Burden of perianal disease in Crohn's disease: Accelerating medical therapy and high rates of perianal surgery over the last four decades - Results from a population-based study over four decades

BACKGROUND

Few population-based studies have investigated the prevalence and disease course of perianal manifestation in Crohn's disease.

AIMS

To analyse the prevalence and outcomes of perianal Crohn's disease including medical therapies and need for perianal surgery, over different therapeutic eras based on the time of diagnosis; cohort A (1977-1995), cohort B (1996-2008), and cohort C (2009-2018) METHODS: Patient inclusion lasted between 1977 and 2018. We followed patients prospectively, and regularly reviewed both in-hospital and outpatient records. We defined a perianal surgical procedure as any perianal incision and excision, fistulotomy, or abscess drainage.

RESULTS

We included 946 incident patients. Perianal disease at diagnosis was present in 17.4% (n = 165) of the total cohort, with a declining prevalence in cohorts A/B/C, respectively (24.7%/18.5%/13.2%; p = 0.001). By the end of follow-up, an additional 9.3% (n = 88) of the total cohort developed perianal disease. Cumulative immunosuppressive and biologic exposure increased over time; biologic use was higher in patients with perianal disease [pLog Rank < 0.001]. The overall rate of perianal surgery was 44.7% (113/253), with a probability of 28.3% (95% CI: 25.4-31.2) after 10 years, 41.0% (95% CI: 37.5-44.5) after 20 years, and 64.1% (95% CI: 59-69.2) after 30 years. There was no statistically significant difference in the probability of first perianal surgery among cohorts A/B/C [Log Rank = 0.594].

CONCLUSIONS

The burden of perianal disease and perianal surgery rates were high in this cohort. Therapeutic strategy was accelerated in patients with perianal Crohn's over time with higher exposure to immunosuppressives and biologics. Surgical management of perianal disease remained unchanged amongst the cohorts.

Declining Trends of Reoperations and Disease Behaviour Progression in Crohn's Disease over Different Therapeutic Eras-A Prospective, Population-Based Study from Western Hungary between 1977-2020, Data from the Veszprem Cohort

BACKGROUND AND AIMS

Few population-based studies have investigated long-term surgery rates for Crohn's disease [CD]. Our aim was to analyse disease progression and surgery rates in a population-based cohort over different therapeutic eras, based on the time of diagnosis: cohort-A [1977-1995], cohort-B [1996-2008], and cohort-C [2009-2018].

METHODS

A total of 946 incident CD patients were analysed (male/female: 496/450; median age at diagnosis: 28 years [y]; interquartile range [IQR]: 22-40]). Patient inclusion lasted between 1977 and 2018. Immunomodulators have become widespread in Hungary since the mid-1990s and biologic therapies since 2008. Patients were followed prospectively, with both in-hospital and outpatient records reviewed regularly.

RESULTS

The probability of disease behaviour progression from inflammatory [B1] to stenosing or penetrating phenotype [B2/B3] significantly decreased (27.1 ± 5.3%/21.5 ± 2.5%/11.3 ± 2.2% in cohorts A/B/C, respectively, after 5 years; 44.3 ± 5.9%/30.6 ± 2.8%/16.1 ± 2.9% after 10 years, respectively; [pLogRank <0.001]). The probability of first resective surgery between cohorts A/B/C were 33.3 ± 3.8%/26.5 ± 2.1%/28.1 ± 2.4%, respectively, after 5 years; 46.1 ± 4.1%/32.6 ± 2.2%/33.0 ± 2.7% after 10 years, respectively; and 59.1 ± 4.0%/41.4 ± 2.6% [cohorts A/B] after 20 years. There was a significant decrease in first resective surgery risk between cohorts A and B [plog rank = 0.002]; however, no further decrease between cohorts B and C [plog rank = 0.665]. The cumulative probability of re-resection in cohorts A/B/C was decreasing over time (17.3 ± 4.1%/12.6 ± 2.6%/4.7 ± 2.0%, respectively, after 5 years [plog rank = 0.001]).

CONCLUSION

We report a continuous decline in reoperation rates and disease behaviour progression in CD over time, with the lowest values in the biologic era. In contrast, there was no further decrease in the probability of first major resective surgery after the immunosuppressive era.

Time Trends of Environmental and Socioeconomic Risk Factors in Patients with Inflammatory Bowel Disease over 40 Years: A Population-Based Inception Cohort 1977-2020

BACKGROUND

Data from population-based studies investigating trends in environmental factors associated with inflammatory bowel disease (IBD) is lacking. We aimed to assess long-term time trends of environmental and socioeconomic factors in IBD patients from a well-defined population-based cohort from Veszprem, Hungary.

METHODS

Patients were included between 1 January 1977, and 31 December 2020. Trends of environmental and socioeconomic factors were evaluated in three periods based on the decade of diagnosis, representing different therapeutic eras: cohort-A,1977-1995; cohort-B,1996-2008 (immunomodulator era); and cohort-C, 2009-2020 (biological era).

RESULTS

A total of 2240 incident patients with IBD were included (ulcerative colitis (UC) 61.2%, male 51.2%, median age at diagnosis: 35 years (IQR 29-49)). Rates of active smoking significantly decreased over time in Crohn's disease (CD): 60.2%, 49.9%, and 38.6% in cohorts A/B/C (p < 0.001). In UC, the rates were low and stable: 15.4%, 15.4%, and 14.5% in cohorts A/B/C (p = 0.981). Oral contraceptive use was more common in CD compared to UC (25.0% vs. 11.6%, p < 0.001). In UC, prevalence of appendectomy before diagnosis decreased over time: 6.4%, 5.5%, and 2.3% in cohorts A/B/C (p = 0.013). No significant changes were found in the socio-geographic characteristics of the IBD population (urban living: UC, 59.8%/64.8%/ 62.5% (p = 0.309) and CD, 62.5%/ 62.0%/ 59.0% (p = 0.636), in cohorts A/B/C). A greater percentage of patients had completed secondary school as the highest education level in later cohorts in both UC (42.9%/50.2%/51.6%, p < 0.001) and CD (49.2%/51.7%/59.5%, p = 0.002). A higher percentage of skilled workers (34.4%/36.2%/38.9%, p = 0.027) was found in UC, but not in CD (p = 0.454).

CONCLUSION

The association between trends of known environmental factors and IBD is complex. Smoking has become less prevalent in CD, but no other major changes occurred in socioeconomic factors over the last four decades that could explain the sharp increase in IBD incidence.

Incidence, Prevalence, Disease Course, and Treatment Strategy of Crohn's Disease Patients from the Veszprem Cohort, Western Hungary: A Population-based Inception Cohort Study Between 2007 and 2018

BACKGROUND AND AIMS

The number of prospective population-based studies on Crohn's disease[CD] is still limited from Eastern Europe. The present study is a continuation of the Veszprem IBD cohort. Our aim was to analyse incidence, prevalence, disease phenotype, treatment strategy, disease course, and surgical outcomes in a prospective population-based inception cohort including CD patients diagnosed between 2007 and 2018.

METHODS

A total of 421 consecutive inception patients were included [male/female:237/184; mean age at diagnosis: 33.3 ± 16.2years]. Both in-hospital and outpatient records were collected and comprehensively reviewed. Demographic data were derived from the Hungarian Central Statistical Office.

RESULTS

Mean incidence rate was 9.9 [95% CI: 9.0-10.9]/105 person-years in this 12-year period. Prevalence rate was 236.8 [95% CI: 220.8-252.8] in 2015; 17.6% and 20.0% of the patients had stenosing[B2] and penetrating[B3] disease behavior at diagnosis,respectively. The probability of disease behaviour progression from luminal to B2/B3 phenotype was 14.7% (standard error [SE]: 2.2) at 5 years after diagnosis. Distribution of maximal therapeutic steps during the total follow-up (8.5 years [8.5y], standard deviation [SD]: 3.3) was 5-aminosalicylic acid [5-ASA] in 15.7%, corticosteroids in 14.3%, immunosuppressives in 42.5%, and biologic therapy in 26.2%. The probability of receiving biologictherapy after diagnosis was 20.9% [SE: 2.0] at 5 years. The probability of first resective surgery was 20.7% [SE: 2.0] at 1 year, 26.1% [SE: 2.2] at 5 years, and 30.7% [SE: 2.4] at 10 years. The perianal surgery rate was 31.3% among patients with perianal involvement.

CONCLUSIONS

The incidence of CD in Hungary was high, similar to high-incidence areas in Western Europe. Treatment strategies are reflecting the biologic era. Disease behaviour progression was lower, as well as long-term [10y] surgery rates decreasing compared with data from previous decades.

Long-term Colectomy rates of Ulcerative Colitis over 40-year of Different Therapeutic eras - Results from Western Hungarian Population-based Inception Cohort between 1977-2020

BACKGROUND AND AIMS

Few population-based studies have investigated the long-term colectomy rates of ulcerative colitis (UC). We aimed to assess the colectomy rates over 40 years of different therapeutic eras in a prospective population-based inception cohort from Veszprem Province, Western Hungary.

METHODS

Patient inclusion lasted between January1,1977, and December31, 2018. Patient follow-up ended December 31,2020. Colectomy rates and disease course were examined in three different eras based on the time of UC diagnosis; cohort-A(1977-1995),cohort-B(1996-2008), and cohort C(2009-2018).

RESULTS

A total of 1,370 incident UC patients were included (male 51.2%,median age at diagnosis:37 years). Median follow-up was 17 years (IQR 9-24); 87 patients(6.4%) underwent colectomy. The cumulative probability of colectomy in the total population was 2.6%(95%CI 2.2-3.0), 4.2%(95%CI 3.6-4.8), 7.0%(95%CI 6.2-7.8), and 10.4%(95%CI 9.1-11.7) after 5, 10, 20, and 30 years, respectively. The proportion of extensive colitis at diagnosis increased over time (24.2%/24.3%/34.9% in cohorts A/B/C,p=0.001). Overall exposure to immunomodulators (11.3%/20.9%/34.4% in cohorts A/B/C,p<0.001), as well as the probability for biological therapy initiation increased over time; 0%/ 3.3%(95%CI 2.6-4.0)/ 13.9%(95%CI 12.1-15.7), p<0.001. There were no statistically significant differences in the cumulative probability of colectomies between cohorts A/B/C; 1.7% (95%CI 1.0-2.4), 2.5%(95%CI 1.9-3.1), and 3.7%(95%CI 2.7-4.7) after 5 years; 3.5%(95%CI 2.5-4.5), 4.2%(95%CI 3.4-5.0), and 4.5%(95%CI 3.3-5.7) after 10 years; and 7.5% (95%CI 6.1-8.9) and 6.3% (95%CI 5.2-7.4) in cohorts A/B after 20 years (Log-rank=0.588). Extensive colitis (HR;2.24:95%CI 1.55-3.23) and continuous active disease activity (HR;6.36:95%CI 3.46-11.67) were independent predictors for colectomy.

CONCLUSION

No differences in colectomy rates have been observed in the incident UC patients over 40 years despite increasing use of immunomodulators and biological therapies.

Epidemiology, treatment strategy, natural disease course and surgical outcomes of patients with ulcerative colitis in Western Hungary - a population-based study between 2007 and 2018, data from the Veszprem County cohort

BACKGROUND

The number of population-based studies in ulcerative colitis (UC) is limited from Eastern Europe. Our aim was to analyze incidence, prevalence, disease phenotype, treatment strategy, disease course and colectomy rates in a prospective population-based inception cohort including UC patients diagnosed between 2007 and 2018. The present study is a continuation of the Veszprem IBD cohort since 1977.

METHODS

467 UC patients were included (male/female: 236/231; median age at diagnosis: 36 years(y) [IQR: 25-54]). Both in-hospital and outpatient records were collected and comprehensively reviewed. The mean length of follow-up was 8.34±3.6y. The source of demographic data was derived from the Hungarian Central Statistical Office.

RESULTS

Mean incidence rate was 11.02/10 5 person-years in this 12-year period. Prevalence was 317.79/10 5 persons in 2015. Disease extent at diagnosis was proctitis(E1) in 22.3%, left-sided colitis(E2) in 43.9%, and extensive colitis(E3) in 33.8%. The probability of disease extent progression was 11.6%(SE:1.8) after 5 years. Distribution of maximal therapeutic steps were 5-ASA in 46.9%, corticosteroids in 16.3%, immunosuppressives in 19.3%, and biologicals in 16.5%. The probability of receiving biological therapy after diagnosis was 9.9%(SE:1.4) at 3 years. Overall colectomy rate was 4.1% in the population. The probability of colectomy was 1.5%(SE:0.6) at 1 year, 3.6%(SE:0.9) at 5 years and 4.4%(SE:1.0) at 10 years.

CONCLUSIONS

The incidence of UC was high in Hungary, similar to high-incidence areas in Western Europe. Treatment strategies are in line with the biologic era. The probability of progressing into proximal disease, and the medium- and long-term colectomy rates were both lower compared data from Western European centers.

Does Hospitalization Predict the Disease Course in Ulcerative Colitis? Prevalence and Predictors of Hospitalization and Re-Hospitalization in Ulcerative Colitis in a Population-based Inception Cohort (2000-2012)

BACKGROUND AND AIMS

Limited data are available on the hospitalization rates in population-based studies. Since this is a very important outcome measure, the aim of this study was to analyze prospectively if early hospitalization is associated with the later disease course as well as to determine the prevalence and predictors of hospitalization and re-hospitalization in the population-based ulcerative colitis (UC) inception cohort in the Veszprem province database between 2000 and 2012.

METHODS

Data of 347 incident UC patients diagnosed between January 1, 2000 and December 31, 2010 were analyzed (M/F: 200/147, median age at diagnosis: 36, IQR: 26-50 years, follow-up duration: 7, IQR 4-10 years). Both in- and outpatient records were collected and comprehensively reviewed.

RESULTS

Probabilities of first UC-related hospitalization were 28.6%, 53.7% and 66.2% and of first re-hospitalization were 23.7%, 55.8% and 74.6% after 1-, 5- and 10- years of follow-up, respectively. Main UC-related causes for first hospitalization were diagnostic procedures (26.7%), disease activity (22.4%) or UC-related surgery (4.8%), but a significant percentage was unrelated to IBD (44.8%). In Kaplan-Meier and Cox-regression analysis disease extent at diagnosis (HR extensive: 1.79, p=0.02) or at last follow-up (HR: 1.56, p=0.001), need for steroids (HR: 1.98, p<0.001), azathioprine (HR: 1.55, p=0.038) and anti-TNF (HR: 2.28, p<0.001) were associated with the risk of UC-related hospitalization. Early hospitalization was not associated with a specific disease phenotype or outcome; however, 46.2% of all colectomies were performed in the year of diagnosis.

CONCLUSION

Hospitalization and re-hospitalization rates were relatively high in this population-based UC cohort. Early hospitalization was not predictive for the later disease course.

Association of extraintestinal manifestations and anaemia with disease outcomes in patients with inflammatory bowel disease

OBJECTIVE

The association between extraintestinal manifestations (EIMs) and disease activity suggest a common pathogenetic link with inflammatory bowel disease (IBD). We report on the association of EIMs and anaemia with long-term disease outcomes, including treatment steps, hospitalization, and surgery in the prospective population-based IBD inception cohort from Veszprem province.

METHODS

Data of 678 incident IBD patients (Crohn's disease/ulcerative colitis(CD/UC): 331/347) diagnosed from 1st January 2000 to 31st December 2012 were analyzed (CD: m/f: 176/155, median age at diagnosis: 28, IQR: 21-40 years, disease duration: 6, IQR: 2-9 years; UC: m/f: 200/147, median age at diagnosis: 36, IQR: 26-50 years, duration: 7, IQR: 4-10 years).

RESULTS

EIMs were present in 30% of the CD and 17.3% of the UC patients. In CD, female gender (p = 0.02) need for steroid (p  < 0.001) and azathioprine (AZA) (p = 0.02), while in UC, young age at onset (p = 0.03), extensive disease (p = 0.003), female gender (p = 0.07), need for steroids (p < 0.001) and AZA (p = 0.004) and need for IBD-related hospitalization (p = 0.01) were associated with the presence of EIMs. Anaemia was present in 56.7% of the CD and 30.2% of the UC patients. In both CD and UC anaemia was associated with age at onset (pCD = 0.001, pUC = 0.04), disease location/extent (pCD = 0.02, pUC < 0.001), steroid and AZA use (for both pCD,UC < 0.001), need for surgery/colectomy (pCD < 0.001, pUC = 0.002) and hospitalization (pCD = 0.004, pUC < 0.001) and in CD, it was associated with anti TNF therapy(p = 0.002).

CONCLUSIONS

The presence of EIMs was associated with disease phenotype in UC and with treatment strategy in both CD and UC. Additionally, anaemia was associated with hospitalization and surgery in both CD and UC, suggesting that EIMs and anaemia may be helpful in stratifying disease severity in IBD.

Prevalence and predictors of hospitalization in Crohn’s disease in a prospective population-based inception cohort from 2000-2012

AIM: To analyze the prevalence, length and predictors of hospitalization in the biological era in the population-based inception cohort from Veszprem province.

METHODS: Data of 331 incident Crohn’s disease (CD) patients diagnosed between January 1, 2000 and December 31, 2010 were analyzed (median age at diagnosis: 28; IQR: 21-40 years). Both in- and outpatient records were collected and comprehensively reviewed.

RESULTS: Probabilities of first CD-related hospitalization and re-hospitalization were 32.3%, 45.5%, 53.7% and 13.6%, 23.9%, 29.8%, respectively after one, three and five years of follow-up in Kaplan-Meier analysis. First-year hospitalizations were related to diagnostic procedures (37%), surgery or disease activity (27% and 21%). Non-inflammatory disease behavior at diagnosis (HR = 1.32, P = 0.001) and perianal disease (HR = 1.47, P = 0.04) were associated with time to first CD-related hospitalization, while disease behavior change (HR = 2.38, P = 0.002) and need for steroids (HR = 3.14, P = 0.003) were associated with time to first re-hospitalization in multivariate analyses. Early CD-related hospitalization (within the year of diagnosis) was independently associated with need for immunosuppressives (OR = 2.08, P = 0.001) and need for surgeries (OR = 7.25, P < 0.001) during the disease course.

CONCLUSION: Hospitalization and re-hospitalization rates are still high in this cohort, especially during the first-year after the diagnosis. Non-inflammatory disease behavior at diagnosis was identified as the pivotal predictive factor of both hospitalization and re-hospitalization.

Low incidence of venous thromboembolism in inflammatory bowel diseases: prevalence and predictors from a population-based inception cohort

OBJECTIVE

Patients with inflammatory bowel diseases (IBD) are considered to have an increased risk for venous thromboembolism (VTE). The aim of the present study was to analyze the incidence and risk factors of VTE in a population-based inception cohort in the Veszprem province database between 1977 and 2012.

MATERIAL AND METHODS

A total of 1708 incepted IBD patients were included (male/female: 879/829; CD (Crohn's disease): 648, age at onset: 29, interquartile range (IQR): 22-39; UC (ulcerative colitis): 1060, age at onset: 36, IQR: 26-50 years). Both in- and outpatient records were collected and comprehensively reviewed and followed up for a total of 21,369 patient-years.

RESULTS

Twenty-two VTE events were identified in 19 patients (6 events in 5 CD and 16 in 14 UC patients). The incidence rate of VTE in IBD was 1.03 per 1000 patient-years. The risk of VTE in UC was associated with extensive location (odds ratio (OR): 3.25, 95% confidence interval (CI): 1.13-9.35), presence of fulminant episode during the disease course (OR: 4.15, 95% CI: 1.28-13.5), smoking (OR: 3.46, 95% CI: 1.14-10.5), and need for steroids (OR: 2.97, 95% CI: 0.99-8.92).

CONCLUSION

The incidence of VTE was lower than previously reported. The incidence was higher in males and in UC it was associated with extensive disease, fulminant episodes, corticosteroids-requiring disease and smoking, but not with age at onset.

Is early limited surgery associated with a more benign disease course in Crohn’s disease?

AIM: To analyze the difference in disease course and need for surgery in patients with Crohn’s disease (CD).

METHODS: Data of 506 patients with incident CD were analyzed (age at diagnosis: 31.5 ± 13.8 years). Both hospital and outpatient records were collected prospectively with a complete clinical follow-up and comprehensively reviewed in the population-based Veszprem province database, which includes incident CD patients diagnosed between January 1, 1977 and December 31, 2008. Follow-up data were collected until December 31, 2009. All patients included had at least 1 year of follow-up available. Patients with indeterminate colitis at diagnosis were excluded from the analysis.

RESULTS: Overall, 73 patients (14.4%) required resective surgery within 1 year of diagnosis. Steroid exposure and need for biological therapy were lower in patients with early limited surgery (P < 0.001 and P = 0.09). In addition, surgery rates during follow-up in patients with and without early surgery differed significantly after matching on propensity scores (P < 0.001, HR = 0.23). The need for reoperation was also lower in patients with early limited resective surgery (P = 0.038, HR = 0.42) in a Kaplan-Meier and multivariate Cox regression (P = 0.04) analysis. However, this advantage was not observed after matching on propensity scores (P_Logrank = 0.656, P_Breslow = 0.498).

CONCLUSION: Long-term surgery rates and overall exposure to steroids and biological agents were lower in patients with early limited resective surgery, but reoperation rates did not differ.

Risk of colorectal cancer in Crohn's disease patients with colonic involvement and stenosing disease in a population-based cohort from Hungary

BACKGROUND & AIMS

Since data is limited regarding the risk of colorectal cancer (CRC) in Crohn's disease (CD) patients who present with stenosing disease in the colon, this study was undertaken to assess CRC risk in such patients, using a population-based, Veszprem province database, which includes incidental patients diagnosed between January 1, 1977 and December 31, 2011.

METHODS

Data from 640 incidental CD patients were analyzed (M/F ratio: 321/319, age-at-diagnosis: 28 years (IQR: 22-38)). Both hospital and outpatient records were collected and comprehensively reviewed.

RESULTS

CRC was diagnosed in six CD patients during a follow-up of 7759 person-years. Sixty-two patients presented with colonic/ileocolonic disease and a stenotic lesion in the colon with a follow-up of 702 person-years (median: 10.5, IQR: 5-16years). Colorectal cancer developed in 6.5% (equalling 0.57/100 person-years), the SIR (6.53, 95% CI: 2.45-17.4) was increased with four patients observed versus 0.61 expected. In a Kaplan-Meier analysis, the probability of developing CRC was 5.5% and 7.5% after 5- and 10 years, respectively, versus 0.4% in patients with other phenotypes (HR: 18.8, p<0.001). A sensitivity analysis included patients with stenosing colonic lesion at diagnosis or during follow-up (n=91, follow-up: 1180 person-years, median: 12, IQR: 6-17years). The probability of developing CRC was 3.6% and 4.9% after 5- and 10 years, respectively.

CONCLUSIONS

The risk of CRC in CD patients presenting with or developing a stenotic lesion in the colon is high even after a short disease duration, suggesting the need for careful surveillance.

Evolution of disease phenotype in adult and pediatric onset Crohn’s disease in a population-based cohort

AIM: To investigate the evolution of disease phenotype in adult and pediatric onset Crohn’s disease (CD) populations, diagnosed between 1977 and 2008.

METHODS: Data of 506 incident CD patients were analyzed (age at diagnosis: 28.5 years, interquartile range: 22-38 years). Both in- and outpatient records were collected prospectively with a complete clinical follow-up and comprehensively reviewed in the population-based Veszprem province database, which included incident patients diagnosed between January 1, 1977 and December 31, 2008 in adult and pediatric onset CD populations. Disease phenotype according to the Montreal classification and long-term disease course was analysed according to the age at onset in time-dependent univariate and multivariate analysis.

RESULTS: Among this population-based cohort, seventy-four (12.8%) pediatric-onset CD patients were identified (diagnosed ≤ 17 years of age). There was no significant difference in the distribution of disease behavior between pediatric (B1: 62%, B2: 15%, B3: 23%) and adult-onset CD patients (B1: 56%, B2: 21%, B3: 23%) at diagnosis, or during follow-up. Overall, the probability of developing complicated disease behaviour was 49.7% and 61.3% in the pediatric and 55.1% and 62.4% in the adult onset patients after 5- and 10-years of follow-up. Similarly, time to change in disease behaviour from non stricturing, non penetrating (B1) to complicated, stricturing or penetrating (B2/B3) disease was not significantly different between pediatric and adult onset CD in a Kaplan-Meier analysis. Calendar year of diagnosis (P = 0.04), ileal location (P < 0.001), perianal disease (P < 0.001), smoking (P = 0.038) and need for steroids (P < 0.001) were associated with presence of, or progression to, complicated disease behavior at diagnosis and during follow-up. A change in disease location was observed in 8.9% of patients and it was associated with smoking status (P = 0.01), but not with age at diagnosis.

CONCLUSION: Long-term evolution of disease behavior was not different in pediatric- and adult-onset CD patients in this population-based cohort but was associated to location, perianal disease and smoking status.

Learning to Play Using Low-Complexity Rule-Based Policies: Illustrations through Ms. Pac-Man

In this article we propose a method that can deal with certain combinatorial reinforcement learning tasks. We demonstrate the approach in the popular Ms. Pac-Man game. We define a set of high-level observation and action modules, from which rule-based policies are constructed automatically. In these policies, actions are temporally extended, and may work concurrently. The policy of the agent is encoded by a compact decision list. The components of the list are selected from a large pool of rules, which can be either hand-crafted or generated automatically. A suitable selection of rules is learnt by the cross-entropy method, a recent global optimization algorithm that fits our framework smoothly. Cross-entropy-optimized policies perform better than our hand-crafted policy, and reach the score of average human players. We argue that learning is successful mainly because (i) policies may apply concurrent actions and thus the policy space is sufficiently rich, (ii) the search is biased towards low-complexity policies and therefore, solutions with a compact description can be found quickly if they exist.