Recruitment difficulties in brain tumour patients cause participation bias: findings from a neuropsychological study of adult inpatients with supratentorial intracranial tumours.

Purpose: Patients who participate in questionnaire surveys, clinical studies and clinical trials can be different from patients who do not participate. The occurrence and direction of this response, participation or ascertainment bias is unpredictable, and can harm the external validity of medical research.

Methods: We compared the characteristics of patients with intracranial tumours who participated in a psychological study of inspection time with the characteristics of patients who did not participate for a number of reasons.

Results: Of 178 newly diagnosed adults with intracranial tumours, 136 (76%) were eligible, of whom 76 (56%) participated and 34 (25%) declined. There were no significant differences in terms of age and sex of the patients who participated and those who declined. When the participation group was combined with those who were ineligible and those who declined, the majority of patients in the combined cohort (n = 152) had a WHO grade III or IV glioma (high-grade glioma) (48.0%), and only 13.2% had a WHO grade I or II glioma (low-grade glioma). However, only 38.2% of those who participated had a WHO grade III or IV glioma, and 23.7% had a WHO grade I or II glioma. Comparisons of the participation vs. ineligible and declined groups revealed there was a significant difference (p = 0.002) between the ratio of high-grade to low-grade gliomas in the total and recruited cohorts. Comparisons of only the participation vs. declined groups approached significance (p = 0.051). WHO grade III and IV glioma patients were under-represented, and WHO grade I or II glioma patients were over-represented in the study group.

Conclusions: Noninterventional, non-therapeutic applied neuropsychological studies in neuro-oncology are susceptible to bias since the spectrum of neuropathologies in recruited patients can be significantly different from that of the total cohort. These data could help anticipate recruitment rates for applied neuropsychological studies in clinical neuro-oncology and may help anticipate likely selection biases amongst those who participate.