INTRODUCTION: The impact of diagnostic delay on colorectal cancer mortality has never been conclusively evaluated. Most studies show either no association or find that rapidly diagnosed patients have higher mortality rates than patients with longer waits in the primary and secondary health care sector. The disparate and contradictory findings may be due to use of different outcome measures, use of arbitrary cut-off points in measuring delay, and failure to adjust for important clinical conditions at presentation, i.e. confounding factors such as comorbidity or symptom manifestation. AIM: To analyse mortality from colorectal cancer as a function of the diagnostic interval in more comprehensive models that make symptomatic cancer patients more comparable at the time of their first encounter with the health care system. METHODS: A total of 262 consecutive patients with a first-time diagnosis of colorectal cancer were included in a prospective, population-based study in the former Aarhus County, Denmark, from 2004 to 2005. Patients were identified from medical databases, and each patient’s diagnosis was confirmed by the general practitioner (GP) who also provided a detailed description of the diagnostic pathway. The diagnostic interval was defined as the time from first presentation of symptoms to the GP to the date of diagnosis. All patients were followed up for three years after diagnosis. Colon and rectal cancer were analysed separately. We used conditional logistic regression to estimate three-year mortality odds ratios as a function of the diagnostic interval stratifying for gender, age, comorbidity, symptom interpretation and cancer history. The diagnostic interval was treated as a continuous variable using restricted cubic splines with four knots that permitted mortality to increase or decrease repeatedly. RESULTS: Forty-two percent of patients died within three years of diagnosis. Three-year mortality odds ratios decreased with longer diagnostic interval until the reference point of 30 days and then increased with a diagnostic interval longer than 30 days. The adjusted model was statistically significant for colon cancer (p=0.02). DISCUSSION/CONCLUSION: After adjusting we observed the well-known waiting list paradox, but also a significantly poorer prognosis for patients with diagnostic intervals longer than 30 days. The shifting association would have been hidden if the diagnostic interval had been treated as a categorical variable with standard cut-off points.