Predictors of Complicated Disease Course in Children and Adults With Ulcerative Colitis: A Nationwide Study From the epi-IIRN

Abstract

Background

Data on predictors of complicated ulcerative colitis (UC) course from unselected populations cohorts are scarce. We aimed to utilize a nationwide cohort to explore predictors at diagnosis of disease course in children and adults with UC.

Methods

Data of patients diagnosed with UC since 2005 were retrieved from the nationwide epi-IIRN cohort. Complicated disease course was defined as colectomy, steroid-dependency, or the need for biologic drugs. Hierarchical clustering categorized disease severity at diagnosis based on complete blood count, albumin, C-reactive protein and erythrocyte sedimentation rate (ESR), analyzed together.

Results

A total of 13 471 patients with UC (1427 [11%] pediatric-onset) including 103 212 person-years of follow-up were included. Complicated disease course was recorded in 2829 (21%) patients: 1052 (7.9%) escalated to biologics, 1357 (10%) experienced steroid-dependency, and 420 (3.1%) underwent colectomy. Probabilities of complicated disease course at 1 and 5 years from diagnosis were higher in pediatric-onset (11% and 32%, respectively) than adult-onset disease (4% and 16%; P < .001). In a Cox multivariate model, complicated course was predicted by induction therapy with steroids (hazard ratio [HR], 1.5; 95% CI, 1.2-2.0), extraintestinal manifestations (HR, 1.3; 95% CI, 1.03-1.5) and the disease severity clusters of blood tests (HR, 1.8; 95% CI, 1.01-3.1), while induction therapy with enemas (HR, 0.6; 95% CI, 0.5-0.7) and older age (HR, 0.99; 95% CI, 0.98-0.99) were associated with noncomplicated course.

Conclusion

In this nationwide cohort, the probability of complicated disease course during the first 5 years from diagnosis was 32% in pediatric-onset and 16% in adults with UC and was associated with more severe clusters of routinely collected laboratory tests, younger age at diagnosis, extraintestinal manifestations, and type of induction therapy.

Introduction

While some patients with ulcerative colitis (UC) may rapidly respond to therapy with 5 aminosalicylic acids (5-ASA),¹ others may follow a severe or resistant disease course requiring multiple immunomodulators and biologics or eventually, colectomy. In adults, during the first 5 years from diagnosis, approximately half of the patients require hospitalization, and 4% to 10% undergo colectomy.^2-4 Most guidelines in children and adults recommend biologics in patients who have inadequate response to conventional therapy including 5-ASA and thiopurines.^5,6 However, predictors of disease outcomes may facilitate a timely choice of the most appropriate treatment in patients early during the disease course.

Previous studies, summarized in systematic reviews of the pediatric PIBD-Ahead⁷ and adult IBD-Ahead⁸ projects, have suggested several variables to predict high-risk patients, including younger age and disease extension at diagnosis in adults, disease severity, hemoglobin, and white blood cell (WBC) count at diagnosis in children. However, of the various proposed prediction models, only a few were consistently reproduced, such as the need for steroids.^9,10 Moreover, we previously failed to validate the main predictive models in children using a prospective inception cohort.¹¹ Administrative studies of large-unselected population cohorts reported inconsistent results of predicting long-term disease course by using ICD codes of anemia,^9,12use of steroids,⁹ younger age,^9,13 sex,^12,14,15 disease extention,^13,16 and comorbidities.¹⁴ Most studies did not include children, and laboratory results were not included systematically in any, likely due to the inconsistent timing of laboratory tests and a high rate of missing data in administrative cohorts.

In this nationwide study, we aimed to explore predictors of complicated disease course in children and adults with UC from a large nationwide administrative cohort, while applying advanced modeling to optimize the use of laboratory tests.

Methods

This study utilized the epidemiology cohort of the Israeli inflammatory bowel disease (IBD) Research Nucleus (epi-IIRN). It includes all patients in Israel with IBD (n = 58 640 as of July 2021) from the 4 health maintenance organizations (HMOs), which insure 98% of the country’s residents.¹⁷ The cohort includes information about medication purchases, procedures, diagnoses, physician’s visits, blood tests, and other ambulatory health services. Medication purchase records are very accurate since they are dispensed and heavily subsidized by the HMOs. We previously developed and validated case ascertainment algorithm to identify patients with IBD with high accuracy (99% specificity, 89% sensitivity, 92% positive predictive value [PPV], and 99% negative predictive value [NPV]).¹⁷ Briefly, the algorithm uses a combination of IBD-related International Statistical Classification of Diseases, Ninth Revision (ICD-9) codes, alone if more than 5 to 6 codes exist for the patients (depending on the HMO) or combined with purchases of IBD-related medications if fewer codes exist. Inflammatory bowel disease type is determined by the majority of UC or Crohn’s disease specific codes of the 3 most recent healthcare contacts, or the most recent codes when <3 are recorded (sensitivity 92%, specificity 97%, PPV 97%, NPV 92%). Data obtained from the HMOs were linked with the Ministry of Health’s national registries that maintain prospective records on surgeries and admissions.

Outcomes

The primary outcome was complicated disease course, defined as the first occurrence of one of the following: colectomy (Supplementary Table 1), the need for biologics drugs, or steroid dependency (ie, >90 cumulative days of steroids in a given disease year, independent of other treatments). We previously validated the indication of colectomy within the epi-IIRN cohort showing high accuracy.³ Use of immunomodulators was not included as one of the outcomes, since the need for biologics is a more robust outcome of severe disease course in UC.

Eligibility Criteria

We included an inception cohort of all patients diagnosed with UC from January 2005 to July 2020, allowing for 1 year of look-forward period. The HMOs shifted from paper to electronic records around 2000, and subsequently the year 2005 was previously validated as the cutoff for determining incidence,¹⁷ allowing a sufficient “look-back” period. Hence, we excluded those with IBD-related code/medication prior to 2005, as it could not be determined if the first code/medication reflects the diagnosis or the first record that appears in the newly established computerized system. We also excluded patients with complicated disease during the first 3 months post diagnosis to allow for a true prediction analysis. The analyses were stratified by pediatric-onset (ie, 0 to <18 years of age), adult-onset (18-65 years) and elderly-onset (>65 years) disease.

Predictors

Potential predictors were chosen a priori for analysis including demographic data, routinely collected laboratory tests, medications used as induction treatment during the first 3 months from diagnosis, extraintestinal manifestations (EIMs, Supplementary Table 2), diagnostic delay from start of symptoms and, in children, also anthropometrics. We estimated the diagnostic delay of each patient from the first code of gastrointestinal-related symptom (ie, with ICD9 codes, similar to the reported in previous Canadian administrative study¹⁸) or abnormal hemoglobin or albumin, as defined by age- and sex-specific normal reference, during the 5 years prior to UC diagnosis. Demographic data included year of diagnosis, age at diagnosis, sex, ethnicity (Jews or Arab), residence type (urban or rural), district of residence and social economic status (SES; captured on a 10-point, standardized scale based on Israel Central Bureau of Statistics socioeconomic data). Laboratory results obtained prior to the diagnosis date included erythrocyte sedimentation rate (ESR), C-reactive protein (CRP), albumin, hemoglobin, platelets, white blood cell count, and neutrophils count (Supplementary Table 3). Given the heterogeneity of available laboratory tests, we standardized the analysis through hierarchical clustering into 3 disease severity groups that ranged from mild to severe, according to any available values (see details in the Supplementary Material). To include all patients with UC and avoid selection bias, we imputed the most commonly used laboratory tests, namely platelets, WBC count, and hemoglobin in those without any available values by multiple imputation based on demographic and other clinical variables (sex, age at diagnosis, year of diagnosis, district of residence, total number of gastroenterology visits, total number of other health contacts, number of gastrointestinal-related ICD9 codes, and total number of emergency department visits), as previously developed and validated on the epi-IIRN cohort.^19-22

Induction therapy was defined as the first treatment within 3 months after diagnosis, including oral 5-ASA, rectal therapy, systemic steroids, and antibiotics. In patients with more than 1 induction treatment, we referred to the most intensive (eg, a patient who initiated 5-ASA and shortly thereafter escalated to steroids was categorized in the steroids group).

Statistical Analyses

Data are compared by Student’s t test, Wilcoxon rank sum test, and χ², as appropriate. To explore univariable differences between groups, we used standardized mean or median difference (SMD) for continues variables, and odds ratio (OR) with 95% confidence intervals (CIs) for categorical variables. This was used, rather than P values, to account for the large sample size, reporting the effect size and not merely statistical significance. The SMD is considered significant when the confidence interval does not include 0; positive value indicates higher mean/median in the “favorable outcomes” group, while negative values indicated otherwise. For OR, significance is determined when the confidence interval does not include 1.

Time to complicated disease was evaluated using Kaplan Meier survival curves and compared by the log-rank test. A Cox proportional hazards model with Bonferroni correction to adjust for multiple comparisons was fitted to assess the relationship between predictors and time to outcomes. The proportional hazards assumption was verified using the R function cox.zph() from the survival package, and a global P value < .05 was considered as violation of the assumption; P < .05 was considered significant and all computations were made in R version 4.2.2.

Results

Study Population

A total of 13 471 patients with UC met the eligibility criteria and were included in the study (1427 [11%] pediatric-onset, 10 646 [79%] adult-onset, and 1398 [10%] elderly-onset; Table 1). The median follow-up period was 7.5 years (interquartile range [IQR] 3.9-11.5 years) which translates into 103 212 person-years of follow-up. Complicated disease course was recorded in 2829 (21%) patients, of whom 1052 (7.9%) escalated to biologics, 1357 (10%) experienced steroid-dependency and 420 (3.1%) underwent colectomy. The probabilities for complicated disease course at 1, 3, and 5 years from diagnosis were 11%, 25%, and 32%, respectively, in children, and 4%, 11%, and 16%, respectively, in adults (P < .001; Figure 1).

Univariate Analysis Among the Entire Cohort

Numerous variables predicted a complicated disease course in a univariate analysis (Table 1), but considering the different follow-up duration across the 2 groups, we explored their probabilities for complicated disease at 1, 3, and 5 years from diagnosis from the survival analyses. The probabilities were higher in patients younger than 40 years vs older ones (6%, 15%, and 19%, vs 4%, 10%, and 15%, respectively; P < .001), in patients receiving induction treatment with systemic steroids vs patients without any induction therapy (8%, 21%, 29% vs 5%, 13%, 19%; P < .001), in patients with low vs high SES level (6%, 14%, 19% vs 4%, 12%, 16%; P < .001), in Arabs vs Jews (7%, 16%. 22% vs 5%, 12%, 17%; P < .001) and in those with extraintestinal manifestations vs those without (5%, 13%, 19% vs 5%, 13%, 17%; P = .02), while the other categorical variables did not achieve statistical significance after adjustment to follow-up period. The probability of complicated disease course was higher in patients with abnormal results of all included blood tests at diagnosis (ie, CRP, ESR, platelets, hemoglobin and WBC count) compared with those with normal results (Table 1 and Table 2). In addition, the median values of all individual blood tests included in the cluster analysis were significantly worse in those who developed complicated disease course (Table 1).

All 13 471 patients were grouped into 3 distinct clusters of disease severity based on laboratory tests at diagnosis from mild to severe (Table 3). Supporting internal validity, there was a gradual worsening in the median values of the individual tests across all clusters. For instance, patients in the highest cluster had ESR value 3 times greater than those of the lowest cluster and CRP values almost 6 times greater (Table 3). The clusters predicted disease course with a gradual increase in the proportions of patients who had complicated disease course amongst the disease severity clusters (P < .001; Figure 2). The lowest cluster had also the least frequent complicated disease (19%), while the highest cluster the most frequent (29%). The probabilities for complicated disease at 1, 3, and 5 years from diagnosis were 8%, 18%, and 25%, respectively, in patients with severe disease compared with 4%, 11%, and 16%, respectively, in patients with mild disease (P < .001).

Multivariate Analysis Among the Entire Cohort

Most predictors retained their significance in a multivariate Cox proportional hazards model. Induction treatment with steroids (HR, 1.5; 95% CI, 1.2-2.0), extraintestinal manifestations (HR, 1.3; 95% CI, 1.03-1.5), and the cluster of severity (HR, 1.8; 95% CI, 1.01-3.1) were independently associated with complicated disease (Figure 3). On the other hand, induction therapy with enemas of 5-ASA (HR, 0.6; 95% CI, 0.5-0.7) and older age at diagnosis (HR, 0.99; 95% CI, 0.98-0.99) were associated with noncomplicated disease course (Figure 3).

Predictors of Pediatric-onset Disease

In univariate analyses, predictors of complicated disease course in the pediatric-onset cohort were broadly similar to adults but without statistical significance likely due to the smaller sample size (Supplementary Table 4). The laboratory severity clusters displayed progressively worsening median values for all included tests (Supplementary Table 5) but without a clear predictive utility for complicated disease course (Figure 2). In a multivariate Cox proportional-hazards model, induction treatment with 5-ASA was protective (HR, 0.5; 95% CI, 0.3-0.9; Supplementary Figure 1).

Discussion

In this nationwide study of over 13 000 patients with UC, we found that one-fifth of patients with UC developed complicated disease course during the follow-up period, and the probabilities for complicated disease course were twice as high in pediatric-onset disease than in adults (32% vs 16% at 5 years from diagnosis). Complicated disease was predicted by younger age at diagnosis, diagnostic delay, low SES level, ethnicity, extraintestinal manifestations, laboratory test prior to diagnosis (ie, CRP, ESR, hemoglobin, platelets and albumin), and induction therapy with oral steroids. In a multivariable adjusted model, type of induction therapy, extraintestinal manifestations, and laboratory tests measured at diagnosis (assessed independently or when grouped in clusters) retained their significance, probably since these variables reflect more severe disease. Previous studies repeatedly showed that EIMs are associated with poor disease outcomes,^24-26 possibly due to genetic or microbiome-related factors. Of note, the laboratory-based severity clusters predicted disease-related outcomes in a gradual and intuitive gradual escalation, lending support to their validity. Predictors in children generally mirrored the findings in adults, but given the small sample size, most results did not reach statistical significance.

Three previous population-based studies evaluated predictors of disease course in patients with UC, all of which reported association between steroids induction^27,28 and disease extension^13,27,28 and complicated disease course. We also showed association with induction therapy, extending the findings from only steroid to the use of 5-ASA enemas, alluding to limited distal disease, as data on disease extent were not available in our cohort. One of the 3 previous studies reported higher risk in females,²⁷ one reported higher risk in males,²⁸ and the third similar risk in both sexes.¹³ We did not find an association between sex and complicated disease course in both the univariate and multivariable analyses. We, in agreement to one of the previous studies¹³but in contradition to another,²⁸ found an association between complicated disease course and younger age at diagnosis (the third study did not include age as one of the potential predictors). Indeed, pediatric-onset disease has been consistently associated with more extensive disease, need for biologics more frequently, and higher rate of admissions for acute severe colitis.^3,29,30 As we previously showed in the epi-IIRN cohort, low SES was associated with worse outcomes in CD, probably due to lower availability of quality health care and lower adherence to medical therapy.^31,32 This likely explains also the worse outcomes found in the Arabs population that suffers from lower SES in Israel compared with Jews.³³ One major advantage of our study is the use of hierarchical modeling to cluster all patients in severity groups to account for the expected significant heterogeneity of sampling. While in general there was a high proportion of missing values, almost all had some values available that allowed the clustering. We thus managed to show, for the first time in a nationwide study, that routinely collected laboratory tests at disease onset are strong predictors of disease course in adults with UC. Furthermore, we showed that the median values of the laboratory tests were consistent across the severity prediction groups.

The adult IBD-ahead systematic review suggested that younger age at diagnosis, extensive disease, steroid-dependency, and disease duration increased the risk for complicated disease course in UC.⁸ The corresponding pediatric IBD-ahead review added laboratory tests at diagnosis including hemoglobin, hematocrit, and albumin as possible predictors of colectomy.⁷ Similarly, 2 administrative studies (not population-based) suggested that anemia was associated with hospitalizations,^9,12 but these defined anemia by ICD codes and not by results of blood tests, as done here. In addition to administrative studies, previous cohort studies supported the predictive utility of routinely collected laboratory tests in UC. The PROTECT prospective study concluded that albumin, hemoglobin, vitamin D, and calprotectin were associated with disease course of pediatric patients with UC.³⁴ Numerous retrospective single-center studies evaluated the association between blood tests and outcomes. Some suggested an association between higher CRP^35-38 or platelets³⁸ and worse outcomes, while others proposed the ratio between neutrophil to lymphocyte counts^39,40 or between CRP and albumin.⁴¹

As a limitation of the study, we included only predictors available in the epi-IIRN administrative database, while some relevant variables such as disease extent, endoscopic severity, smoking, and clinical scores such as PUCAI and Mayo score, were not included. In addition, our administrative data set largely lacks data about results of biopsies and fecal calprotectin, which have been widely available through the HMOs only during the last 3 years. The epi-IIRN lacks clinical details of hospitalized patients, and thus data regarding cytomegalovirus and Clostridium difficile infections were not included in our study. Similarly, we did not include hospitalizations or the occurrence of acute severe colitis in our definition of complicated disease course given the limited accuracy of the main indication for admission in our cohort.

In conclusion, we demonstrate that complicated disease course is common in patients with UC, particularly in children. Our study suggests that extraintestinal manifestations and the need for systemic steroids as induction treatment are strong predictors of complicated disease course, as well as severity of laboratory values, including hemoglobin, albumin, platelets, WBC, CRP, and ESR. Our results suggest that baseline variables at diagnosis, and particularly laboratory tests, can be used in the clinical decision-making of early escalation to more intensive therapy. However, further research is needed to validate our results and verify how these should be taken into account when considering treatment for active disease.

Acknowledgments

The authors wish to thank Chagit Friss and Adi Mendelovici for study design, data cleaning, preparation, and formulation and Gili Focht for study design and epidemiological support.

Author Contributions

O.A.—Study concept and design, data acquisition, manuscript drafting, statistical analysis, data interpretation

R.L.—Study concept and design, data acquisition, statistical analysis, manuscript revision, data interpretation

S.G.—Data acquisition, manuscript revision

R.K.—Data acquisition, manuscript revision

Y.L.W.—Data acquisition, manuscript revision

N.L.—Data acquisition, manuscript revision

E.M.—Data acquisition, manuscript revision

O.L.—Data acquisition, manuscript revision

E.Z.—Data acquisition, manuscript revision

H.Y.– Data acquisition, manuscript revision

D.S.—Data acquisition, manuscript revision

I.D.—Data acquisition, manuscript revision

D.N.—Study supervision, study concept and design, data acquisition, statistical analysis, manuscript revision, data interpretation

D.T.—Study supervision, study concept and design, data acquisition, statistical analysis, manuscript revision, data interpretation

Funding

The epi-IIRN project was funded by a grant from the Leona M. and Harry B. Helmsley Charitable Trust.

Conflicts of Interest

O.L.—In the last 3 years, received consultation fees from Trendlines

H.Y.—Reports institutional research grants from Pfizer and the ISF; consulting fees from AbbVie, Janssen, Pfizer, Takeda, and Bristol Myers Squibb and Elly Lilli; honoraria for lectures from AbbVie, Janssen, Pfizer, and Takeda; participation in a Data Safety Monitoring Board or Advisory Board for AbbVie, Pfizer, Takeda, and Bristol Myers Squibb.

I.D.—In the last 3 years, received consultation fee(s), research grant(s), or honorarium(s) from AbbVie, Altman Research, Athos, Arena, BMS/Celgene, Celltrion, Eli-Lilly, Falk Pharma, Food Industries Organization, Gilead, Galapagos, Genentech/Roche, Iterative Scopes, Janssen, Neopharm, Pfizer, Rafa Laboratories, Sublimity, Sangamo, Sandoz, Takeda, and Wilbio. Share holder: Gutreat, Harp Diagnostics

D.T.—In the last 3 years, received consultation fee(s), research grant(s), royalties, or honorarium(s) from Janssen, Pfizer, Hospital for Sick Children, Ferring, Abbvie, Takeda, Atlantic Health, Shire, Celgene, Lilly, Roche, ThermoFisher, BMS

All other authors have nothing to declare.

Data Availability

Access to the data underlying this article will be granted on reasonable request to the corresponding author.

References