Per protocol, patients with primary CNS non-Hodgkin's lymphoma in an intergroup phase II trial conducted by the North Central Cancer Treatment Group and the Eastern Cooperative Oncology Group had their cognitive functions measured using the Folstein and Folstein Mini-Mental Status Examination (MMSE) and their physical functions measured using the Eastern Cooperative Oncology Group Performance Score (PS) at study entry, at each treatment evaluation, and at quarterly intervals thereafter until disease progression or death. Of the 53 eligible participants who began therapy, 46 (87%) had baseline MMSE scores recorded, 36 (68%) had at least one follow-up MMSE, and 32 (60%) had both, while 52 (98%) had baseline PS, 49 (92%) had at least one follow-up PS, and 48 (91%) had both. Patterns of MMSE and PS values over time were studied in each individual, in the group as a whole, in the 20 patients who completed the study regimen, in the 23 who survived more than a year, and in patients who were classified as nonprogressors at each key evaluation. For each patient, all recorded values were plotted versus time, with dates of disease progression and death included, to look for signs of decline in cognitive or physical function preceding adverse events. Long-term declines in scores of both cognitive and physical function were observed in many treated patients with primary CNS non-Hodgkin's lymphoma. Nearly all patients who were alive more than 52 weeks after study entry had a demonstrable decline in cognitive and physical functionality. Such declines may occur before disease progression is documented; they may also occur in some patients who have long-term follow-up without evidence of disease progression. Declining MMSE and PS was a poor predictor of disease progression. There was no association of PS and toxicity. The data from this study demonstrated the considerable difficulties we encountered conducting an ancillary study such as this within a multicenter clinical trial. Firstly, the test instruments written into the protocol were unable to tell if the declines seen were due to disease, treatment, co-morbidity, or other factors. Secondly, the missing data created difficulties in interpreting outcome.