Associations between healthy eating patterns and immune function or inflammation in overweight or obese postmenopausal women

¹ From the Fred Hutchinson Cancer Research Center, Cancer Prevention Program, Seattle, WA (AB, MLN, BS, KLAC, AMCT, and CMU); the Departments of Epidemiology (AB, EAK, AMCT, and CMU), Laboratory Medicine (MHW, BW, and ZCL), and Pathobiology (EAK), University of Washington, Seattle, WA; and the Department of Public Health Sciences, University of Alberta, Edmonton, AB (YY)

² Supported by grants from the National Institutes of Health (CA 69334, DK 02860, DK 035816).

³ Address reprint requests to CM Ulrich, Fred Hutchinson Cancer Research Center, Cancer Prevention Program, 1100 Fairview Avenue N, M4-B402, PO Box 19024, Seattle, WA 98109-1024. E-mail: nulrich{at}fhcrc.org.

ABSTRACT  
Background: The link between poor nutritional status and impaired immune function is well established; however, most studies have focused on individual nutrients instead of overall dietary patterns.

Objective: Our objective was to investigate associations between 3 indexes of overall diet quality [the Diet Quality Index (DQI), the DQI including supplementary calcium (DQI-Ca), and the Healthy Eating Index (HEI)] and biomarkers of inflammation and immunity.

Design: This cross-sectional study included 110 overweight or obese postmenopausal women. Dietary intake measured by food-frequency questionnaire was used to calculate diet quality scores. C-reactive protein (CRP) and serum amyloid A (SAA) were measured by latex-enhanced nephelometry. Flow cytometry was used to measure natural killer (NK) cell cytotoxicity and to enumerate and phenotype lymphocyte subsets. T lymphocyte proliferation was assessed by ³H-thymidine incorporation as well as by the carboxyfluorescein–succinimidyl ester method of cell division tracking. Multivariable-adjusted linear regression analysis was used to investigate associations between diet quality scores and markers of inflammation and immune function.

Results: Higher diet quality was associated with increased proportions of cytotoxic and decreased proportions of helper T lymphocytes. CRP and SAA concentrations were higher among women with a lower-quality diet; these associations became nonsignificant after adjustment for body mass index or percentage body fat. We observed limited evidence for an association between healthy eating patterns and greater lymphocyte proliferation and no evidence for an association with NK cell cytotoxicity.

Conclusion: Our results provide limited evidence that healthy eating patterns contribute to enhanced immune function and reduced inflammation in overweight and obese postmenopausal women.

Key Words: Diet quality • women • immune function • inflammation • lymphocyte proliferation

INTRODUCTION  
The link between poor nutritional status and impaired immune function is well established (1-8). However, the standard approach taken when studying the relation between nutrition and immune function has been to focus on the effects of a particular macro- or micronutrient (7, 9-12). Although this strategy is important in evaluating specific biologic mechanisms, humans consume a wide variety of foods, not isolated nutrients. Further, single-nutrient studies do not consider their joint participation with other nutrients in numerous metabolic reactions (13).

Indexes of dietary quality have been developed in recent years to address this shortcoming in nutrition research. Two of these, the Diet Quality Index (DQI) and the Healthy Eating Index (HEI), were developed to measure adherence to dietary guidelines and were shown to adequately measure overall diet quality (14, 15). The DQI is a composite score of an overall healthy diet that reflects a person's adherence to the 8 Diet and Health recommendations of the National Academy of Sciences (16), whereas the HEI measures adherence to the Food Guide Pyramid developed in 1992 (17).

Few studies to date have considered the overall quality of the diet and its relation to immune function in healthy populations; these have been largely limited to markers of inflammation (18-20). For example, Ford et al (18) reported that the HEI score was inversely associated with C-reactive protein (CRP) concentration, which was largely attributed to grain consumption. Fung et al (19), however, found no association between HEI score and various markers of inflammation and endothelial dysfunction. Al-Zahrani et al (20) showed a possible connection between a lower-quality diet and periodontitis, an inflammatory condition of the gums and bones surrounding the teeth. Thus, diet quality may be associated with inflammatory conditions.

A comprehensive assessment of healthy eating patterns in relation to immune markers has not previously been done. Thus, in this study, we determined the DQI (21) and HEI (17) scores of 110 overweight or obese but otherwise healthy postmenopausal women, along with >10 markers of innate and acquired immunity to examine the association between overall diet quality and measures of immune function and inflammation. The immune function markers included lymphocyte subsets and proportions; lymphocyte proliferation, which serves to increase the number of lymphocytes available to fight an infection; and natural killer (NK) cell cytotoxicity (innate immunity), which reflects the ability to lyse cells that have become abnormal or that have been infected by a virus (8, 22). Inflammatory markers included acute-phase proteins that are thought to be linked to cardiovascular disease (23) and cancer (24, 25). We hypothesized that higher-quality diets would be positively associated with measures of improved immune function (ie, greater NK cell cytotoxicity; greater numbers of T lymphocytes, greater T lymphocyte proliferation, or both) and reduced inflammation [ie, lower serum concentrations of CRP, serum amyloid A (SAA), and interleukin-6 (IL-6)].

SUBJECTS AND METHODS  
Study design
This was a cross-sectional investigation of the associations between indexes of healthy eating and immune markers.

Participants
Study participants included 114 women aged 51–75 y who participated in an exercise intervention trial conducted at the Fred Hutchinson Cancer Research Center and the University of Washington (26) and who agreed to participate in an ancillary study of immune function (27). Eligibility criteria were as follows: in sufficiently good health to participate in a moderate-exercise intervention (26); postmenopausal; nonsmoking; alcohol consumption < 2 drinks/d (26 g/d); overweight or obese as defined by body mass index (BMI; in kg/m²) 25–45 (women with BMI 24.0–24.9 were included if body fat > 33%); weight stable for at least 3 mo; sedentary (<60 min/wk of moderate or higher intensity exercise); not taking postmenopausal hormones for at least the past 6 mo; fasting blood glucose 126 mg/dL; no history of diabetes, invasive cancer, cardiovascular disease, lung disease, asthma, or serious allergies; and no regular use of drugs that are known to affect the immune system, including nonsteroidal antiinflammatory medications and corticosteroids (26, 27). We excluded study participants with a reported energy intake < 600 kcal/d or > 4000 kcal/d (n = 4; final sample size: n = 110), because nutrient calculations in this range are not reliable. All study procedures were approved by the institutional review board at the Fred Hutchinson Cancer Research Center. We used baseline data from the exercise intervention trial for the analyses in this report.

Data collection
At the baseline clinic visit, study staff measured body weight to the nearest 0.1 kg with the use of a balance-beam scale (Detecto, Jericho, NY) and height to the nearest 0.1 cm with the use of a stadiometer. The average of duplicate measures was used to compute the BMI. Total body fat was assessed with the use of dual-energy X-ray absorptiometry (Hologic QDR 1500; Hologic Inc, Waltham, MA). Information, including age, education, income, employment status, marital status, race or ethnicity, and smoking history, was collected by self-administered questionnaire (27).

Dietary intake measures
Dietary intake was assessed with a validated food-frequency questionnaire (FFQ) (28) administered at baseline.

Diet Quality Index
The DQI is a composite score of overall healthy diet that reflects a person's adherence to the 8 Diet and Health recommendations of the National Academy of Sciences (reduce total fat intake to 30% of energy; reduce saturated fat intake to <10% of energy; reduce cholesterol intake to 300 mg/d; eat 5 servings of fruit and vegetables per day; eat 6 servings of breads, cereals, and legumes per day; maintain moderate protein intake; limit sodium to 2400 mg/d; and maintain adequate calcium intake) (16). Because of the analytic complexity of measuring servings of grains from an FFQ, these analyses used grams of fiber per 1000 kcal as a proxy measure for servings of breads, cereals, and legumes (15, 29). The mean estimates of nutrient intake from the FFQs were used to score participants for each of the 8 recommendations as previously published (21). Scores were given as follows: 0 if participants met a given recommendation, 1 if consumption was within 30% of a recommendation, and 2 if consumption differed by >30%. Scores were summed to give a total DQI score (range: 0–16), with a lower score indicating a healthy diet (15).

DQI with calcium from supplements
Because the DQI includes a measure of dietary calcium and because the majority (60%) of the study population was taking a calcium-containing supplement, we also calculated DQI scores that included calcium intake from supplements (DQI-Ca). Thus, only the calcium component of the total DQI score was altered.

Healthy Eating Index
The HEI measures adherence to the Food Guide Pyramid, developed in 1992 (17). It comprises 10 components (saturated fat, total fat, cholesterol, sodium, grain, fruit, dairy, meat, and vegetable intakes, as well as a measure of dietary variety), each of which contributes 10 points to the maximum possible score of 100. The intakes of saturated fat, total fat, cholesterol, and sodium were scored with 10 points if saturated fat 10% of energy, total fat 30% of energy, cholesterol 300 mg, and sodium 2400 mg. A zero score was given for 15% of energy for saturated fat, 45% of energy for total fat, cholesterol 450 mg, and sodium 4800 mg. Between these 2 cutoffs, scores were scaled proportionately. A similar process was used for food groups (grains, fruit, vegetables, dairy, and meat); consumption of the recommended number of servings for the person's age and sex resulted in a score of 10, with a score of 0 if no servings were consumed. The other component of the HEI is a measure of dietary variety, adapted for use with the FFQ (30), which is based on the total number of unique foods consumed per month. With the exception of milk and fruit juices, no beverages were included in the analyses.

Dietary supplements
Information about nutritional supplements was collected during an in-person interview with each participant, and supplement labels were photocopied. Participants were classified as regular multivitamin users if, for 12 consecutive months, they had taken at least 1 pill/wk of a supplement containing 3 vitamins.

Immune function measures
Fasting blood samples were drawn between 0730 and 0830 at the University of Washington Department of Laboratory Medicine and processed within 1 hour. Participants complied with the following blood drawing criteria: adequate sleep the previous night (6–9 h), no exercise or alcohol in the previous 24 h, no topical corticosteroids or aspirin for the previous 48 h, no symptoms of infection or systemic antihistamines or corticosteroids for 1 wk before the blood drawing, and no immunizations during the previous 3 wk. Serum, plasma, and urine were collected and stored at –70 °C (27; Meyers J, Ulrich CM, et al, unpublished observations, 2007). Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood by Ficoll-Hypaque separation and cryopreserved in liquid nitrogen. All immune assays were conducted at the University of Washington Clinical Immunology Laboratory.

CRP and SAA were measured by latex-enhanced nephelometry with the use of high-sensitivity assays on the Behring Nephelometer II analyzer (Dade-Behring Diagnostics, Deerfield, IL) with a lower detection limit of 0.2 mg/L for CRP and 0.7 mg/L for SAA. Interassay CVs were 5–9% for CRP and 4–8% for SAA.

IL-6 was measured with the use of solid-phase sandwich enzyme-linked immunosorbent assay with the Biosource Human IL-6 Immunoassay kit (Biosource, Camarillo, CA). The interassay CVs were <10% for concentrations > 35pg/mL, and the analytic sensitivity was 8 pg/mL.

Lymphocytes were isolated from whole blood with the use of a whole-blood lysis technique. A 4-color flow cytometer (XL-MCL; Beckman Coulter, Miami, FL) was used to enumerate subsets of lymphocytes in blood samples, as described previously (27). T lymphocyte proliferation studies were performed by 2 methods: tritiated ³H-thymidine incorporation into PBMCs stimulated by phytohemagglutinin (PHA) and by carboxyfluorescein succinimidyl ester (CFSE) tracking of cell division in PBMCs stimulated by anti-CD3. Cells from 2 control subjects were included in every experiment. For the ³H-thymidine incorporation, 200 µL of 5 x 10⁴ PBMCs/well were incubated in microtiter plates with 25 µL PHA of 0.1 and 0.5 µg/mL in 5 replicates each. After incubation for 72 h at 37 °C, cells were pulsed for 24 h with 25 µg ³H-thymidine at 2 µCi/well, harvested, and counted with a ß-counter. For the method of cell division tracking, 170 µL of 10⁷ cells/mL in RPMI was used for each sample. Carboxyfluorescein diacetate (CFDA)–SE (Molecular Probes, Carlsbad, CA), a precursor of CFSE, was added to the cell suspension at a final concentration of 10 µmol/L. Cells were incubated for 10 min, washed twice, and resuspended in 3.0 mL Complete Medium, and 180 µL was then transferred by pipettes into 16 wells of a microtiter plate (100 000 cells/well). Next, 20 µg of 2 ng/mL anti-CD3 antibody (BD Biosciences, San Jose, CA) was added to 8 of the wells to specifically stimulate T lymphocytes. The remaining 8 wells were used as control unstimulated cells. After incubation for 3 d at 37 °C, identical wells were pooled into 5-mL sterile tubes containing 2 mL Complete Medium and incubated for 3 more days. On the sixth day, cells were harvested, and the CFSE–fluorescein isothiocyanate intensity of viable lymphocytes was measured with a flow cytometer (XL-MCL; Beckman Coulter, Miami, FL).

The flow-cytometric assay for measuring NK cell cytotoxicity used by our group was described previously (27). To prepare the target cells, K562 cells (in log phase of growth) were washed twice with phosphate-buffered saline and bovine serum albumin and incubated with 3-3'-dioctadecyloxacarbocyanine perchlorate (DiO; Live/Dead cytotoxicity kit no. L7010; Molecular Probes, Eugene, OR) at a concentration of 2 x 10⁶ cells/mL for 20 min at 37 °C with 5% CO₂. The cells were then washed twice with phosphate-buffered saline and bovine serum albumin, resuspended in RPMI to a concentration of 1 x 10⁶ cells/mL, and filtered through a 35-µm strainer. Mononuclear cells were prepared by Ficoll-Hypaque differential centrifugation of blood effector cells; diluted corresponding to final effector-to-target cell (E:T) ratios of 50:1, 25:1, 12.5:1, and 6.25:1; and incubated with the DiO-labeled K562 cell suspension (target cells) for 4 h at 37 °C with 5% CO₂. After incubation, propidium iodide (0.03 mg/mL final concentration) was added to each tube to identify dead cells. The percentage of dead target cells (ie, dual positive for DiO and propidium iodide) of total DiO-identified target cells was used as the measure of NK cell cytotoxicity. Each assay was performed in duplicate and with appropriate controls. We repeated the NK cell cytotoxicity assay in 13 study participants who underwent additional blood drawings between 1 wk and 9 mo after the initial blood drawing, under identical blood drawing criteria. Intraclass correlation coefficients between the initial and repeat blood drawings were r = 0.84 (E:T: 6.25:1), r = 0.91 (E:T: 12.5:1), r = 0.90 (E:T: 25:1), and r = 0.79 (E:T: 50:1).

Statistical analysis
Pearson's correlation coefficients and multivariable linear regression analyses were used to analyze the associations between dietary quality scores and measures of inflammation and immune function. Immune outcomes were log-transformed where appropriate (CRP, SAA, IL-6). Because some values of IL-6 were zero (n = 16), these were first changed to 0.5 before log transformation; inclusion or exclusion of these participants did not alter the results appreciably. The DQI and DQI-Ca scores were divided into 4 categories of diet quality (excellent = 0–5; good = 6–7; fair = 8–10; poor = 11–16) (15). For the HEI, 3 categories are typically used, as defined by the US Department of Agriculture (good = 81; needs improvement = 51–80; poor 50) (31). However, because few women in this population had scores 81 (n = 7), we divided HEI scores into approximate quartiles (excellent = 71; good = 62–70; fair = 53–61; poor = 52). For all of the diet indexes, poor diet quality was used as the reference group in regression analyses. Analysis was performed with the use of SAS version 9.1.3 (SAS institute, Cary, NC). A 2-sided P value < 0.05 was considered statistically significant.

Exclusions from specific analyses included 3 participants from CRP, SAA, and IL-6 (CRP > 10 mg/L, suggesting an inflammatory condition) (23); 1 participant from cytotoxic T cell analyses (cell counts 3.7 SD below the mean); and 1 participant from lymphocyte proliferation (³H-thymidine incorporation method; assay values 100-fold lower). The number of missing values for various assays ranged from 1 to 4, except for lymphocyte proliferation by the method of cell division tracking in which sufficient cells were available only for 89 participants.

Potential confounding variables considered were age, education, employment, race or ethnicity, ever lost 10 pounds intentionally, regular multivitamin use, statin use, smoking history, season of blood drawing, and alcohol intake. Any covariate that substantially (>20%) altered the ß-coefficient associated with diet quality score was retained in all models. The final multivariable-adjusted models included age, race or ethnicity, and season of blood drawing. Because biomarkers of inflammation are known to be associated with adiposity (23, 32), we also present multivariable-adjusted models for inflammatory biomarkers with inclusion of the ratio of BMI to percentage body fat (continuous adjustment variables) to evaluate whether the associations are independent of adiposity.

Lymphocyte subpopulation variables included number and percentage of total T cells (CD3⁺CD45⁺), helper T cells (CD3⁺CD4⁺CD45⁺), cytotoxic T cells (CD3⁺CD8⁺CD45⁺), ratio of CD4⁺ to CD8⁺ cells, number and percentage of B cells (CD19⁺CD45⁺), and number and percentage of NK cells (CD3^–CD16⁺CD45⁺CD56⁺).

PHA-stimulated lymphocyte proliferation was evaluated either directly as counts per minute (CPM; CPM of stimulated cells – CPM of unstimulated cells) or as a proliferation index (PI; CPM of stimulated cells/CPM of unstimulated cells). For lymphocyte proliferation results by flow cytometry (cell division tracking), a multifactorial proliferation analysis was generated with the MODFIT LT Program (Verity Software House LT, Topsham, ME). Variables included PI (total number of proliferating and parent cells/number of back-calculated original parent cells), parent percentage (percentage of parent cells that did not divide), precursor frequency (fraction of original parent cells that went on to divide 3 times), and upper generation PI (total number of proliferating and parent cells from generations 3 and above/number of back-calculated original parent cells).

NK cytotoxicity was investigated first individually at 2 of the 4 original E:T dilutions (25:1 and 12.5:1) and subsequently as nonindependent-repeated measures. These intermediate dilutions were in the linear range of the NK cell cytotoxicity curve and showed the greatest reproducibility (27). Generalized estimating equation regression accounted for within-person correlation of the 2 measures of NK cell cytotoxicity (33).

RESULTS  
Study participants
Characteristics of the study participants are described in Table 1. The average age of the participants was 60.6 ± 6.7 y. The participants were overweight or obese, with a mean (±SD) BMI of 30.3 ± 3.9, and a mean percentage body fat of 47.1 ± 4.7%, as measured by dual-energy X-ray absorptiometry. The study participants were highly educated, with 86% having at least attended some college, and the majority (88%) of the participants were white. The mean score on the DQI was 8.3 ± 2.8 (whereby a lower score is indicative of a healthier diet). The average HEI score (whereby a higher score indicates a higher-quality diet) was 62.3 ± 11.7; this is slightly lower than what has been observed for the general US population of women aged 51 y (mean HEI score: 66.6) (34).

View this table:
TABLE 1. Characteristics of the study participants (n = 110)¹

 
Inflammatory biomarkers
The DQI and DQI-Ca were significantly correlated with both CRP and SAA but not with IL-6 (Table 2). A higher-quality diet (measured by the DQI) was generally associated with lower concentrations of CRP or SAA (Table 3). Because CRP is elevated among obese persons (23, 32), we explored whether the observed association between diet quality and CRP or SAA was independent of BMI or percentage body fat. The association did not remain statistically significant after adjusting for either of those variables, which suggests that healthy eating patterns are linked to biomarkers of inflammation largely by effects on adiposity (Table 3; adjustment for BMI gave virtually identical results to adjustment for percentage body fat).

View this table:
TABLE 2. Partial correlation coefficients (adjusted for age) between diet indexes and immune outcomes¹

 

View this table:
TABLE 3. Association between Diet Quality Index (DQI) or Healthy Eating Index (HEI) and biomarkers of inflammation¹

 
T cell, B cell, and NK cell numbers and proportions
We did not observe any significant associations between DQI, DQI-Ca, or HEI scores and the absolute number or proportion of all CD3⁺ T cells. However, all 3 indexes were correlated with the proportions of cytotoxic and helper T cells in total lymphocytes (Table 2). In regression analyses, a significant trend was observed toward lower proportions of helper T cells and higher proportions of cytotoxic T cells with increasing diet quality (Table 4). Proportions of helper T cells were 7–10% lower and proportions of cytotoxic T cells were 31–36% higher in persons consuming an excellent diet than were those with a poor diet. A better-quality diet, measured by the HEI, was associated with a trend toward higher absolute numbers of cytotoxic T cells; this trend was also seen for the other indexes, although it did not reach statistical significance in multivariable analyses.

View this table:
TABLE 4. Association between Diet Quality Index (DQI), DQI including supplementary calcium (DQI-Ca), or Healthy Eating Index (HEI) score and T, B, or natural killer (NK) cell proportions

 
We did not observe any associations between DQI or DQI-Ca scores and B cell or NK cell numbers or proportions (Table 2). A better-quality diet (HEI) was correlated with increased counts of B cells (Table 2), but this did not remain statistically significant in the regression analyses (data not shown).

Lymphocyte proliferation
We determined lymphocyte proliferation after 1) stimulation by PHA (measured by tritiated-thymidine uptake) and 2) stimulation by anti-CD3 [measured by cell division tracking (35)]. We observed no associations between any of the 3 diet indexes and lymphocyte proliferation as determined by the tritiated-thymidine assay (data not shown). However, with the use of cell division tracking, higher diet quality (measured by the DQI, whereby a lower score indicates a higher-quality diet) correlated positively with greater lymphocyte proliferation (PI, r = –0.23; percentage of parent cells that did not divide, r = 0.26; Table 2). This relation was marginally statistically significant in the age-adjusted regression analysis with 4 categories of diet quality (P for trend = 0.08–0.09; data not shown), and was statistically significant for the DQI-Ca in age-adjusted models only (parent percentage, P for trend = 0.049; data not shown). Similarly, women with a good compared with a poor diet tended to have a lower parent percentage for the HEI in the age-adjusted model (parent percentage, P for trend = 0.06; data not shown). Because of greater variability in this measure of immune function, which may have affected our ability to observe associations in 4-category regression analyses, we also compared women with a fair or poor diet to women with a good or excellent diet. For the HEI, better-quality diet was associated with statistically significantly higher proliferation (Table 5). In multivariable regression models, these associations did not remain statistically significant. Regression analyses showed associations between DQI or DQI-Ca and indexes of proliferation that were borderline significant (P = 0.06–0.08), although multivariable adjustments reduced the associations (P = 0.11–0.15; data not shown).

View this table:
TABLE 5. Association between Healthy Eating Index score (2 categories) and measures of T lymphocyte proliferation in response to anti-CD3 by flow cytometric cell division tracking¹

 
NK cell cytotoxicity
No associations were observed between the DQI or DQI-Ca and NK cell cytotoxicity at E:T dilutions of 25:1 or 12.5:1 or when analyzed as nonindependent repeated measures (data not shown). For the HEI we observed a significant decrease in NK cell cytotoxicity among women with a fair diet compared with a poor diet (multivariable adjusted generalized estimating equation model, P = 0.02; data not shown). However, no clear trends were observed; thus, we believe that this finding may have occurred by chance.

DISCUSSION  
In this study, we aimed to determine whether diet quality, as measured by the DQI, DQI-Ca, and HEI, was associated with a large set of markers of inflammation and immune function. We observed strong evidence for an association between a healthy diet and proportions of cytotoxic or helper T lymphocytes, little evidence of an association with increased lymphocyte proliferation or decreased inflammatory biomarkers, and no evidence for a relation with NK cell cytotoxicity.

Previous research has shown an inverse association between the HEI and serum CRP concentrations (18), even after adjusting for markers of obesity (BMI). In our study, we did not see an association with the HEI. However, we observed that a higher-quality diet (DQI) was associated with lower concentrations of CRP or SAA. This association was attenuated and not statistically significant after adjusting for adiposity, indicating that the decrease in CRP or SAA seen with a higher-quality diet is most likely mediated by obesity (23, 32). That is, consuming a healthier diet may lead to decreased adiposity, which in turn could lead to decreased amounts of these inflammatory markers.

Interestingly, we observed that the proportion of cytotoxic T lymphocytes increased by 30%, whereas the proportion of helper T lymphocytes decreased by 10% as the quality of the diet increased. This was consistent across all 3 indexes. Cytotoxic T cells are responsible for recognizing and destroying body cells infected with viruses, whereas helper T cells are involved in activating macrophages (Th1 cells) or the humoral response (Th2 cells) (36). It has generally been observed that immune function declines with age (2, 5, 37, 38). Aging is associated with changes in T cell subsets; in general, both T helper and T cytotoxic subsets tend to decrease (22, 39, 40). These changes may be mediated by altered nutritional status (5), because the elderly tend to have a high prevalence of micronutrient deficiencies (4, 41).

Several human and animal studies have reported relations between various nutritional manipulations and T cell subsets (4, 39, 42-46). For example, n–3 fatty acids may affect the activity of interleukin-2, which induces proliferation and differentiation of lymphocytes (45), and vitamin E was suggested to play a role in T cell differentiation or maturation in the thymus (44). None of the diet indexes considered vitamin E intake or the proportions of various fatty acids. Thus, it is unlikely that our results are due to effects of vitamin E or fat composition. Although Mazari and Lesourd (4) found that lower folate status was associated with lower T helper cell counts, our study population was generally adequate in folate status; indeed, unmetabolized folic acid was found in the circulation of 78% of these participants (47). Nevertheless, the clinical relevance of this change in T lymphocyte proportions remains unclear. We observed that CD8⁺ cells were higher among participants with a higher-quality diet, which would indicate that these persons may be able to more easily destroy abnormal or infected endogenous cells. Regardless, it is unclear whether these subset changes would lead to differences in a person's ability to fight infections.

Assays that capture cell function, such as lymphocyte proliferation and NK cell cytotoxicity assays, may provide a more biologically and clinically meaningful measure of immunity (48). We found no associations between healthy eating patterns and NK cell cytotoxicity. However, we observed limited evidence for a relation between dietary quality and lymphocyte proliferation. A higher-quality diet, as measured by the DQI, was significantly correlated with greater lymphocyte proliferation, yet this association did not remain statistically significant in regression analyses adjusted for age, race, and season of blood drawing. This may have been due to the smaller number of women in the 4 categories of diet quality and the relatively greater assay variability. Participants with an excellent or good diet (as measured by the HEI) had greater lymphocyte proliferation than did participants with a fair or poor diet in the age-adjusted analyses, but not in the multivariable model. Thus, our results indicate possible evidence for a relation between diet quality and lymphocyte proliferation that warrants further investigation. We only observed these suggestive results for the method of cell division tracking; the mitogenic activation with anti-CD3 in this assay resembles the physiologic activation of T lymphocytes more closely than does stimulation with PHA (22). Further, the 2 methods measure different aspects of lymphocyte proliferation; one measures a PI by determining the incorporation of labeled thymidine into the DNA of dividing cells, whereas the flow-cytometric assay determines proliferative ability by measuring the dilution of fluorescent dye as the cells progress through generations. No significant correlation was observed between the 2 PIs (r = 0.11 for the PHA PI measured at 0.1 µg/mL compared with the anti-CD3 PI); thus, it is not surprising that we observed divergent results.

Previous studies have shown increases in lymphocyte proliferation with increased intake of folate (49), vitamin B-6 (50), carotenoids (51), polyphenols (52), or better nutritional status (4); all of these dietary components would be expected to be higher in persons with better-quality diets. Interestingly, increased intake of n–3 fatty acids, which would likely be considered a component of a high-quality diet, is thought to decrease lymphocyte proliferation (53-55). This relation may explain why we failed to observe clearer trends between diet quality and lymphocyte proliferation; the indexes that we used did not distinguish between different types of fatty acids, suggesting a promising area for further research.

To our knowledge, this is the first study to investigate associations between overall diet quality and a comprehensive set of immune markers. Strengths of the study design were the strict exclusion and blood drawing criteria, excluding women with underlying infection and inflammation, and eliminating by design most other confounding factors, such as smoking or exercise before the blood drawing. We used several state-of-the-art immune measures, and our study size was relatively large for this type of investigation.

Several limitations must also be acknowledged. The study was cross-sectional, so changes over time could not be assessed. The study population was more homogeneous than the general population; nevertheless, we still observed heterogeneity in diet quality and in the immune outcomes studied, which gave us the variability required to detect associations. Although the results of our study may not extend to the general population, the BMIs of participants in this study reflect 66% of the female US population in this age group (56). Finally, because of the number of outcomes that were tested, it is possible that some of the statistically significant results could have occurred by chance, and findings should be interpreted with caution until they have been replicated in other populations.

The principles of good nutrition and advice to adhere to healthy eating patterns have been consistent messages given to the public for the past 50 years. Still, the US Dietary Guidelines are revised every 5 y to reflect new information on diet, health, and disease prevention. A modified version of the HEI based on the 2005 Dietary Guidelines is currently being developed. The DQI also has a modified version, the DQI-Revised, which is similar to the DQI but has not been as extensively studied as the original DQI. Other eating patterns, such as the Mediterranean diet, have also been recently emphasized as contributing to decreased disease risk. Other researchers have developed alternate diet indexes that measure these trends and have sometimes found them to be better predictors of chronic disease risk or inflammatory conditions (19, 30, 57). We chose to use the DQI and HEI for this investigation because they have been extensively studied and validated. Further, the indexes that we used still encompass established aspects of healthy eating, and inclusion of slight modifications of these indexes would not be likely to significantly alter the interpretation or relevance of our results. Future investigations may observe stronger associations as diet quality indexes are updated.

In conclusion, our results provide limited evidence that healthy eating patterns contribute to improved immune function and reduced inflammation in overweight and obese postmenopausal women. Further research should explore mechanisms for the observed associations with T cell subsets.

ACKNOWLEDGMENTS  
The author's responsibilities were as follows—AMCT, CMU, and YY: obtained funding for the study; CMU, AMCT, YY, and AB: designed the study; KLAC: collected the data; MHW, BW, and ZCL: performed laboratory analyses; AB, CMU, and BS: analyzed the data; AB: wrote the first draft of the manuscript; all authors participated in the revision of subsequent drafts and approved the final version of the manuscript; CMU, MLN, EAK, and AMCT: provided significant guidance and advice. None of the authors had a personal or financial conflict of interest.

REFERENCES