | 1.4 |
| Unprocessed red meat, serving/wk | 2.7 | 4.1 | 4.3 | 5.2 | 3.9 | 2.3 |
| Processed red meat, serving/wk | 1.5 | 1.8 | 1.5 | 2.6 | 1.7 | 0.9 |
| Chicken, serving/wk | 1.6 | 2.7 | 3.9 | 2.9 | 2.7 | 2.6 |
| Fish, serving/wk | 1.5 | 2.1 | 3.0 | 1.9 | 2.2 | 2.4 |
| Egg, serving/wk | 1.2 | 1.5 | 1.6 | 2.0 | 1.5 | 1.0 |
| Low-fat dairy product, serving/wk | 6.2 | 9.7 | 12.2 | 10.9 | 9.3 | 7.9 |
| High-fat dairy product, serving/wk | 6.5 | 6.6 | 5.0 | 9.5 | 6.2 | 3.6 |
| Alternate Healthy Eating Index | 52.5 | 51.4 | 53.5 | 44.9 | 51.6 | 60.7 |
Values are means (SD) for continuous variables; percentages for categorical variables, and are standardized to the age distribution of the study population. The median and 5th and 95th percentiles of animal protein (% of total energy): 14% and 9–20% for women; 12% and 7–17% for men; 13% and 8–19% for total population. The median and 5th and 95th percentiles of plant protein (% of total energy): 5% and 4–7% for women; 5% and 4–8% for men; 5% and 4–7% for total population.
Animal and plant protein intake and biomarkers
Higher intake of animal protein was associated with higher concentration of IGF-1 and lower concentration of IGFBP-2 and possibly IGFBP-1, after adjusting for confounders including BMI and other non-dietary factors (Table 2). Compared to participants in the lowest quintile, those in the highest quintile of animal protein had 5.3% (95% CI: 2.4, 8.3; P-trend<.001) higher concentration of IGF-1 and −7.3% (95% CI: −14.8, 0.8; P-trend=0.02) and −9% (95% CI: −17.9, 0.7; P-trend=0.004) lower concentration of IGFBP-1 and IGFBP-2, respectively. In contrast, no associations were observed between animal protein intake and C-peptide, insulin and IGFBP-3. Additional adjustment for diet quality did not materially change the association for the IGF axis. In a secondary analysis of comparing animal protein from meat and non-meat sources, higher intake of animal meat protein was associated with higher IGF-1 and lower IGFBP-1 and IGFBP-2, while higher intake of animal non-meat protein was associated with lower C-peptide and higher IGF-1 and IGFBP-1 (Supplementary Table 4–5).
Table 2.
Association between animal protein intake and biomarkers of insulin and IGF axis (pooled results of NHS and HPFS)
| Quintiles of animal protein intake (%) | Percentage difference Q5 vs. Q1 | Ptrend | |||||
|---|---|---|---|---|---|---|---|
| Q1 ≤10.4 |
Q2 >10.4 to 12.2 |
Q3 >12.2–13.9 |
Q4 >13.9 to 16.0 |
Q5 >16.0 |
|||
|
C-peptide
(N=11146) |
|||||||
| Model 1 | 2.15 (0.03) | 2.14 (0.03) | 2.2 (0.03) | 2.24 (0.03) | 2.36 (0.03)* | 9.8 (4.8, 14.9) | <.001 |
| Model 2 | 2.55 (0.09) | 2.46 (0.09) | 2.46 (0.09) | 2.48 (0.09) | 2.50 (0.09) | −1.9 (−6.3, 2.7) | 0.47 |
| Model 3 | 2.52 (0.09) | 2.45 (0.09) | 2.46 (0.09) | 2.49 (0.09) | 2.54 (0.09) | 0.9 (−3.7, 5.8) | 0.31 |
|
Insulin
(N=1750) |
|||||||
| Model 1 | 7.48 (0.28) | 8.03 (0.28) | 8.42 (0.31) | 7.91 (0.27) | 8.84 (0.33)* | 18.1 (3.8, 34.4) | 0.004 |
| Model 2 | 9.34 (0.81) | 9.60 (0.82) | 9.86 (0.84) | 9.22 (0.77) | 9.93 (0.85) | 6.3 (−6.8, 21.2) | 0.37 |
| Model 3 | 9.27 (0.80) | 9.54 (0.81) | 9.86 (0.84) | 9.29 (0.77) | 10.07 (0.86) | 8.7 (−4.9, 24.3) | 0.14 |
|
IGF-1
(N=10499) |
|||||||
| Model 1 | 132 (0.99) | 134 (1.00) | 136 (1.02)* | 136 (1.00)* | 135 (1.00) | 2.3 (−0.3, 5.0) | 0.07 |
| Model 2 | 131 (2.67) | 134 (2.73) | 137 (2.79)* | 137 (2.78)* | 138 (2.82)* | 5.3 (2.4, 8.3) | <.001 |
| Model 3 | 131 (2.68) | 134 (2.73) | 137 (2.79)* | 137 (2.78)* | 137 (2.81)* | 4.8 (1.9, 7.9) | <.001 |
|
IGFBP-1
(N=6641) |
|||||||
| Model 1 | 17.8 (0.46) | 16.9 (0.42) | 15.7 (0.39)* | 15.4 (0.38)* | 13.9 (0.34)* | −21.9 (−28.4, −14.8) | <.001 |
| Model 2 | 12.5 (0.79) | 12.4 (0.78) | 12.0 (0.76) | 12.1 (0.76) | 11.6 (0.73) | −7.3 (−14.8, 0.8) | 0.02 |
| Model 3 | 12.5 (0.79) | 12.4 (0.78) | 12.0 (0.76) | 12.1 (0.76) | 11.6 (0.73) | −7.2 (−14.8, 1.1) | 0.03 |
|
IGFBP-2
(N=2000) |
|||||||
| Model 1 | 320 (10.4) | 311 (9.5) | 289 (8.8) | 270 (7.5)* | 260 (7.4)* | −18.7 (−26.9, −9.6) | <.001 |
| Model 2 | 251 (16.5) | 254 (16.6) | 236 (15.3) | 229 (14.5) | 229 (14.6) | −9.0 (−17.9, 0.7) | 0.004 |
| Model 3 | 255 (16.8) | 256 (16.7) | 236 (15.2) | 227 (14.4)* | 225 (14.4)* | −11.8 (−20.6, −1.9) | <.001 |
|
IGFBP-3
(N=10501) |
|||||||
| Model 1 | 4031 (21.7) | 4049 (21.7) | 4063 (21.8) | 4048 (21.3) | 4047 (21.5) | 0.4 (−1.4, 2.3) | 0.80 |
| Model 2 | 4094 (59.5) | 4116 (59.6) | 4133 (60.1) | 4118 (59.6) | 4117 (59.9) | 0.6 (−1.4, 2.6) | 0.86 |
| Model 3 | 4099 (59.7) | 4118 (59.7) | 4132 (60.1) | 4115 (59.6) | 4107 (60.0) | 0.2 (−1.8, 2.3) | 0.46 |
Different from quintile 1 (P<0.05)
Units of biomarkers: ng/mL for C-peptide; mIU/mL for insulin; ng/mL for IGF-1; ng/mL for IGFBP-1; ng/mL for IGFBP-2; ng/mL for IGFBP3.
Model 1: adjusted for age (continuous) and sex. Model 2: Model 1 + race (white or nonwhite), fasting status (yes or no), smoking (never, past or current), NSAIDS (yes or no), chronic disease score (0, 1, 2, 3+), case-control status, physical activity (continuous) and BMI (continuous). Model 3: Model 2 + Alternate Healthy Eating Index (continuous). Animal protein and plant protein were mutually adjusted.
Higher intake of plant protein was associated with lower concentration of C-peptide and higher concentration of IGF-1 and IGFBP-1 after adjusting for BMI and other non-dietary factors (Table 3). Compared to participants in the lowest quintile, those in the highest quintile of plant protein had −6.7% (95% CI: −10.9, −2.4; P-trend<.001) lower concentration of C-peptide and 4.9% (95% CI: 2.0, 7.9; P-trend<.001) and 6.3% (95% CI: −2.1, 15.5; P-trend=0.009) higher concentration of IGF-1 and IGFBP-1, respectively. After adjustment for diet quality, the associations remained significant for IGF-1 (P-trend<.001) and suggestive for IGFBP-1 (P-trend=0.02) while the association for C-peptide became null. In contrast, there was no association between plant protein intake and insulin, IGFBP-2 and IGFBP-3. When we examined the association of animal-to-plant protein intake ratio with the biomarkers, we found that higher ratio of animal to protein intake was associated with lower IGFBP-1 and IGFBP-2 (Supplementary Table 6). Using deciles of animal and plant protein intake instead of quintiles did not change the results (Supplementary Tables 7–8). The separate results for men and women were consistent with the pooled results (Supplementary Tables 9–12).
Table 3.
Association between plant protein intake and biomarkers of insulin and IGF axis (pooled results of NHS and HPFS)
| Quintiles of plant protein intake (%) | Percentage difference Q5 vs. Q1 | Ptrend | |||||
|---|---|---|---|---|---|---|---|
| Q1 ≤4.3 |
Q2 >14.3 to 4.9 |
Q3 >4.9 to 5.4 |
Q4 >5.4 to 6.1 |
Q5 >6.1 |
|||
|
C-peptide
(N=11146) |
|||||||
| Model 1 | 2.35 (0.03) | 2.29 (0.03) | 2.21 (0.03)* | 2.17 (0.03)* | 2.08 (0.03)* | −11.7 (−15.6, −7.6) | <.001 |
| Model 2 | 2.57 (0.09) | 2.54 (0.09) | 2.48 (0.09) | 2.46 (0.09)* | 2.40 (0.09)* | −6.7 (−10.9, −2.4) | <.001 |
| Model 3 | 2.50 (0.09) | 2.50 (0.09) | 2.47 (0.09) | 2.49 (0.09) | 2.49 (0.09) | −0.3 (−5.4, 5.1) | 0.66 |
|
Insulin
(N=1750) |
|||||||
| Model 1 | 8.44 (0.30) | 8.45 (0.30) | 8.00 (0.29) | 7.79 (0.29) | 7.90 (0.29) | −6.4 (−17.5, 6.2) | 0.04 |
| Model 2 | 9.49 (0.81) | 9.68 (0.82) | 9.43 (0.80) | 9.45 (0.81) | 9.57 (0.82) | 0.8 (−11.2, 14.6) | 0.96 |
| Model 3 | 9.21 (0.80) | 9.57 (0.81) | 9.42 (0.80) | 9.58 (0.82) | 9.94 (0.87) | 7.9 (−7.0, 25.2) | 0.20 |
|
IGF-1
(N=10499) |
|||||||
| Model 1 | 132 (0.98) | 134 (0.99) | 135 (1.00) | 137 (1.03)* | 137 (1.03)* | 3.8 (1.1, 6.5) | <.001 |
| Model 2 | 132 (2.70) | 134 (2.73) | 135 (2.73) | 137 (2.79)* | 138 (2.83)* | 4.9 (2.0, 7.9) | <.001 |
| Model 3 | 132 (2.72) | 135 (2.76) | 135 (2.73) | 137 (2.79)* | 137 (2.86)* | 3.6 (0.3, 7.0) | 0.002 |
|
IGFBP-1
(N=6641) |
|||||||
| Model 1 | 14.4 (0.36) | 15.2 (0.38) | 15.7 (0.39)* | 16.3 (0.41)* | 17.8 (0.44)* | 23.1 (13.1, 34.1) | <.001 |
| Model 2 | 11.8 (0.75) | 12.0 (0.76) | 12.2 (0.77) | 12.1 (0.76) | 12.5 (0.79) | 6.3 (−2.1, 15.5) | 0.009 |
| Model 3 | 11.8 (0.75) | 12.0 (0.76) | 12.2 (0.77) | 12.1 (0.76) | 12.4 (0.79) | 5.5 (−4.1, 16.1) | 0.02 |
|
IGFBP-2
(N=2000) |
|||||||
| Model 1 | 270 (7.9) | 285 (8.4) | 283 (8.4) | 290 (9.0) | 308 (9.4)* | 13.9 (2.7, 26.4) | <.001 |
| Model 2 | 234 (15.3) | 237 (15.2) | 244 (15.5) | 233 (15.0) | 242 (15.6) | 3.4 (−6.6, 14.5) | 0.38 |
| Model 3 | 241 (15.9) | 240 (15.4) | 243 (15.5) | 230 (14.9) | 232 (15.3) | −3.9 (−14.6, 8.1) | 0.34 |
|
IGFBP-3
(N=10501) |
|||||||
| Model 1 | 4015 (21.3) | 4037 (21.4) | 4050 (21.5) | 4052 (21.9) | 4085 (22.0) | 1.8 (−0.1, 3.7) | 0.06 |
| Model 2 | 4089 (59.6) | 4107 (59.7) | 4114 (59.4) | 4116 (59.6) | 4147 (60.5) | 1.4 (−0.6, 3.5) | 0.22 |
| Model 3 | 4103 (60.3) | 4112 (59.8) | 4115 (59.4) | 4110 (59.6) | 4128 (61.3) | 0.6 (−1.7, 2.9) | 0.94 |
Different from quintile 1 (P<0.05)
Units of biomarkers: ng/mL for C-peptide; mIU/mL for insulin; ng/mL for IGF-1; ng/mL for IGFBP-1; ng/mL for IGFBP-2; ng/mL for IGFBP3.
Model 1: adjusted for age (continuous) and sex. Model 2: Model 1 + race (white or nonwhite), fasting status (yes or no), smoking (never, past or current), NSAIDS (yes or no), chronic disease score (0, 1, 2, 3+), case-control status, physical activity (continuous) and BMI (continuous). Model 3: Model 2 + Alternate Healthy Eating Index (continuous). Animal protein and plant protein were mutually adjusted.
Stratified analyses by lifestyle factors
In stratified analysis, higher intake of animal protein was associated with higher concentration of IGF-1 and lower concentration of IGFBP-1 and IGFBP-2 among participants with unhealthy lifestyle behaviors (Table 4). Compared to participants in the lowest quintile, those in the highest quintile of animal protein had 5.1% (95% CI: 1.9, 8.4; P-trend<.001) higher concentration of IGF-1 and −10.2% (95% CI: −18.2, −1.4; P-trend=0.02) and −13.1% (95% CI: −22.4, −2.7; P-trend<.001) lower concentration of IGFBP-1 and IGFBP-2, respectively. In contrast, there was no association between animal protein intake and biomarkers of insulin and IGF axis among those with healthy lifestyle behaviors but these associations were generally in opposite directions except for IGF-1. However, there was no significant interaction by lifestyle behaviors. When we stratified by age and individual lifestyle factors, the associations between animal protein intake and IGF-1, IGFBP-1, and IGFBP-2 tended to be stronger in participants who were younger (<60 years), overweight/obese, or ever-smoker (Supplementary Tables 13–17).
Table 4.
Association between animal and plant protein and biomarkers of insulin and IGF axis, stratified by healthy and unhealthy lifestyle groups# (pooled results of NHS and HPFS)
| Quintiles of animal protein (%) | Percentage difference Q5 vs. Q1 | Ptrend | Pinteraction | |||||
|---|---|---|---|---|---|---|---|---|
| Q1 ≤10.4 |
Q2 >10.4 to 12.2 |
Q3 >12.2–13.9 |
Q4 >13.9 to 16.0 |
Q5 >16.0 |
||||
| C-peptide | ||||||||
| Healthy (N=2539) | 2.20 (0.49) | 2.07 (0.46) | 2.14 (0.48) | 2.01 (0.45) | 2.11 (0.48) | −3.8 (−15.5, 9.6) | 0.33 | 0.14 |
| Unhealthy (N=8607) | 2.59 (0.10) | 2.54 (0.10) | 2.54 (0.10) | 2.61 (0.10) | 2.65 (0.10) | 2.2 (−2.9, 7.5) | 0.09 | |
| Insulin | ||||||||
| Healthy (N=354) | 8.17 (2.10) | 8.05 (2.19) | 8.18 (2.19) | 7.98 (2.08) | 7.65 (2.00) | −6.3 (−36.4, 37.9) | 0.30 | 0.79 |
| Unhealthy (N=1396) | 9.61 (0.88) | 10.06 (0.89) | 10.37 (0.92) | 9.76 (0.85) | 10.8 (0.97) | 12.4 (−2.7, 29.8) | 0.04 | |
| IGF-1 | ||||||||
| Healthy (N=2385) | 131 (11.27) | 135 (11.67) | 146 (12.66)* | 138 (12.07) | 138 (12.15) | 5.0 (−2.9, 13.5) | 0.02 | 0.90 |
| Unhealthy (N=8114) | 131 (2.92) | 134 (2.98) | 135 (3.00)* | 137 (3.03)* | 137 (3.05)* | 5.1 (1.9, 8.4) | <.001 | |
| IGFBP-1 | ||||||||
| Healthy (N=1492) | 19.4 (5.58) | 21.5 (6.26) | 20.3 (5.97) | 18.9 (5.56) | 21.7 (6.43) | 11.8 (−10.6, 39.7) | 0.96 | 0.37 |
| Unhealthy (N=5149) | 11.7 (0.83) | 11.3 (0.80) | 11.0 (0.77) | 11.2 (0.78) | 10.5 (0.74)* | −10.2 (−18.2, −1.4) | 0.02 | |
| IGFBP-2 | ||||||||
| Healthy (N=385) | 354 (89.0) | 416 (112.2) | 394 (104.8) | 359 (95.4) | 361 (95.1) | 2.1 (−24.4, 37.7) | 0.76 | 0.21 |
| Unhealthy (N=1615) | 251 (17.5) | 247 (16.9) | 229 (15.6) | 221 (14.7)* | 218 (14.7)* | −13.1 (−22.4, −2.7) | <.001 | |
| IGFBP-3 | ||||||||
| Healthy (N=2386) | 4082 (241) | 4174 (247.7) | 4271 (254.2) | 4188 (251.4) | 4090 (247.1) | 0.2 (−5.1, 5.7) | 0.73 | 0.50 |
| Unhealthy (N=8115) | 4089 (65.3) | 4097 (65.4) | 4099 (65.4) | 4094 (64.9) | 4100 (65.5) | 0.3 (−1.9, 2.5) | 0.49 | |
| Quintiles of plant protein (%) | Percentage difference Q5 vs. Q1 | Ptrend | Pinteraction | |||||
| Q1 ≤4.3 |
Q2 >14.3 to 4.9 |
Q3 >4.9 to 5.4 |
Q4 >5.4 to 6.1 |
Q5 >6.1 |
||||
| C-peptide | ||||||||
| Healthy (N=2539) | 2.20 (0.50) | 2.21 (0.50) | 2.15 (0.48) | 2.11 (0.47) | 2.03 (0.46) | −7.4 (−19.6, 6.6) | 0.69 | 0.17 |
| Unhealthy (N=8607) | 2.58 (0.10) | 2.59 (0.10) | 2.57 (0.10) | 2.59 (0.10) | 2.61 (0.10) | 1.5 (−4.1, 7.4) | 0.85 | |
| Insulin | ||||||||
| Healthy (N=354) | 7.68 (2.12) | 9.83 (2.64) | 7.97 (2.05) | 8.26 (2.16) | 7.10 (1.80) | −7.6 (−39.4, 40.9) | 0.20 | 0.45 |
| Unhealthy (N=1396) | 9.65 (0.87) | 9.91 (0.87) | 9.90 (0.88) | 10.04 (0.9) | 10.75 (0.99) | 11.4 (−5.2, 30.9) | 0.05 | |
| IGF-1 | ||||||||
| Healthy (N=2385) | 128 (11.3) | 134 (11.7) | 133 (11.5) | 141 (12.2)* | 142 (12.5)* | 11.0 (1.9, 21.0) | <.001 | 0.04 |
| Unhealthy (N=8114) | 132 (2.96) | 134 (2.97) | 135 (2.98) | 136 (3.02) | 136 (3.09) | 2.5 (−1.0, 6.1) | 0.04 | |
| IGFBP-1 | ||||||||
| Healthy (N=1492) | 19.5 (5.75) | 20.2 (5.94) | 19.4 (5.60) | 20.4 (5.96) | 23.2 (6.83) | 18.8 (−6.0, 50.0) | 0.31 | 0.48 |
| Unhealthy (N=5149) | 10.9 (0.77) | 11.1 (0.78) | 11.4 (0.80) | 11.2 (0.79) | 11.2 (0.80) | 2.6 (−7.6, 14.0) | 0.07 | |
| IGFBP-2 | ||||||||
| Healthy (N=385) | 327 (87.0) | 386 (103.8) | 355 (91.9) | 337 (85.9) | 361 (96.4) | 10.3 (−19.7, 51.6) | 0.15 | 0.18 |
| Unhealthy (N=1615) | 238 (16.5) | 233 (15.7) | 237 (15.8) | 224 (15.3) | 223 (15.6) | −6.3 (−17.5, 6.4) | 0.51 | |
| IGFBP-3 | ||||||||
| Healthy (N=2386) | 4122 (249.2) | 4063 (242.9) | 4095 (243.5) | 4183 (248.0) | 4259 (257.5) | 3.3 (−2.6, 9.6) | 0.03 | 0.04 |
| Unhealthy (N=8115) | 4086 (65.8) | 4106 (65.4) | 4107 (65.0) | 4085 (65.2) | 4091 (66.7) | 0.1 (−2.4, 2.7) | 0.38 | |
Different from quintile 1 (P<0.05)
Healthy lifestyle was defined as never smoking or ever smoking less than 5 pack-year, never or moderate alcohol drinking (<14 g/d in women and <28 g/d in men), BMI of 18.5 to less than 25, and physical activity of at least 150 min/wk at moderate level or at least 75 min/wk at vigorous level (equivalent to ≥7.5 MET-h/wk).
Units of biomarkers: ng/mL for C-peptide; mIU/mL for insulin; ng/mL for IGF-1; ng/mL for IGFBP-1; ng/mL for IGFBP-2; ng/mL for IGFBP3.
All models adjusted for age (continuous), sex, race (white or nonwhite), fasting status (yes or no), smoking (never, past or current), NSAIDS (yes or no), chronic disease score (0, 1, 2, 3+), case-control status, physical activity (continuous), BMI (quintiles) and Alternate Healthy Eating Index (continuous). Animal and plant proteins were mutually adjusted in all models.
Isocaloric nutrient-substitution models
Substituting 3% of plant protein with animal protein was suggestively inversely associated with IGFBP-1 (P=0.03) (Table 5). Moreover, substituting 3% of plant protein with animal protein sources such as poultry or fish was suggestively positively associated with C-peptide and inversely associated with IGFBP-1. Higher Insulin and IGF-1 were tended to be observed when substituting 3% of plant protein with eggs (P=0.04) and dairy products (P=0.01), respectively. Lower IGFBP-3 was tended to be shown when substituting 3% of plant protein with processed red meat (P=0.03). In secondary analysis, the observed higher C-peptide for fish and IGF-1 for dairy were possibly mainly driven by non-fatty fish (P=0.05 for C-peptide) and other dairy products excluding milk, yogurt, and cheese (P<.001 for IGF-1) (Supplementary Table 18). Moreover, substituting 3% of plant protein with cheese was suggestively associated with lower insulin (P<.001) and higher IGFBP-1 (P=0.03). Higher IGFBP-2 and lower IGFBP-3 were observed when substituting 3% of plant protein with yogurt (P<.001), and other dairy (P=0.04), respectively.
Table 5.
Effect estimates for changes in biomarkers by substitution of 3% energy from plant protein with different sources of animal protein (pooled results of NHS and HPFS)
| C-peptide | Insulin | IGF-1 | ||||
|---|---|---|---|---|---|---|
| β ± SE | P | β ± SE | P | β ± SE | P | |
| Plant protein with | ||||||
| Animal protein | 0.031 ± 0.019 | 0.10 | −0.003 ± 0.052 | 0.95 | 0.005 ± 0.011 | 0.65 |
| Processed red meat | −0.035 ± 0.071 | 0.62 | −0.230 ± 0.202 | 0.25 | −0.053 ± 0.042 | 0.21 |
| Unprocessed red meat | 0.031 ± 0.022 | 0.16 | −0.001 ± 0.06 | 0.98 | 0.002 ± 0.013 | 0.88 |
| Poultry | 0.040 ± 0.02 | 0.05 | −0.024 ± 0.056 | 0.67 | −0.010 ± 0.012 | 0.41 |
| Fish | 0.068 ± 0.022 | 0.002 | 0.063 ± 0.062 | 0.31 | 0.005 ± 0.014 | 0.72 |
| Egg | −0.002 ± 0.052 | 0.97 | 0.313 ± 0.153 | 0.04 | −0.007 ± 0.031 | 0.82 |
| Dairy | 0.005 ± 0.02 | 0.80 | −0.022 ± 0.058 | 0.70 | 0.035 ± 0.013 | 0.01 |
| IGFBP-1 | IGFBP-2 | IGFBP-3 | ||||
| β ± SE | P | β ± SE | P | β ± SE | P | |
| Plant protein with | ||||||
| Animal protein | −0.072 ± 0.033 | 0.03 | 0.006 ± 0.042 | 0.89 | 0.001 ± 0.008 | 0.70 |
| Processed red meat | 0.048 ± 0.125 | 0.70 | −0.310 ± 0.166 | 0.06 | −0.065 ± 0.03 | 0.03 |
| Unprocessed red meat | −0.052 ± 0.039 | 0.18 | −0.048 ± 0.049 | 0.33 | −0.006 ± 0.009 | 0.54 |
| Poultry | −0.105 ± 0.035 | 0.003 | 0.024 ± 0.045 | 0.60 | −0.012 ± 0.009 | 0.18 |
| Fish | −0.124 ± 0.039 | 0.001 | −0.011 ± 0.049 | 0.83 | 0.011 ± 0.01 | 0.27 |
| Egg | 0.060 ± 0.096 | 0.53 | −0.105 ± 0.117 | 0.37 | −0.042 ± 0.022 | 0.06 |
| Dairy | −0.009 ± 0.037 | 0.81 | −0.002 ± 0.045 | 0.97 | 0.009 ± 0.009 | 0.54 |
Biomarkers were log-transformed. Multivariable models included protein intake from plant sources and from all the animal food items, other macro nutrients (saturated, polyunsaturated, monounsaturated and trans fats and carbohydrate), total calorie intake (all continuous), age (continuous), sex, race (white or nonwhite), fasting status (yes or no), smoking (never, past or current), NSAIDS (yes or no), chronic disease score (0, 1, 2, 3+), case-control status, physical activity (continuous), BMI (continuous) and Alternate Healthy Eating Index (continuous). The differences in β coefficients, variances, and covariance were used to estimate the substitution effect.[4, 16]
Mediation analysis
Among participants with biomarkers, multivariable adjusted HRs (95% CI) of type 2 diabetes per 1-SD increase of animal and plant protein intake were 1.18 (1.12–1.25) and 0.89 (0.83–0.94), respectively (Supplementary Table 19). After adjustment for BMI and AHEI, we observed a positive association for animal protein (HR=1.10; 95% CI=1.03–1.16) but no association for plant protein (HR=1.03; 95% CI=0.95–1.11). The results were similar in the full cohort participants with and without biomarkers. In the mediation analysis, additional adjustment for biomarkers of insulin and IGF axis partially attenuated the association between animal protein and type 2 diabetes. For animal protein, the proportion of association mediated by the biomarkers were 4.5% (1.2%−14.9%, P=0.05) for C-peptide, 7.8% (2.6%−21.2%, P=0.02) for insulin, 4.9% (1.5%−14.7%, P=0.02) for IGF-1, 14.2% (6.4%−28.4%, P<.001) for IGFBP-1, 9.2% (2.4%−29.2%, P=0.046) for IGFBP-2, and 20.0% (9.1%−38.3%, P<.001) for IGFBP-3.
DISCUSSION
In 14,709 US adults, we found that higher intake of animal protein was associated with higher levels of IGF-1 and lower levels of IGFBP-2. In contrast, higher intake of plant protein intake was associated with higher levels of IGF-1. Moreover, higher animal protein intake and lower plant protein intake was suggestively associated with lower levels of IGFBP-1 and substitution of plant protein with animal protein sources was associated with lower levels of IGFBP-1.
Comparison with previous studies
Insulin resistance and hyperinsulinemia is one of the major biological mechanisms linking diet and a number of chronic diseases. Recent studies provided supporting evidence that diets with high insulinemic potential were associated with increased risk of colorectal cancer, type 2 diabetes and multiple myeloma.[32–34] Diets with high insulinemic potential included foods with high animal proteins such as red meat, processed meat, poultry, non-fatty fish, eggs and low-fat dairy products.[14] These foods were highly correlated with circulating C-peptide in the previous studies. In the current study, we also found a positive association between animal protein and C-peptide, but the association was nullified after adjustment for BMI and other major lifestyle factors. This suggests that the higher C-peptide associated with higher intake of animal protein is largely mediated or confounded by weight gain/lifestyle factors, which cannot be distinguished through our study design. Participants with higher intake of animal protein had higher BMI while those with higher intake of plant protein had lower BMI. Unlike animal protein, plant protein was inversely associated with C-peptide after adjustment for BMI and other major lifestyle factors, but the association became null when we further adjusted for overall diet quality. Although plant protein was not independently associated with C-peptide, substitution of plant protein with animal protein sources, particularly poultry and non-fatty fish was suggestively associated with higher C-peptide levels. The overall results for insulin were generally consistent with C-peptide.
IGF-1 signaling pathway plays an important role in regulation of cell proliferation, apoptosis, survival, metabolism, and migration. IGFBPs function as transport proteins for circulating IGF-1 and regulate the activity of IGF-1.[35] It is postulated that IGF-1 signaling pathway is associated with higher risk of cancer.[36, 37] Several studies have examined the association of diets with IGF-1 and IGFBPs. A cross-sectional study of 1,037 healthy women found no association between fat and carbohydrate intake with IGF-1, but there was a positive association between protein intake with circulating IGF-1 concentration.[12] However, protein intake was not associated with IGFBP-3. Similar results were found among middle aged and elderly men suggesting that higher intake of protein, red meat, fish and seafood are associated with higher IGF-1 concentrations.[11] Another study of 292 women also showed evidence that a plant based diet is inversely associated with IGF-1 while positively associated with IGFBP-1 and IGBP-2 when they compared meat eaters, vegetarians and vegans. However, there were no differences in C-peptide and IGFBP-3 among the diet groups.[13] Consistent with the previous studies, our study showed that protein intake from animal and plant sources are positively associated with IGF-1. Moreover, animal protein was inversely associated with IGFBP-1 and IGFBP-2 while plant protein was positively associated with IGFBP-1 after adjusting for lifestyle factors and diet quality. The observed differences in the biomarker concentrations showed potentially clinically relevant values ranging from 4% to 12% (highest vs. lowest quintiles in the fully adjusted models) that are comparable to previous diet and biomarker studies.[16, 38–40]
In addition, we found that the inverse associations between animal protein and IGFBP-1 and IGFBP-2 biomarkers were only observed in participants with unhealthy behaviors (i.e., obesity, heavy alcohol intake, smoking and physical inactivity). In contrast, participants with healthy lifestyle behaviors showed no associations between animal protein and the biomarkers and the direction of associations were generally opposite except for IGF-1 and IGFBP-3. There are several explanations for this finding. First, the influence of high animal protein intake on IGFBP-1 and IGFBP-2 may be enhanced by other unhealthy behaviors that may already have upregulated insulin and IGF-1 and downregulated IGFBP-1 and IGFBP-2. Second, participants with healthy versus unhealthy behaviors may be consuming different sources of animal proteins even if total animal protein was the same. In our secondary analysis, we found that animal meat protein tends to be more adversely associated with biomarkers of insulin and IGF axis, compared to non-meat animal protein.
In the mediation analysis, we found that IGF-1 and IGFBPs were estimated to mediate approximately 5–20% of the association between animal protein and type 2 diabetes. To our knowledge, no studies have investigated the mediation effects of insulin/IGF biomarkers in the relationship between dietary protein and risk of chronic diseases including diabetes. However, our group (NHS) [41] and others [42–45] have previously conducted prospective observational studies and reported that lower IGFBP-1 and 2 [41, 43, 44] and higher IGFBP-3 [41, 42] were associated with increased risk of type 2 diabetes. The results for total IGF-1 [41, 42, 45] was less consistent but free IGF-1 (most bioactive component of total IGF-1) [41] was positively associated with risk of type 2 diabetes. Our study provides more direct evidence supporting that higher animal protein intake may increase the risk of type 2 diabetes partially mediated through IGF-1 signaling pathway. However, our mediation analysis included rather restricted and smaller number of participants with varying numbers of the biomarkers measured once over a decade of follow-up. Thus, it is important to interpret our results with cautions as the results may have been less precise and underestimated to some degree. More large studies, preferably with repeated measures of biomarkers, are required to comprehensively analyze whether the associations between dietary protein and chronic diseases are mediated through insulin and IGF pathways.
Implications and future directions
Our study has several important implications. First, our findings provide plausible evidence for recent dietary protein studies that insulin and IGF-1 pathways are potential underlying biological mechanisms that explain the observed link between dietary protein intake and risk of diseases and mortality. Second, beyond the amount of protein intake, sources of protein are important determinants of circulating insulin and IGFs. We observed clear associations for animal protein sources and provide suggestive evidence that substitution of plant protein with animal protein was associated with unfavorable insulinemic biomarker profiles. Our study suggests that different protein sources act differently on circulating insulin and IGFs and reduction of animal protein sources and replacement with plant protein may be recommended to improve insulin resistance. However, it is also important to note that the observed findings may not all directly related to protein per se but also other factors that come along with either animal or plant sources of protein. Given the complex relationships of diets, insulin, IGFs and diseases development, more studies are needed to better understand the influence of various protein sources on insulin/IGF system and how it is related to subsequent diseases. Utilizing metabolomics data can also provide further insights to understand the biological mechanisms between dietary protein, diseases and longevity. Lastly, we provide some evidence that unhealthy lifestyles may enhance the changes in circulating insulin and IGFs by higher animal protein intake. Participants with healthy lifestyles had better insulinemic biomarkers regardless of protein intake. Although our results need confirmation, it highlights the importance of adopting healthy lifestyles to maintain healthier biomarker profile.
Limitations
There are several limitations. First, given the nature of observational design of the study, we cannot draw a causal relationship between protein intake and biomarkers. Second, diet was measured using a self-reported questionnaire and thus measurement error is inevitable. However, our previous validation studies using 7-day dietary records and recovery biomarkers showed that our FFQs provide reasonably valid measurements of nutrients and foods.[18–22] Third, there is potential measurement error in biomarkers as they were measured once. To reduce measurement error, we recalibrated the biomarkers, and measurement error is likely to be random, which may result in more conservative results. Fourth, although our study had reasonably wide ranges of protein intake (5th-95th percentiles, 8–19% of calories for animal and 4–7% of calories for plant), we were not able to examine the association in the extreme ranges of protein intake; thus, it is possible that the effects of animal and plant protein on biomarkers may have been underestimated. Fifth, although we thoroughly adjusted for potential confounders, we cannot completely rule out potential residual confounding by unmeasured factors. Lastly, our cohorts are primarily composed of white health professionals, which may limit the generalizability of findings, but it strengthens internal validity; and the characteristics of our participants were comparable to those of large multiethnic US cohorts.
Conclusions
Animal protein intake was positively associated with IGF-1 and inversely associated with IGFBP-1 and IGFBP-2. On the other hand, plant protein intake was positively associated with IGF-1 and IGFBP-1 and substitution of plant protein with animal protein sources was associated with lower IGFBP-1.
Supplementary Material
Funding:
This work was supported by the National Institutes of Health (UM1 CA186107, U01 CA167552 and P01 CA87969).
Footnotes
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Conflict of interest: The authors have declared no conflicts of interest.
References
- [1].Moughan PJ. Dietary protein for human health. British Journal of Nutrition. 2012;108:S1–S2. [DOI] [PubMed] [Google Scholar]
- [2].Phillips SM. Current concepts and unresolved questions in dietary protein requirements and supplements in adults. Frontiers in Nutrition. 2017;4:13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [3].Santesso Na, Akl E, Bianchi M, Mente A, Mustafa R, Heels-Ansdell D, et al. Effects of higher-versus lower-protein diets on health outcomes: a systematic review and meta-analysis. European journal of clinical nutrition. 2012;66:780–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Bernstein AM, Sun Q, Hu FB, Stampfer MJ, Manson JE, Willett WC. Major dietary protein sources and risk of coronary heart disease in women. Circulation. 2010;122:876–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Malik VS, Li Y, Tobias DK, Pan A, Hu FB. Dietary Protein Intake and Risk of Type 2 Diabetes in US Men and Women. American journal of epidemiology. 2016;183:715–28. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [6].Naghshi S, Sadeghi O, Willett WC, Esmaillzadeh A. Dietary intake of total, animal, and plant proteins and risk of all cause, cardiovascular, and cancer mortality: systematic review and dose-response meta-analysis of prospective cohort studies. Bmj. 2020;370:m2412. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [7].Preis SR, Stampfer MJ, Spiegelman D, Willett WC, Rimm EB. Dietary protein and risk of ischemic heart disease in middle-aged men. The American journal of clinical nutrition. 2010;92:1265–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [8].Song M, Fung TT, Hu FB, Willett WC, Longo VD, Chan AT, et al. Association of Animal and Plant Protein Intake With All-Cause and Cause-Specific Mortality. JAMA Intern Med. 2016;176:1453–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Adeva-Andany MM, González-Lucán M, Fernández-Fernández C, Carneiro-Freire N, Seco-Filgueira M, Pedre-Piñeiro AM. Effect of diet composition on insulin sensitivity in humans. Clinical nutrition ESPEN. 2019;33:29–38. [DOI] [PubMed] [Google Scholar]
- [10].Rietman A, Schwarz J, Tomé D, Kok FJ, Mensink M. High dietary protein intake, reducing or eliciting insulin resistance? European journal of clinical nutrition. 2014;68:973–9. [DOI] [PubMed] [Google Scholar]
- [11].Larsson SC, Wolk K, Brismar K, Wolk A. Association of diet with serum insulin-like growth factor I in middle-aged and elderly men. The American journal of clinical nutrition. 2005;81:1163–7. [DOI] [PubMed] [Google Scholar]
- [12].Holmes MD, Pollak MN, Willett WC, Hankinson SE. Dietary correlates of plasma insulin-like growth factor I and insulin-like growth factor binding protein 3 concentrations. Cancer Epidemiology and Prevention Biomarkers. 2002;11:852–61. [PubMed] [Google Scholar]
- [13].Allen NE, Appleby PN, Davey GK, Kaaks R, Rinaldi S, Key TJ. The associations of diet with serum insulin-like growth factor I and its main binding proteins in 292 women meat-eaters, vegetarians, and vegans. Cancer Epidemiology and Prevention Biomarkers. 2002;11:1441–8. [PubMed] [Google Scholar]
- [14].Tabung FK, Wang W, Fung TT, Hu FB, Smith-Warner SA, Chavarro JE, et al. Development and validation of empirical indices to assess the insulinaemic potential of diet and lifestyle. Br J Nutr. 2016;116:1787–98. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Wu K, Feskanich D, Fuchs CS, Chan AT, Willett WC, Hollis BW, et al. Interactions between plasma levels of 25-hydroxyvitamin D, insulin-like growth factor (IGF)-1 and C-peptide with risk of colorectal cancer. PLoS One. 2011;6:e28520. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [16].Yu Z, Malik VS, Keum N, Hu FB, Giovannucci EL, Stampfer MJ, et al. Associations between nut consumption and inflammatory biomarkers. The American journal of clinical nutrition. 2016;104:722–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [17].Agricultural Research Service, US Department of Agriculture. Welcome to the USDA National Nutrient Database for Standard Reference. https://ndb.nal.usda.gov/.
- [18].Willett WC, Sampson L, Stampfer MJ, Rosner B, Bain C, Witschi J, et al. Reproducibility and validity of a semiquantitative food frequency questionnaire. American journal of epidemiology. 1985;122:51–65. [DOI] [PubMed] [Google Scholar]
- [19].Salvini S, Hunter DJ, Sampson L, Stampfer MJ, Colditz GA, Rosner B, et al. Food-based validation of a dietary questionnaire: the effects of week-to-week variation in food consumption. International journal of epidemiology. 1989;18:858–67. [DOI] [PubMed] [Google Scholar]
- [20].Feskanich D, Rimm EB, Giovannucci EL, Colditz GA, Stampfer MJ, Litin LB, et al. Reproducibility and validity of food intake measurements from a semiquantitative food frequency questionnaire. Journal of the American Dietetic Association. 1993;93:790–6. [DOI] [PubMed] [Google Scholar]
- [21].Yuan C, Spiegelman D, Rimm EB, Rosner BA, Stampfer MJ, Barnett JB, et al. Relative Validity of Nutrient Intakes Assessed by Questionnaire, 24-Hour Recalls, and Diet Records as Compared With Urinary Recovery and Plasma Concentration Biomarkers: Findings for Women. American journal of epidemiology. 2018;187:1051–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [22].Yuan C, Spiegelman D, Rimm EB, Rosner BA, Stampfer MJ, Barnett JB, et al. Validity of a Dietary Questionnaire Assessed by Comparison With Multiple Weighed Dietary Records or 24 Hour Recalls. American journal of epidemiology. 2017;185:570–84. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [23].Lee DH, de Rezende LFM, Eluf-Neto J, Wu K, Tabung FK, Giovannucci EL. Association of type and intensity of physical activity with plasma biomarkers of inflammation and insulin response. Int J Cancer. 2019;145:360–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [24].Tabung FK, Fung TT, Chavarro JE, Smith-Warner SA, Willett WC, Giovannucci EL. Associations between adherence to the World Cancer Research Fund/American Institute for Cancer Research cancer prevention recommendations and biomarkers of inflammation, hormonal, and insulin response. International journal of cancer. 2017;140:764–76. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Rosner B, Cook N, Portman R, Daniels S, Falkner B. Determination of blood pressure percentiles in normal-weight children: some methodological issues. American journal of epidemiology. 2008;167:653–66. [DOI] [PubMed] [Google Scholar]
- [26].Willett W Nutritional epidemiology: Oxford University Press; 2012. [Google Scholar]
- [27].Song M, Giovannucci E. Substitution analysis in nutritional epidemiology: proceed with caution. Eur J Epidemiol. 2018;33:137–40. [DOI] [PubMed] [Google Scholar]
- [28].Kim MN, Lo C-H, Corey KE, Luo X, Long L, Zhang X, et al. Red meat consumption, obesity, and the risk of nonalcoholic fatty liver disease among women: Evidence from mediation analysis. Clinical Nutrition. 2022;41:356–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [29].Tobias DK, Hu FB, Chavarro J, Rosner B, Mozaffarian D, Zhang C. Healthful dietary patterns and type 2 diabetes mellitus risk among women with a history of gestational diabetes mellitus. Archives of internal medicine. 2012;172:1566–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [30].Flandre P, Saidi Y. Estimating the proportion of treatment effect explained by a surrogate marker. Stat Med. 1999;18:107–9. [DOI] [PubMed] [Google Scholar]
- [31].Benjamin DJ, Berger JO, Johannesson M, Nosek BA, Wagenmakers E-J, Berk R, et al. Redefine statistical significance. Nature human behaviour. 2018;2:6–10. [DOI] [PubMed] [Google Scholar]
- [32].Tabung FK, Wang W, Fung TT, Smith-Warner SA, Keum N, Wu K, et al. Association of dietary insulinemic potential and colorectal cancer risk in men and women. The American journal of clinical nutrition. 2018;108:363–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [33].Lee DH, Fung TT, Tabung FK, Colditz GA, Ghobrial IM, Rosner BA, et al. Dietary pattern and risk of multiple myeloma in two large prospective US cohort studies. JNCI Cancer Spectr. 2019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [34].Lee DH, Li J, Li Y, Liu G, Wu K, Bhupathiraju S, et al. Dietary Inflammatory and Insulinemic Potential and Risk of Type 2 Diabetes: Results From Three Prospective US Cohort Studies. Diabetes care. 2020;43:2675–83. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [35].Baxter RC. IGF binding proteins in cancer: mechanistic and clinical insights. Nature Reviews Cancer. 2014;14:329–41. [DOI] [PubMed] [Google Scholar]
- [36].Giovannucci E. Insulin, insulin-like growth factors and colon cancer: a review of the evidence. J Nutr. 2001;131:3109S–20S. [DOI] [PubMed] [Google Scholar]
- [37].Giovannucci E. Nutrition, insulin, insulin-like growth factors and cancer. Hormone and metabolic research= Hormon-und Stoffwechselforschung= Hormones et metabolisme. 2002;35:694–704. [DOI] [PubMed] [Google Scholar]
- [38].Fung TT, McCullough ML, Newby P, Manson JE, Meigs JB, Rifai N, et al. Diet-quality scores and plasma concentrations of markers of inflammation and endothelial dysfunction. Am J Clin Nutr. 2005;82:163–73. [DOI] [PubMed] [Google Scholar]
- [39].Hang D, Kværner AS, Ma W, Hu Y, Tabung FK, Nan H, et al. Coffee consumption and plasma biomarkers of metabolic and inflammatory pathways in US health professionals. The American journal of clinical nutrition. 2019;109:635–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [40].Yu Z, Ley SH, Sun Q, Hu FB, Malik VS. Cross-sectional association between sugar sweetened beverage intake and cardiometabolic biomarkers in US women. British Journal of Nutrition. 2018;119:570–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [41].Rajpathak SN, He M, Sun Q, Kaplan RC, Muzumdar R, Rohan TE, et al. Insulin-like growth factor axis and risk of type 2 diabetes in women. Diabetes. 2012;61:2248–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [42].Drogan D, Schulze MB, Boeing H, Pischon T. Insulin-like growth factor 1 and insulin-like growth factor–binding protein 3 in relation to the risk of type 2 diabetes mellitus: results from the EPIC–potsdam study. American journal of epidemiology. 2016;183:553–60. [DOI] [PubMed] [Google Scholar]
- [43].Lewitt MS, Hilding A, Brismar K, Efendic S, Östenson C-G, Hall K. IGF-binding protein 1 and abdominal obesity in the development of type 2 diabetes in women. European journal of endocrinology. 2010;163:233–42. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [44].Petersson U, Ostgren CJ, Brudin L, Brismar K, Nilsson PM. Low levels of insulin-like growth-factor-binding protein-1 (IGFBP-1) are prospectively associated with the incidence of type 2 diabetes and impaired glucose tolerance (IGT): the Söderåkra Cardiovascular Risk Factor Study. Diabetes & metabolism. 2009;35:198–205. [DOI] [PubMed] [Google Scholar]
- [45].Sandhu MS, Heald AH, Gibson JM, Cruickshank JK, Dunger DB, Wareham NJ. Circulating concentrations of insulin-like growth factor-I and development of glucose intolerance: a prospective observational study. Lancet. 2002;359:1740–5. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.