Patterns of missing mini mental status exam (MMSE) in radiation therapy oncology group (RTOG) brain cancer trials

The Mini Mental Status Exam (MMSE) instrument has been commonly used in the Radiation Therapy Oncology Group (RTOG) to assess mental status in brain cancer patients. Evaluating patient factors in relation to patterns of incomplete MMSE assessments can provide insight into predictors of missingness and optimal MMSE collection schedules in brain cancer clinical trials. This study examined eight RTOG brain cancer trials with ten treatment arms and 1,957 eligible patients. Patient data compliance patterns were categorized as: (1) evaluated at all time points (Complete), (2) not evaluated from a given time point or any subsequent time points but evaluated at all the previous time points (Monotone drop-out), (3) not evaluated at any time point (All missing), and (4) all other patterns (Mixed). Patient characteristics and reasons for missingness were summarized and compared among the missing pattern groups. Baseline MMSE scores and change scores after radiation therapy (RT) were compared between these groups, adjusting for differences in other characteristics. There were significant differences in frequency of missing patterns by age, treatment type, education, and Zubrod performance status (ZPS; P < 0.001). Ninety-two percent of patients were evaluated at least once: seven percent of patients were complete pattern, 49% were Monotone pattern, and 36% were mixed pattern. Patients who received RT only regimens were evaluated at a higher rate than patients who received RT + other treatments (49-64% vs. 27-45%). Institutional error and request to not be contacted were the most frequent known reasons for missing data, but most often, reasons for missing MMSE was unspecified. Differences in baseline mean MMSE scores by missing pattern (Complete, Monotone dropout, Mixed) were statistically significant (P < 0.001) but differences were small (<1.5 points) and significance did not persist after adjustment for age, ZPS, and other factors related to missingness. Post-RT change scores did not differ significantly by missing pattern. While baseline and change scores did not differ widely by missing pattern for available measurements, incomplete data was common and of unknown reason, and has potential to substantially bias conclusions. Higher compliance rates may be achievable by addressing institutional compliance with assessment schedules and patient refusal issues, and further exploration of how educational and health status barriers influence compliance with MMSE and other tools used in modern neurocognitive batteries.