Completeness and registration bias in PROCARE, a Belgian multidisciplinary project on cancer of the rectum with participation on a voluntary basis.

Background: PROCARE, a Belgian multidisciplinary project on rectal cancer, started in 2006 with participation on a voluntary basis. Completeness and bias of registration in PROCARE were assessed.

Methods: Data from 6353 patients with rectal cancer were extracted from the population based Belgian Cancer Registry for the period 2006-2008. Registration bias was studied by comparing patient, tumour and treatment characteristics of cases registered and non-registered in PROCARE. Relative survival (RS) of patient subgroups was analysed.

Results: PROCARE included 37% of all Belgian rectal cancer patients. Registration was highly variable between participating centres which recorded on average 56% of their patients. Significant differences in patient, tumour and treatment related characteristics were observed between registered and non-registered patients. The 5-year RS was 77% (95% confidence interval (CI): 74-80%) for registered patients and 56% (95% CI: 53-59%) for non-registered patients. After adjustment for patient, tumour characteristics and volume of centre, the relative excess risk of dying (RER) between registered and non-registered patients was 2.15 (95% CI: 1.85-2.50, p<0.001). The 5-year RS of patients treated in centres that never participated in the project was 59% (95% CI: 55-63%) and, after adjustment, the RER was 1.16 (95% CI: 1.00-1.35, p<0.050) compared to patients of the participating centres.

Conclusion: Registration of PROCARE patient data was incomplete, biased and variable between centres. Participation on a voluntary basis should be avoided for further projects. Quality assurance on a centre level requires compulsory and complete registration with a minimal but relevant data set for all patients treated in all centres.