Cumulative toxicity endpoint opinion

I am a radiation oncologist specializing in GI cancers. I am interested in your brief opinion on a trial endpoint if you have time. We are considering a multicenter trial using proton therapy for anal cancer. which is a novel Radiation for anal cancer. Standard treatment (chemoRadiation) causes significant acute toxicity. Planning data indicates proton radiation offers decreased dose to normal organs compared to standard RT (IMRT).

We are running a small feasibility study currently and are interested in opening this to more sites with a toxicity endpoint as a single arm study. The disease is rare enough that large randomized trials are quite challenging (but this maybe a future goal). Based on your experience, I wondered if you might be able to render an opinion on the following approach.

Rtog 0529 was a national trial using the current standard IMRT with good benchmarks for toxicity rates. The events of interest are GU, GI, Derm, and Hem events that are grade 3 or higher. Looking at IMRT in rtog 0529, there were grade3+ events in 83% of pts(43/52). However, there were actually 53 events in the 52 patients since they had more than one type per patient. All of the events are of interest.

I am interested in a total toxicity endpoint that factors the cumulative number of grade 3+ events. I do not know if there is any precedence for this. For example, in RTOG 0529 there were 1.02 grade3+ events per patient (53/52). Would it seem reasonable to power a trial with hopes of reducing the total grade 3 event per patient rate rather than the proportion of patients that had any grade 3+ event? A reduction of 20% would be clinically meaningful (0.82 grade3+ events per patient). It seems like this approach would be more inclusive of all grade 3+ events in contrast but I have not seen it used before. Any other approaches to account for all grade + events and not simply the proportion of patients experiencing any grade3+ event? I am curious if you think this alternate approach has any merit or if there are major flaws with this.


since there was only one patient with multiple events, any gain in statistical power obtained by accounting for multiple events is likely small. The events are different types ie GI, GU, derm and hem, but you ignore type and just consider any of these as an event?

if so, there was a good paper by Mascha et al. written for a clinical audience that compared methods for analysing multiple binary outcomes, as you have. I’m not sure if it’s apt but he also made available programming code for estimating power i think … Statistical grand rounds: design and analysis of studies with binary- event composite endpoints: guidelines for anesthesia research. He showed that a GEE model has superior statistical power compared to the ‘any-versus-none’ type composite you are using

Thank you for the response! I will look at the paper. All of the event types would be of clinical interest. Since proton therapy is more costly, it would need to be demonstrated that the dosimetric gains in planning could actually translate into meaningful differences in toxicity. For this reason I am interested in considering an endpoint that accounts for all clinically relevant toxicities of grade 3 or higher in some fashion. My concern with the any vs none approach is that in a theoretical model of two treatment modalities, there may be a similar proportion of patients experiencing any grade3+ events, but absolute number of events may still be different if patients have more than one. Hoping to account for this in some way in the design.

This toxicity index endpoint was brought to my attention. Maybe appropriate for these goals.

Rogatko A, Babb JS, Wang H, Slifker MJ, Hudes GR. Patient characteristics compete with dose as predictors of acute treatment toxicity in early phase clinical trials. Clin Cancer Res. 2004;10(14):4645-51. Epub 2004/07/23. doi: 10.1158/1078-0432.CCR-03-0535. PubMed PMID: 15269136.