Prematurity, Delivery Method, and Infant Feeding Type Are Not Associated with Paediatric-onset Inflammatory Bowel Disease Risk: A Scottish Retrospective Birth Cohort Study

Abstract

Background and Aims

The incidence of paediatric-onset inflammatory bowel disease [PIBD] continues to rise globally. We aimed to determine whether mode of delivery, gestational age at birth, or type of infant feeding contribute to the development of PIBD in a nationwide cohort of Scottish children.

Methods

All children born in Scotland between 1981 and 2017 were identified using linked health administrative data to determine mode of delivery, gestational age at birth, and type of infant feeding. PIBD cases were defined as onset of Crohn’s disease [CD], ulcerative colitis [UC], or IBD-unclassified [IBDU] before age 16 years. Validation was performed within an entire Scottish health board [16% of total population] via individual case-note verification. Hazard ratios [HR] were calculated for each exposure using Cox proportional hazards models.

Results

A study population of 2 013 851 children was identified including 1721 PIBD cases. Validation of 261 PIBD patients coded as CD and/or UC identified 242 [93%] as true positive. Children delivered vaginally did not have an altered risk of developing PIBD compared with those delivered by caesarean section, adjusted HR 0.95 [95% CI 0.84-1.08] [p = 0.46]. Compared with children born at term [≥37 weeks], children born prematurely did not have an altered risk of developing PIBD, i.e., at 24-31 weeks of gestation, HR 0.99 [95% CI 0.57-1.71] [p = 0.97] and at 32-36 weeks of gestation, HR 0.96 [95% CI 0.76-1.20] [p = 0.71]. Compared with children exclusively breastfed at age 6 weeks, children exclusively formula fed did not have an altered risk of developing PIBD: adjusted HR 0.97 [95% CI 0.81-1.15] [p = 0.69].

Conclusions

This population-based study demonstrates no association between mode of delivery, gestational age, or exclusive formula feeding at 6 weeks, and the development of PIBD.

Epidemiology, inflammatory bowel disease, paediatrics

1. Introduction

The incidence of paediatric-onset inflammatory bowel disease [PIBD] is rising globally.¹ Children in Scotland demonstrate the highest incidence of PIBD [IBD diagnosed <16 years of age] in the UK and one of the highest worldwide, with epidemiological data over four decades confirming a 9-fold increase in the incidence of paediatric Crohn’s disease [CD].^2–6 Despite recent advances in studying the genetics of IBD, with over 240 risk loci now identified,⁷ such a significant rise in PIBD incidence over only a few decades cannot be explained by genetics alone and multiple other factors, including environmental aspects, must play an important role.

Gestation and early infancy are periods potentially sensitive to environmental influences,⁸ emphasising the need for population studies where early life and familial exposures can be captured. In particular, from conception to age 2 years [‘the first 1000 days’] appears to be an important window in which genetic and environmental interactions have the greatest impact on future health outcomes.⁹ Many perinatal risk factors have therefore been postulated to contribute to the development of IBD, possibly through alterations in epigenetics and/or the microbiome. However, in general, high interest in perinatal and environmental factors has not been accompanied by studies of high methodological quality. A systematic review on breastfeeding and PIBD risk by Barclay et al.¹⁰ identified issues of recall bias,^11–18 under-powered studies,^{11,13,14,16,17} inappropriate or barely appropriate control populations [not age-, sex-, or population-matched,^14,17 or if sibling-matched, not ensured to be disease-free^15,16], and failure to appropriately define either the exposure [e.g., breastfeeding] or the outcome [IBD].^13,14 Only robust, population based epidemiological data with nationwide data linkage will allow appropriate investigation of perinatal risk factors. Fortunately, Scotland is the only area in the UK which can provide such high-quality, consistent epidemiological data for all geographical areas and for all parts of the IBD severity spectrum.

We aimed to determine whether any of mode of delivery, gestational age at birth, or type of infant feeding at age 6 weeks contribute to the development of PIBD in a nationwide cohort of Scottish children, utilising health administrative data.

2. Methods

2.1. Setting

Scotland covers an area of 30 981 square miles [80 240 km²] and has a population of approximately 5.4 million people, with the vast majority of inhabitants Caucasian, and approximately 915 000 [17%] aged less than 16 years.¹⁹ In Scotland, specialist paediatric gastroenterology, hepatology, and nutrition [PGHAN] services are coordinated through three tertiary academic centres [Glasgow, Edinburgh, and Aberdeen] which act as regional referral hubs, covering all district general hospitals nationwide. Paediatric gastroenterology patients within Scotland are managed within the publicly funded National Health Service [NHS] and all patients are provided with a unique community health index [CHI] number to ensure accurate identification and linkage across health services.²⁰ A dedicated electronic data and innovation research service [eDRIS], part of Public Health Scotland, provides access to high-quality health administrative data with national coverage back to 1981, including specific data-linkage services and a designated Safe Haven within which to analyse these data.

2.2. Study design

In a nationwide, population-based cohort study, we identified all children born in Scotland throughout 1981 to 2017 within the Scottish Morbidity Record maternity inpatient dataset [SMR02]. SMR02 contains both maternal and infant social demographic and clinical data on all deliveries in women discharged from Scottish maternity hospitals, capturing 98.5% of all live births recorded by the National Records of Scotland [NRS].²¹ All multiple births were excluded as these could not be linked to individual outcomes, as was any person with missing data within SMR02. A small number of outliers, whose data were considered potentially inaccurate and might skew the final analysis, were also excluded, including children born to mothers age <15 years or >56 years and births at gestational age <24 weeks or >43 weeks [Figure 1].

Figure 1.

Scottish birth cohort 1981–2017. ^#All children born in Scotland throughout 1981–2017 were identified within the Scottish Morbidity Record maternity inpatient dataset [SMR02]. *Any patient who met more than one exclusion criterion could only be excluded once. SMR02 = Scottish Morbidity Record 02 [maternity inpatient and day case].

Individual-level deterministic data linkage was then applied to the Scottish Morbidity Record general/acute inpatient dataset [SMR01] and child health preschool dataset, using each person’s unique CHI number. PIBD cases were defined as children with any general/acute day case or inpatient discharge record with main diagnosis coded to International Classification of Diseases [ICD] code for Crohn’s disease [ICD-9 555; ICD-10 K50], ulcerative colitis [UC] [ICD-9 556; ICD-10 K51], or indeterminate colitis [IC] [ICD-9 558.9; ICD-10 K52.3] before age 16 years. Age cut-off less than 16 years has been consistently used within Scottish PIBD epidemiology studies to minimize the possibility that adolescent patients who initially present to adult IBD centres might be investigated and monitored by adult physicians without referral to paediatric services. Broadening the search to include all instances of ICD codes for IBD in any billing position was also assessed. This allowed complete accrual and paediatric follow-up for all patients with PIBD born in the 21 years from 1981 to 2001, with then progressively censored follow-up information on those born from January 2002 onwards until end of study on 31 December 2017.

Mode of delivery, gestational age at birth, and type of infant feeding were determined for all children born within the study period where these data were available. Vaginal deliveries [including spontaneous and assisted vaginal deliveries] against caesarean section [both elective and emergency] were obtained from SMR02, as were gestational age at birth data. SMR02 quality assurance assessment demonstrates accuracy of 98% for date of delivery, 97% for caesarean section as mode of delivery, and 92% for estimated gestation.^21,22

Although infant feeding data are also available within SMR02, their accuracy has not been validated. Feeding type is documented as ‘feed first given’ and ‘feed at discharge’ which, given many healthy infants are discharged within 24 h of delivery, is unlikely to represent established feeding type. Final analysis of infant feeding type within our study was therefore performed on child health preschool data, using feeding type documented by a child health nurse at the 6-week infant health check offered to all children. Feeding type is classified as exclusively breastfed, exclusively formula fed, mixed feeding type [both breast milk and formula feeding], and unknown [no feeding type recorded].

2.3. Validation

Validation of ICD coding accuracy was performed within an entire Scottish Health Board. NHS Lothian covers 16% of the national population and includes Edinburgh [Scotland’s second largest city] as well as its surrounding areas. Within this strict geographical area [determined by postcode], individual case-note verification was undertaken for all PIBD coded cases identified through SMR01, augmented as necessary by searching endoscopy and histopathology records of gastro-intestinal [GI] departments [paediatric and/or adult] which performed endoscopy on potential patients with PIBD to 31 December 2017.²³ Although cases for validation were only identified within a specialised non-primary care setting, it would be highly unusual for any patient <16 years of age with PIBD to have follow-up outside a tertiary paediatric gastroenterology centre. Cases were considered true positive if they met revised Porto criteria for PIBD,²⁴ allowing the positive predictive value [PPV] of PIBD coding within Scottish health administrative data to be determined. IBD subtype was further validated within this true-positive NHS Lothian cohort using IBD classes criteria to assess IBD subtype accuracy.²⁵

2.4. Statistical analysis

All analyses were performed using R statistical software [version 3.6.1; R Foundation for Statistical Computing, Vienna, Austria]. Hazard ratios [HR] were calculated for each exposure using Cox proportional hazards models and controlling for potential confounding factors: sex, maternal age at delivery, year of birth, postcode sector, and deprivation. These pre-specified covariates were based on published literature as well as on the expert opinion of the research team. Postcode sector includes approximately 3000 registered addresses, and is required as a covariate due to the recognised north-south gradient of incident patients with PIBD within Scotland and other Northern Hemisphere countries.^5,26 The Scottish Index of Multiple Deprivation [SIMD] is an area-based measure of relative deprivation across seven domains, calculated and weighted for 6976 data-zones, with results graded into quintiles from 1 ‘most deprived’ to 5 ‘least deprived’.²⁷ Fractional polynomials were used to model year of birth with all other continuous variables categorised. IBD subtype grouping was determined for each patient as any ICD code for CD and/or UC to minimise misclassification bias, resulting in a proportion of patients being included within both analyses.

All hypotheses were tested using likelihood-ratio [LR] tests. Dependent [two-way] interactions were explored between the three main exposures. Sensitivity analysis was performed for all risk factors, using previously validated true-positive Scotland-wide PIBD cases identified within prospective Scotland-wide PIBD incidence studies 1981-1995 and 2003-2017.^{2,4–6,28,29} Sensitivity analysis was also performed to determine any difference in outcomes following exclusion of ‘unknown’ feeding type at the 6-week child health check.

This study was approved by the Privacy Advisory Committee, the Caldicott Guardians for NHS Lothian, and the South-East Scotland Research Ethics Committee [REC:18/LO/1062]. The data underlying this article were provided by Public Health Scotland by permission. Data will be shared on request to the corresponding author with permission of Public Health Scotland. Data are reported according to the reporting of studies conducted using observational routinely collected data [RECORD] statement.³⁰

3. Results

3.1. Cohort

An initial cohort of 2 081 601 children born in Scotland 1981-2017 was identified within SMR02. All multiple births [n = 27 696] and any person with missing or inaccurate data within SMR02 [n = 32 417] were excluded, as were outliers with potentially inaccurate data whose results might skew the final analysis [n = 8446]. With a small number of patients excluded within more than one category, this resulted in a final study population of 2 013 851 children [97% of cohort] [Figure 1].

In total, 1721 patients with PIBD [CD/UC only] were identified using data-linkage to SMR01 primary ICD code for inpatient admission. The baseline characteristics of this cohort demonstrate male predominance of 58%, median age at first admission with a relevant PIBD code 12 years, and approximately two-thirds of cases as CD, consistent with our previous nationwide validated Scottish PIBD cohorts [Table 1].^5,6,31

Table 1.

Characteristics of paediatric-onset inflammatory bowel disease cases.

Number of PIBD cases [1981- 2017]	1721
Sex [male]	998/1721 [58%]
Median age at first admission with a relevant PIBD code [interquartile range]	12 years [10–14 years]
IBD subtype	1205 [70%] CD; 614 [36%] UC

Patients with both Crohn’s disease [CD] and ulcerative colitis [UC] codes included within both IBD subtype analyses [n = 98].

PIBD, paediatric-onset inflammatory bowel disease <16 years of age; IBD, inflammatory bowel disease.

3.2. Mode of delivery

Children delivered vaginally did not have an altered risk of developing PIBD when compared with those delivered by caesarean section [LR p-value = 0.46]. Within our cohort, 1 609 198 [80%] children were delivered vaginally, including 1400 patients with PIBD: adjusted HR 0.95 [95% CI 0.84-1.08] [p = 0.46] [Figure 2, Table 2]. No significant differences were noted during IBD subtype analysis [Tables 3 and 4].

Table 2.

Perinatal environmental risk factors and likelihood of developing paediatric-onset inflammatory bowel disease.

				Univariable		Multivariable
Variable	Level	n	Cases	HR univariable [95% CI]	p-value	adjusted HRa [95% CI]	p-value	LR p-value
Delivery	Vaginal	1 609 198	1400	0.87 [0.77-0.98]	0.022	0.95 [0.84-1.08]	0.460	0.462
	Caesarean	404 653	321	1	-	1	-
Gestation	24–31 weeks	15 560	13	1.01 [0.58-1.74]	0.978	0.99 [0.57-1.71]	0.966	0.932
	32–36 weeks	96 915	79	0.98 [0.78-1.23]	0.858	0.96 [0.76-1.20]	0.710
	37–43 weeks	1 901 376	1629	1	-	1	-
Feed type	Formula	631 429	439	0.93 [0.78-1.10]	0.391	0.97 [0.81-1.15]	0.692	0.006
	Mixed	90 043	78	1.36 [1.04-1.76]	0.023	1.33 [1.03-1.74]	0.032
	Unknown	1 023 265	1011	0.62 [0.54-0.73]	<0.001	0.82 [0.67-1.00]	0.054
	Breast	269 114	193	1	-	1	-

HR, hazard ratio; LR, likelihood ratio; CI, confidence interval.

The bold values represents significant p-value <0.05.

^aAdjusted model controlling for sex, maternal age, postcode, deprivation, year of birth.

Table 3.

Perinatal environmental risk factors and likelihood of developing paediatric-onset Crohn’s disease.

				Univariable		Multivariable
Variable	Level	n	Cases	HR univariable [95% CI]	p-value	Adjusted HRa [95% CI]	p-value	LR p-value
Delivery	Vaginal	1 609 198	964	1.01 [0.87-1.16]	0.935	0.87 [0.77-1.02]	0.102	0.106
	Caesarean	404 653	241	1	-	1	-
Gestation	24–31 weeks	15 560	7	0.75 [0.36-1.57]	0.445	0.73 [0.35-1.53]	0.402	0.594
	32–36 weeks	96 915	55	0.94 [0.72-1.24]	0.677	0.93 [0.71-1.22]	0.598
	37–43 weeks	1 901 376	1143	1	-	1	-
Feed type	Formula	631 429	333	1.02 [0.84-1.24]	0.837	1.01 [0.83-1.24]	0.894	0.080
	Mixed	90 043	56	1.20 [0.88-1.64]	0.240	1.32 [0.97-1.80]	0.078
	Unknown	1 023 265	677	1.28 [1.07-1.54]	0.008	0.87 [0.69-1.10]	0.250
	Breast	269 114	139	1	-	1	-

HR, hazard ratio; LR, likelihood ratio; CI, confidence interval.

The bold values represents significant p-value <0.05.

^aAdjusted model controlling for sex, maternal age, postcode, deprivation, year of birth.

Table 4.

Perinatal environmental risk factors and likelihood of developing paediatric-onset ulcerative colitis.

				Univariable		Multivariable
Variable	Level	n	Cases	HR univariable [95% CI]	p-value	Adjusted HRa [95% CI]	p-value	LR p-value
Delivery	Vaginal	1 609 198	511	1.25 [1.01-1.54]	0.041	1.04 [0.84-1.29]	0.735	0.734
	Caesarean	404 653	103	1	-	1	-
Gestation	24–31 weeks	15 560	6	1.26 [0.79-0.57]	0.568	1.36 [0.61-3.05]	0.456	0.774
	32–36 weeks	96 915	28	0.95 [0.65-1.39]	0.779	0.99 [0.68-1.45]	0.957
	37–43 weeks	1 901 376	580	1	-	1	-
Feed type	Formula	631 429	124	0.78 [0.58-1.05]	0.095	0.76 [0.56-1.02]	0.072	0.010
	Mixed	90 043	27	1.19 [0.76-1.85]	0.452	1.33 [0.85-2.07]	0.217
	Unknown	1 023 265	395	1.53 [1.18-1.98]	0.001	0.67 [0.48-0.93]	0.017
	Breast	269 114	68	1	-	1	-

HR, hazard ratio; LR, likelihood ratio; CI, confidence interval.

The bold values represents significant p-value <0.05.

^aAdjusted model controlling for sex, maternal age, postcode, deprivation, year of birth.

Figure 2.

Forest plot of hazard ratios for paediatric-onset inflammatory bowel disease perinatal risk factors. LR, likelihood ratio; HR, hazard ratio.

3.3. Gestational age at birth

Gestational age at delivery was not associated with an altered risk of developing PIBD [LR p-value = 0.93]. Within our cohort 15 560 [0.8%] children were delivered at 24-31 weeks of gestation, including 13 patients with PIBD. Compared with children delivered at term [≥37 weeks], children delivered at 24-31 weeks did not have a different hazard of developing PIBD: adjusted HR 0.99 [95% CI 0.57-1.71] [p = 0.97]; 96 915 [4.7%] children were delivered at 32-36 weeks of gestation, including 79 patients with PIBD. Compared with children delivered at term [>37 weeks], children delivered at 32-37 weeks did not have a different hazard of developing PIBD: adjusted HR 0.96 [95% CI 0.76-1.20] [p = 0.71] [Figure 2]. No significant differences were noted during IBD subtype analysis [Tables 3 and 4] and no difference in outcomes was demonstrated using gestational age as a continuous variable.

3.4. Type of infant feeding

Within our cohort, 631 429 [31%] children, including 439 patients with PIBD, were exclusively formula fed. Compared with the 269 114 [13%] children exclusively breastfed, formula-fed children did not have an altered hazard of developing PIBD: adjusted HR 0.97 [95% CI 0.81-1.15] [p = 0.69]; 90 043 [5%] children, including 78 patients with PIBD, were classed as mixed feeding type and, compared with children exclusively breastfed, this group did demonstrate an increased hazard of developing PIBD: adjusted HR 1.33 [95% CI 1.03-1.74] [p = 0.03]. A total of 1 023 265 [51%] children did not have established feeding type recorded, including 1011 patients with PIBD, and compared with children exclusively breastfed, did not have an altered hazard of developing PIBD: adjusted HR 0.82 [95% CI 0.67-1.00] [p = 0.05] [Figure 2]. Despite no difference in PIBD risk between exclusive breastfeeding or formula feeding at the mandatory 6-week health check, overall infant feeding type was associated with an altered risk of developing PIBD [LR p-value <0.01] [Table 2].

3.5. Interaction terms and sensitivity analysis

Proportionality assumption was tested using scaled Schoenfeld residuals, with no indication of violation from proportional hazards [p = 0.271]. No significant two-way interactions between mode of delivery, type of infant feeding, or gestational age at birth were identified using likelihood ratio testing. Sensitivity analysis was performed for all risk factors, using previously validated true-positive Scotland-wide patients with PIBD identified within prospective Scotland-wide PIBD incidence studies 1981-1995 and 2003-2017 [n = 1144]. Mode of delivery and gestational age at birth did not demonstrate any significant differences from the full cohort. The estimated effect of mixed feeding compared with breastfeeding demonstrated a similar hazard: HR 1.26 [95% CI 0.94-1.70] [p = 0.12]; however, unlike the full model, this was not statistically significant. A second sensitivity analysis to evaluate the impact of feeding on PIBD risk removed all ‘unknown’ cases without 6-week feeding data recorded. The estimated effect of formula feeding, HR 0.95 [95% CI 0.79-1.13] [p = 0.54], and mixed feeding, HR 1.32 [95% CI 1.01-1.72] [p = 0.04], compared with breastfeeding did not change with this analysis.

3.6. Validation cohort

Individual case-note validation was performed on all 298 cases coded for PIBD as the main condition within NHS Lothian, using strict post-code criteria to identify health board boundaries. Among these patients, 1602 coded admissions for CD, 405 coded admissions for UC, and 45 for indeterminate colitis were recorded; 37 patients coded as indeterminate colitis [558.9; K52.3] had never been identified as CD or UC. At validation these patients were all false-positive cases, primarily infants with non-specific enterocolitis and failing to meet revised Porto criteria. Indeterminate colitis was therefore excluded from all further analysis and was not able to be used as a surrogate ICD code to identify inflammatory bowel disease unclassified [IBDU] or patients with IBD in general. Broadening the search to include all ICD codes for IBD in any billing position, increased PIBD case numbers to 355; however, 49 of the extra 57 cases identified were validated to be false-positive and hence full analysis was limited to main diagnostic code only.

Of 261 PIBD patients coded as CD and/or UC, 242 were validated as true-positive: PPV 93% [95% CI 89%-95%], specificity 99% [95% CI 99%-99%]. Of those true-positive patients, 167 met IBD classes criteria for CD with 150 [90%] coded exclusively as CD, 13 [8%] as both CD/UC, and four [2%] exclusively as UC; 65 were classed as UC with 61 [94%] coded exclusively as UC and four [6%] exclusively as CD only. Ten patients were classed as IBDU with 8 [80%] coded exclusively as UC and 2 [20%] exclusively as CD. Overall, IBD subtype PPV was 96% for CD [169 patients ever coded for CD and 163 true-positive] and 71% for UC [86 patients ever coded for UC with 61 true-positive]; 41 false-negative patients were identified within the prospective NHS Lothian IBD database who were not identified through ICD coding, providing overall sensitivity of 85% [95% CI 81%-89%].

4. Discussion

This nationwide population-based study of over 2 million Scottish births between 1981 and 2017 has demonstrated no association of any of mode of delivery, gestational age at birth, or exclusive infant formula feeding at age 6 weeks with the development of PIBD before 16 years of age. A signal suggesting increased PIBD risk in infants with mixed breast and formula feeding has been identified and requires further exploration. Utilisation of Scotland’s extensive health administrative data in a setting of universal public health care has enabled complete ascertainment of patients with IBD diagnosed <16 years of age and accurate linkage across health care episodes. Scotland is unique as the only area in the UK which could feasibly provide such high-quality epidemiological data within a robust birth-cohort design³² for all geographical areas and as continuous high-quality data for nearly four decades.

We have used validation of PIBD coding on multiple levels to overcome the potential for misclassification bias associated with health administrative data. First, we have demonstrated that the baseline characteristics of this cohort are consistent with previous prospective Scottish PIBD studies. Second, patients with PIBD within an entire Scottish health board have been validated via individual case-note verification, demonstrating a PPV of 93%. Finally, a sensitivity analysis was performed including only those nationwide cases validated as true-positive within our previous peer-reviewed Scottish incidence studies 1981-1995 and 2003-2017, totalling 1144 patients with PIBD [66% of the current cohort] and demonstrating no difference in the estimated hazards.

Previous algorithm studies performed within Israel, Sweden, and Canada to predict PIBD cases have used varying combinations of health administrative data [including number of IBD codes, IBD-related medications, IBD-related procedures] to maximise accuracy and avoid repeated validation studies.^33–35 Although our Scottish study has not performed algorithm testing and is based on a single ICD code, we have demonstrated a PPV of 93% which is consistent with established algorithms [Israel PPV 92%; Sweden PPV 93%; Ontario PPV 76%]. The high accuracy of Scottish health administrative data is likely multifactorial, including Scotland’s almost universal public health care with high-quality health administrative data and national coverage extending back to 1981, its unique individual CHI number assigned to every person at birth allowing easy longitudinal data capture, and its small land mass with a well-established paediatric gastroenterology network including three tertiary centres covering all 14 health boards and district general hospitals nationwide, allowing complete data capture. A likely trade-off of this high PPV and low false-positive classification [with lowest limit of 95% confidence interval at 89%] is the greater potential for false-negative misclassification. Although the overall sensitivity of our Scottish study at 85% is comparable to previous algorithm-based studies [Israel 89%, Canada 90%], any false-negative cases remain a potential source of bias. Those patients with IBD but not coded within SMR01 may represent patients recently diagnosed or with milder phenotypes who have not yet experienced complications or required inpatient follow-up.

Potentially key epigenetic and exposomal triggers related to caesarean section and inability to exclusively breastfeed infants are frequently postulated to increase the risk of developing IBD. However, previous studies have demonstrated mixed results, and few have focused on the risks of developing IBD in childhood. A 2014 systematic review and meta-analysis of caesarean section and IBD risk suggested a positive association for CD [OR 1.38, 95% CI 1.12-1.70], but not UC [OR 1.07, 95% CI 0.87-1.32].³⁶ Although large case numbers were included overall, Li et al.³⁶ highlight that meta-analysis was hindered by heterogeneity in study design and population characteristics, lack of statistical control for confounders within some studies,^37–39 and the resulting disregard of the major issue of selection bias. Further, only three studies were of paediatric IBD, and findings were negative for UC⁴⁰ but positive for CD.^40–42 Within Scotland, a previous review of planned caesarean delivery at term and adverse outcomes in childhood health included IBD risk [non-validated diagnoses] as a secondary outcome with no significant difference found relative to vaginal delivery [adjusted HR 1.01, 95%CI 0.97-1.05].²² The null findings for mode of delivery within our robust nationwide study, in the context of previous mixed results, suggests that caesarean delivery is likely a single contributor to PIBD risk within a multi-hit model affecting the epigenome, the gut microbiome, and the immune system.

A 2017 meta-analysis of breastfeeding and IBD risk demonstrated a lower risk of developing both CD [OR 0.71, 95% CI 0.59-0.85] and UC [OR 0.78, 95% CI 0.67-0.91] for any breastfeeding, as well as a dose-dependent association.⁴³ However, an earlier study confined to PIBD risk did not show any significant difference.¹⁰ Within our study, although overall feeding type LR was statistically significant, this was related to the effect of mixed feeding and not to exclusive breast vs formula feeding types which are hypothesised to most significantly effect potential exposomal triggers in the development of PIBD. It is possible that mixed feeding has specific influences on the gut microbiome and immune responses different from those of exclusive formula feeding; however, understanding this relationship when using health administrative data is complicated by a lack of clarity as to what constitutes mixed feeding and the potential confounding reasons and motivations for mixed feeding. Most recently a North American study, using 1119 CD patients from the RISK inception cohort, moved beyond incidence effects to investigate perinatal exposures on CD phenotype and disease severity in children.⁴⁴ Although no effect was found related to mode of delivery, children who were breastfed in infancy were significantly less likely to develop complicated stricturing or penetrating disease at diagnosis or by 3 years compared with those who had never been breastfed [OR 0.65, 95% CI 0.44-96, p = 0.03]. Although disease phenotype could not be analysed within our current study, this association between feeding type and disease severity again highlights that despite our null findings, infant feeding type may play a role in the development of PIBD as part of a multi-hit model rather than as a single causative factor.

Gestational age has previously been suggested as a potential risk factor for the development of IBD, possibly due to the higher likelihood of bacterial vaginosis and altered microbial flora found in preterm deliveries.⁴⁵ A 2007 case-control study did suggest an association between preterm birth and the risk of CD [OR 1.5, 95% CI 1.1-2.0] and UC [OR 1.3, 95% CI 0.9-1.9]; however, neither the exposure nor the diagnosis were validated.⁴⁵ Data obtained from the Nurses’ Health Study, including almost 150 000 women and 552 cases, have also been reported, with no significant association between preterm birth and IBD identified.⁴⁶ More recently, antibiotic usage and infection in the first year of life have shown to be predictive of the development of IBD at any age, with the strongest association before 10 years of age.^47,48 Within our study cohort we were unable to obtain data on antibiotic use, which can only be accessed within Scotland by linking community-based primary care prescriptions and does not include hospital-based treatment of infection in infancy.

We recognise the inherent weaknesses of using health administrative data in clinical and epidemiological research studies, and that the development of a local algorithm would benefit future studies within Scotland to improve the accuracy of identifying patients with PIBD and IBD subtype, as performed elsewhere.^33–35 These weaknesses were minimised through use of a birth-cohort design, case validation, multivariable analysis, and sensitivity analyses as previously described. Within our study, separate analyses for CD and UC necessarily included all patients with an ICD code for each subtype and resulted in a proportion of patients being included within both analyses. Despite approximately 10% of our prevalent cases known to have a diagnosis of IBDU, this subtype could not be separately analysed, with 558.9/K52.3 found not to be a valid PIBD code. Low case numbers within our subanalysis of prematurity, especially for those born extremely premature [n = 13], promoted wide confidence intervals which can only be overcome with international collaborations.

The accuracy of feeding data within our study has not been quantified, and national records from the 6-week check were only established in most NHS Board areas across Scotland by the mid-1990s [range 1991-2010]; hence feeding status at 6 weeks is missing for children born early in our cohort. Any potential bias related to the high missing data rate for type of infant feeding is likely related to time and postcode, with patients missing data more likely to have been born in earlier years or in study areas which delayed commencement of child health database recording. Nevertheless, this is the most robust national data source available on established feeding practices and breastfeeding exclusivity, and potential bias is minimised as part of our multivariable model controlling for year of birth and postcode.

In conclusion, this robust, validated, nationwide, population-based study of over 2 million Scottish births, using health administrative data, has demonstrated no association between mode of delivery, gestational age at birth, exclusive infant formula feeding at age 6 weeks, and the development of PIBD. Although none of these risk factors was independently significant for the development of PIBD, a signal suggestive of increased PIBD risk was noted for mixed ,and it remains possible that all factors could contribute to risk as environmental triggers within a multi-hit model affecting the epigenome, the gut microbiome, and the immune system. Future international collaborations could be used to confirm these data, further explore the effects of mixed feeding on PIBD risk, and investigate other potential perinatal risk factors for both paediatric- and adult-onset IBD.

Acknowledgements

The authors would like to acknowledge the support of the eDRIS Team [Public Health Scotland] for their involvement in obtaining approvals; also provisioning and linking data and the use of the secure analytical platform within the National Safe Haven.

Funding

This work was supported by an Edinburgh Children’s Hospital Charity Research Fellowship award [to C.J.B.] and by funding from Crohn’s and Colitis UK [Edinburgh network] and a joint CORE [Guts UK] and British Society of Paediatric Gastroenterology, Hepatology and Nutrition [BSPGHAN] Development Grant. P.H. is supported by an NHS Research Scotland Career Researcher Fellowship.

Conflict of Interest

D.C.W. has received consultancy fees, speaker fees, and/or travel support from Abbvie, Nestle Health Sciences, and Celltrion.

Author Contributions

C.B. was involved in the conception and design of the study, data collection, and data analysis, wrote the original and subsequent revisions of the manuscript, and approved the final manuscript. R.W. and P.H. were involved in the conception and design of the study, provided critical review of original and subsequent manuscript drafts, and approved the final manuscript. C.S. assisted in data analysis, provided critical review of original and subsequent manuscript drafts, and approved the final manuscript. D.W. serves as guarantor of the article, acquired all funding, conceptualised and supervised this study, was involved in data collection and analysis, provided critical review of original and subsequent manuscript drafts, and approved the final manuscript.