Introduction

Chronic pancreatitis (CP) is a disease characterized by chronic inflammation and irreversible fibrosis of the pancreatic parenchyma. The lifetime prevalence of disease-related complications (e.g., diabetes mellitus, exocrine pancreatic insufficiency, osteopathy) with metabolic and nutritional consequences is extremely high in CP [1,2,3]. Moreover, symptoms, such as abdominal pain and nausea, may further interfere with dietary intake contributing to additional nutritional compromise and potential loss of endocrine and exocrine function requiring medical management and aggressive nutritional interventions [1,2,3,4]. The chronic inflammation and dysfunctional glycemic control present in CP leads to the development and severity of malnutrition often detected in this patient population, contributing to more frequent hospitalizations, increased length of stay, and reduced survival [1,2,3]. With a variety of medical complications such as abdominal pain, diabetes mellitus, and altered bone metabolism, helping patients manage their CP is a significant challenge facing medical and dietetic professionals [1, 2, 5]. Identification of interventions that can be used to reduce inflammation and slow disease progression is needed.

Optimizing dietary intake is believed to be essential to the successful management of CP. Current standard of care includes pancreatic enzyme replacement therapy (PERT), reduction of dietary fat, and elimination of alcohol [1,2,3, 6]. However, these therapies have limited success, and there is a need to enhance medical treatment to reduce the recurring exacerbations in CP that increase severity of malnutrition and the associated sequelae of the disease [1, 2].

Many studies report that patients with CP have suboptimal dietary intake and poor diet quality [3, 6]. Because of the inflammatory nature of CP [7], understanding characteristics of the diet that may reduce inflammation could provide insight into dietary interventions to reverse or delay CP progression. (Poly)phenols are hypothesized to have anti-inflammatory properties and are abundantly found in plant-derived foods and beverages [8]. Therefore, estimation of dietary (poly)phenol intake in those with CP may provide an opportunity to further understand and modulate food sources of (poly)phenols in the diet with the ultimate goal of reducing inflammation and overall disease progression. However, estimating (poly)phenol intake is difficult, as (poly)phenol content of foods and beverages is not routinely included in food composition databases [9]. Therefore, detailed and improved methods are needed to help clinicians and researchers better assess dietary intake of these bioactive phytochemicals. Replicable methods will allow for comparable data between studies and ultimately aid in elucidating the impact of (poly)phenol-rich foods on chronic inflammatory conditions such as CP.

Food frequency questionnaires (FFQ) are dietary assessment tools that measure habitual intake by querying participants on their typical portion size and frequency of intake of a finite list of foods and beverages over a specific period (e.g., the previous 90 days). FFQs are commonly used in human nutrition studies because they are cost-effective, typically self-administered, and their output can be utilized to quantify usual intake of dietary patterns, foods, nutrients, and bioactives. However, FFQs have been subjected to scrutiny due to the reliance on self-reported estimates of portion sizes. To improve the accuracy of participant responses, FFQs with added food and portion size graphics, graphical FFQs, have been designed to provide a more accurate presentation of portion sizes [10]. Further, computerized versions of graphical FFQs allow for complex skip algorithms to reduce participant burden and remove the risk of missed questions and multiple marks present on paper forms [10].

FFQs are the best dietary assessment tool to measure habitual intake. However, they often group similar foods with a comparable micronutrient and macronutrient profile to minimize participant burden. For example, an FFQ may query participants on frequency of intake of “berries” rather than several questions on individual types of berries. However, when translating FFQ data into intakes of specific dietary components such as (poly)phenols, deconstruction of food group categories into individual foods would improve accuracy of the dietary assessment. Further, mixed dishes with multiple ingredients must be deconstructed into individual ingredients to calculate total (poly)phenol content for the item. Therefore, quantitating (poly)phenol intake from a FFQ requires multiple steps and decisions that must be reported in detail for replication [11].

In the present study, our primary goal was to develop and report in detail methods for derivation of (poly)phenol intake using data collected from a computerized web based graphical FFQ. With this methodology, estimations of habitual (poly)phenol intake and specific (poly)phenol subclasses could be quantified and compared between CP participants and controls. We then determined whether (poly)phenol intake was associated with existing dietary indices that could be targets for improving diet quality in future research. These indices include the Healthy Eating Index (HEI), alternative Mediterranean dietary score (aMED), Empirical Dietary Inflammatory Pattern (EDIP), and Empirical Dietary Index for Hyperinsulinemia (EDIH).

Experimental Methods

Study Design

The current study is a cross-sectional analysis of previously collected data from cases with CP and controls without CP at a large midwestern academic medical center. Data collection included electronic medical record (EMR) review (pancreatic disease characteristics, laboratory parameters, medication use), standard anthropometric measurements, a computerized web-based graphical FFQ with a recall period of 90 days (VioScreen™, Viocare Inc., Princeton, NJ), and a standardized health questionnaire [6]. To assess the feasibility of the proposed methodology for estimating dietary (poly)phenol exposure, a retrospective analysis of FFQ data was completed. This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects/patients were approved by the Institutional Review Board at the Ohio State University Wexner Medical Center (2020H0287). Written, informed consent was obtained from all subjects/patients. Recruitment and eligibility criteria for the primary analysis have been previously published and include those diagnoses with CP without concurrent gastrointestinal illnesses and were ≥ 18 years of age [6]. The diagnosis of CP was confirmed by a gastroenterologist, using published diagnostic guidelines from the American Pancreatic Association [12].

Derivation of an Enhanced Method to Estimate (Poly)Phenol Intake from FFQ Data (Fig. 1)

Fig. 1
figure 1

Methodology implemented to convert food frequency questionnaire data into food and beverage conversions compatible with the Phenol-Explorer database

Full size image

To quantitate (poly)phenol intake from FFQ output, a database containing (poly)phenol content in each FFQ item was built in Microsoft Excel. First, a list was compiled of all reported FFQ items which included foods, beverages, mixed dishes (i.e., items with more than 1 ingredient such as “coleslaw” or “pizza”), and all the individual foods included within FFQ groups. Foods of animal origin with presumed negligible (poly)phenol content were removed from the list. Mixed dishes were deconstructed into their single ingredient foods by matching to a US representative recipe in the Food Commodity Intake Database (FCID) that was selected by a registered dietitian nutritionist (RDN), which was accessed electronically (https://fcid.foodrisk.org/recipes/). Then, all items on the list were assigned matches with foods and beverages within the most comprehensive (poly)phenol database, the Phenol-Explorer database (PED, http://phenol-explorer.eu/) version 3.6 [13] while following a previously established series of steps (Fig. 1) [14, 15].

Content values in the PED are estimated using a variety of experimental methods and values to include in the database were chosen based on the appropriate method for the food matrix and/or (poly)phenol subclass following previously published methods [15,16,17]. For our database, (poly)phenol values calculated via chromatography without previous hydrolysis were selected for use. However, there were several exceptions [15,16,17]. For example, chromatography after hydrolysis was used for assessment of lignans in all items, ellagic acid in walnuts, hydroxycinnamic acids in cereal and cereal products, legumes, peanut products, soy and soy products, and olives [15]. Total anthocyanins were determined via pH differential method when available. If pH differential method was not available in the PED, anthocyanin content was calculated as the sum of individual anthocyanins via chromatography without hydrolysis [18]. If this method was unavailable, (poly)phenol content was not calculated. (Poly)phenol values were maintained in their original form for estimation and not translated to aglycone equivalents if presented in conjugate form in the PED, as this is the form they are typically found in nature. Additionally, other research groups commonly report intake values summed without aglycone equivalent translation, which will allow for comparison between cohorts [15, 19,20,21,22,23,24,25,26]. Retention factors were not applied to content values, as it is recommended that retention factors are specific for each food, (poly)phenol, and process, and this is not discernable from FFQs which typically provide data on generic processes (e.g., “cooked”) [27].

The PED content values are reported in mg/100 g fresh weight for foods or mg/100 mL for beverages. The Food Data Central (https://fdc.nal.usda.gov/) database was used to convert (poly)phenol content data in beverages from milliliters to gram weight. To determine (poly)phenol content of each FFQ item, the gram weight was multiplied by the (poly)phenol content in mg/100 g. To calculate total (poly)phenol and subclass (e.g., flavonoid) content values for each FFQ item, individual (poly)phenols were summed.

To determine the (poly)phenol content of an FFQ group of items, the (poly)phenol content of the individual foods comprising the group were calculated from the PED individually. Then, a weighted estimate for each group of foods was calculated using consumption data from the nationally representative 2005–2018 National Health and Nutrition Examination Survey (NHANES) to determine the proportional intakes of each individual food to the total intakes for the FFQ group. Weighted averages were determined using the Individual Foods Files NHANES 2005–2018 for adults over the age of twenty years. The aggregated gram intakes were computed for each unique Food and Nutrient Database for Dietary Studies (FNDDS) food code. The total intake for each individual food item (g) were divided by the intake of the FFQ group (g) to determine the proportional contribution of each food to the group’s (poly)phenol content. (Poly)phenol content values for each food were determined, then weights were applied to estimate the total (poly)phenol content of the FFQ group. For example, in NHANES, strawberries accounted for 58% of the total intakes of berries, and thus the (poly)phenol content of strawberries was used to account for 58% of the total (poly)phenol content for the FFQ group “berries such as strawberries and blueberries.” NHANES data was also utilized to calculate a weighted average for items which had several matches in the PED. However, some items which had detailed varieties in the PED were not differentiated in the NHANES dietary intake data and therefore all varieties from the PED were assumed to have equal contribution to the dietary assessment group. For example, each variety of beer in the PED was assumed to have equal contribution to the FFQ group “beer (all types)” due to the lack of adequate differentiation between beers in the FNDDS (Fig. 1).

In addition to summation of individual (poly)phenol compounds to calculate polyphenol content of FFQ items, a separate analysis was also conducted using PED values measured via the Folin assay. The Folin assay is a spectrophotometric assay used to determine a crude estimate of total phenolic compounds and these data were used to compare with the computed summed values.

All data entered in the database were checked for quality assurance by two research assistants. Participant daily (poly)phenol intake estimations were calculated by summing (poly)phenol content values for all reported FFQ items.

Assessment of Dietary Patterns

Dietary patterns were assessed via two a-priori methods and two empirical hypothesis-oriented approaches. HEI-2015 and the aMED scores were tabulated following previously published methods [6, 28, 29]. The inflammatory and insulinemic potentials of the diet were assessed via EDIP and EDIH scores, respectively [30, 31]. The EDIP score assigns weighted values to daily intakes of different food groups classified as proinflammatory (e.g., processed meat, fish other than dark-meat fish) or anti-inflammatory (e.g., tea, coffee, beer, wine, green-leafy vegetables, pizza), based on plasma CRP, IL6, and TNFα-R concentrations [30]. The EDIH is derived from dietary and biomarker data which weights food groups according to insulinemic potential as defined by fasting plasma C-peptide [31]. In previous studies, both scores were calculated such that higher scores were indicative of more pro-inflammatory or hyperinsulinemic dietary patterns. However, to ease comparisons with HEI-2015 and aMED, the scores were inverted so that higher overall EDIP and EDIH scores represented dietary patterns with more anti-inflammatory or lower insulinemic dietary potential, respectively.

Statistical Analyses

In the present study, median daily total (poly)phenol intake as well as median daily (poly)phenol intake of each of the five (poly)phenol subclasses as identified in the PED were calculated [16, 17]. Intakes of (poly)phenols and (poly)phenol subclasses (expressed as mg per 1000 kcal/day) as well as energy-adjusted dietary index scores (HEI-2015, aMED, EDIH and EDIP) were summarized by medians and interquartile ranges (IQR), and Mann–Whitney U tests were used to compare intakes between CP cases and controls. Foods and food groups contributing to total (poly)phenol intake were summarized as percentages of total (poly)phenol intake in each group. Spearman’s rank correlations were used to determine correlations between (poly)phenol intake and each dietary pattern score among the full cohort. The relationships between (poly)phenol intake and dietary pattern indices were further evaluated using log-linear regression models that adjusted for predefined confounders (diagnosis, age, gender, BMI, and total energy intake). Because of the high contribution of coffee to total (poly)phenol intake, all analyses were conducted both with and without inclusion of coffee. Data were also analyzed to determine the top five food contributors to (poly)phenol intake.

Subgroup analyses were done comparing intake of (poly)phenols and (poly)phenol subclasses between the CP and control groups when stratified by BMI classification (underweight- BMI < 18.5kg/m2, normal weight- BMI ≥ 18.5–24.9 kg/m2, overweight- BMI ≥ 25–29.9 kg/m2, and obese- ≥ 30kg/m2). Analyses were conducted with SAS 9.4 (SAS Institute, Cary, NC).

Results

Study Population

Estimating dietary (poly)phenol intake using the described methodology supports estimates of (poly)phenol intake in other US populations [26]. The study population included 52 cases with CP and 48 controls. As previously described, the mean age was 52 ± 14 years with 67% male [6]. The mean BMI in the CP group was lower than controls, due to the lack of matching for this variable (24 vs. 31 mg/kg2, p < 0.001) [6].

(Poly)Phenol Intake

Estimates of median daily energy-adjusted (poly)phenol intake were significantly lower in the CP cohort compared to controls (Fig. 2). Total (poly)phenol intake remained significantly lower among those with CP when estimates were calculated excluding coffee (Fig. 3). Similarly, the same three (poly)phenol subclasses were found to be suboptimal in those with CP (e.g., flavonoids, stilbenes, and lignans). Further assessment of BMI as a continuous variable and the association with total polyphenol between groups was not significant suggesting BMI is not significantly associated with total polyphenol intake (p = 0.284; supplemental Table 1).

Fig. 2
figure 2

Energy-adjusted total (poly)phenol and (poly)phenol subclass intake between those with (n = 52) and without chronic pancreatitis (n = 48). Diamonds represent means, bars represent medians, boxes represent interquartile ranges that extent to the minimum and maximum, and outliers are represented by circles. *p-value ≤ 0.05; **p-value ≤ 0.01; ***p-value ≤ 0.001; not significant (NS) is defined as p-value > 0.05

Full size image
Fig. 3
figure 3

Energy-adjusted total (poly)phenol excluding coffee consumption and (poly)phenol subclass intake between those with (n = 52) and without chronic pancreatitis (n = 48). Diamonds represent means, bars represent medians, boxes represent interquartile ranges that extent to the minimum and maximum, and outliers are represented by circles. *p-value ≤ 0.05; **p-value ≤ 0.01; ***p-value ≤ 0.001; not significant (NS) is defined as p-value > 0.05

Full size image
Table 1 Percent contribution of the top five foods to daily intake of (poly)phenols and (poly)phenol classes per 1000 kcal
Full size table

In addition to total (poly)phenol intake, intake of three of the five (poly)phenol classes was significantly different between groups. The CP group had significantly lower intake of flavonoids, lignans and stilbenes. The Folin assay showed median total (poly)phenol intake per day was nearly 1.7-fold greater than the median calculated from the combination of other analytical methods among all participants (Supplementary Table 2).

“Coffee (not including lattes or mochas)” was the primary contributor to total (poly)phenol and phenolic acid intake in both the CP and control groups (Table 1). Among both groups, tea was the top contributor to flavonoid intake, broccoli was the major contributor to lignan intake, and red wine contributed most to stilbene intake. The top contributor to other (poly)phenol intake in the CP group was pizza, while the primary contributor to other (poly)phenols in the control group was 100% whole grain breads.

Correlation of (Poly)Phenol and Dietary Intake Patterns

Though the mean EDIH score was more than 2.7 times higher in controls compared to CP patients, the difference was not statistically significant. However, the EDIP score was significantly (> 8 times) higher in controls compared to those with CP (p = 0.052) (Table 2). In the overall sample, total (poly)phenol intake was associated with all dietary pattern scores included in the analysis (Table 3). When controlling for age, gender, BMI, and caloric intake, these relationships remained statistically significant (Table 4). While EDIH scores were not statistically significantly different in both groups, a weak positive association was observed between EDIH scores and total (poly)phenol intake in those with CP, but not controls; whereas EDIP scores were significantly different among both groups and weak positive correlations were observed when stratified by disease state (Supplemental Table 3).

Table 2 Comparisons of dietary pattern scores between those with and without chronic pancreatitis
Full size table
Table 3 Correlation between dietary pattern scores and total (poly)phenol intake with and without coffee for the overall sample of CP cases and controls (N = 100)
Full size table
Table 4 Coefficients (p-values) from log-linear regression models for (poly)phenol intake including and excluding coffee
Full size table

Subgroup Analyses

Participants with CP who were categorized as overweight and obese had lower daily intake of total (poly)phenols (excluding (poly)phenols from coffee beverages) than those in the overweight and obese categories in the control group (Table 5). Those with CP who were normal weight also reported lower daily intake of flavonoids and stilbenes (data not shown). Lower intake of lignans were reported among participants with CP categorized as overweight or obese when compared to controls within these BMI categories, and there were no differences when assessed with and without coffee consumption (data not shown).

Table 5 Estimated total (poly)phenol intake (mg/1000 kcal) among those with and without chronic pancreatitis stratified by body mass index category
Full size table

Discussion

We present an enhanced and feasible method for estimating dietary (poly)phenol intake using a computerized web-based graphical FFQ that can be administered in the clinic and coupled to the PED. Using this methodology, we estimated (poly)phenol intake in CP compared to healthy controls and described the relationship between (poly)phenol intake and common indices used to define dietary quality and patterns of intake and compare these findings between patients with CP and a control sample. Our findings suggest those with CP report lower intake of total (poly)phenols and several subclasses of (poly)phenols when compared to healthy controls, identifying key differences in typical dietary intake that may provide insight into potential targets for those designing (poly)phenol-rich dietary pattern interventions aimed at modulating inflammatory pathways in those with CP [32,33,34]. Additionally, there was a correlation between polyphenol intake and empirical dietary indices that are associated with biomarkers of insulinemic and inflammatory pathways. Collectively, these results identify opportunities for further investigation into dietary interventions that would increase (poly)phenol exposure in CP, which could potentially alter the inflammatory state and disease course. Polyphenol-rich dietary interventions could potentially impact the spectrum of symptoms related to CP without the side effects of pharmacotherapy, though future clinical testing is required. However, we have completed an important step in understanding (poly)phenol intake in this population.

(Poly)phenol intake in a US-based population of patients with CP is lower than controls and thus is associated with an inflammatory dietary pattern. Published data from large epidemiologic cohorts have demonstrated significant variation in (poly)phenol intake, with estimates ranging from 400–1500 mg/day [15, 35]. Some of these differences may be due to the dietary assessment method employed, while others may be directly influenced by cultural differences within each population. Huang et al. estimated (poly)phenol intake at 884 mg per 1000 kcal/day when assessed by 24-h recall in approximately 10,000 U.S. adults in 2013–2016 as part of NHANES [26]. This value is somewhat greater than findings in this study, potentially due to their inclusion of the Folin assay, which likely contributed to overestimation of (poly)phenol intake because the assay is not specific for phenolic compounds [26]. Energy-adjusted total (poly)phenol intake among those with CP may also be lower than Huang et al.’s estimations of the general US population (e.g., 884 mg/day) due to self-restriction of both total energy intake and intake of (poly)phenol-rich foods in order to minimize meal-related gastrointestinal symptoms such as fruits and vegetables which are relatively rich in fiber and (poly)phenols [6, 11, 36, 37]. Energy adjustment of (poly)phenol intake is a superior method of estimating exposure within this population due to the variation in energy intakes observed between those with differing levels of disease severity [38].

Coffee is identified as the major contributor to total (poly)phenol intake in both groups within this study, indicating a potential dietary target to increase (poly)phenol consumption that would be relatively easy to implement in patients who can tolerate coffee intake [24, 35]. In addition, significant differences in intakes of (poly)phenols from the flavonoid, lignan, and stilbene subclasses may relate to the typical food and beverage sources that are uncommon or more difficult to tolerate in CP. Lignans are present in foods such as flaxseeds, legumes, fruits, and vegetables, while stilbenes are predominantly found in wines, berries, and grapes. Fiber content is high in many lignan-rich foods, which may explain the lower intake since dietary fiber has been shown to contribute to an increase in fecal weight and fecal fat and may be self-restricted in CP due to gastrointestinal discomfort [37]. Designing interventions to improve tolerance to these foods may increase intake of these (poly)phenol subclasses while simultaneously improving diet quality. Additionally, though flavonoid intakes appeared to be similar in both groups when coffee beverages were included in dietary estimates, flavonoid intake from other sources was significantly lower in CP, indicating that the high concentration of (poly)phenols in coffee may mask poorer diet quality in relation to dietary flavonoids.

Greater reported intakes of several food groups comprised of (poly)phenol-rich foods, such as whole-grain products, fruits, vegetables, and legumes result in higher HEI and aMED scores [29, 39]. Therefore, the positive correlation between (poly)phenol intake and HEI and aMED scores was expected. Previous research from independent cohorts suggests that those with greater diet quality exhibit higher (poly)phenol intake and higher fiber intake [26, 40, 41]. For the first time, we demonstrate a significant association between the EDIP and EDIH scores with total (poly)phenol intake, indicating that consumption of (poly)phenol-containing foods may directly influence not only diet quality, but measurable biological indicators of dietary inflammatory or insulinemic potential. Together, these results support further investigation of the role of (poly)phenol-rich dietary interventions to modulate inflammation in CP.

Dietary interventions to modulate inflammatory pathways in CP could dramatically alter patient outcomes and lead to a shift in standard of nutrition care within this population. As such, further study of food-based approaches to reduce inflammation are of significant interest to our group and others [22, 42,43,44,45]. To accurately design clinical trials aimed at delivering a (poly)phenol-rich food product, dietary exposure must be quantified and controlled to ensure that interval changes are due to the intervention delivered. Data reported within this CP cohort will therefore inform the development of a low-(poly)phenol control diet tailored to those with CP as constructed in other (poly)phenol-rich food interventions [22, 32, 42,43,44,45]. Perhaps more importantly, the identification of typical (poly)phenol-rich foods sources consumed and tolerated by those with CP will allow researchers to target these foods as well as those that are not regularly consumed to enhance (poly)phenol exposure and improve dietary patterns.

Strengths and Limitations

Calculating (poly)phenol content values for individual FFQ food categories remains a significant challenge, yet the methodological framework employed within this study improves confidence in estimates [20, 46]. The inclusion of mixed dishes, appropriate analytical methods in the PED, and using US dietary data to improve the accuracy of assessments enhance polyphenol assessment from FFQ data. However, limitations of dietary analysis from FFQ may have over- or underestimated serving sizes or excluded foods due to inaccurate recall or individual perception [10, 47]. Validation of this dietary information against diet records would enhance confidence. Furthermore, (poly)phenols were added to the database as presented in the PED and were not converted to their aglycone form, as is published by others in the literature[21, 23, 26, 48]. Therefore, our findings are comparable to many other cohorts that also present (poly)phenol intakes in this fashion [21, 23, 26, 48]. However, the reader should understand that intake values presented for (poly)phenols include the weight of their conjugates for some (poly)phenols. Future work in our group will expand the database to include aglycone equivalents to present intake values in both forms. Lastly, differences in BMI between CP and controls may have influenced analyses, though we accounted for this by adjusting for BMI in the log-linear regression analyses both as a continuous and a categorical variable. Future investigations of larger samples will assist in determining the impact of BMI and other factors such as age, sex, socioeconomic status, and smoking on (poly)phenol estimates and the relationship to dietary pattern indices [49]. Despite these limitations, both groups had intake estimates that are comparable to previously reported populations, increasing confidence in our results [21, 23, 25, 48].

Conclusion

We demonstrated an enhanced method for estimating (poly)phenol intake using an established FFQ. Using this method, we report reduced consumption of total (poly)phenols and several (poly)phenol subclasses in patients with CP. Assessment of the (poly)phenol content and the inflammatory and insulinemic potential of the diet as well as associations between these factors among those with CP may inform the design of future dietary pattern intervention studies in CP. Results will provide evidence that can be used by RDNs to formulate recommendations for specific dietary modifications with the goal of minimizing the inflammatory potential of the diet in CP and other inflammatory conditions [8, 50].