Association Between Circulating Levels of C-reactive Protein and Interleukin-6 and Risk of Inflammatory Bowel Disease

Abstract

Background & Aims

There is evidence that immune dysfunction precedes symptoms of inflammatory bowel disease (IBD) by several years. Characterization of preclinical systemic inflammation could contribute to our understanding of the biology of IBD and, ultimately, facilitate development of strategies for early disease detection and intervention. We evaluated associations between circulating levels of interleukin-6 (IL6) and high sensitivity C-reactive protein (hsCRP) and diagnosis of incident Crohn’s disease (CD) or ulcerative colitis (UC).

Methods

We conducted a nested case–control study of participants enrolled in 2 population-based, nationwide, prospective cohort studies (the Nurses’ Health Study and the Nurses’ Health Study II). We analyzed blood specimens, collected before diagnosis, from 83 persons with CD, 90 persons with UC, and 344 matched individuals without IBD (controls). Plasma levels of hsCRP and IL6 were measured. We investigated associations between each inflammatory marker and IBD risk using multivariable logistic regression models to adjust for potential confounding exposures.

Results

Compared to the lowest quintile of IL6 level, the highest quintile was associated with an odds ratio (OR) of 4.68 (95% confidence interval, 1.91–11.46) for CD (Ptrend<.001) and an OR of 3.43 (95% confidence interval, 1.44–8.15) for UC (Ptrend=.004). The highest quintile of hsCRP level, compared to the lowest quintile, was associated with an OR of 2.82 (95% confidence interval, 1.15–6.87) for CD (Ptrend = 0.019) and an OR of 1.79 (95% confidence interval, 0.80–3.99) for UC (Ptrend = 0.015).

Conclusions

Plasma levels of IL6 and hsCRP before diagnosis are associated with risk of incident CD and UC. Subclinical levels of systemic inflammation may be a feature of an early disease state that precedes the development of symptomatic IBD.

INTRODUCTION

Crohn’s disease (CD) and ulcerative colitis (UC), collectively termed inflammatory bowel disease (IBD), are chronic inflammatory conditions of the gastrointestinal tract that are estimated to affect over one million individuals in the US alone.¹ While the etiology of IBD remains uncertain, CD and UC are believed to arise as a result of inappropriate mucosal immune responses in genetically predisposed individuals.² Our appreciation of the natural history of CD and UC is limited largely to events that occur after disease diagnosis. However, the presence of serologic markers such as perinuclear anti-neutrophil cytoplasmic antibody (pANCA) and antimicrobial antibodies in the circulation of IBD patients, many years prior to presentation,^{3, 4} suggests that a considerable time lag exists between mucosal immune dysfunction and the development of symptoms. Preclinical disease phases have been described for other immune-mediated inflammatory disorders, such as rheumatoid arthritis,⁵ where inflammatory markers, such as interleukin-6 (IL6), also appear to be elevated several years in advance of symptomatic disease.⁶

To our knowledge, no previous study has examined pre-diagnostic inflammatory markers in relation to IBD risk. Characterizing preclinical inflammation in IBD could give insights into the natural history of CD and UC, and might help identify potential windows for early therapeutic or preventive interventions in high-risk individuals. We therefore conducted a nested case-control study using prospectively collected data from the Nurses’ Health Study (NHS) cohorts to examine the association between pre-diagnostic plasma high-sensitivity C-reactive protein (hsCRP) and interleukin-6 (IL6) in relation to the risk of incident CD and UC.

METHODS

Study populations

The NHS is a nationwide prospective cohort that was initiated in 1976 with the enrollment of 121,700 female registered nurses, aged 30–55 years, who had completed a postal health questionnaire. Follow-up questionnaires have been mailed every two years to update health and lifestyle information. The Nurses’ Health Study II (NHSII) is a parallel cohort established in 1989 with the enrollment of 116,686 female nurses, aged 25–42 years. Women in the NHSII have been followed-up similarly, by means of biennial questionnaires. Follow-up in both cohorts exceeds 90%.

Ascertainment of cases and controls

We have previously reported in detail the ascertainment of IBD cases within the NHS and NHSII cohorts.^{7, 8} In brief, from baseline in both cohorts, participants have been able to report diagnoses of UC or CD through an open-ended questionnaire response. In the NHS, a diagnosis of UC has been specifically queried from 1982 onward, and from 1992 onward for CD. In the NHSII, diagnoses of UC and CD have both been queried specifically since 1993. When a participant in either cohort reported a diagnosis of CD or UC, a supplementary questionnaire was mailed and permission requested for access to related medical records. Two gastroenterologists, blinded to exposure information, reviewed case records. Data on diagnostic investigations, histopathology, and disease extent and behavior were extracted. Using standardized criteria,^{9, 10} participants were classified as having UC based on typical clinical presentation of ≥ 4 weeks plus endoscopic or surgical pathological findings compatible with UC. A diagnosis of CD was made based on typical clinical history of ≥ 4 weeks in conjunction with characteristic small bowel changes on endoscopic, surgical, or radiologic investigations, or an endoscopic or surgical pathologic specimen consistent with CD. Disagreements were resolved through consensus. The IBD case confirmation rate among those with available records was 78% in the NHS and 74% in the NHSII.⁸

Blood collection in the NHS cohorts has been described previously.^{11, 12} Briefly, between 1989 and 1990, 32,826 NHS participants (aged 43–69 years) provided blood specimens, along with a short questionnaire. Similarly, between 1996 and 1999, 29,611 NHSII participants (aged 32–54 years) provided blood specimens and questionnaires. Blood specimens were returned on ice packs by overnight courier. On arrival, blood specimens were centrifuged without delay, and aliquots of plasma were transferred to continuously-monitored liquid nitrogen freezers. We have previously demonstrated the stability of plasma IL6 and hsCRP collected and processed using this method,¹³ and additional data suggest that these biomarkers remain stable over prolonged periods in frozen specimens.^{14, 15}

The present study is a nested case–control study conducted among women with available blood specimens. We identified 83 cases of incident CD (59 in the NHS, and 24 in the NHSII), and 90 cases of UC (58 in the NHS and 32 in the NHSII) diagnosed after blood collection. Each IBD case was matched to two controls based on month of blood collection, menopausal status, fasting status, and use of menopausal hormone therapy (MHT) at the time of blood collection.⁸ Our final analysis included 83 CD cases and 90 UC cases matched to 344 controls (165 for CD and 179 for UC).

This study was approved by the institutional review boards of the Brigham and Women’s Hospital and Harvard Medical School (protocol number 2001P001128).

Ascertainment of Exposures

Our primary exposures of interest were plasma concentrations of IL6 and hsCRP, assays for which were conducted in the laboratory of Dr Nader Rifai (Children’s Hospital, Boston, MA). Masked replicate quality control specimens comprised 10% of each batch. Laboratory personnel were blinded to specimen case, control, and quality control status. IL6 concentration was determined by solid phase sandwich ELISA (R&D systems, Minneapolis, MN); hsCRP was assayed by immunoturbidimetry (Roche Diagnostics, Indianapolis, IN). Overall within-batch coefficients of variation were 2.7% to 6.7% for IL6 and 1.1% to 4.3% for hsCRP.

Ascertainment of covariates

Information regarding pertinent health and lifestyle factors was obtained by self-reporting through postal questionnaires. Data on weight, menopausal status, and use of menopausal hormone therapy (MHT) were taken from the short questionnaires returned at the time of blood collection; data for other covariates, including smoking status, oral contraceptive (OC) use, leisure-time physical activity, and use of aspirin or non-steroidal anti-inflammatory drugs (NSAIDs) were obtained from the standard questionnaires completed closest to the time of blood collection. Body mass index (BMI) was derived from weight in kilograms divided by the square of baseline height in meters. Participants' self-reporting of body weight, height, OC use, and physical activity has been previously validated.^16–18

Statistical analysis

We first examined the possibility of a non-linear relationship between plasma IL6 or hsCRP and IBD risk using a previously reported non-parametric method with restricted cubic splines.¹⁹ Using the likelihood ratio test to compare models with only the linear term to models with the linear and the cubic spline terms, we did not observe evidence of non-linearity (all P_comparison > 0.10), suggesting linear relationships between inflammatory maker levels and risk of CD and UC. We therefore chose to use quintiles derived from the continuous values for plasma IL6 (in pg/ml) and hsCRP (in mg/l). We calculated quintile cut points separately for each cohort, based on the distribution of each inflammatory marker among controls. We used both conditional and unconditional logistic regression models to estimate odds ratios (OR) and 95% confidence intervals (CI) for risk of CD or UC. Because both approaches yielded similar effect estimates, we used unconditional models, so that power was maximized for stratified analyses. In multivariable models, beyond age and cohort, we adjusted for BMI, cumulative physical activity, OC use, and smoking status, since these factors have previously been reported to be associated with the risk of CD or UC.²⁰ Additional adjustment of multivariate models for NSAID or aspirin use at the time of blood collection, history of appendectomy, menopausal status/MHT use, predicted vitamin D status, or ethnicity did not substantially alter our effect estimates; thus, to avoid overfitting, we did not include these variables in our final models. P values for linear trend across quintiles of inflammatory markers were determined using the median IL6 or hsCRP value for each quintile as an ordinal variable in logistic models.

We performed analyses stratified by BMI (<25 or ≥25 kg/m2), smoking status (ever or never), and, using the median value for all participants as a cut point, physical activity (<11 MET-h/week or ≥11 MET-h/week) and interval between blood collection and diagnosis (or indexing as control; <6.7 years or ≥ 6.7 years). Exploratory stratified analyses were conducted according to categories of time interval between blood collection and diagnosis (0 to 5 years, > 5 to 10 years, or > 10 years), documented disease distribution at diagnosis (ileal, colonic, or ileocolonic for CD; left colonic or extending beyond left colon for UC), and documented family history of IBD (present or absent). Multiplicative interaction was assessed using the Wald test based on interaction terms that were the cross product of the stratifying variable and the inflammatory marker of interest. All analyses were conducted using SAS version 9.3 (Cary, NC). All P values were two-sided and the threshold for statistical significance was set at 0.05.

RESULTS

The median time interval between blood collection and diagnosis of CD or UC was 6.6 and 6.8 years, respectively (range, 1 month to 20.4 years). The age range of participants at the time of blood sampling was 35 to 69 years. Characteristics of cases and controls at the time of blood collection are shown in Table 1. Cases and controls did not differ significantly with respect to age, BMI, physical activity, menopausal status or exogenous hormone use (OC or MHT). Compared to controls, the proportion of former smokers was greater among UC cases (P_difference = 0.047). Compared to the median concentration in controls (1.0 pg/ml), pre-diagnostic IL6 concentration was statistically significantly higher in CD cases (1.7 pg/ml; P_difference < 0.001) and UC cases (1.2 pg/ml P_difference = 0.002). Similarly, compared to the mean concentration in controls (1.5 mg/l), pre-diagnostic hsCRP concentration was higher in CD cases (2.3 mg/l, P_difference = 0.037) and UC cases (2.2 mg/l P_difference = 0.030).

Table 1.

Baseline characteristics of cases and controls at time of blood collection

	Controls (N=344)	CD		UC
		Cases (N=83)	Pdifference*	Cases (N=90)	Pdifference*
Number of participants
NHS	232	59		58
NHSII	112	24		32
Median age, years (IQR)	51.7 (46.7)	52.7 (46.3–58.9)	0.41	50.4 (46.6–57.4)	0.39
Median BMI, Kg/m2 (IQR)	24.9 (22.1–28.4)	24.4 (21.7–29.0)	0.39	25.4 (22.9–29.6)	0.24
Median physical activity, MET-h/wk (IQR)	10.7 (5.2–21.0)	10.9 (5.2–19.4)	0.60	10.3 (4.4–20.4)	0.56
Smoking status, %
Never	51	48		37
Former	39	35	0.19	49	0.047
Current	10	17		14
Premenopausal, %	35	31	0.53	38	0.61
Oral contraceptive use, %#	66	59	0.23	72	0.25
Menopausal hormone therapy, %†	25	24	0.88	24	0.95
Median Plasma IL6, pg/ml (IQR)	1.0 (0.7–1.7)	1.7 (0.9–3.1)	<0.001	1.2 (0.9–2.7)	0.002
Median Plasma hsCRP, mg/l (IQR)	1.5 (0.7–3.7)	2.3 (0.8–4.7)	0.037	2.2 (0.9–5.2)	0.030

Percentages are given for categorical variables. For continuous variables, the median value is presented with interquartile range (IQR) in parentheses.

^*The P value for difference was computed by chi-square for categorical variables and by Wilcoxon rank sum test for continuous variables.

^#Ever use of oral contraceptives.

^†Current use of hormonal therapy among post-menopausal women.

BMI, body mass index; CD, Crohn’s disease; MET, metabolic equivalent of task; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; SD, standard deviation; UC, ulcerative colitis.

In multivariable-adjusted logistic regression models, compared to the lowest quintile of pre-diagnostic plasma IL6, the highest quintile was associated with an OR of 4.68 (95% CI, 1.91–11.46) for CD (P_trend < 0.001; Table 2), and an OR of 3.43 (95% CI, 1.44–8.15) for UC (P_trend = 0.004; Table 3). For pre-diagnostic hsCRP, compared to the lowest quintile, the highest quintile was associated with an OR of 2.82 (95% CI, 1.15–6.87) for CD (P_trend = 0.019; Table 2), and an OR of 1.79 (95% CI, 0.80–3.99) for UC (P_trend = 0.015; Table 3).

Table 2.

Risk of Crohn’s disease according to quintiles of plasma inflammatory markers

	Q1	Q2	Q3	Q4	Q5	Ptrend*
IL6 (pg/ml)
NHS, median (range)	0.56 (0.19–0.69)	0.82 (0.69–0.96)	1.15 (0.96–1.33)	1.71 (1.34–2.22)	3.48 (2.23–78.2)
NHSII, median (range)	0.46 (0.34–0.58)	0.68 (0.59–0.81)	0.95 (0.82–1.08)	1.29 (1.08–1.72)	2.44 (1.75–27.9)
Number of cases/controls	9/65	11/72	10/69	22/69	31/69
Unadjusted OR (95% CI)	Referent	1.10 (0.43–2.83)	1.05 (0.40–2.74)	2.30 (0.99–5.37)	3.25 (1.44–7.34)	<0.001
Adjusted OR# (95% CI)	Referent	1.18 (0.45–3.08)	1.22 (0.46–3.28)	2.92 (1.19–7.19)	4.68 (1.91–11.46)	<0.001
hsCRP (mg/l)
NHS, median (range)	0.37 (0.07–0.59)	0.89 (0.60–1.26)	1.91 (1.29–2.54)	3.66 (2.57–4.81)	6.97 (4.93–36.7)
NHSII, median (range)	0.24 (0.03–0.38)	0.62 (0.43–0.87)	1.12 (0.89–1.49)	2.38 (1.50–4.04)	8.60 (4.05–23.3)
Number of cases/controls	11/68	15/70	15/69	20/69	22/68
Unadjusted OR (95% CI)	Referent	1.33 (0.57–3.09)	1.34 (0.58–3.14)	1.79 (0.80–4.02)	2.00 (0.90–4.44)	0.079
Adjusted OR# (95% CI)	Referent	1.36 (0.58–3.22)	1.51 (0.62–3.63)	2.22 (0.94–5.27)	2.82 (1.15–6.87)	0.019

^*The P value for linear trend was generated using an ordinal variable based on the median value for each quintile of IL6 or hsCRP.

^#Multivariable models adjusted for cohort, age, smoking status, body mass index, oral contraceptive use, and cumulative physical activity.

CI, confidence interval; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; OR, odds ratio; Q1–5, inflammatory marker quintiles.

Table 3.

Risk of ulcerative colitis according to quintiles of plasma inflammatory markers

	Q1	Q2	Q3	Q4	Q5	Ptrend*
IL6 (pg/ml)
NHS, median (range)	0.56 (0.19–0.69)	0.82 (0.69–0.96)	1.15 (0.96–1.33)	1.71 (1.34–2.22)	3.48 (2.23–78.2)
NHSII, median (range)	0.46 (0.34–0.58)	0.68 (0.59–0.81)	0.95 (0.82–1.08)	1.29 (1.08–1.72)	2.44 (1.75–27.9)
Number of cases/controls	11/65	10/72	25/69	14/69	30/69
Unadjusted OR (95% CI)	Referent	0.82 (0.33–2.06)	2.14 (0.98–4.70)	1.20 (0.51–2.83)	2.57 (1.19–5.55)	0.012
Adjusted OR# (95% CI)	Referent	0.92 (0.36–2.33)	2.38 (1.06–5.37)	1.37 (0.56–3.36)	3.43 (1.44–8.15)	0.004
hsCRP (mg/l)
NHS, median (range)	0.37 (0.07–0.59)	0.89 (0.60–1.26)	1.91 (1.29–2.54)	3.66 (2.57–4.81)	6.97 (4.93–36.7)
NHSII, median (range)	0.24 (0.03–0.38)	0.62 (0.43–0.87)	1.12 (0.89–1.49)	2.38 (1.50–4.04)	8.60 (4.05–23.3)
Number of cases/controls	17/68	9/70	16/69	21/69	27/68
Unadjusted OR (95% CI)	Referent	0.51 (0.22–1.23)	0.93 (0.43–2.98)	1.22 (0.59–2.51)	1.59 (0.79–3.18)	0.014
Adjusted OR# (95% CI)	Referent	0.51 (0.21–1.24)	0.99 (0.45–2.18)	1.35 (0.63–2.91)	1.79 (0.80–3.99)	0.015

^*The P value for linear trend was generated using an ordinal variable based on the median value for each quintile of IL6 or hsCRP.

^#Multivariable models adjusted for cohort, age, smoking status, body mass index, oral contraceptive use, and cumulative physical activity.

CI, confidence interval; NHS, Nurses’ Health Study; NHSII, Nurses’ Health Study II; OR, odds ratio; Q1–5, inflammatory marker quintiles.

We found that pre-diagnostic plasma levels of IL6 and hsCRP were moderately correlated among controls, CD cases, and UC cases (r = 0.52, 0.55, and 0.40, respectively; all P < 0.001). In models co-adjusted for both markers, the effect estimates for IL6 were generally similar to those without adjustment for hsCRP (P_trend ≤ 0.016 for UC and CD). In contrast, for hsCRP, the association with CD risk was no longer statistically significant after adjustment for IL6 (P_trend = 0.34) and only a non-significant trend was apparent for UC risk (P_trend = 0.072). In an analysis of combined inflammatory marker levels (Supplementary Table 1), compared to individuals with both IL6 and hsCRP in the lowest three quintiles, the multivariable-adjusted OR for those with both markers in the highest two quintiles was 4.24 (95% CI, 2.11–8.51) for CD, and 2.22 (95% CI, 1.14–4.35) for UC. We did not observe evidence of multiplicative interaction between IL6 and hsCRP in their associations with CD or UC risk (both P_interaction > 0.43).

We examined the possibility that previously reported risk factors for CD or UC may modify the association between IL6 or hsCRP and disease risk. We found no evidence of statistically significant interaction between inflammatory marker levels and cohort, BMI, physical activity, or smoking status, and risk of CD or UC (all P_interaction ≥ 0.056; Supplementary Tables 2 & 3).

We considered the possibility that the observed associations may have been driven undiagnosed, symptomatic cases of IBD at the time of blood collection. We therefore performed sensitivity analyses excluding IBD cases diagnosed within two years of blood collection. Following the exclusion of 12 cases of CD and 16 cases of UC, the median levels of IL6 and hsCRP among CD cases were 1.5 (IQR, 0.9–3.1) and 2.2 (IQR, 0.7–4.7), respectively (P_difference both < 0.004, compared to controls). For UC cases, the median levels of IL6 and hsCRP were 1.3 (IQR, 0.9–2.5) and 2.0 (IQR, 0.8–5.3), respectively (P_difference both ≥ 0.097, compared to controls). Effect estimates for the risk of CD and UC in relation to IL6 were broadly similar to those of our main analysis (Supplementary Tables 4; both multivariable-adjusted P_trend ≤ 0.01). For hsCRP, a statistically significant association remained across quintiles in relation to CD risk (P_trend = 0.026; Supplementary Table 4); however, the association between hsCRP and UC risk was attenuated (P_trend = 0.067; Supplementary Table 4), perhaps as a result of diminished power.

We did not observe statistically significant correlations between IL6 or hsCRP levels and time interval between blood collection and diagnosis of CD or UC (all P ≥ 0.58; Supplementary Figures 1 & 2), and median time interval between blood collection and diagnosis did not differ significantly between quintiles of IL6 or hsCRP in CD or UC cases (all P_difference ≥ 0.058). In analyses stratified by categories of time interval between blood collection and diagnosis, we found no evidence that time interval modified the associations between IL6 or hsCRP and IBD risk (all P_interaction ≥ 0.33; Supplementary Tables 2 & 3 and Supplementary Figures 3 & 4). Furthermore, adjustment of our multivariable models for time elapsed between blood draw and diagnosis, as a continuous variable, did not alter our effect estimates.

Given that some participants reported comorbid conditions that could have potentially influenced inflammatory marker levels (Supplementary Table 5), we performed additional adjustment for presence or absence of any of these chronic conditions. Our effect estimates were not materially altered (data not shown).

We examined whether additional participant or disease characteristics (Supplementary Tables 6 & 7) modified the associations between inflammatory marker levels and IBD risk. There appeared to be differences in the magnitude of the association between IL6 and CD risk according to the presence of documented stricturing disease or a family history of IBD (Supplementary Tables 8 & 9; P_{heterogeneity} ≥ 0.011). However, given multiple comparisons testing and limitations in statistical power, we considered these findings exploratory.

DISCUSSION

In our nested case-control study, conducted within two US nationwide prospective cohorts, we found that pre-diagnostic circulating levels of IL6 and hsCRP were associated with the risk of CD and UC. These observations suggest that subclinical inflammation may be a feature of a pre-disease state in IBD. Preclinical disease phases, defined by positive serologic markers in the absence of symptoms, have been described for other immune-mediated inflammatory disorders such as rheumatoid arthritis (RA)⁵ and systemic lupus erythematosus.²¹ Data suggest that elevated inflammatory markers may be associated with the preclinical phase of RA. In a nested case-control study conducted using pooled data from the Women’s Health Study (WHS) and NHS, plasma soluble tumor necrosis factor receptor II (a surrogate for TNFα) was found to be elevated up to 12 years prior to diagnosis, and was positively associated with disease risk.⁶ In the same analysis, a modest association was observed between plasma IL6 and RA risk, although this was only statistical significant in the NHS.⁶

Although less well characterized, the concept of a preclinical phase in IBD has attracted increasing attention,²² fostered, in part, by studies that have demonstrated the presence of circulating serologic markers several years before IBD diagnosis.^{3, 4, 23} In a seminal study based on Israeli Defense Force recruits, and a more recent, larger analysis of participants in the European Investigation into Nutrition and Cancer (EPIC), positivity for serologic markers, including anti-Saccharomyces cerevisiae antibody (ASCA), perinuclear anti-neutrophil cytoplasmic antibody (pANCA), and antibodies against bacterial components, such as the anti-flagellin antibody, anti-CBir1, appeared to predicted development of CD and UC in low-risk individuals. Preliminary findings from an analysis of 100 US military personnel, the ‘PREDICTS’ study,²³ also suggest that antimicrobial serologic markers are elevated up to six years before CD diagnosis, and may be associated with disease phenotype at presentation.²³ Whether the pathophysiologic processes responsible for the generation of these antibodies in IBD patients could also contribute to elevated inflammatory markers in not known; however, it is interesting that positivity for anti-CBir1 was found to be associated with higher CRP among patients with ankylosing spondylitis who lacked clinically evident IBD.²⁴

Although considerable evidence implicates IL6 signaling in gut mucosal immune homeostasis and IBD pathogenesis,^25–27 it remains uncertain whether IL6 or hsCRP would be specific enough to screen for preclinical IBD. Nonetheless, it is conceivable that biomarkers of subclinical inflammation could be used in conjunction with genetic and serologic markers to help risk stratify individuals who are at increased risk, perhaps on account of a family history of IBD.

Our study possesses several strengths. All IBD cases were confirmed by physician case record review, rather than by self-reporting alone, or diagnostic coding in health information databases. All blood samples predated the diagnosis of IBD and we were able to control for multiple potential confounding exposures using prospectively-collected data.

We recognize that our study also has some limitations. All participants were female nurses, and, although age-specific incidence of IBD in our cohorts is largely similar to that observed in other U.S. populations,⁸ participants tended to be older at the time of blood collection than peak age-specific incidence estimates. These factors may influence the generalizability of our findings. We were mindful that the observed associations could have been influenced by symptoms associated with undiagnosed IBD in cases at the time of blood collection. We attempted to reduce bias from undiagnosed IBD by excluding cases diagnosed within 2 years of blood collection, and found that the effect estimates were generally similar to those of our main analysis. Additionally, we did not observe statistical evidence of interaction between time since blood collection and inflammatory marker levels in their associations with CD or UC risk, nor did additional adjustment for time interval impact on our multivariable effect estimates. Although the longest interval between blood collection and diagnosis was 20 years, the majority of cases were diagnosed within 10 years of blood sampling (upper quartile 10.6 years), and we cannot exclude that the associations between inflammatory markers and IBD risk may diminish over greater time intervals. Finally, the differences in overall median inflammatory marker levels between cases and controls in our study were small; however, differences of similar magnitude have been reported between groups with disparate outcomes in studies of cardiovascular disease.^{28, 29} Moreover, when comparing extreme quintiles of median inflammatory marker levels, where risk of CD or UC was most evident, the differences were more substantial.

In summary, to our knowledge, this is the first report of an association between pre-diagnostic circulating inflammatory markers and risk of IBD. Although the biologic mechanisms underlying these observations require elucidation, our data suggest that subclinical inflammation precedes IBD diagnosis by several years. Our data contribute to our understanding of the natural history of IBD and may ultimately facilitate identification of individuals with preclinical IBD who may benefit from preventive or therapeutic interventions. We believe that validation of our findings in additional prospective series is warranted.

Abbreviations

BMI

body mass index

CD

Crohn’s disease

CI

confidence interval

hsCRP

high-sensitivity C-reactive protein

IBD

inflammatory bowel disease

IL6

interleukin-6

NHS

Nurses’ Health Study

NHSII

Nurses’ Health Study II

MHT

menopausal hormone therapy

NSAID

non-steroidal anti-inflammatory drug

OC

oral contraceptive

OR

odds ratio

TNFα

tumor necrosis factor alpha

UC

ulcerative colitis