Chemotherapy for Primary Central Nervous System Lymphoma

Antonio M. P. Omuro, MD; Lauren E. Abrey, MD


Neurosurg Focus. 2006;21(5) 

In This Article

Interpreting the Literature on PCNSL

In the absence of Phase III trials, interpreting results of retrospective studies, patient series, pilot studies, and Phase II trials becomes essential for making evidence-based treatment decisions. However, such studies are frequently small, and large confidence intervals are observed around the reported outcome measures.[1] Commonly reported end points such as response rates and survival are not adequate measures of efficacy in PCNSL. Response rates are frequently reported because they are easy to compare and can also be assessed in regimens that subsequently use WBRT. However, most regimens for PCNSL achieve similarly high response rates to chemotherapy and steroid agents, and therefore do not discriminate between various modalities. Moreover, if responses are not durable, the timing of the response assessment may significantly influence results. Shorter treatments may overestimate response rates in comparison with longer treatments because response assessment and confirmatory scans are obtained earlier in the course of the disease, when transient responses may still be seen. Likewise, when combined chemo- and radiotherapy regimens are used, response rates to chemotherapy obtained prior to WBRT are over estimated and cannot be confirmed, because final response rates assessed after WBRT are confounded by the activity of radiotherapy.

The OS is another frequently reported measure of efficacy that is reliable and easy to compare. However, this measure may be significantly influenced by salvage treatments, which are often effective in PCNSL, particularly after chemotherapy-only regimens; thus, considerable attention must be given to which salvage strategies were used when interpreting survival data. Because of the limitations in interpretation of response rates and OS, intent-to-treat EFS, time to treatment failure, or PFS are more adequate measures of efficacy in PCNSL and allow more reliable and appropriate comparisons among different regimens.

The risk of neurotoxicity is another important aspect to be considered. Various definitions of neurotoxicity have been used, ranging from subtle changes in the results of neuropsychological evaluation to severe and incapacitating dementia. Therefore, it is difficult to compare different studies for risk of neurotoxicity. Moreover, delayed neurotoxicity is a time-dependent variable because the incidence increases over time and is cumulative. Thus, variability in the extent of follow up among reported studies may render a comparison of data on neurotoxicity impossible. As a time-dependent variable, Kaplan-Meier analysis with competing risk estimates would be the optimal way to overcome differences in follow-up duration, but using such methodology is difficult in the absence of a reproducible and standardized definition of neurotoxicity.[34] Studies encompassing long-term follow-up data and updated reports of previously published trials may provide more reliable information on long-term outcomes and neurotoxicity rates.

Finally, it is important to consider the institutions participating in a study, and to take into account whether the study is a single-center or multicenter one. There is growing evidence that PCNSL treatment delivered in large referral centers is associated with less toxicity and higher efficacy than in smaller institutions that enroll fewer patients and have limited resources.[29] Therefore, comparison with the same institution's historical controls may provide more reliable conclusions than comparison with literature-derived data.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.
Post as: