Zinc intake and risk of Crohn’s disease and ulcerative colitis: a prospective cohort study

Abstract

Background: Diet plays a role in the pathogenesis of Crohn’s disease (CD) and ulcerative colitis (UC). Dietary zinc may influence risk of disease through effects on autophagy, innate and adaptive immune response and maintenance of the intestinal barrier.

Methods: We analysed data from 170 776 women from the Nurses Health Study I and Nurses Health Study II, who were followed for 26 years. Zinc intake was assessed using semi-quantitative food frequency questionnaires administered every 4 years. Incident CD and UC were ascertained by medical record review. Cox proportional hazards models adjusting for potential confounders determined the independent association between zinc intake and incident disease.

Results: Over 3 317 550 person-years (p-y) of follow-up, we identified 269 incident cases of CD and 338 incident cases of UC. Zinc intake ranged from 9 mg/day in the lowest quintile to 27 mg/day in the highest quintile. Compared with women with the lowest quintile of intake, the multivariate hazard ratios (HR) for CD were 0.92 [95% confidence interval (CI), 0.65 – 1.29) for women in the second quintile of intake, 0.60 (95% CI, 0.40 – 0.89) for the third quintile, 0.57 (95% CI, 0.38 – 0.86) for fourth quintile and 0.74 (95% CI, 0.50 – 1.10) for the highest quintile ( P_trend  = 0.003). The association was stronger for dietary zinc (HR 0.63, 95% CI, 0.43 – 0.93, comparing extreme quintiles) than for zinc intake from supplements. Neither dietary nor supplemental zinc modified risk of UC.

Conclusions: In two large prospective cohorts of women, intake of zinc was inversely associated with risk of CD but not UC.

Introduction

Diet has long been purported to play a role in the pathogenesis of Crohn’s disease (CD) and ulcerative colitis (UC). ^1–4 Diet may plausibly influence risk of disease through its well-recognized effect on the composition of the gut microbiome, a key player in the pathogenesis of inflammatory bowel diseases (IBD), or through direct effects of various dietary components on the intestinal immune response. ¹ Yet this remains one of the most understudied exposures, due to the lack of high-quality, prospectively collected data to minimize the influence of biased recall associated with dietary ascertainment among individuals suffering from the symptoms of CD and UC that precede formal diagnosis. Moreover, studies should ideally account for the significant variations in an individual’s diet over long-term follow-up. Recent observational studies in large cohorts with prospectively ascertained and updated diet information have provided valuable insights into several macronutrients as risk factors for IBD. ^5–10 For example, a diet high in fibre, particularly from fruits and vegetables, is inversely associated with risk of CD, whereas intake of polyunsaturated fats may modify risk of UC. ^5–10 Dietary protein has also been variably reported to modify risk of both diseases. ^2,3,10

Most previous studies of diet have focused on macronutrient food groups. ^2,3 . However, it is well recognized that micronutrients play an important role in various biological processes relevant to IBD pathogenesis, including innate and adaptive immune responses, pathogen sensing and response and maintenance of the integrity of the intestinal barrier. Few studies have prospectively examined the role of such micronutrient factors in large cohorts. One particular micronutrient supported by strong biological plausibility is zinc. In cell culture experiments and animal models of colitis, administration of zinc improves intestinal barrier function and reduces expression of pro-inflammatory cytokines. ^11–14 In contrast, zinc deficiency is associated with more severe colitis and a larger inflammatory burden. ¹⁵ In human studies, zinc administration reduces intestinal permeability in CD ¹⁶ and is effective in the treatment of various diarrhoeal diseases ranging from tropical enteropathy to acute infectious diarrhoea. Zinc also is a co-factor for various enzymes involved in maintenance of intestinal integrity and regulates autophagy and bacterial clearance in macrophages. ¹⁷

Using the strengths of two large cohorts with prospective, repeated assessments of diet and other lifestyle exposures associated with risk, long duration of follow-up and physician-adjudication of incident diagnoses of disease, we examined the association between dietary and supplemental intakes of zinc and the incidence of CD and UC.

Methods

Study population

Our study included women participating in the Nurses Health Study (NHS I) or Nurses Health Study II (NHS II) cohorts. The NHS I, established in 1976, recruited 121 700 female registered nurses between the ages of 30 and 55 years. The NHS II, initiated in 1989, recruited a younger cohort of 116 686 female registered nurses between the ages of 25 and 42 years at enrolment. Both cohorts continue to follow the women biennially through questionnaires assessing personal and environmental exposures, diet and updated medical history, with a response rate consistently exceeding 90%. Included in this study were women who had completed a detailed semi-quantitative food frequency questionnaire (SFFQ) in 1984 in NHS I and in 1991 in NHS II. Women who had a diagnosis of IBD before this baseline questionnaire, had died or had a previous diagnosis of cancer except for non-melanoma skin cancer were excluded. The study was approved by the Institutional Review Board of Partners Healthcare.

Ascertainment of zinc intake

All participants of NHS were administered a 121-item SFFQ in 1984, which was expanded to a 131-item SFFQ in 1986 that was then administered every 4 years in 1990, 1994, 1998, 2002 and 2006. Women in NHS II were administered a similar SFFQ first in 1991 and subsequently in 1995, 1999, 2003 and 2007. Each SFFQ enquired about consumption of various foods over the previous 1 year. For each item, a commonly used portion size was specified and participants were asked regarding the frequency of consumption of each food item on an average, with responses ranging from ‘never or less than once per month’ to ‘six or more times per day’. Participants were also asked about consumption of multivitamins and zinc supplements including brand, dose, duration and frequency of use. Nutrient intakes were calculated by multiplying the frequency of the food consumed and the nutrient content of specific portion sizes based on tables provided by the Department of Agriculture. Nutrient intakes were adjusted for total energy intake by the residual method. ¹⁸ Total intake of zinc included that consumed from diet and through supplements and multivitamins. In previous studies, zinc intake assessed from the SFFQ in our cohorts and assessed against the average of four 1-week diet records demonstrated strong correlation with a Pearson coefficient of 0.63. ^19–21 The foods accounting for the highest proportion of dietary zinc included red meat, cold breakfast cereal, chicken without skin, hamburgers and skimmed milk with the correlation coefficients for self-report of each of these food items against the 1-week diet record ranging from 0.38 to 0.81. ^19–21

Identification of CD and UC

Details regarding our ascertainment of cases of CD and UC have been reported previously. ^5,6,22–27 In brief, since the inception of the cohort, participants of both NHS I and NHS II have been asked to self-report a diagnosis of CD or UC. Participants reporting this diagnosis were sent a supplemental questionnaire ascertaining history of disease in greater detail including date of diagnosis, disease complications and need for specific medications and surgery. Permission was requested to review medical records and, when obtained, all medical records were reviewed independently by two Board-certified gastroenterologists. A diagnosis of CD or UC was made based on standard criteria encompassing typical symptoms of 4 weeks’ duration or longer and supporting endoscopic, radiological and histological findings. Disagreements, which were infrequent, were settled through consensus. Among those with CD, we noted whether the disease involved small intestine alone, colon alone or both locations. All case ascertainment was blinded to exposure information. As reported previously, during this study period 2735 women in NHS I and 2541 women in NHS II self-reported a diagnosis through 2010 in NHS I and 2009 in NHS II. ⁶ Among 3415 women who were still alive, reported a date of diagnosis after cohort inception and could be contacted, permission to review medical records was obtained from 82% of women ( n  = 1532) who confirmed a diagnosis of IBD on the supplemental questionnaire. From among those with sufficient medical records for review, we excluded women who did not have chronic colitis ( n  = 312) or those with non-IBD chronic colitis ( n  = 192), resulting in our final cohort of 269 incident cases of CD and 338 cases of UC ( Figure 1 ).

Covariates

We gathered biennially updated information on other potential lifestyle influences on IBD including smoking, ²⁴ use of menopausal hormone therapy ²⁶ or oral contraceptives, ²⁷ and non-steroidal anti-inflammatory drugs. ²² Dietary intake of fibre and vitamin D was assessed using the SFFQ. ^23,28 Physical activity was assessed every 4 years as reported previously. ²⁵ Body mass index (BMI) was calculated by dividing weight (in kilograms) by the square of height (in metres).

Statistical analysis

All women contributed person-time, beginning from the date of return of the baseline questionnaire (1984 in NHS I and 1991 in NHS II) until the diagnosis of CD or UC, death or date of return of the last questionnaire. After ensuring proportionality of hazards, a Cox proportional hazards model adjusting for potential covariates was used to identify the multivariate hazard ratios (HR) and 95% confidence intervals (CI) between zinc intake and incidence of CD and UC. We modelled BMI as calculated at baseline to minimize the influence of pre-diagnosis symptoms on weight. We modelled all other covariates as time-varying, updating information with responses from each biennial questionnaire cycle. For our primary analysis, we modelled the total intake of zinc as reported on the questionnaire cycle before the diagnosis of CD or UC, thus with a lag of at least 2–4 years. In a sensitivity analysis, we increased the lag to two questionnaire cycles, or 4–8 years. We separately examined the association with dietary zinc and zinc intake from supplements. To examine the effect of long-term zinc consumption, we cumulatively averaged the intake of zinc from all preceding questionnaires before each biennial period. In sensitivity analysis, we additionally adjusted for intake of vitamin C, phytates and calcium, as they may modify bioavailability of dietary zinc. ^20,21 We observed no heterogeneity in the effect of dietary zinc between NHS I and NHS II ( P  > 0.30); thus, we pooled data from both cohorts for all analyses. In a separate analysis, we stratified zinc intake into three categories based on total intake of less than the recommended daily allowance (RDA) (< 8 mg/day), up to twice the RDA (8–16 mg/day), and more than twice the RDA (> 16 mg/day). To ensure that the association was not confounded by other nutrients present in food items with high zinc content, we repeated the multivariate Cox models substituting zinc intake with quintiles of intake of each of the five most common food items contributing to dietary zinc.

We tested for interaction between dietary zinc and other covariates (body mass index, smoking status, oral contraceptive use, dietary fibre and fat) by introducing a cross-product interaction term in the multivariate model, examining additive as well as multiplicative effects. For linear associations, tests for trend were conducted using the median values for each quintile of zinc intake as a continuous variable in the regression models. We also examined the possibility of a non-linear association between zinc and IBD risk, using restricted cubic splines. We selected three-knots splines, consistent with previous analyses in this cohort. We tested for overall significance of the spline curve using a likelihood ratio test that compared the −2 log likelihood of a model with zinc intake expressed as spline terms with that of a model without these terms.

Results

Study population

Our final cohort included 76 738 women from NHS and 94 071 women from NHS II. Over 3 317 550 person-years (p-y) of follow-up, we identified 269 incident cases of CD (incidence 8 per 100 000 p-y) and 338 cases of UC (incidence 10 per 100 000 p-y). Zinc intake at baseline ranged from a median of 9.0 mg/day among women in the lowest quintile [interquartile range (IQR) 8.0–9.1 mg/day) to 27.0 mg/day (IQR 21.3–36.0 mg/day) among women in the highest quintile. Table 1 compares the baseline characteristics of the study population stratified by quintile of zinc intake. There were no significant differences in age distribution across the quintiles; most women were of Caucasian ethnicity. A similar proportion of women in each quintile had a history of smoking and use of oral contraceptives or menopausal hormone replacement therapy. There were a higher proportion of women with a BMI of 30 or greater in the highest quintile when compared with those in the lowest quintile of zinc intake. Women in the highest quintile of zinc intake also had greater protein intake, but were similar in their consumption of total calories and fat and had similar levels of physical activity. Total zinc intake comprised almost exclusively dietary zinc in the lowest three quintiles of intake, whereas zinc intake from supplements comprised a larger proportion of total intake in women belonging to the highest quintile.

Women in higher quintiles of total zinc intake had reduced risk of CD. Compared with a CD incidence of 11 per 100 000 p-y in the quintile with the lowest zinc intake, the incidence of CD was 6 per 100 000 p-y in the third and fourth quintiles and 8 per 100 000 p-y in the highest quintile of intake ( Table 2 ). Adjusting for potential confounders, women in the third (HR 0.60, 95% CI, 0.40 – 0.89), fourth (HR 0.57, 95% CI, 0.38 – 0.86) and fifth (HR 0.74, 95% CI, 0.50 – 1.10) quintiles had lower risk of CD ( P_trend  = 0.003) ( Figure 2 ). The reduction in risk with increasing zinc intake was seen up to a daily intake of 15 mg/day, after which the risk reduction reached a plateau. Compared with individuals with intake of zinc less than the recommended daily allowance (8 mg/day), those with intake of 8–16 mg/day (HR 0.69, 0.44 – 1.08) and > 16 mg/day (HR 0.52, 0.32 – 0.86) had a reduced risk of CD. We performed subgroup analyses examining if the effect of zinc intake varied by location of CD. Comparing extreme quintiles of intake, the association with lower disease risk was more prominent for colonic CD (HR 0.40, 95% CI, 0.19 – 0.85) than for ileal CD (HR 1.23, 95% CI, 0.72 – 2.21). Compared with women in the lowest quintile, those in the highest quintile of total zinc intake had a non-significant increase in risk of UC (HR 1.40, 95% CI, 0.98 – 2.00) ( Table 2 ).

We then separately examined the effects of dietary and supplemental zinc. There was a stronger inverse association with dietary zinc intake than for zinc intake from supplements. Compared with women in the lowest quintile of dietary zinc, women in the highest quintile had a reduced risk of CD (HR 0.63, 95% CI, 0.43 – 0.93; P_trend  = 0.04) ( Table 3 ). Neither dietary zinc nor supplemental zinc intake modified risk of UC.

Sensitivity analysis

We performed various sensitivity analyses to examine the robustness of our findings. First, we examined the impact of long-term zinc intake by cumulatively averaging intake of zinc as assessed on each dietary survey before each 4-year questionnaire interval ( Table 4 ). Similar to our primary analysis, increasing quintiles of cumulatively averaged zinc intake was also associated with a reduced risk of CD. To minimize the potential influence of reverse causation whereby subclinical symptoms prior to disease diagnosis may modify diet or zinc absorption, we introduced a lag of two questionnaire cycles (4–8 years) between last assessment of zinc intake and disease diagnosis. We similarly observed higher dietary zinc to be associated with reduced risk of CD but not of UC. The HR for the highest quintile of intake compared with the lowest quintile was 0.63 (95% CI, 0.42 – 0.95). As dietary intake of phytates, calcium or vitamin C may influence bioavailability of dietary zinc, we repeated our Cox proportional hazards models additionally adjusting for quintile (Q) of intake of each of these three nutrient groups. The inverse association between total (Q4 vs Q1 HR 0.56, 95% CI, 0.37 – 0.85; Q5 vs Q1 HR 0.71, 95% CI, 0.47 – 1.06) and dietary zinc (Q5 vs Q1 HR 0.62, 95% CI, 0.42 – 0.92) and incidence of CD were not materially changed. None of the five main contributors to dietary zinc (red meat (Q5 vs., Q1 HR 0.61, 95% CI, 0.35 – 1.06), chicken without skin (Q5 vs Q1, HR 0.90, 95% CI, 0.58 – 1.42), cold breakfast cereal (Q5 vs Q1 HR 1.06, 95% CI, 0.72 – 1.56), skimmed milk (Q5 vs Q1, HR 0.80, 95% CI, 0.55 – 1.17), and hamburgers (Q5 vs Q1 HR 0.68, 95% CI, 0.473 – 1.09) were independently associated with risk of CD, supporting a specificity of the association with overall dietary zinc rather than specific foods with high zinc content. A cumulative sum of daily servings of all five foods also was not independently associated with risk of CD or UC. We observed no significant interactions between zinc intake and body mass index, smoking, oral contraceptive use or dietary macronutrients.

Discussion

Diet may influence the risk of CD and UC through a variety of plausible mechanisms, including modulation of the gut microbiome and regulation of host immune responses. Additionally, the effect of diet on intestinal disorders may be distinct from the effect of systemic disease due to effect of dietary factors, both systemically after absorption and locally through direct effects on the luminal barrier, intestinal immune response and the microbiome. In two large prospective cohorts of women followed over 26 years, we demonstrate that higher intake of zinc was associated with a reduced risk of CD but not of UC. The inverse association was more striking for zinc derived from diet compared with supplements. Our findings remained robust to a variety of assumptions in sensitivity analyses.

Zinc is recognized to have panoply of effects on immune function and may influence the pathogenesis of CD through several mechanisms. ^{11,12,29–32} It plays a role in innate immunity mediated through its effect on the function of natural killer cells, macrophages and neutrophils and on adaptive immune responses by influencing the function of T- and B-lymphocytes. ³² . Prasad et al . examined the production of tumour necrosis factor-α (TNFα) and interleukin-1β (IL-1β) from human umbilical vein endothelial cells in a zinc=deficient compared with a zinc-sufficient environment. ^33,34 The cell lines in the zinc-deficient medium demonstrated reduced production of inflammatory cytokines and decreased signalling through the NF-κβ pathway; similar findings were observed in lung epithelial cell cultures by Bao et al.²⁹ The inhibitory effect of zinc on the NF-κB pathway may be mediated through direct inhibition of NF-κB-dependent transcripts, in particular the enzyme-inducible-κB kinase (IKK). ^34,35 In animal models of colitis, administration of zinc via enema reduced mucosal inflammation and myeloperoxidase activity; such rats had lower diarrhoea and colonic weight than untreated rats. ¹⁴ A second mechanism of effect of zinc is through its role in maintaining integrity of the intestinal barrier. In an in vitro intestinal cell model ( Caco2 cells), zinc deficiency resulted in reduced transepithelial electrical resistance, indicating greater intestinal permeability. ³⁰ Disruption of tight junctions associated with zinc deficiency was mediated through delocalization of E-cadherin and β–catenin in intestinal epithelial cells.

Deficiency of zinc is common in patients with IBD. However, no previous studies have examined zinc levels before diagnosis. In a cohort of 154 IBD patients, mean zinc levels were lower in those with CD than in those with UC or controls, ³⁶ suggesting a specificity of association with CD, consistent with our findings. In a study of 12 patients with quiescent CD, zinc supplementation reduced intestinal permeability, which in turn was associated with reduced rate of relapse of disease activity. ¹⁶ Further support of the specificity of the association with CD comes from Lahiri et al.¹⁷ In monocyte-derived macrophages (MDM), chronic stimulation through the pattern recognition receptor, nucleotide-binding oligomerization domain-2 (NOD2), a key CD genetic susceptibility locus, resulted in increased intracellular zinc through upregulation of metallothioneines. This led to increased bacterial clearance and autophagy, another key mechanism underlying the development of CD. Addition of zinc to such models restored autophagy and enhanced bacterial clearance by intestinal macrophages. Finally, the zinc transporter (ZIP8 or SLC39A8) is a transcriptional target of NF-κB and negatively regulates pro-inflammatory response, further supporting the potential role of zinc in the pathogenesis of IBD. ³⁵

Our findings showed a more robust association with total intake of zinc as well as zinc intake from diet when compared with intake from supplements. This could be explained by several possibilities. First, the range of distribution of daily total or dietary zinc was more uniform across quintiles, allowing for sufficient discrimination to examine a dose-response relationship. In contrast, supplemental zinc intake was more skewed, with a mean baseline intake being less than 0.7 mg/day for the lowest four quintiles. Thus, it is plausible that we did not have sufficient discriminatory power to tease out the independent effect of supplemental zinc in our study. Second, it is possible that zinc exerts its effects in conjunction with the food mixture with which it is consumed. Finally, it is also possible that supplemental zinc is consumed more frequently in conjunction with other nutrient components that may interfere with its effect. Nevertheless, future studies on micronutrient risk factors may need to independently examine their role when consumed as diet compared with as supplements. In contrast to the inverse association with CD, we observed a non-significant increase in risk of UC in the highest quintile of recent zinc intake, though this effect was not observed when examining dietary zinc alone or total cumulative zinc intake. Thus there is less confidence in the association between zinc intake and risk of UC, and this merits examination in further studies. Despite overlapping genetic risk, several environmental factors, most prominently smoking and appendectomy, have demonstrated divergent directions of association with CD and UC, ^4,37–40 and such effects may be observed for diet as well.

Our study has several strengths. Previous studies examining the association of diet with incident IBD have several limitations, including retrospective assessment of diet which is susceptible to bias from pre-diagnosis modification of diet and assessment of diet at a single time point, which does not account for changes in zinc intake over time. To our knowledge, ours is the first study to examine the association between dietary zinc and incidence of CD and UC. Our findings were robust to a number of sensitivity analyses and adjusting for relevant potential confounders, and are supported in their mechanistic plausibility by strong laboratory evidence. Further studies are essential to identify whether specific subgroups of patients, perhaps influenced by underlying host genotype, are more susceptible to the effects of dietary zinc. In parallel, it is important to study the role of zinc in the natural history of these diseases and whether it plays a role in the maintenance of remission or reduction of inflammation in those with active disease.

We acknowledge limitations associated with our study. First, our cohort consisted entirely of women, most of whom were Caucasian in ethnicity. Second, the median age of our cohort was older than many patients diagnosed with IBD. However, our analysis did include a broad spectrum of individuals diagnosed between the third and ninth decade of life. Furthermore, to our knowledge, no lifestyle risk factors (other than oral contraceptive use or menopausal hormone therapy) have demonstrated age- or sex-specific associations; thereby the above limitations are unlikely to affect our internal validity or generalizability of our findings. Erroneous misclassification of cases as controls owing to incomplete information is likely to be non-differential with respect to exposure and consequently to result in a bias towards the null, making our estimate more conservative than the true effect size. Additionally, the age-adjusted incidences of CD and UC in our cohorts are similar to those reported from the Olmsted county and Kaiser Permanente population-based IBD cohorts, and the magnitude of associations for environmental risk factors identified in our cohorts is similar to those previously reported. ⁴¹ Third, whereas it is plausible that recent diet may reflect modified eating due to pre-diagnosis symptoms of CD or UC in other population-based cohorts, the median latency period between onset of symptoms and diagnosis was less than 2 years and few patients were diagnosed outside this window. In our primary analysis, the gap between diet and diagnosis was at least 2–4 years and our findings remained robust, and were indeed strengthened, on introducing a gap of 4–8 years between diet and diagnosis. Our associations also remained unchanged with long-term cumulatively averaged zinc intake. Fourth, the number of cases within each quintile limited our ability to perform more detailed subgroup analyses. Finally, as in any observational study, one cannot exclude the possibility of unmeasured confounders. However, our analysis adjusted for several previously described risk factors and observed no attenuation of the association with zinc intake.

In conclusion, we demonstrated total intake of zinc to be inversely associated with risk of CD but not of UC. This association was particularly strong for dietary zinc and persisted with long-term zinc intake. Considerable mechanistic plausibility supports this association, including the effect of zinc on autophagy and microbial clearance, innate immune response and maintenance of the integrity of the epithelial barrier. Further studies on the role of zinc in mediating disease activity and outcomes are warranted.

Funding

This work was supported by a Research Scholars Award of the American Gastroenterological Association (A.N.A., H.K), a Senior Investigator Award from the Crohn’s and Colitis Foundation of America (A.T.C), and the National Institutes of Health (K24 DK098311, P01 CA87969, P30 DK043351, K08 DK064256, K23 DK091742, K23 DK099681, and UM1 CA176276).

The research presented in this manuscript is original. The contents of this article are solely the responsibility of the authors. The AGA, CCFA and the NIH had no role in the collection, management, analysis, or interpretation of the data and had no role in the preparation, review or approval of the manuscript.

Acknowledgments

The authors acknowledge the dedication of the Nurses’ Health Study I and II participants and members of Channing Division of Network Medicine.

Author contributions

A.A.: study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; obtained funding; study supervision. H.K.: acquisition of data; analysis and interpretation of data; critical revision of the manuscript for important intellectual content. M.S.: analysis and interpretation of data; critical revision of the manuscript for important intellectual content. L.H.: acquisition of data; critical revision of the manuscript for important intellectual content. J.R.: study concept and design; acquisition of data; critical revision of the manuscript for important intellectual content. A.C.: study concept and design; acquisition of data; analysis and interpretation of data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; statistical analysis; obtained funding; study supervision.

Conflict of interest: A.A.: scientific advisory board for Cubist and Abbvie. J.R.: consultant for Policy Analysis, Inc. A.C.: consultant for Bayer HealthCare, Millennium Pharmaceuticals, Pfizer Inc., Pozen Inc.

References