The Rate of Development of Complicated Crohnʼs Disease (Penetrating or Stricturing) Increases with Calendar Decade of Diagnosis: Reality or Bias?

Abstract

Purpose: The rate of development of complicated (penetrating or stricturing) Crohn's disease (CD) may be influenced by patient characteristics but also bias of patient selection or inaccurate classification in historical cohort studies. We determined if calendar decade of diagnosis influenced rate of development of complicated disease. Methods: 207 patients with non-stricturing non-penetrating CD were followed from diagnosis to development of complications or last clinic follow-up. The influences of decade of diagnosis and patient specific characteristics including Montreal classification and medication use on the rate of development of complicated disease in a historical cohort were determined. Results: Duration of disease was far longer in patients diagnosed before 1990 compared to after 1990 (24.6 vs. 6.8 years; p<0.001). The proportion diagnosed before age 40 was lower (30 vs. 48%, p=0.02) and surgery rates higher (68 vs. 30%; p<0.001) in the earlier diagnosis group. Immunomodulator use was equal in both groups (34%) yet time to first immunomodulator prescription was significantly longer in the earlier group (17.6 vs. 3.5 yrs; p<0.001). Kaplan Meier analysis was performed to study the rate of development of complications by decade of diagnosis (Figure). Rates of development of complicated disease were unusually low among patients diagnosed before 1990. In a Cox proportional hazards model, only calendar decade of diagnosis (HR 6.9; p<0.001) and age at diagnosis after age 40 (HR 0.29; p=0.006) influenced the hazard of development of complications. Conclusion: Earlier calendar decade of diagnosis of CD is associated with a lower rate of progression to complicated disease. While this might reflect changes in disease course with calendar time, it likely represents a bias in patient selection (survival bias) or inaccurate behavior classification because of differences in technology to detect and define disease complications. Such biases must be taken into account in historical cohort studies.Figure