Changes in Dietary Inflammatory Index Score over Time and Cancer Development in Rural Post-Menopausal Women

Inflammation plays a key role in cancer development. As an important modulator of inflammation, the role of diet should be explored. The purpose of this study was to determine the association between diets with a higher inflammatory potential, as measured by the Dietary Inflammatory Index (DII®), and cancer development in a cohort of rural post-menopausal women. Dietary intake from a randomized controlled trial cohort of rural, post-menopausal women in Nebraska was used to compute energy-adjusted DII (E-DIITM) scores at baseline and four years later (visit 9). A linear mixed model analysis and multivariate logistic regression evaluated the association between E-DII scores (baseline, visit 9, change score) and cancer status. Of 1977 eligible participants, those who developed cancer (n = 91, 4.6%) had a significantly larger, pro-inflammatory change in E-DII scores (Non-cancer: Δ 0.19 ± 1.43 vs. Cancer: Δ 0.55 ± 1.43, p = 0.02). After adjustment, odds of cancer development were over 20% higher in those with a larger change (more pro-inflammatory) in E-DII scores than those with smaller E-DII changes (OR = 1.21, 95% CI [1.02, 1.42], p = 0.02). Shifting to a more pro-inflammatory diet pattern over four years was associated with increased odds of cancer development, but not with E-DII at baseline or visit 9 alone.


Introduction
Chronic inflammation is a central feature of cancer, known to promote genetic and epigenetic changes and tumor progression [1][2][3]. Overall diet patterns have shown robust associations with the development of cancer and have been investigated as a modifiable risk factor of inflammation. Mediterranean diet patterns have been extensively studied for their role in inflammation prevention and found to be linked to anti-inflammatory properties [4][5][6][7]. In contrast, Western-type diet patterns have been associated with increased pro-inflammatory markers, including C-reactive protein (CRP), interleukin-6 (IL-6), and fibrinogen [8][9][10]. The Dietary Inflammatory Index (DII ® ) was developed and validated to characterize and quantify cumulative dietary inflammatory potential [11]. As opposed to previous studies that have evaluated certain dietary patterns, the DII itself is not a dietary pattern. Rather, the DII takes into consideration the synergistic effect of multiple individual nutrients and the inflammatory potential of the diet. DII scores have been shown to positively correlate with markers of inflammation such as hs-CRP, as well as cancer outcomes across a variety of populations, including post-menopausal women in the

Materials and Methods
Study population: The present study uses an existing cohort of participants previously enrolled in a completed randomized controlled trial (RCT) evaluating the effect of a fouryear vitamin D and calcium supplementation intervention on cancer development in rural, post-menopausal women in Nebraska from June 2009 to August 2015. Further details of the study design have been previously detailed [22]. In short, independently living post-menopausal women aged 55 years or older from 31 rural Nebraskan counties were screened for inclusion into the RCT. Women aged 55 and older were targeted as a population likely to be post-menopausal and an at-risk group for cancer due to age within the parent study. Post-menopausal was defined as at least four years past the last menses. Women were excluded for (1) any history of cancer diagnosis, unless it was a superficial basal or squamous cell carcinoma of the skin or a previous curatively treated malignancy that was resolved 10 or more years prior to the study; (2) history of chronic kidney disease; or (3) previous participation in a calcium/vitamin D study [22]. After meeting inclusion and exclusion criteria, participants were consented and randomized into either a placebo group or active treatment group of vitamin D (2000 IU) and calcium (1500 mg divided into 500 mg, three times daily). The primary outcome was any cancer assessed at 6-month intervals over four years. Within the RCT, a food frequency questionnaire (FFQ) was completed at the baseline visit and at the end of the four-year trial (visit 9). The RCT study was approved by the Institutional Review Board (IRB) of Creighton University (#624917-9-03; 8/5/2008), obtained informed consent from all participants, and is registered on clinicaltrials.gov (NCT01052051). Utilizing this trial, we conducted a secondary analysis, examining the association between dietary inflammatory potential over time and cancer development. In the present study, participants were included if they completed both FFQs in order to compute baseline and visit 9 DII scores.
Dietary Inflammatory Index: The Block Food Frequency Questionnaire (FFQ) was administered at baseline and at the last visit (Year 4; visit 9). From these FFQs, absolute values of nutrient intakes (including supplements) were calculated by NutritionQuest, utilizing primarily USDA data [23]. The DII has been previously validated with various inflammatory markers and calculated within over 40 populations, including post-menopausal women; complete descriptions of the DII are available elsewhere [11]. In short, the DII includes up to 45 parameters with individual inflammatory effect scores. A total of 29 parameters (all Antioxidants 2023, 12, 946 3 of 12 were nutrients) were available from the FFQs for score calculation ( Table 1). The calculation for the DII is based on dietary data from a world database. This world database provides an estimate of mean and standard deviations (SD) for each inflammatory food parameter. Dietary data collected from study participants is used to calculate z-scores and centered proportions based on the world average intake, to minimize "right skewing" [11]. This proportion is centered on zero by multiplying by 2 and subtracting 1. The resulting value is multiplied by the respective food parameter effect score to create a food parameter-specific DII score. The centered proportion scores are multiplied by the corresponding food parameter effect score, creating a food parameter-specific DII score, which are summed for the overall DII score for each participant. Possible DII scores range from -8.87 to 7.98, with a more pro-inflammatory diet indicated by a higher (more positive) DII score [11]. In this study, energy-adjusted DII (E-DII™) scores also were calculated for each participant using the available assessed nutrients from the FFQs. E-DII scores are calculated per 1000 calories of food consumed, utilizing the energy-standardized version of the world database [24]. A higher, more positive, E-DII score designates a more pro-inflammatory diet whereas smaller, more negative E-DII values indicate more anti-inflammatory diets. In this analysis, the E-DII scores were the primary exposures and were analyzed as a continuous variable. E-DII scores were evaluated at both baseline and visit 9 (last visit) E-DII scores. The change in E-DII scores was calculated (Visit 9 score-Baseline score). Cancer development: The primary outcome measured in this study was first diagnosis of cancer (all types) during the study time frame, creating a dichotomous outcome (cancer yes/no). Participants were asked at each visit for any new cancer diagnoses, which were then verified by medical records, ICD-9 cancer codes, and pathology reports. New cancer status in participants was only counted once (i.e., no second primaries).
Other measures: Weights and heights were measured at each visit during the study and were used to calculate body mass index [BMI = weight (kg)/height (m) 2 ]. BMI was categorized as <25, 25-29.9, and ≥30.0 kg/m 2 for analysis. Other demographic information, such as age and race (White/non-White) were collected at baseline. Hormone replacement therapies (HRT) (yes/no), smoking status (never/ever), and physical activity (</≥ 150 min of moderate activity equivalents per week) were chosen as additional covariates to analyze from the RCT, for their known impact on cancer development. Physical activity data were evaluated as a combined activity-equivalent variable of moderate and vigorous reported physical activity, based off the Physical Activity Guidelines for Americans recommendation of meeting 150 min a week of moderate-intensity activity, 75 min a week of vigorousintensity activity, or an equivalent combination of moderate-and vigorous-intensity aerobic activity [25]. HRT status was assessed from baseline medication reports, grouping together estrogen, estrogen agonist, and estrogen antagonist therapies (yes/no). Use of therapy with estrogen agonists and antagonists (such as tamoxifen or raloxifene) during the study, were primarily for treatment or prevention of osteoporosis and not for prevention of breast cancer [22].
Statistical analysis: Descriptive statistics were reported for all continuous and categorical variables (means, standard deviations, counts and percentages). E-DII scores were used for analysis to control for calorie intake. A linear mixed model (LMM) analysis on E-DII measurements with fixed effects for cancer status and time and a random effect for subjects was conducted and adjusted for multiple comparisons. Univariate and multivariate logistic regression were conducted, controlling for pertinent confounders (age, smoking status, BMI, physical activity, HRT). Race was not accounted for as a covariate due to limited group size of non-White participants (n = 9). Baseline E-DII scores and the E-DII change scores were included in the multivariate logistic regression to investigate the change in E-DII measurements, accounting for the baseline measurements. A power analysis was conducted for the parent study to detect reduction of cancer incidence between the intervention and control groups. They determined recruiting a sample size of 2300 and ending the study with a minimum of 1000 participants per group would be at least 80% powered, accounting for attrition [22]. No additional power analyses were performed for the present secondary analysis; analysis was performed on all available and eligible study participants. All statistical analyses were done using SAS, Version 9.4. Significance level was set at p < 0.05, two tailed.

Characteristics of the Study Population
Out of the total cohort of 2303 post-menopausal women, there were 2221 participants with baseline dietary data adequate to compute E-DII scores. Among them, 1977 participant records were linked to visit 9 E-DII follow-up data (see Figure 1). Therefore, this study included 1977 participants with baseline and visit 9 E-DII scores, including 1886 participants without cancer and 91 (4.6%) who developed cancer. The majority of the population was White (99.5%), with a mean age of 65 years old and mean BMI of 29.9 kg/m 2 ( Table 2). About two-thirds (67.7%) of the population had never smoked, the majority were not taking hormone replacement therapy (HRT) (81.6%), and 41.6% of participants met a physical activity level of ≥ 150 minutes of moderate activity equivalents per week. Of note, those who developed cancer were older than those who did not develop cancer (68.1 ± 7.8 vs. 65.1 ± 6.8, respectively).
Antioxidants 2023, 12, x FOR PEER REVIEW 4 Use of therapy with estrogen agonists and antagonists (such as tamoxifen or raloxi during the study, were primarily for treatment or prevention of osteoporosis and no prevention of breast cancer [22]. Statistical analysis: Descriptive statistics were reported for all continuous and cat ical variables (means, standard deviations, counts and percentages). E-DII scores used for analysis to control for calorie intake. A linear mixed model (LMM) analysis DII measurements with fixed effects for cancer status and time and a random effe subjects was conducted and adjusted for multiple comparisons. Univariate and mul iate logistic regression were conducted, controlling for pertinent confounders (age, s ing status, BMI, physical activity, HRT). Race was not accounted for as a covariate d limited group size of non-White participants (n = 9). Baseline E-DII scores and the E change scores were included in the multivariate logistic regression to investigat change in E-DII measurements, accounting for the baseline measurements. A power ysis was conducted for the parent study to detect reduction of cancer incidence bet the intervention and control groups. They determined recruiting a sample size of 230 ending the study with a minimum of 1000 participants per group would be at least powered, accounting for attrition [22]. No additional power analyses were performe the present secondary analysis; analysis was performed on all available and eligible s participants. All statistical analyses were done using SAS, Version 9.4. Significance was set at p < 0.05, two tailed.

Characteristics of the Study Population
Out of the total cohort of 2303 post-menopausal women, there were 2221 particip with baseline dietary data adequate to compute E-DII scores. Among them, 1977 pa pant records were linked to visit 9 E-DII follow-up data (see Figure 1). Therefore study included 1977 participants with baseline and visit 9 E-DII scores, including participants without cancer and 91 (4.6%) who developed cancer. The majority of the ulation was White (99.5%), with a mean age of 65 years old and mean BMI of 29.9 k ( Table 2). About two-thirds (67.7%) of the population had never smoked, the majority not taking hormone replacement therapy (HRT) (81.6%), and 41.6% of participants m physical activity level of ≥ 150 minutes of moderate activity equivalents per week. Of those who developed cancer were older than those who did not develop cancer (68.1 vs. 65.1 ± 6.8, respectively).

E-DII Scores and Cancer Development
The mean baseline E-DII score for the entire cohort was −1.49 ± 1.74 and the mean visit 9 E-DII score was −1.29 ± 1.72. The difference in mean E-DII scores were tested at baseline and visit 9, as well as the difference in mean E-DII change scores (visit 9−baseline mean), between the cancer and non-cancer groups (Table 3). Mean E-DII scores were not significantly different between cancer groups at baseline (Non-cancer: −1.48 ± 1.74 vs. Cancer: −1.65 ± 1.62, p = 0.78) or visit 9 (Non-cancer: −1.29 ± 1.72 vs. Cancer: −1.11 ± 1.70, p = 0.75). There was a significant difference in the change in E-DII over time between cancer groups, with those who developed cancer having a larger, i.e., proinflammatory change (Figure 2; Non-cancer: ∆ 0.19 ± 1.43 vs Cancer: ∆ 0.55 ± 1.43; p = 0.02).  Logistic regression was used to further explore the association between the ch in E-DII scores and cancer development (Table 4). There was a significant associa between change in E-DII scores and cancer status, after controlling for the sele covariates. Specifically, the odds of cancer development in those with a larger ch (more pro-inflammatory) in E-DII scores was 1.21 times the odds of the group w smaller change (more anti-inflammatory) in E-DII scores (OR = 1. 21   Logistic regression was used to further explore the association between the change in E-DII scores and cancer development (Table 4). There was a significant association between change in E-DII scores and cancer status, after controlling for the selected covariates. Specifically, the odds of cancer development in those with a larger change (more proinflammatory) in E-DII scores was 1.21 times the odds of the group with a smaller change (more anti-inflammatory) in E-DII scores (OR = 1.21, 95% CI = [1.02, 1.42], p= 0.02). Age was significantly related to cancer development, with those in the oldest age group (≥75 years) having the highest increased risk (OR = 2.45, 95% CI = [1.22, 4.92]; p = 0.02; data not shown). There were no significant associations between cancer groups and baseline E-DII scores, after adjustment (OR = 1.04, 95% CI = [0.90, 1.19]; p = 0.64).

Discussion
Identification and treatment of specific needs of those with cancer, including dietary interventions, is a major challenge for health care professionals facing a growing cancer survivor population coinciding with the epidemic of a typical pro-inflammatory Western dietary pattern. Our findings show in a population of rural, post-menopausal women, there was a significantly larger change in E-DII scores in the participants who developed cancer, shifting to higher (more pro-inflammatory) scores after four years. Additionally, those with a higher, more pro-inflammatory change in E-DII scores had higher odds of developing cancer. While we found an increase in inflammatory potential in the diets of cancer patients over time, there was no significant association between E-DII scores and cancer status at either baseline or visit 9 time points.
Changes in DII or E-DII scores over time have not been extensively examined and those studies that have been conducted produced equivocal results. Two studies have been conducted utilizing the Women's Health Initiative (WHI) cohort of post-menopausal women, looking at change patterns in DII scores over three years. The first study evaluated the risk of invasive breast cancer and the other, the risk of colorectal cancer. When examining breast cancer risk, there was no significant association between a DII pattern of pro-inflammatory change and risk of invasive breast cancer, after multivariate adjustment [26]. When examining colorectal cancer risk in post-menopausal women, there was a significantly higher risk of proximal colon cancer in those with a pro-inflammatory change in DII score over three years, but not in overall colorectal cancer risk [27]. A Swedish study by Bodén et al. assessed DII change over ten years and cancer risk in over 35,000 men and women with follow-up data [28]. This study found a 10-year change in DII scores was not associated with cancer risk, even in women who changed from an anti-inflammatory pattern to more pro-inflammatory pattern [28]. While the size of the total cohort was relatively large, looking at individual diet patterns severely limited the number of participants per group, with only 132 women changing from an anti-inflammatory pattern to a pro-inflammatory pattern. This may have limited the ability to observe associations as significant. These studies highlight the uncertainty of the association between change in DII scores and the association with cancer.
While the use of the DII/E-DII to assess diet changes in relation to cancer diagnosis is novel, the idea that people change their diet with a cancer diagnosis is more researched. Previous literature demonstrates a wide range in the prevalence of diet changes, which may differ by population type. Those with evident diet changes after being diagnosed with cancer ranges from 28-60%, depending on cancer type and sex [29][30][31][32][33][34]. Within the WHI cohort, 28% of post-menopausal women made diet changes after a breast cancer diagnosis [34]. Of these women, those who experienced a decrease in diet quality, assessed by the Healthy Eating Index-2010 (HEI-2010), had a significantly higher risk of death from breast cancer, after adjustment [34]. A higher HEI-2010 score would be indicative of higher intakes of fruits, vegetables, seafood and plant-proteins, and whole grains, while also minimizing intake of refined grains, saturated fat, added sugars, and sodium, which aligns with a more anti-inflammatory diet pattern. Indeed, several dietary indices, including the HEI-2010, are inversely associated with DII scores, in that as DII scores become more anti-inflammatory with lower scores, HEI-2010 scores increase, showing healthier diet patterns [35]. While over 70% of WHI cohort did not change their diet after a cancer diagnosis, it does indicate that those whose changes reflect unhealthy or pro-inflammatory eating patterns may also experience poorer health outcomes. Although the present study was unable to account for time of cancer diagnosis within the four-year timeframe, evidence from the WHI cohort suggests it may be plausible that the observed shifts to more proinflammatory diets affected the development of cancer. Conversely, it is still possible that a cancer diagnosis may lead to diet pattern changes related to treatment side effects and disease progression. Therefore, it would be important to establish a timeline of events, with yearly diet assessments, in future studies to distinguish a causal relationship.
In contrast to our study, DII scores with higher inflammatory potential have been associated with several types of cancers, including but not limited to colorectal cancer [36,37], prostate cancer [38], renal cancer [38], ovarian cancer [15,39], and breast cancer [13,38,40]. A recent meta-analysis reviewing 21 studies of DII and breast cancer showed while there was an overall 16% increased risk of breast cancer in those with the most pro-inflammatory DII scores, the six studies that used E-DII scores revealed no association with breast cancer risk [41]. Authors noted that the use of E-DII as the independent variable has been utilized infrequently, which could lead to this finding [41]. As energy intake is strongly associated with DII scores, the E-DII score was created to control for energy intake at the source of the data that is used to create the DII scores [24]. Outside of breast cancer, the E-DII score has been associated with increased risk of other cancers [42] and other disease states [43].
Several mechanisms for how diet impacts cancer development have been explored, including the interdependent relationship between inflammatory markers and oxidative stress [44,45]. Oxidative stress is broadly defined as the imbalance of oxidants and antioxidants, where the level of free radicals, including reactive oxygen species (ROS) is not overcome by the antioxidant system. The inflammatory process encompasses an array of physiologic responses, ultimately to resolve cell injury in the acute setting. However, chronic inflammation has been associated with damage to DNA and tissues, promoting cancer development [3]. Conditions driven by oxidative stress in turn influence systemic inflammation through enhancing pro-inflammatory gene expression, including the production of the pro-inflammatory cytokine TNF-α [44,45]. Conversely, conditions instigated by an inflammatory process also contributes to oxidative stress through the creation of ROS and antioxidant depletion [45]. This interconnected relationship between inflammation and oxidative stress is important to understand in the context of the role of diet.
For example, diets high in fiber, a DII component, have been associated with cancer prevention and reduced cancer mortality [46][47][48]. Through colonic microbial fermentation of fiber, short-chain fatty acids (SCFA) are produced [49]. SCFA in turn can influence systemic inflammation through activation of G-protein receptors and promotion of histone acetylation [44], as well as impact oxidative stress through the regulation of oxidoreductase and restoration of the antioxidant, glutathione [50,51]. Through these mechanisms, SCFA have been associated with inhibitory effects on colon cancer [52]. The DII components also include several known antioxidants including vitamins A, C, E, beta-carotene, zinc, and several polyphenols. High-antioxidant dietary patterns, like plant-based diets, are often sought as a cancer prevention intervention to impact inflammation and oxidative stress. The Meat and Three Greens (M3G) Feasibility Trial examined the impact of daily leafy green consumption in those at risk of colorectal cancer on oxidative DNA damage and inflammation. After 4 weeks of consuming leafy green vegetables, both markers of oxidative DNA damage (plasma and fecal 8OHdG) and inflammation (TNF-α) significantly decreased, compared to controls [53]. A recent RCT by Gualtieri et al., examined the impact of supplementing healthy subjects with an antioxidant-rich, food-derived juice drink while following a Mediterranean diet, compared to the control of just the Mediterranean diet over two weeks. Those with the addition of the antioxidant-rich juice positively improved the oxidative and inflammatory gene expression, including Superoxide dismutase (SOD1) and Peroxisome Proliferator-Activated Receptor γ (PPARγ) [54]. These studies illustrate the biological plausibility of using realistic dietary interventions to impact inflammation and oxidative stress mechanisms in the short term. Further investigation should be conducted to determine the ultimate impact on health status and disease prevention over time. While individual dietary components may contribute to cancer prevention, the DII integrates the synergistic effect of multiple nutrients and foods together that have been associated with inflammatory markers. The DII has additional preliminary associations with markers of oxidative stress including serum malondialdehyde and total antioxidant capacity [55]. However, a greater understanding of the mechanisms associated with diet changes indicating high inflammatory potential can impact cancer development is needed.
Our study did have a sufficiently large cohort and the original RCT was well powered to see cancer incidence but was comparatively smaller than other post-menopausal women cohorts examining DII and cancer. For example, a study utilizing the Iowa Women's Health Study included 34,700 women and 2910 breast cancer cases [56]. Our present study was only able to examine 91 cancer cases among 1977 women, which may have affected our ability to see an association. A rural population may also introduce additional factors that were not accounted for in this study, including education status and other social determinants of health, which have been associated with higher DII scores further limiting our ability to see an association [13,27]. Further research should be conducted to compare changes in DII scores in rural versus urban populations, as they may face distinct challenges in improving health and preventing chronic disease. Additionally, other studies focusing on DII in post-menopausal women have shown higher mean inflammatory potential, compared to our mean E-DII scores [13,27,56]. This suggests our study population had more antiinflammatory diets overall, which could have prevented us from seeing an association between cancer development and E-DII scores.
This study is strengthened by having baseline and follow up data but limited by its small number of cancer diagnoses and diversity, with almost all participants being White, affecting generalizability to other populations. Timepoints of when cancer was diagnosed were not available for this analysis. This limits our understanding of when cancer was diagnosed in relation to the shift in dietary changes and should be assessed in future studies. Recall bias could also affect the reports in the FFQs, which are known to be associated with response set biases. Among these, social desirability seems to be more strongly expressed in women than in men, skewing nutrient intakes from which E-DII scores were calculated [57][58][59]. Our study may be a biased representation of cancer cases compared to the original study, as our study does not include those who developed cancer but did not have follow-up data.

Conclusions
Analyzing the change in E-DII scores in association with cancer risk has seldom been evaluated. Our pilot data shows that a significant change in an individual's diet toward a more pro-inflammatory pattern increases the odds of a diagnosis of cancer. However, how pro-inflammatory shifts in E-DII scores affects health outcomes during cancer survivorship is relatively unknown. Therefore, this research can serve as a foundation for interventional trials in cancer survivorship to determine if implementing anti-inflammatory diet patterns can improve health outcomes and contribute to new body of evidence for health care professionals, transforming diet and lifestyle education for cancer survivors. Funding: The original study was funded by the National Cancer Institute (NCI-R01CA129488-01A2) and Creighton University internal funding. The present study has no funding to disclose.

Institutional Review Board Statement:
The original study was conducted in accordance with the Declaration of Helsinki, approved by the Institutional Review Board of Creighton University (#624917-9-03; 8/5/2008), and is registered on clinicaltrials.gov (NCT01052051).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the original study.
Data Availability Statement: Not applicable.

Acknowledgments: Figures created with BioRender.com.
Conflicts of Interest: J.L. reports having joined the scientific panel for GrassrootsHealth after the original RCT study was underway. GrassrootsHealth is a nonprofit 501(c)(3) organization based in San Diego, California, with public health promotion as its primary endeavor. The focus is on vitamin D testing and education. J.L. has advised GrassrootsHealth on research design and data interpretation, but she is not employed by them, and has never received any funding from them. The Dietary Inflammatory Index (DII ® ) is a registered trademark of the University of South Carolina. J.R.H. owns controlling interest in Connecting Health Innovations LLC (CHI), a company planning to license the right to his invention of the DII from the University of South Carolina in order to develop computer and smart phone applications for patient counseling and dietary intervention in clinical settings. N.S. was an employee of CHI. These activities have no direct bearing on the use of the DII as a research tool. C.H. is funded by a Veteran's Affairs Merit Award. L.G-A., D.T.G., J.M., E.R.L. and M.K.J. have no conflict or disclosures to report. The funders of the original study had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.