Is Vaccination Against COVID-19 Associated With Inflammatory Bowel Disease Flare? Self-Controlled Case Series Analysis Using the UK CPRD

INTRODUCTION: To investigate the association between vaccination against coronavirus disease 2019 (COVID-19) and inflammatory bowel disease (IBD) flare. METHODS: Patients with IBD vaccinated against COVID-19 who consulted for disease flare between December 1, 2020, and December 31, 2021, were ascertained from the Clinical Practice Research Datalink. IBD flares were identified using consultation and corticosteroid prescription records. Vaccinations were identified using product codes and vaccination dates. The study period was partitioned into vaccine-exposed (vaccination date and 21 days immediately after), prevaccination (7 days immediately before vaccination), and the remaining vaccine-unexposed periods. Participants contributed data with multiple vaccinations and IBD flares. Season-adjusted incidence rate ratios (aIRR) and 95% confidence intervals (CI) were calculated using self-controlled case series analysis. RESULTS: Data for 1911 cases with IBD were included; 52% of them were female, and their mean age was 49 years. Approximately 63% of participants had ulcerative colitis (UC). COVID-19 vaccination was not associated with increased IBD flares in the vaccine-exposed period when all vaccinations were considered (aIRR [95% CI] 0.89 [0.77–1.02], 0.79 [0.66–0.95], and 1.00 [0.79–1.27] in IBD overall, UC, and Crohn's disease, respectively). Analyses stratified to include only first, second, or third COVID-19 vaccinations found no significant association between vaccination and IBD flares in the vaccine-exposed period (aIRR [95% CI] 0.87 [0.71–1.06], 0.93 [0.75–1.15], and 0.86 [0.63–1.17], respectively). Similarly, stratification by COVID-19 before vaccination and by vaccination with vectored DNA or messenger RNA vaccine did not reveal an increased risk of flare in any of these subgroups. DISCUSSION: Vaccination against COVID-19 was not associated with IBD flares regardless of prior COVID-19 infection and whether messenger RNA or DNA vaccines were used.


INTRODUCTION
It has been recognized for some years that patients with inflammatory bowel disease (IBD) receiving immunosuppressive medication should be advised to receive vaccination against a number of infections (1,2). These commonly include influenza and pneumococcus, but there are variations around the world (3).
Despite this, the uptake of vaccination has often been suboptimal (4)(5)(6)(7). With rates of vaccination at 60%-80% being reported, seasonal flu vaccine is relatively well accepted, but 1 UK hospital reported only 32.5% vaccinated during the H1N1 pandemic of 2009 (4), and less than 50% have been recorded as receiving Hepatitis B or pneumococcal vaccines, as recommended (7). A variety of reasons for this have been proposed and have included worries about safety and the risk of IBD flare demonstrated by surveys of patients (8). This has led to interest in the safety of vaccinations in people with IBD, with specific reference to their effect on disease activity (9).
The recognition that some patients with IBD, including those treated with steroids for active flares, are at particular risk of coronavirus disease 2019 (COVID- 19), caused inevitable concern (10,11). The rapid development of vaccinations offered the prospect of alleviating this risk. However, given the previous experience of vaccination in IBD, it is unsurprising that although most patients were willing to accept vaccination, it also caused concerns about safety and efficacy among others (12)(13)(14). These may have been added to by limited reports of possible exacerbation of IBD post-vaccination (15,16).
A small number of subsequent studies have now shown evidence, specifically in IBD, of immunogenicity (17,18), clinical efficacy, and safety (19) of COVID-19 vaccinations. The review by Bhurwal et al (19) did not evaluate the association between vaccination against COVID-19 and IBD flares. These studies have, however, mainly been case series or surveys from individual or small groups of centers and/or have related to specific vaccines. The risk of IBD flare with vaccination against COVID-19 originally suggested (15,16) has not to date been conclusively excluded in a large representative population. We have therefore conducted a study in the Clinical Practice Research Datalink (CPRD) population, which is representative of the overall UK population to clarify whether COVID-19 vaccination is associated with an increased risk of IBD flare.

Data
Source data were extracted from CPRD Aurum, a longitudinal anonymized electronic database of health records from 19 million patients registered with 738 general practices that dates back to 1995 (20). It includes information on demographic details, lifestyle factors, diagnoses, results of investigations, consultations, primary care prescription, and vaccinations. Diagnostic and prescription data are recorded using medical codes (a combination of Read 2, SNOMED, and local EMIS codes) and product codes, respectively. Data for vaccination against COVID-19, including date of vaccination and vaccine brand, are provided by National Health Service Digital. COVID-19 is defined using primary care diagnosis, serology, or polymerase chain reaction result.

Approvals
CPRD Research Data Governance (reference: 21_000670). This study used anonymized patient health records from the CPRD and did not require individual participant consent.

Study design
This is a self-controlled case series analysis. This method quantifies the association between exposure and outcome using data from exposed participants who developed an outcome and is extensively used in vaccine safety studies (21,22). It has the advantage of implicitly controlling for all between-person confounding, by conditioning on the time of events and analyzing when exposures occur in relation to this within each individual.

Population
Adults aged 18 years or older with $1 primary care consultation for IBD, $1 prescription for 5-amino salicylate drugs (mesalazine, balsalazide, and olsalazine), or any conventional immunosuppressing drugs, that is, azathioprine, 5-mercaptopurine, methotrexate, mycophenolate mofetil, ciclosporin, tacrolimus, and sirolimus before December 1, 2020, were eligible to be included in the study, provided they also received $1 vaccination against COVID-19 and consulted their primary care provider for $1 IBD flare in the study period. A primary care diagnosis of IBD recorded in the CPRD has been validated to have a 92% positive predictive value for probable, or highly probably, diagnosis of IBD (23). Codes are provided in supplemental data, as Appendix 1 (see Supplementary Digital Content 1, http:// links.lww.com/AJG/C879).

Study period
The study period was from December 1, 2020, to December 31, 2021. Follow-up was censored if death, emigration from participating general practice, or last collection of data from general practice occurred before December 31, 2021.

Exposure
Vaccination against COVID-19 was the exposure of interest and was defined using product codes for vaccines and vaccination dates. Product codes were used to define the vaccine type and brand, specifically vectored DNA (AZD1222) and messenger RNA (mRNA) (mRNA-1273, BNT1262b2).

Outcome
IBD flare was the outcome of interest. It was defined as primary care consultation with a diagnostic coding for IBD, diarrhea, abdominal pain, or rectal bleeding entered on that date and accompanied with a corticosteroid prescription on the same or the subsequent date. Date of primary care consultation for IBD flares was used to define the outcome date, and participants could contribute data with multiple flares.

Exposed and unexposed periods
The study period was divided into a prevaccination period that immediately preceded vaccination, a vaccine-exposed period that immediately followed vaccination, and the remaining vaccineunexposed baseline period ( Figure 1). The vaccine-exposed period was defined as the date of vaccination and the 21 days immediately after the date of vaccination because it takes approximately 2-3 weeks for primary COVID-19 immunization to induce an immunological response (24,25). We hypothesized that this period of immune reconstitution was most likely to be associated with an increased disease activity. Because patients with disease flare or acute illnesses may delay vaccination, the 7 days immediately preceding vaccination was considered separate from the vaccine-unexposed baseline period to minimize potential confounding. The vaccine-unexposed baseline period comprised the remaining follow-up time postcohort entry and before cohort exit. As illustrated in Figure 1, the vaccine-unexposed baseline period comprised follow-up time either before or after vaccination against COVID-19.
The study started on December 1, 2020, 1 week before the first COVID-19 vaccine was administered outside of trial setting in the UK to allow each potential vaccinated participant to have 7 days prevaccination period.

Statistical analyses
A multinomial Poisson model conditioned on the number of events and adjusted for the 4 seasons as categories defined in line with the Meteorological Office description (26) was fitted to calculate the adjusted incidence rate ratios (aIRR) and 95% confidence interval (CI) for association between vaccination and IBD flares. The analyses were adjusted for season because vaccination against COVID-19 predominantly occurred in the winter and spring months in the UK and there is a seasonal pattern to UC (27,28). The 7 days before and 21 days after COVID-19 vaccination were the prevaccination and vaccine-exposed periods, respectively. The remaining study period was considered as the vaccine-unexposed baseline period. A sensitivity analysis to account for bias due to late presentation of IBD flares considered 6-week post-vaccination exposed period. Stratified analysis considered 1st, 2nd, or 3rd vaccine doses and IBD type in the entire data set. Stratified analysis according to vaccine type (AZD1222 vs BNT1262b2) and prior COVID-19 considered the first vaccination against COVID-19. P values ,0.05 (2-sided) were considered as statistically significant. Data analyses were conducted using Stata v.16.

RESULTS
Data for 1,911 cases with IBD were included ( Figure 2). Most of them were female (52%), and their mean (SD) age was 49 (17) years. A total of 1,209 (63%) had UC, 604 (32%) had Crohn's disease, and 98 (5%) had IBD without any specific coding for subtype. A total of 754 (40%), 1,132 (59%), and 23 (1%) participants received BNT162b2, AZD1222, and mRNA-1273 vaccines, respectively, as their first vaccine dose. One hundred thirty-four (7%) participants had COVID-19 before their first vaccine dose. A total of 1,005 (53%), 809 (42%), and 97 (5%) participants received 3, 2, and 1 vaccination doses against COVID-19, respectively, in the study period. A total of 1,754 (91.8%), 137 (7.2%), and 20 (1%) participants had 1, 2, and more than 2 IBD flares in the study period. Seventy-four (3.9%) participants did not contribute data for the entire follow-up period due to death (n 5 16 [0.8%]) or transfer out of general practice surgery (n 5 58 [3%]). One hundred one of the 108 (93.5%) patients who had an IBD flare in the 3-week vaccineexposed period immediately after their first vaccination against Figure 1. Schematic representation of self-controlled case series analysis periods. The vaccine unexposed baseline, prevaccination, and vaccine-exposed periods are shaded speckled, solid, and lined, respectively. Vaccinations against coronavirus disease 2019 are represented by solid arrows. Unfilled arrows below indicate the start and end of the study period. Not all participants received all 3 vaccinations. Follow-up began on the latest of current registration date in general practice surgery or December 1, 2020, and was censored on the earliest of December 31, 2021, death date, transfer out date, or date of last data collection from the general practice surgery. Figure 2. Study population selection criteria for SCCS analysis. *Rationale for exclusion: Primary vaccination against COVID-19 in the UK were administered $4 weeks apart. Thus, vaccination dates ,4 week apart may potentially be incorrect entries. In addition, vaccine-exposed period was 3 weeks after vaccination in this study. If the vaccinations are administered less than 4 weeks apart, the prevaccination period of the second vaccine truncates the postvaccine exposed period of the earlier vaccine, thus potentially misclassifying outcomes. COVID-19, coronavirus disease 2019; GP, general practitioner; IBD, inflammatory bowel disease; SARS-CoV-2, severe acute respiratory syndrome coronavirus 2; SCCS, self-controlled case series.
The American Journal of GASTROENTEROLOGY VOLUME 118 | AUGUST 2023 www.amjgastro.com COVID-19 received another dose of COVID-19 vaccine. Similarly, 1,713 of the 1,803 (95%) patients who did not have an IBD flare in the 3-week vaccine exposed period after their first vaccination against COVID-19 received another dose of COVID-19 vaccine.
Vaccinations against COVID-19 were not associated with IBD flares in the 21-day vaccine-exposed period when all vaccinations were analyzed together in a single data set or separately ( Table 1). The aIRR (95% CI) for flare in the vaccine-exposed period in those with ulcerative colitis (UC) was significantly reduced at 0.79 (0.66-0.95), whereas in Crohn's disease, it was unaltered (aIRR 1.00 [0.79-1.27]) ( Table 2). Data for 98 patients who could only be classified as IBD were excluded from this analysis.

DISCUSSION
Our study has demonstrated in a population representative of patients with IBD in the UK vaccinated with the COVID-19 vaccines commonly in use in the UK that COVID-19 vaccination was not associated with an increase in flares of IBD. This remained true in subgroups of the data defined by the vaccine technology received, the type of IBD (Crohn's disease or UC), and the presence or absence of prior COVID-19. It is similarly true no matter which of 3 doses of the vaccine are studied. In fact, for patients with UC, the rate of flare was significantly reduced during the 3 weeks after vaccination.
Strengths of our study are its power, the generalizability of its results, and the confidence we are able to have that our results are not influenced by confounding factors that might affect the choice to be vaccinated because we used the selfcontrolled case series methodology that is widely used in vaccine safety studies (29). The power of the study is derived from the large base population of CPRD from which it is drawn, and its importance in this instance is that it permits our relative risk estimates to be quite precise and so to confidently exclude any large increase in flares post-vaccination. To illustrate this overall, our adjusted incidence rate ratio for IBD flares was 0.89, and our 95% CI of 0.77-1.02 allows us to state that our data are unlikely to have arisen in a population where there was an excess of flares of more than 2% above baseline after vaccination. Our confidence in generalizability of our result to The American Journal of GASTROENTEROLOGY patients with IBD in the UK likewise is derived from our data source because CPRD is representative of the over 98% of the UK population registered with a general practitioner (20,30), and we included all adults in this population who received COVID-19 vaccination and experienced a coded IBD flare within the study period.
Finally, our use of a self-controlled case series design ensures that non time-dependent between-person confounding was excluded because participants were compared only with themselves at different time points (21). Because all patients who had received both vaccination and experienced an IBD flare were included in our study, there was no selection bias. As with all studies, though, ours has limitations. First, we have been obliged due to a lack of availability of linked inpatient data to adjust our flare definition compared with that which we have previously used in a manner, which effectively excludes flares presenting first to hospital for admission. We have performed this because previous experience suggests to us that the recording of hospital admission dates in primary care data may not be adequately precise in this setting and could affect results by causing misclassification bias (31,32). It seems, however, very unlikely that vaccination would preferentially precipitate this small subset of severe flares without any effect on milder flares, and so we do not think this will have biased out results. Similarly, IBD flares that were managed exclusively in hospital outpatient clinics were excluded from this study. However, general practitioners serve as the first point of contact for patients with IBD experiencing a flare and participate in their initial outpatient management, including with corticosteroid prescription, unless the disease flare is severe enough to warrant hospital admission (33). Any bias from missing data on IBD flares requiring hospitalization or those that were managed in gastroenterology outpatient clinics is unlikely to affect the validity of our findings because any resulting bias will be nondifferential in nature.
Second, because general practice records do not reliably code for biologics, we are unable to examine whether the subset of patients receiving them have an altered risk of adverse effects from vaccination. We see no reason though to expect more extreme immunologically driven side effects in these groups in whom the vaccine is less immunogenic (17,18). Similarly, we are unable to examine subgroups by the extent or distribution of IBD because this information is not in general coded in primary care records. Again, we see no reason to believe though that the effect of the vaccine regarding this would be differential between these groups. Another limitation of our method is that because we require steroid prescription to define flare, it is possible that there may be an association with more minor flares treated with 5aminosalicylates alone. Although we cannot exclude this, we feel that such minor effects would be unlikely to greatly discourage vaccination uptake and that it is the more significant flares which we have studies which are the primary concern. Patients who experience an IBD flare soon after vaccination against COVID-19 may be discouraged from seeking future vaccinations against COVID-19. This has the potential to bias any association between vaccination and disease flare when data from multiple vaccinations are analyzed together. To minimize such a bias, we presented data on association between vaccination and disease flares according to sequential vaccine doses. Furthermore, our results show that IBD flares temporally associated with first dose of vaccination against COVID-19 did not deter patients from getting further vaccinations against COVID-19. Finally, we can of course study only the vaccinations that have been widely used in the UK National Health Service because we have no data relevant to other vaccine technologies, which may limit the generalizability of our findings in settings where other vaccine technologies are in use.
Our findings are consistent with those of the recent metaanalysis of studies of the safety of severe acute respiratory syndrome coronavirus 2 vaccination in patients with IBD (34), showing as they do no increase in flare risk. In contrast to the 6 small cohorts comprising 4,537 patients and 75 flares reported there though, our report represents the experience of 73,626 patients with IBD with documented vaccination among whom 1,940 experienced a flare at some time during the study period. In addition, rather than reporting the absolute flare incidence postvaccination (flare probability 0.01 [95% CI 0.01-0.03] (34)), we have reported an incidence rate ratio comparing the risk in periods after vaccination with patients' experience at other times (IRR 0.89 [95% CI 0.77-1.02]). Of the patients in the studies included in meta-analysis cited earlier, most of them (n 5 3,316) came from a single study reporting a US cohort study, ascertaining data through repeated survey of participants (35). This study, though potentially less representative of patients with typical IBD than is ours, was able to report on biologic and immunomodulator use and therefore to confirm a low absolute risk of disease flare defined using a combination of symptoms and treatment change within 1 month of vaccination against COVID-19 in a population in which most were taking biologics or small molecules before vaccination. However, it reported a high rate of IBD symptoms, e.g., bowel frequency, extraintestinal manifestations, and abdominal pain in 12%, 12%, and 11% of participants, respectively, in this period and did not report comparative estimates, leaving the question of association between COVID-19 vaccination and IBD flares unanswered. A further study not within the meta-analysis, which is based on self-reported flares by patients in a questionnaire (36), gives additional assurance that the lack of association does indeed include minor flares because these would be included in the 147 subjective records of flare which they report.
Our study provides population-based evidence that vaccination against COVID-19 in patients with IBD does not increase the risk of flare. Patients expressing concern regarding this should therefore be reassured and encouraged to take up vaccination if they have not already done so.