Circulating 25-hydroxyvitamin D concentration can predict bowel resection risk among individuals with inflammatory bowel disease in a longitudinal cohort with 13 years of follow-up

Abstract

Background:

Although the beneficial properties of vitamin D in anti-inflammation and immunity-modulation are promising in the management of inflammatory bowel disease (IBD), data were limited for the critical IBD prognosis. The association between serum vitamin D levels and the risk of bowel resection in individuals with IBD remains largely unknown.

Materials and Methods:

The authors performed a longitudinal cohort study among 5474 individuals with IBD in the UK Biobank. Serum 25-hydroxyvitamin D [25(OH)D] was measured using direct competitive chemiluminescent immunoassay. Bowel resection events were ascertained via national inpatient data. Multivariable-adjusted Cox proportional hazard regression was used to examine the association between serum 25(OH)D and bowel resection risk, presented with hazard ratios (HRs) and 95% CIs. Restricted cubic spline (RCS) was used to evaluate dose-response associations.

Results:

During a mean follow-up of 13.1 years, the authors documented 513 incident bowel resection cases. Compared to participants with vitamin D deficiency, nondeficient participants showed a significantly reduced bowel resection risk in IBD (HR 0.72, 95% CI: 0.59–0.87, P=0.001), Crohn’s disease (CD, HR 0.74, 95% CI: 0.56–0.98, P=0.038), and ulcerative colitis (UC, HR 0.73, 95% CI: 0.57–0.95, P=0.020). When comparing extreme quintiles of 25(OH)D level, participants with IBD showed a 34% reduced risk of bowel resection (95% CI: 11–51%, P=0.007) and participants with UC showed a 46% reduced risk (95% CI: 19–64%, P=0.003), while this association was not significant in CD (HR 0.93, 95% CI: 0.59–1.45, P=0.740). Linear dose-response associations were observed using the RCS curve (all P-nonlinearity>0.05).

Conclusion:

Increased serum level of 25(OH)D is independently associated with reduced bowel resection risk in IBD. This association was significant in UC but may not be stable in CD. Vitamin D deficiency is a risk factor for bowel resection in individuals with IBD, and may be an effective metric in predicting and risk-screening surgical events.

Introduction

Highlights

• Higher vitamin D level is independently associated with reduced bowel resection risk in individuals with inflammatory bowel disease (IBD), and this association remains stable after dealing with season variations, reverse causality, and potential confounders.

• The study identified a linear dose-response relationship between vitamin D level and the bowel resection risk in individuals with IBD.

• Vitamin D deficient status with serum 25-hydroxyvitamin D concentration under 50 nmol/l is a strong risk factor and promising to predict the bowel resection risk in the management of IBD.

• The findings of our study provide encouraging evidence for clinical practice for monitoring and modulating the vitamin D level in IBD.

Inflammatory bowel disease (IBD), including Crohn’s disease (CD) and ulcerative colitis (UC), is a chronic gastrointestinal disease with relapse-remitting course. With currently no cure, rapid growth in global prevalence and high burden of complications is raising challenges in IBD management¹. Bowel resection is a crucial clinical outcome in improving the long-term prognosis of IBD². A meta-analysis showed nearly half of patients with CD and one-fifth of patients with UC would require surgical intervention in the first 10 years after diagnosis³. Previous studies have identified some modifiable environmental factors like air pollution⁴, ultra-processed food⁵, and dietary fibers⁶ that may influence the risk of IBD-related surgery. This suggests the importance of uncovering risk factors to identify individuals with high risk of bowel resection during the long-term management of IBD.

Vitamin D deficiency, commonly referred to as serum 25-hydroxyvitamin D [25(OH)D] concentration under 50 nmol/l⁷, is prevalent among individuals with IBD⁸. Vitamin D can influence intestinal homeostasis and decrease serum C-reactive protein (CRP) levels in individuals with IBD^9,10, suggesting its promising effect against adverse IBD-related outcomes¹¹. However, real-world data was limited for the impact of serum 25(OH)D on the IBD-related surgery risk. A previous study has observed an increased surgery risk among IBD patients with vitamin D deficiency compared to individuals with 25(OH)D ≥75 nmol/l¹². Although encouraging as it is, the study did not report the dose-response associations nor consider the impact of residual confounders such as diets and lifestyle. Therefore, large population-based studies are needed to assess the potential role of 25(OH)D in the bowel resection risk among IBD.

We conducted a longitudinal cohort study to investigate the associations between serum 25(OH)D concentration and vitamin D deficiency with the risk of bowel resection in IBD. We also evaluated dose-response associations to explore whether there is an optimal 25(OH)D concentration for IBD.

Materials and methods

Study population

This study is a longitudinal cohort study leveraged data from the UK Biobank, a large-scale cohort incorporated over 500 000 participants from 22 assessment centers across the UK recruited between 2006 and 2010. We conducted a retrospective analysis based on the prospectively collected data. Statistical analysis was conducted from 10th September 2023 to 15th November 2023. Upon recruitment, each participant signed an informed consent form, finished touchscreen questionnaires, and underwent physical examination and blood sample collection. Health-related outcomes can be followed by linkage to national health-related records.

In the current study, we identified individuals with IBD at baseline from three sources: primary care data, hospital inpatient data, and self-reported diagnostic information. Individuals with IBD were identified in hospital inpatient data using ICD-9 codes 555 and 556 and ICD-10 codes K50 and K51 and in primary care data using specific codes mapping to ICD-10. Self-report data was from the verbal interview by trained staff in the UK Biobank. We included participants with baseline IBD (N=6096), and IBD phenotypes by disease extent and behavior were also collected¹³. We further excluded participants without baseline serum 25(OH)D information (N=619) and with 25(OH)D concentration outside the normal distribution by Rosner’s test (N=3)¹⁴. After exclusions, we included 5474 individuals with IBD (Fig. 1). A sample size of 5000 participants with IBD had 80% statistical power to identify a minimal difference in the effect of 1.29 (0.77) for vitamin D deficiency on bowel resection using the Epi Info software (https://www.cdc.gov/epiinfo/pc.html). This work has been reported in line with the strengthening the reporting of cohort, cross-sectional, and case–control studies in surgery (STROCSS) criteria¹⁵ (Supplemental Digital Content 1, http://links.lww.com/JS9/C267).

Figure 1.

Flowchart of inclusion and exclusion.

Measurement of serum 25(OH)D concentration

The serum 25(OH)D concentration was tested from the blood samples collected at the initial recruitment in assessment centers. The concentration is measured by direct competitive chemiluminescent immunoassay on the platform Diasorin Liaison XL (detection range: 10–375 nmol/l) at the central laboratory of UK Biobank. The variation for internal quality control samples showed coefficients between 5.04 and 6.14%, indicating low variation and high credibility¹⁶. We leveraged data of serum 25(OH)D concentration of baseline collected blood sample at recruitment centers. The mean lag of two serum 25(OH)D measurements was 4.4 years, with the intraclass correlation coefficient (ICC) as 0.77. The widely-recognized standard of vitamin D deficiency is serum 25(OH)D concentration under 50 nmol/l, associated with many unfavorable health conditions⁷. In our study, 73.9% of baseline deficient participants were still in deficient status.

Assessment of outcome

The bowel resection was identified through operation records from hospital inpatient data of the National Health Service available in the UK Biobank. Operation procedures were recorded by the Offices of Population Censuses and Surveys Classification of Interventions and Procedures (OPCS4) codes (Supplementary Table S1, Supplemental Digital Content 2, http://links.lww.com/JS9/C268). We also identified bowel resection from elective and emergency surgery via the hospital admission records. The Audit Commission review of 2009 to 2010 concluded procedural coding OPCS-4 overall accuracy of 90% and diagnostic coding ICD-10 overall accuracy of 89%¹⁷.

The study period was from the baseline recruitment date of UK Biobank to the end of follow-up. The follow-up time (in person-years) was defined as from the recruited time to the first recording time of bowel resection, the end of follow-up (Hospital Episode Statistics for England in 2022-10-31, Scottish Morbidity Record in 2022-08-31, Patient Episode Database for Wales in 2022-03-31), loss to follow-up, or death, whichever occurred first.

Assessment of covariates

We selected the following covariates in our analysis. The sociodemographic factors include sex (male or female), BMI (continuous in kg/m²), ethnic background (white or non-white), age (continuous), Townsend Deprivation Index (TDI, continuous), education level (college and above, high school, and below). TDI was to measure the material deprivation within the population¹⁸. Lifestyle factors include smoking status (never or ever), alcohol consumption (heavy or non-to-moderate), current drinking status (current drinking or not), diets (healthy or unhealthy), and physical activities (adequate or inadequate). Physical activity was collected by a short form of International Physical Activity Questionnaire and adequate levels were defined based on the recommendation from American Heart Association¹⁹. Factors directly associated with serum 25(OH)D concentration include sunlight exposure (measured by time spent outdoors in summer and winter), blood-sampling seasons, vitamin D supplement use (yes or no), and multivitamin supplement use (yes or no). Other variables include inflammatory indicators (serum C-reaction protein [CRP], continuous), baseline comorbidities (measured by Charlson comorbidity index [CCI]), and regular IBD-related medication use of 5-aminosalicylic acid (5-ASA), corticosteroids, and immunosuppressors. Directed acyclic graphs (DAG) were used to identify and control confounders in the statistical model²⁰. Details about processing the covariates are presented in Supplementary Table S2 (Supplemental Digital Content 2, http://links.lww.com/JS9/C268). When covariate information was missing or recorded as ‘unknown’, we used the median values for continuous variables and the most frequently occurred category for categorical variables.

Statistical analysis

Baseline characteristics of participants were shown according to the groups of serum 25(OH)D concentration. Continuous variables were displayed as means (SD) and categorical variables were summarized as numbers (percentages).

In the primary analysis, Kaplan–Meier curve was drawn to evaluate the associations of serum 25(OH)D concentration with risk of bowel resection among individuals with IBD, CD, and UC. Cox proportional hazards model was used to evaluate the hazard ratios (HRs) and 95% CIs. We constructed five multivariable models to perform a stepwise regression: (1) the minimally-adjusted model adjusted for age, sex, and blood-sampling seasons; (2) Model 1 was further adjusted for factors directly associated with serum 25(OH)D level, including multivitamin and vitamin D supplement use; (3) Model 2 was further adjusted for education level and TDI; (4) Model 3 was further adjusted for lifestyles, including smoking status, alcohol consumption, physical activity, BMI, and diets. (5) the fully-adjusted model was further adjusted for IBD-related medication use. The proportional hazard assumption was tested using the weighted residual method. The multicollinearity among the covariates was evaluated by variance inflation factor (VIF) and no violation was observed (all values <2). We also calculated the population attributable fraction (PAF) to assess the possible avoidable bowel resection risk provided that vitamin D deficiency is corrected assuming causal associations exist. We used the restricted cubic spline (RCS) with three knots at the 10th, 50th, and 90th percentiles to visualize the dose-response association between serum 25(OH)D concentration and bowel resection risk in IBD, CD, and UC²¹. The nonlinearity of the associations was tested using the likelihood ratio test. To evaluate the effect of unmeasured residual confounding, we calculated E-values derived from the main models²². To address regression dilution bias, we further used ICC of serum 25(OH)D measurements between measurements to recalibrate the estimates^23,24.

In the secondary analysis, we conducted analysis (1) recategorizing serum 25(OH)D concentration into four groups (≤25, 25 to≤50, 50 to≤75, and >75 nmol/l) with clinical relevance^25,26; (2) investigating associations between serum 25(OH)D and bowel resection risk by admission source; (3) exploring whether there is a difference in associations stratified by baseline bowel resection history; (4) investigating associations among IBD with the different Montreal phenotypes based on the ICD-10 codes¹³ (Supplementary Table S1, Supplemental Digital Content 2, http://links.lww.com/JS9/C268); (5) assessing the associations stratified by median disease duration (≤15, >15 years) of participants when recruited.

In the subgroup analysis, we assessed the association between serum 25(OH)D concentration and the risk of bowel resection stratified by age (<60, ≥60 years), sex, TDI (in tertiles), smoking status, current alcohol drinking status, and the medication use. To verify the robustness of our results, we performed several sensitivity analyses based on the fully-adjusted model. We further: (1) excluded incident outcomes in the first 1 years, 2 years, and 3 years of follow-up; (2) additionally adjusting for CRP level; (3) additionally adjusting for outdoor activity time; (4) additionally adjusting for CCI; (5) stratified by white (n=5248) and black (n=37) participants, as vitamin D level is generally lower in black individuals compared to broader population, but level of bioavailable vitamin D is comparable²⁷; (6) using multiple imputations to reprocess the missing variables; (7) using competing risk model (R package ‘cmprsk’) to calculate the cumulative incidence of bowel resection in 1, 3, 5, and 10 years of follow-up. Considering sampling season is a major factor influencing 25(OH)D level, we used two methods to deal with the seasonal variations. The first is using season-specific cut-points, which recategorized participants with different 25(OH)D concentrations into season-specific quintiles. The second is using May-standardized serum 25(OH)D concentration. We did a regression based on participants’ blood sampling month, using the 25(OH)D concentration in May as reference, and adjusted for age and sex to calculate the May-standardized 25(OH)D concentration²⁸.

All statistical analyses were performed using R 4.3.1, and two-sided P-value <0.05 was considered as statistically significant.

Results

Baseline characteristics

Baseline characteristics of participants were shown according to the quintiles of serum 25(OH)D concentration and vitamin D deficiency status in Table 1 and Supplementary Table S3 (Supplemental Digital Content 2, http://links.lww.com/JS9/C268). We identified 5474 patients with IBD, of which 2828 (51.7%) were female and 3586 (65.5%) with UC. The mean (SD) age was 57.3 (7.9) years. The overall prevalence of vitamin D deficiency was 55.2%. Among all participants, participants with lower 25(OH)D were more likely to have less outdoor time, and less likely to have adequate physical activities and use vitamin supplements.

Table 1.

Baseline characteristics of participants at baseline visit.

		Serum 25-hydroxyvitamin D concentration in quintiles, nmol/l
Characteristics	Overall n=5474	Quintiles 1 (10.0 to ≤29.3), n=1095	Quintiles 2 (>29.3 to≤41.4), n=1100	Quintiles 3 (>41.4 to≤53.1), n=1094	Quintiles 4 (>53.1 to≤66.5), n=1090	Quintiles 5 (>66.5 to≤148.0), n=1095
Age, year (mean SDa)	57.27 (7.93)	56.15 (8.05)	56.59 (7.99)	57.67 (8.00)	57.87 (7.87)	58.06 (7.58)
Sex (%)
Female	2828 (51.7)	549 (50.1)	566 (51.5)	571 (52.2)	573 (52.6)	569 (52.0)
Male	2646 (48.3)	546 (49.9)	534 (48.5)	523 (47.8)	517 (47.4)	526 (48.0)
Townsend deprivation index (mean SD)	−1.25 (3.09)	−0.52 (3.37)	−1.15 (3.16)	−1.28 (3.06)	−1.59 (2.91)	−1.71 (2.78)
Education level (%)
High school and below	3881 (71.7)	730 (67.6)	760 (69.8)	799 (74.1)	786 (72.7)	806 (74.1)
College and above	1535 (28.3)	350 (32.4)	329 (30.2)	280 (25.9)	295 (27.3)	281 (25.9)
Ethic background (%)
White	5248 (96.3)	986 (90.7)	1047 (95.8)	1063 (97.6)	1067 (98.1)	1085 (99.3)
Others	202 (3.7)	101 (9.3)	46 (4.2)	26 (2.4)	21 (1.9)	8 (0.7)
BMI, kg/m2 (mean SD)	27.19 (4.70)	28.30 (5.59)	27.67 (4.84)	27.47 (4.47)	26.74 (4.18)	25.78 (3.81)
Physical activity (%)
Adequatea	3620 (66.1)	640 (58.4)	695 (63.2)	740 (67.6)	757 (69.4)	788 (72.0)
Inadequate	1854 (33.9)	455 (41.6)	405 (36.8)	354 (32.4)	333 (30.6)	307 (28.0)
Smoking status (%)
Never smoke	2557 (46.9)	504 (46.2)	519 (47.4)	510 (46.8)	519 (47.7)	505 (46.4)
Smokers or past-smokers	2895 (53.1)	588 (53.8)	576 (52.6)	579 (53.2)	568 (52.3)	584 (53.6)
Alcohol drinking status (%)
Non-to-moderate drinkersb	4540 (83.1)	936 (85.6)	920 (83.9)	911 (83.4)	897 (82.4)	876 (80.1)
Heavy drinkers	923 (16.9)	158 (14.4)	176 (16.1)	181 (16.6)	191 (17.6)	217 (19.9)
Current drinkers (%)	4915 (89.8)	942 (86.0)	977 (88.8)	985 (90.0)	1000 (91.7)	1011 (92.3)
Diets (%)
Healthyb	3391 (65.1)	572 (56.2)	657 (62.9)	683 (65.7)	730 (69.7)	749 (70.7)
Unhealthy	1819 (34.9)	446 (43.8)	388 (37.1)	357 (34.3)	317 (30.3)	311 (29.3)
Multivitamin supplement users (%)	1224 (22.4)	121 (11.1)	203 (18.5)	241 (22.0)	312 (28.6)	347 (31.7)
Vitamin D supplement users (%)	313 (5.7)	27 (2.5)	43 (3.9)	68 (6.2)	74 (6.8)	101 (9.2)
Sampling season (%)
Spring	1606 (29.3)	463 (42.3)	381 (34.6)	306 (28.0)	255 (23.4)	201 (18.4)
Summer	1443 (26.4)	110 (10.0)	230 (20.9)	323 (29.5)	355 (32.6)	425 (38.8)
Autumn	1299 (23.7)	151 (13.8)	234 (21.3)	269 (24.6)	305 (28.0)	340 (31.1)
Winter	1126 (20.6)	371 (33.9)	255 (23.2)	196 (17.9)	175 (16.1)	129 (11.8)
Outdoor time in summer, h/d (mean SD)	3.84 (2.36)	3.46 (2.18)	3.59 (2.28)	3.87 (2.47)	4.09 (2.43)	4.17 (2.35)
Outdoor time in winter, h/d (mean SD)	1.94 (1.76)	1.79 (1.64)	1.80 (1.66)	1.93 (1.76)	2.07 (1.77)	2.11 (1.91)
Charlson Comorbidity Index, scores	0.44 (1.09)	0.52 (1.27)	0.45 (1.05)	0.42 (1.10)	0.39 (0.98)	0.39 (1.03)
5-Aminosalicylic acid users (%)	1916 (35.0)	362 (33.1)	377 (34.3)	348 (31.8)	408 (37.4)	421 (38.4)
Corticosteroid users (%)	355 (6.5)	60 (5.5)	79 (7.2)	68 (6.2)	62 (5.7)	86 (7.9)
Immunosuppressor users (%)	591 (10.8)	128 (11.7)	110 (10.0)	109 (10.0)	100 (9.2)	144 (13.2)
Baseline bowel resection history (%)
Yes	719 (13.1)	183 (16.7)	154 (14.0)	147 (13.4)	124 (11.4)	111 (10.1)
No	4755 (86.9)	912 (83.3)	946 (86.0)	947 (86.6)	966 (88.6)	984 (89.9)

^aAdequate level of physical activity refers to a. doing moderate or vigorous-intensity activity at least 150 min in metabolic equivalent minutes per week (MET-minutes/week); or b. more than 5 days/week with 10+ min moderate physical activity; c. more than one day/week with 10+ min vigorous physical activity.

^bNon-to-moderate alcohol consumption was defined as 0–14 g/d for women and 0–28 g/d for men according to US dietary guidelines, above which is defined as heavy level.

^cHealthy diet means the intake of at least 4 criteria in the following 7 food groups based on the food frequency questionnaire in the UK Biobank. Details are described in Table S2.

Primary analysis

In the overall 5474 participants, we documented 513 (9.4%) incident bowel resection cases (232 in CD and 281 in UC) over 67 638 person-years, with an average follow-up time of 13.1 years. We observed inverse associations between vitamin D deficiency and bowel resection risk in IBD, CD, and UC (Fig. 2). Compared with vitamin D deficient group, participants in nondeficient group showed significant 28% (95% CI: 13–41%; P=0.001), 27% (95% CI: 3–45%; P=0.032), and 27% (95% CI: 5–44%; P=0.018) reduced bowel resection risk among individuals with IBD, CD, and UC. Compared to individuals with IBD in the lowest quintile, participants in the fifth quintiles of serum 25(OH)D concentration had a 34% (95% CI: 11–51%, P=0.007, P-trend=0.003) reduced risk of bowel resection in the fully-adjusted model (Table 2). A similar association was observed in participants with UC (HR 0.54, 95% CI: 0.36–0.81, P=0.002, P-trend=0.004), but it did not show statistical significance in CD (HR 0.93, 95% CI: 0.59–1.45, P=0.740, P-trend=0.363) (Table 3). The PAF of the vitamin D deficiency on bowel resection risk was 18% in IBD, while comparing the extreme quintiles, this figure was 21%. Stepwise Cox regression showed similar associations after adjusting for multiple covariates (Supplementary Table S4, Supplemental Digital Content 2, http://links.lww.com/JS9/C268).

Figure 2.

Kaplan–Meier curve of IBD, CD, and UC, stratified by two methods. The first is vitamin D deficiency or nondeficiency in (A) IBD, (B) CD, and (C) UC participants. The second method is stratified by quintiles in (D) IBD, (E) CD, and (F) UC participants.

Table 2.

Associations of serum 25-hydroxyvitamin D concentration with the risk of bowel resection in participants with inflammatory bowel disease.

		Crude model		Minimally-adjusted modela		Fully-adjusted modelb		Recalibrated modelc
	Cases/person-years	HR (95% CI)a	P	HR (95% CI)	P	HR (95% CI)	P	HR (95% CI)	P
Per SD		0.89 (0.81–0.97)	0.011	0.89 (0.80–0.97)	0.012	0.87 (0.78–0.96)	0.005	0.82 (0.72–0.94)	0.003
Nondeficiency vs deficiency
≤50	315/36,899	Ref		Ref		Ref		Ref
>50	198/30,739	0.76 (0.63–0.90)	0.002	0.75 (0.62–0.90)	0.002	0.72 (0.60–0.88)	0.001	0.65 (0.51–0.83)	<0.001
Quintiles
Q1	120/13,264	Ref		Ref		Ref		Ref
Q2	112/13,460	0.92 (0.71–1.19)	0.534	0.91 (0.70–1.18)	0.487	0.91 (0.70–1.19)	0.507	0.88 (0.63–1.24)	0.463
Q3	100/13,633	0.81 (0.62–1.06)	0.127	0.81 (0.61–1.06)	0.125	0.81 (0.62–1.07)	0.144	0.75 (0.53–1.08)	0.124
Q4	92/13,635	0.75 (0.57–0.98)	0.037	0.74 (0.56–0.98)	0.035	0.74 (0.55–0.98)	0.039	0.66 (0.45–0.96)	0.031
Q5	89/13,646	0.72 (0.55–0.95)	0.020	0.71 (0.53–0.94)	0.019	0.67 (0.50–0.91)	0.010	0.58 (0.39–0.86)	0.007
P-trend			0.006		0.006		0.004		0.003

^aMinimally-adjusted model is adjusted for age, sex, and blood sampling seasons.

^bFully-adjusted model is adjusted for age, sex, blood sampling seasons, multivitamin supplement use, vitamin D supplement use, Townsend deprivation index, education level, physical activity status, smoking status, current drinking status, BMI, diets, use of 5-ASA, use of corticosteroids, and use of immunosuppressors.

^cBased on the fully-adjusted model after using intraclass correlation coefficient to recalibrate the serum 25-hydroxyvitamin D concentration.

HR, hazard ratio.

Table 3.

Associations of serum 25-hydroxyvitamin D concentration with the risk of bowel resection in Crohn’s disease and ulcerative colitis.

	Crohn’s disease (n=1889)					Ulcerative colitis (n=3585)
	Cases/person-years	HR (95% CI)a	P	HR (95% CI)b	P	Cases/person-years	HR (95% CI)a	P	HR (95% CI)b	P
Per SD		0.93 (0.81–1.08)	0.347	0.91 (0.76–1.09)	0.303		0.83 (0.72–0.95)	0.007	0.78 (0.65–0.93)	0.006
Nondeficiency vs deficiency
≤50	146/12,838	Ref		Ref		169/24,061	Ref		Ref
>50	86/9,970	0.74 (0.56–0.98)	0.038	0.67 (0.46–0.97)	0.032	112/20,769	0.73 (0.57–0.95)	0.020	0.67 (0.48–0.93)	0.018
Quintiles
Q1	48/4,778	Ref		Ref		72/8,485	Ref		Ref
Q2	55/4,570	1.21 (0.82–1.80)	0.340	1.27 (0.76–2.11)	0.366	57/8,891	0.72 (0.51–1.03)	0.074	0.65 (0.41–1.04)	0.071
Q3	51/4,670	1.14 (0.76–1.71)	0.521	1.17 (0.69–1.99)	0.549	49/8,963	0.61 (0.42–0.90)	0.011	0.52 (0.32–0.86)	0.010
Q4	37/4,477	0.88 (0.56–1.38)	0.568	0.83 (0.46–1.48)	0.523	55/9,157	0.65 (0.44–0.95)	0.025	0.56 (0.34–0.92)	0.022
Q5	41/4,312	0.93 (0.59–1.45)	0.740	0.88 (0.49–1.58)	0.681	48/9,334	0.54 (0.36–0.81)	0.003	0.44 (0.26–0.75)	0.002
P-trend			0.363		0.321			0.004		0.004

^aBased on the fully-adjusted model adjusted for age, sex, blood sampling seasons, multivitamin supplement use, vitamin D supplement use, Townsend deprivation index, education level, physical activity status, smoking status, current drinking status, BMI, diets, use of 5-ASA, use of corticosteroids, and use of immunosuppressors.

^bRecalibrated multivariable estimates accounting for dilution bias using the intraclass correlation coefficients calculated in the subsample of participants with repeat measurements of serum 25(OH)D concentration.

HR, hazard ratio.

Dose-response associations between serum 25(OH)D concentration and risk of bowel resection were observed in IBD, CD, and UC by the RCS curves, and we did not observe nonlinear association (Fig. 3, P-nonlinearity 0.646, 0.745, and 0.330). Recalibration for regression dilution led to stronger estimates of the associations. The E-values comparing extreme quintiles were 2.4 for IBD and 3.2 for UC.

Figure 3.

Associations between the serum 25(OH)D concentration and risk of bowel resection in individuals with (A) IBD participants; (B) CD participants; (C) UC participants using restricted cubic curves. The vertical axis represents the risk of bowel resection risk based on the fully-adjusted model. The horizontal axis represents serum 25(OH)D concentration. The solid line in blue represents HRs, and the light-blue shading represents 95% CIs.

Secondary analysis

In the secondary analysis, when categorizing participants with serum 25(OH)D concentrations ranging from ≤25, 25 to≤50, 50 to≤75, and >75 nmol/l, we found a similar association between serum 25(OH)D and risk of bowel resection. Compared to the lowest group, HRs for the second to fourth groups were 0.93 (95% CI: 0.72–1.20; P=0.578), 0.69 (95% CI: 0.52–0.92; P=0.012), and 0.65 (95% CI: 0.45–0.94; P=0.023) in participants with IBD (Supplementary Table S5, Supplemental Digital Content 2, http://links.lww.com/JS9/C268). We did not find statistically significant differences between participants with bowel resection history at baseline and those without in the association between 25(OH)D and bowel resection risk (P-interaction >0.05, Supplementary Table S6, Supplemental Digital Content 2, http://links.lww.com/JS9/C268).

Regarding bowel resection types by admission source, we observed a significant inverse association between serum 25(OH)D concentration and emergency surgery risk in participants with IBD (P-trend=0.046). The association of 25(OH)D with emergency surgery was significant in UC when comparing extreme quintiles but not significant in CD (Supplementary Table S7, Supplemental Digital Content 2, http://links.lww.com/JS9/C268). We did not observe significant associations in subgroups with clear Montreal phenotypes (Supplementary Table S8–S10, Supplemental Digital Content 2, http://links.lww.com/JS9/C268). The association remained consistent when stratified by different disease duration (Supplementary Table S11, Supplemental Digital Content 2, http://links.lww.com/JS9/C268).

Subgroup and sensitivity analysis

The association between serum 25(OH)D concentration and the risk of bowel resection was not influenced by stratifications of age, sex, TDI, smoking status, current drinking status, and the use of IBD-related medication (all P-interaction >0.05, Supplementary Table S12–S15, Supplemental Digital Content 2, http://links.lww.com/JS9/C268).

Significant inverse associations between 25(OH)D concentration and bowel resection risk were observed when excluding bowel resection in the first 1 year, 2 years, or 3 years of follow-up, and after refilling missing values with multiple imputations (P-trend=0.003) (Supplementary Table S16, Supplemental Digital Content 2, http://links.lww.com/JS9/C268). When additionally adjusted for CRP concentration, CCI scores, or outdoor activity time, and dealing with seasonal variations in 25(OH)D level, the results were consistent (Supplementary Table S17–S19, Supplemental Digital Content 2, http://links.lww.com/JS9/C268). The cumulative incidence of bowel resection in participants with vitamin D nondeficiency versus deficiency were 0.8 versus 1.0%, 2.4 versus 3.6%, 4.0 versus 5.5%, 6.6 versus 9.1% at the 1, 3, 5, and 10 years of follow-up (Supplementary Table S20, Supplemental Digital Content 2, http://links.lww.com/JS9/C268).

Considering the races may result in variations in 25(OH)D concentration, we stratified participants into white and black people. The results were similar to the primary findings in white participants (n=5248, P-trend=0.001) (Supplementary Table S17, Supplemental Digital Content 2, http://links.lww.com/JS9/C268). Only 37 participants reporting with black ethnic background (17 with CD and 20 with UC). Among these participants, six participants had baseline bowel resection and five had incident bowel resection. Interestingly, the five black participants who had incident bowel resection all fell in the lowest quintile of 25(OH)D concentration (≤29.3 nmol/l).

Discussions

In this longitudinal cohort study, we presented that higher serum 25(OH)D concentration is independently associated with a decreased risk of bowel resection in IBD, especially in UC but may not be stable in CD. Compared to vitamin D deficient group, participants with 25(OH)D concentration over 50 nmol/l exhibited significant 28, 27, and 27% reduced bowel resection risk in IBD, CD, and UC. When comparing the extreme quintiles, the associations were statistically significant in IBD and UC, but not in CD, indicating this association in CD needed to be cautiously interpreted and further investigated. RCS model showed linear associations between serum 25(OH)D concentration and bowel resection risk, which presented that serum 25(OH)D concentration as well as vitamin D deficiency status may become a promising indicator of bowel resection risk in IBD. When comparing participants with vitamin D deficiency, the PAF was 18% in IBD, indicating that 18% of incident bowel resection events can be avoided by modulating vitamin D deficiency if causal associations exist.

The inverse association between serum 25(OH)D concentration and bowel resection risk in individuals with IBD is biologically plausible. Animal models deficient in vitamin D were more susceptible to colitis, and vitamin D treatment can ameliorate the detrimental effect²⁹. Vitamin D can inhibit Th1/Th17 cell differentiation in the intestinal mucosa, which further reduces intestinal inflammation³⁰. For clinical study, vitamin D supplementation is associated with improvement in disease activity scores of IBD patients³¹, and it may reduce intestinal inflammation in patients with active UC³². Moreover, a decline in serum 25(OH)D was associated with increased risk of clinical relapse in patients with UC, which usually posed higher risk of hospitalization and surgical interventions¹¹.

The findings of our study were in line with previous knowledge, that vitamin deficiency is linked to rising mortality and poor prognosis of diseases^33,34. For the long-term management of IBD, previous research reported nonlinear association between serum 25(OH)D concentration and all-cause mortality³⁵, whereas evidence for bowel resection risk was still limited. A study from two centers using multivariable-adjusted Logistic regression models indicated low plasma vitamin D level is associated with increased risk of related surgery in IBD¹². For studies with long duration of follow-up, the Cox model is more accurate than the Logistic regression model in evaluating the risk of adverse events³⁶. Our findings presented a linear dose-response association between serum 25(OH)D concentration and bowel resection risk in IBD during long-term follow-up, which adds evidence to the current research.

However, controversy exists between vitamin D levels and surgical-related outcomes, as it is unclear whether vitamin D deficiency is a definitive cause for incident surgical events. Previous studies demonstrated the positive effect of vitamin D therapy on IBD, including the improvement of disease activity scores and inflammatory levels^31,37. A recent systematic review of 22 randomized trials did not observe the treatment effect of vitamin D in clinical relapse or adverse events of IBD³⁸. A meta-analysis of 17 trials with a median follow-up of 6 months found oral vitamin D supplementation can reduce CRP levels but not disease activity index and relapse rate¹⁰. For the negative results, these meta-analyses concluded the limited sample size of clinical trials and the low certainty of the evidence as important reasons. The larger sample size and a longer follow-up period of our observational study overcome the biases of sample selection and 25(OH)D concentration measurement¹⁰. Also, given the heterogeneity of vitamin D doses in meta-analysis, our dose-response curves can provide reference to the clinical practice. Due to the low evidence grade of observational studies, it is hard to conclude the causal effect of vitamin D deficiency on adverse surgical outcomes in IBD. Further studies with higher grades of evidence such as randomized controlled trials and Mendelian randomization analysis are needed to confirm the potential causality.

Presently, recommended vitamin D levels for healthy individuals vary between 50 and 100 nmol/l³⁹, while data is limited for ideal serum vitamin D levels in individuals with IBD^11,40, making it challenging to design personalized vitamin D recommendations to prevent adverse clinical outcomes³⁸. As deficiency criteria are derived from the overall population, our dose-response association findings add evidence to this field. A previous study suggests 25(OH)D levels of 44–78 nmol/l can prevent premature death in IBD³⁵. Combined with the linear dose-response associations with bowel resection events, maintaining serum 25(OH)D levels above 50 nmol/l may still be the optimal recommendation for IBD, whereas higher target concentrations require additional evidence.

Regarding the type of surgical events, individuals in the highest quintile of 25(OH)D concentrations have significant 49% lower risk of emergency surgery. Compared with vitamin D deficiency group, increased 25(OH)D is associated with numerically higher reduced risk in emergency surgery than elective surgery, which may indicate the stronger protective effect of vitamin D level on acute intestine impairment than preexisting obstructive symptoms. Additionally, as we stratified participants by white and black people, all black people with incident bowel resection fell in the lowest quintile of 25(OH)D. This may imply the associations between vitamin D deficiency and risk of surgery in black participants that are consistent with the primary finding. Due to limited sample size of black participants, further investigations are still needed.

Our study also showed the single measurement of 25(OH)D under strict quality control⁴¹ may serve as potential indicators with cost-effective benefits for long-term management. Vitamin D deficiency status may persist without intervention. Between two measurements with a mean lag of 4.4 years, the ICC of serum 25(OH)D concentration was 0.77 and 73.9% of baseline deficient participants were still in deficient status. Population-based studies have shown a high correlation among different measurements, which indicates the 25(OH)D would remain in a certain range for each individual regardless of time span⁴². We observed consistent differences in the cumulative incidence of bowel resection between vitamin D deficient and nondeficient groups. This implies that, as a strong risk factor, the initial level of serum 25(OH)D has predictive potential for immediate, intermediate, and long-term surgical risk. Testing vitamin D levels is an exploratory tool to consider for surgical risk-screening in patients with IBD at the time of initial diagnosis or exacerbation.

Our findings further defined the effect of 25(OH)D on bowel resection independent of multiple sociodemographic factors, lifestyles, and other related confounders. We tested the VIF to avoid high collinearity and used the DAG to visualize the possible causal pathway (Supplementary Figure S, Supplemental Digital Content 2, http://links.lww.com/JS9/C2681). Results from models with stepwise adjustment were also demonstrated. The above strategies justify the confounder control in our study, while residual confounders may still exist. For instance, the accelerometer activity may be a more accurate measurement of physical activity compared with questionnaires⁴³. To deal with the potential effects of residual confounders, we tested the E-value in addition to adjusting multiple covariates. We quantify the minimal strength (HR=2.4) of the assumed uncontrolled confounder to make the observed associations null among individuals with IBD. Therefore, our findings are suggestive of clinical practice and can be a starting point for monitoring 25(OH)D levels in individuals with IBD.

To our knowledge, this study is the largest cohort study investigating the associations between serum 25(OH)D concentration and the bowel resection risk in IBD. The strengths of our study include a large sample size, well-administered database, and an extended follow-up duration averaging 13.1 years. Nevertheless, there are some limitations. First, as an observational study, our findings did not establish causal relationship between serum 25(OH)D concentration and the bowel resection risk in IBD. To minimize reverse causation bias, we excluded incident bowel resection in the first 1, 2, and 3 years of follow-up and got similar findings with the main analysis. Second, compared with the overall population, the observed incidence (9.4%) of bowel resection was relatively low. This could be attributed to the electronic medical record algorithms not capturing all events. However, this lowered incidence may be attributed to the mean age with 57.3. Research showed the IBD-related surgery risk was 18.5% in elderly IBD individuals⁴⁴. Given that 13.1% participants had a resection history, the surgery events identified in our study are comparable. Due to the prospective design of UK Biobank, bowel resection cases potentially undetected should be nondifferential and consequently attenuated the associations. Third, given that our study is based on a large cohort with over 95% white people aged 40–69 years, the generalization of the results to other populations should be interpreted with caution. Finally, participants recruited in our study were not based on the clinical or inpatient settings. This makes it challenging to capture complex clinical processes prior to receiving bowel resection during the long-term follow-up while providing data closely in real-world settings. However, we have considered several variables related to clinical characteristics/process. Further studies are needed to clarify the clinical trajectories and prognosis endpoints of individuals with IBD (especially in important clinical subgroups) according to different vitamin D levels.

Conclusion

In summary, our study suggests that increased level of serum 25(OH)D is independently associated with reduced risk of bowel resection in IBD, which presents as an inverse dose-response linear association. This association was significant in UC but may not be stable in CD. Vitamin D deficiency is a risk factor and may be an effective predictor of the risk of bowel resection in the management of IBD. Future randomized control studies are needed to confirm the effect and threshold of serum 25(OH)D concentration to reduce the risk of bowel resection in IBD.

Ethical approval and consent to participate

The overall ethical approval for the UK Biobank (REC reference: 21/NW/0157) was provided by the North West-Haydock Research Ethics Committee, U.K. in June 2021. Ethical exemptions for this study were obtained from the IRB of The Third Xiangya Hospital of Central South University, China and the Medical Ethics Committee of Zhejiang Chinese Medical University, China to analyse de-identified data.

Consent for publication

All data in our study were anonymous. Written informed consent was obtained from the patients when they were recruited in the UK Biobank, and the information is available in the UK Biobank (https://www.ukbiobank.ac.uk/). For this study, an ethical exemption was obtained by the IRB of the School of Basic Medical Sciences, Zhejiang Chinese Medical University, and additional written informed consent was not needed.

Source of funding

This work was supported by National Natural Science Foundation of China U23A20492, National Natural Science Foundation of China 8217033803, and National Natural Science Foundation of China 82004353.

Author contribution

L.D. (Conceptualization: equal; Design: leading; Methodology: leading; Formal analysis: leading; Writing – original draft: leading; Writing – review & editing: equal). S.W. (Design: equal; Formal analysis: leading; Visualization: leading; Writing – original draft: equal; Writing – review & editing: equal). X.C. (Visualization: leading; Writing – original draft: equal; Writing – review & editing: equal). Y.S. (Formal analysis: supporting; Methodology: supporting; Investigation: equal; Writing – review & editing: supporting). T.F. (Formal analysis: supporting; Methodology: supporting; Investigation: equal; Writing – review & editing: supporting). M.D. (Methodology: supporting; Resources: supporting; Writing – review & editing: supporting). J.C. (Conceptualization: leading; Design: leading; Methodology: equal; Formal analysis: equal; Writing – review & editing: supporting). Z.D. (Methodology: equal; Investigation: equal; Resources: equal; Funding acquisition: leading; Writing – review & editing: supporting). X.W. (Conceptualization: leading; Funding acquisition: leading; Project administration: leading; Writing – review & editing: supporting).

Conflicts of interest disclosure

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Research registration unique identifying number (UIN)

1. Name of the registry: not applicable.

2. Unique identifying number or registration ID: not applicable.

3. Hyperlink to your specific registration (must be publicly accessible and will be checked): not applicable.

Guarantor

Jie Chen, Prof. Zhongyan Du, and Prof. Xiaoyan Wang.

Data availability statement

The data sets analyzed during the current study are available in a public, open-access repository (https://www.ukbiobank.ac.uk/).

Provenance and peer review

Not commissioned, externally peer-reviewed.