Diet-quality scores and the risk of symptomatic gallstone disease: a prospective cohort study of male US health professionals

Abstract

Introduction

Gallstone disease is a common disease with an average prevalence of 10–25% in Westernized countries.^1,2 About 85% of all gallstones are defined as ‘cholesterol gallstones’ and typically occur as a result of disrupted hepatic cholesterol homeostasis, which leads via cholesterol supersaturation to crystal formation.^3,4 The treatment of choice for gallstones is surgical gallbladder removal (cholecystectomy). Many gallstone carriers do not develop symptoms.⁵ However, serious complications caused by symptomatic gallstone disease (SGD) and cholecystectomy include cholecystitis, cholangitis, pancreatitis and even death. In the USA, SGD and cholecystectomy account for approximately 1.8 million ambulatory care visits and more than 600 000 hospitalizations each year, exceeding 6 billion US dollars in annual medical expenses.^6–8

Dietary guidance might be a promising measure to prevent SGD but prospective data are limited to studies of specific food groups and nutrients; moderate alcohol consumption,^9,10 higher intakes of some plant foods, such as nuts,^11,12 fruits and vegetables,¹³ and certain nutrients, such as fiber,¹⁴ poly-unsaturated and mono-unsaturated fats¹⁵ have been associated with lower SGD risk. On the other hand, higher intakes of saturated fatty acids,¹⁶ trans-fat¹⁷ and heme iron¹⁸ have been associated with higher risk of SGD. As foods are primarily eaten in combination, assessment of the role of overall diet quality may further enhance our understanding of the etiology of SGD. However, no large prospective study has examined the association between diet quality and SGD risk.

In the present study, we used three diet-quality scores [the alternate Mediterranean diet (aMed), the Alternate Healthy Eating Index (AHEI-2010), as well as the Dietary Approaches to Stop Hypertension (DASH) diet scores] to investigate associations between diet quality and the risk of SGD in a large cohort of male US health professionals—the Health Professionals Follow-up Study (HPFS).

Methods

Study population

The HPFS is an ongoing cohort study comprising 51 529 mostly White (95%) US male health-care professionals aged 40–75 years at the time of recruitment (1986). Biennially, participants were contacted via postal questionnaire to update information on lifestyle and medical history. Follow-up rates have exceeded 94% cumulatively and the study was approved by the Institutional Review Board at the Harvard T.H. Chan School of Public Health. After applying exclusion criteria (Figure 1), the final baseline population consisted of 43 635 men.

Figure 1.

Flow chart of participants who entered the study at baseline, including number and reason for exclusion.

Open in new tabDownload slide

Dietary assessment and diet-quality scores

Information on usual dietary intake in the preceding year was obtained at baseline and every 4 years using semi-quantitative food-frequency questionnaires (FFQs) that included 131 food items. Possible responses were scaled from ‘never or less than once per month’ to ‘6+ per day’ based on standard portion sizes. The validity of the FFQ in measuring food and nutrient intake in the HPFS has been well documented by comparisons with dietary records and biochemical indicators and can be considered moderate to good.^19–21

We selected three a priori diet-quality scores, including aMed, AHEI-2010 and DASH, because they emphasize intakes of foods, food groups and/or nutrients that have been associated with better health outcomes and include the following: vegetables (except potatoes; included in all three scores), fruits (all scores), nuts (all scores), legumes (all scores), whole grains (all scores), low-fat dairy products (DASH only), fish (aMed only), omega-3 fatty acids (n-3 FA; AHEI-2010 only), poly-unsaturated fatty acids (PUFA; AHEI-2010 only), the ratio between mono-unsaturated and saturated fats (MUFA/SFA; aMed only) and moderate alcohol consumption [aMed (10–25 g/day) and AHEI-2010 (0.5–2 drinks/day)]. The scores also discourage intakes of foods that have been related to adverse health outcomes, including red and processed meat (all scores), sodium (DASH and AHEI-2010), sugar-sweetened beverages (DASH and AHEI-2010) and trans-fat (AHEI-2010 only). For detailed description of the calculations and compositions of each score, please refer to earlier publications.^22–24

Briefly, aMed was created to capture the diet quality of a traditional Mediterranean diet²⁵ and was modified to fit to the American population.²⁴ For foods and nutrients related to a beneficial health outcome, 1 point is assigned for intakes above median intake and 0 points below/at median intake. For food related to adverse health outcomes, the scoring is reversed. For men, alcohol intake of between 10 and 25 g/d is rewarded 1 point, otherwise 0 points. Thus, the aMed score ranges from 0 to 9 points (minimum to maximum adherence).

The AHEI-2010 was derived using the Healthy Eating Index (HEI).²⁶ In brief, HEI was established to measure adherence to the 1995 USDA Dietary Guidelines for Americans and subsequently revised.^22,27 We modified the HEI-2010 according to current science-based evidence most predictive of chronic disease risk (e.g. sugar-sweetened beverages were added in 2010, whereas refined grains were excluded). The AHEI-2010 comprises seven favourable and four unfavourable food items, each contributing 0–10 points to the overall score, depending on predefined consumption quantities (see Supplementary Table 1, available as Supplementary data at IJE online).^22,27 The total score ranges from 0 to 110 points (minimum to maximum adherence).

The DASH diet score was designed for blood-pressure reduction according to results of the DASH trial.^23,28 Its scoring is based on quintiles (Q) of food intakes: for favourable foods, individuals are assigned 1–5 points from Q1 to Q5, whereas the scoring is reversed for unfavourable foods. Accordingly, the DASH score has a possible range from 8 to 40 points, which reflects minimum to maximum adherence.

Assessment of covariables

Smoking status, body mass index (BMI), leisure-time physical activity, medication use and medical conditions, among many others, were self-reported and evaluated at baseline and biennially afterwards using self-administered questionnaires. Physical activity was expressed as metabolic equivalents [metabolic equivalents from recreational and leisure- time activities (MET-hours/week)] from recreational and leisure-time activities. Reproducibility and validity of anthropometric measures and physical activity have been reported previously.^29,30

Ascertainment of SGD

At baseline and on each follow-up questionnaire (until 2000), participants were asked whether they had undergone a cholecystectomy or had been diagnosed with gallstones. Additionally, the participants were asked to report presence of symptoms (‘Did you have symptoms’) and diagnostic tests (‘How was diagnosis made? X-ray/ultrasonography or other’) leading to the diagnosis of gallstones. After 2000, SGD cases included participants who underwent a cholecystectomy. To verify the self-reports of surgical cholecystectomy and diagnosed but unremoved SGD, a random sample of 441 medical records of participants who reported a cholecystectomy or SGD were reviewed and, of these, the diagnosis was confirmed in all but 5 (99%). Furthermore, medical chart review confirmed all self-reported symptoms.³¹ For our main analysis, we defined SGD cases as participants who underwent cholecystectomy or a self-report of diagnosed gallstones accompanied by symptoms.

Statistical analyses

All statistical analyses were conducted using SAS release 9.4 (SAS Institute Inc, Cary, NC). Person-years of follow-up were calculated for each participant from the return date of the baseline questionnaire until the date of SGD diagnosis, death, loss to follow-up or end of follow-up (31 January 2012), whichever came first. Cox proportional hazards regressions were performed to estimate hazard ratios (HRs) and 95% confidence intervals (CIs) for the association between diet scores (in Qs) and the risk of SGD within 2-year intervals over a period of 26 years. Linear trends across Qs were evaluated by the Wald chi-square test using the median score value in each Q modelled as an ordinal variable. The proportional hazards assumption was validated by including cross-product terms between age (continuous) and each score (Qs) in the multivariable model and performing a Wald test. No violation was observed.

Repeated dietary and lifestyle questionnaires were administered to account for changes in dietary habits and lifestyle. In the main analysis, we used simple-updated diet data that were obtained from the most recent FFQ prior to each follow-up interval. In that analysis, participants without dietary information during follow-up were skipped for the respective time period, whereas missing value indicators were generated for missing covariable information.

To limit the possibility of random error in reported diet data and to better represent long-term dietary intake, we also conducted sensitivity analyses, using cumulative average diet-quality scores and covariable information, calculated as the average from all available measurements up to the beginning of each follow-up cycle.³²

We ran three statistical models: Model 1 was stratified by age and calendar time. Model 2 was additionally adjusted for total energy intake, smoking, physical activity and coffee consumption. In Model 3, we additionally adjusted for BMI, which might act as a potential intermediate (for definition of covariables, see Table 2 footnotes).

Associations between diet-quality scores and SGD risk were also assessed after stratification by selected risk factors and putative intermediates: medical history of hypertension and hypercholesterolemia, regular use of medication [thiazide diuretics, non-steroidal anti-inflammatory drugs (NSAID, including aspirin) and statins], anthropometry (BMI <25/≥25 kg/m²), physical activity (<median/≥median) and coffee consumption (<median/≥median).

In additional analyses, we examined associations between each individual food contributing to each score and risk of SGD separately to determine whether the associations for diet-quality scores were driven by specific foods. Furthermore, numerous sensitivity analyses were conducted to assess the robustness of our findings. We studied associations between diet scores and risk of SGD after (i) separately (one-by-one) adjusting for history of hypertension and hypercholesterolemia (yes/no), regular use of thiazide diuretics, NSAID and statins (yes/no) and (ii) excluding individuals with cardiovascular diseases (CVD) and cancer (except localized prostate cancer) at baseline and during follow-up (CVD: n = 5960 cancer: n = 5757). We also reran the analyses after (i) inclusion of asymptomatic gallstones (n = 428) and after (ii) exclusion of symptomatic gallstone cases who did not undergo cholecystectomy (n = 195).

Results

From 1986 to 2012, 2382 incident SGD cases and 1960 cholecystectomies were reported among 43 635 men. The three diet-quality scores were moderately correlated with each other [Spearman partial correlation coefficient (adjusted for age) for aMed and AHEI-2010: 0.60, for aMed and DASH: 0.67 and for DASH and AHEI-2010: 0.64].

Participants with higher scores were older, less likely to smoke or drink coffee, more physically active and had a lower BMI and a slightly lower waist circumference than those with lower scores (Table 1). Additionally, participants with higher scores were more likely to have hypercholesterolemia and less likely to have a history of hypertension. Overall, those in Q5 of the three scores consumed about six to eight more servings of ‘healthy’ foods per day (mainly fruits and vegetables) and about a half to one less serving of meat compared with those in Q1. More detailed information can be obtained from Supplementary Table 2a–c, available as Supplementary data at IJE online.

In age- and multivariable-adjusted analyses, all three diet-quality scores were associated with lower risk of SGD (Table 2). Additional adjustment for medical history of hypertension, hypercholesterolemia, medication use (statins, thiazide diuretics or NSAID) and exclusion of participants who developed cancer or CVD at baseline and during follow-up did not substantially change the results (data not shown). As a diagnosis of diabetes might cause dietary changes, we excluded men with a history of diabetes. However, risk estimates were similar for diabetics [multivariable HRs (95% CIs) for Q5 vs Q1 of diet scores: DASH: 0.69 (0.43–1.12), aMed: 0.65 (0.39–1.09), AHEI-2010: 0.73 (0.46–1.15)]. Furthermore, inclusion of asymptomatic cases (n = 428) or exclusion of cases who did not undergo a cholecystectomy (n = 195) did not alter the findings [Q5 vs Q1: including asymptomatic gallstones: DASH: 0.66 (0.58–0.76), aMed: 0.66 (0.57–0.77), AHEI-2010: 0.64 (0.56–0.74); and for cholecystectomy only: DASH: 0.64 (0.54–0.75), aMed: 0.64 (0.53–0.76), AHEI-2010: 0.57 (0.49–0.67)]. Inverse associations between the three diet-quality scores and SGD risk were even stronger in the cumulative updated analysis for Q5 vs Q1 [multivariable HRs (95% CIs) for DASH: 0.59 (0.52–0.68), aMed: 0.62 (0.54–0.71) and AHEI-2010: 0.66 (0.58–0.75)].

In the stratified analyses, associations generally did not differ substantially between strata. However, for the AHEI-2010 and DASH diet scores, inverse associations appeared to be somewhat stronger among those with BMI < 25 than among participants with BMI ≥ 25. Furthermore, for the aMed and AHEI-2010 scores, the inverse associations were stronger among participants who were more physically active (≥median = 24.22 MET-hours/week) (Supplementary Figure 1A–C, available as Supplementary data at IJE online).

We examined each nutrient or food that contributed to each score separately to determine whether the association was driven by a few specific foods. After multivariable adjustment for total energy intake, smoking status, physical activity and coffee consumption, HRs [(95% CI) for Q5 vs Q1] were: nuts and legumes: 0.68 (0.59–0.78), fruits: 0.77 (0.67–0.89), n-3 FA: 0.81 (0.71–0.92), fish: 0.83 (0.72–0.94) and for the ratio between mono-unsaturated and saturated fatty acids: 0.83 (0.73–0.95). Similarly, when comparing Q1 with Q5, the risk estimates were: red and processed meat: 0.77 (0.66–0.89), trans-fat: 0.71 (0.62–0.81), sodium: 0.71 (0.59–0.86) and 10–25 g vs <10 or >25 g alcohol per day: 0.90 (0.82–0.99)] (Supplementary Figure 2, available as Supplementary data at IJE online).

Discussion

In this prospective cohort study of 43 635 male health-care professionals, the DASH, the aMed and the AHEI-2010 diet scores were inversely associated with risk of SGD. After multivariable adjustment, participants in Q5 of all three scores had an about 35% lower risk of SGD when compared with those in Q1. The findings were consistent in several sensitivity analyses and after adjustment for additional potential confounders.

In our study, participants in Q5 of dietary scores consumed about six to eight servings per day more of food items generally considered healthy (e.g. fruits, vegetables, nuts, whole-grain products) and a half to one serving less meat than those in Q1. By comparison, in the PREDIMED (‘Prevencion con Dieta Mediterranea’), which is a randomized–controlled trial comparing two Mediterranean diets supplemented with extra-virgin olive oil or nuts with a low-fat control diet, participants changed their food intake similarly: average consumption increased by 40 g/day for nuts, 13–18 g/day for vegetables, 5–15 g/day for fruits and 2–2.5 g/day for fish, whereas participants reduced their intakes by 23–29 g/day for meat and 17–45 g/day for dairy products.³³

Earlier studies have mainly concentrated on the intakes of single food groups or isolated nutrients,^9–18 which does not account for possible interaction between food or nutrient combinations. Furthermore, correlation among nutrients might limit the ability to observe true individual associations.³⁴ Currently, studies relating overall dietary patterns to gallstone disease are scarce and often limited by inadequate dietary assessment, inadequate control of confounding and small numbers of cases.^35–37

To our knowledge, the present study is the first large prospective cohort study that investigated a priori diet-quality scores with regard to SGD risk. Recently, a small prospective cohort study (n = 2848, gallstone cases n = 256) investigated the association between a ‘prudent’, a ‘healthy’ and a ‘Western’ diet and risk of gallstones (including symptomatic and asymptomatic cases, assessed by ultrasonography).³⁷ Surprisingly, results were suggestive of an inverse association between the ‘Western’ and the ‘healthy’ diet pattern and risk of gallstones but the 95% CIs included 1. However, in that study, diet was assessed using a 26-food-item FFQ, which might have affected the ability to differentiate between the three dietary patterns. In a subsequent meta-analysis, the authors of the aforementioned study observed inverse associations between coffee, fish and whole-grain bread and risk of gallstones, which is consistent with our findings.³⁷

Gallstone disease has been linked to higher mortality and increased risks of major chronic diseases [e.g. metabolic syndrome, diabetes, cancer and cardiovascular disease (CVD)^38–44], indicating the possibility of shared pathways with other metabolic disorders. The mechanisms underlying SGD development are not well understood, but emerging evidence suggest an important role of insulin resistance, hyperinsulinemia and an unfavourable lipid profile such as hypertriglyceridemia or low concentrations of high-density lipoprotein (HDL) cholesterol.³ HDL plays an important role in cholesterol elimination by transporting cholesterol from cells to the liver, where it can be converted into bile acids—the most important pathway for cholesterol degradation and homeostasis.⁴⁵ Foods and nutrients have been reported to alter lipid metabolism, such as trans-fat,⁴⁶ nuts,⁴⁷ n-3 FA^48,49 and moderate alcohol consumption.⁵⁰ Thus, one plausible mechanism for our findings includes the beneficial effect of these three diets on lipid and metabolic profiles.^51–54

Interestingly, in stratified analyses, we observed stronger inverse associations with SGD risk among those who were lean/physically active. It is possible that, in lean/physically active individuals, other risk factors (including hypercholesterolemia, fatty acid metabolism) are more dominant than e.g. insulin resistance, which might play a larger role in obesity.³ However, results were similar after adjustment for hypercholesterolemia.

Strengths of our study include the large sample size and the comprehensive data collection, which enabled us to conduct stratified analyses with sufficient power and adjust for a variety of potential confounders. Furthermore, both exposure and covariables were repeatedly assessed by validated questionnaires to account for changes over time that can lower the within-person variability of measurements. Nevertheless, some limitations merit further attention. First, high diet-quality scores might function as a marker of generally healthy behaviour. Even though results were similar after we adjusted for a variety of lifestyle factors including physical activity, smoking and BMI, we cannot exclude the possibility of residual or unmeasured confounding. Misclassification of dietary intake is another concern, as diet was self-reported and prone to recall bias. However, misclassification, if any, should be non-differential, which would have attenuated the results towards a null association. Another limitation is the possibility of undiagnosed gallstones, as most gallstones are asymptomatic. However, possible misclassification is expected to be non-differential, as both cases and non-cases might have had silent gallstones. Furthermore, we examined SGD generally accompanied by pain or other gastrointestinal symptoms and treated with cholecystectomy. Additionally, we performed analysis after inclusion of asymptomatic gallstones and observed consistent results. Further, we cannot exclude reverse causation to play a role in the most recent updated analysis. Nevertheless, using cumulative averages of all reported intakes even strengthened our findings. Finally, our analysis only included men, thus generalizability of our findings to females is limited. Despite these limitations, we observed inverse associations of moderate magnitude that were quite consistent in all sensitivity analyses.

In conclusion, the present investigation suggests that the risk of SGD is lower in men with high adherence to a high-quality diet, the aMed, the AHEI-2010 or the DASH diet. Recommendations focusing on high-quality diet targeting SGD are easier to communicate than those for individual nutrients or foods and may be applicable to the prevention and management of this disease. Our findings warrant further examination of the role of high-quality diets on SGD in both intervention and observational studies, particularly those including more heterogeneous populations and high-risk sub-populations (e.g. diabetics, obese).

Funding

This work was supported by grants from the National Institutes of Health (UM1 CA167552, R01 HL35464, K99 CA215314, K99 CA207736, K24 DK098311). The current study was supported by the German Research Foundation (Grant Number WI 4568/2–2 to J.W.). A.T.C. is a Stuart and Suzanne Steele MGH Research Scholar. The funders had no role in the study design; in the collection, analysis and interpretation of the data; in the preparation, review or approval of the manuscript; and in the decision to submit the article for publication.

Acknowledgements

We acknowledge the participants and staff of the Health Professionals Follow-up Study for their valuable contributions as well as the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. We are, furthermore, indebted to Scott G. Smith for his support in statistical analysis. The authors assume full responsibility for analyses and interpretation of these data.

Conflict of interest: None declared.

References