Today I want to evaluate an important new paper published in Cancer by Bill Wierda. http://www3.interscience.wiley.com/cgi-bin/abstract/112214594/ABSTRACT?CRETRY=1&SRETRY=0
It is important because it deals with overall survival rather than length of remission. It compares overall survival in relapsed and refractory patients who were treated with one of the following: fludarabine, fludarabine plus cyclophosphamide (FC) or FC plus rituximab (FCR). This information has been available on the lecture circuit for some time, but Bill has been very helpful in publishing it in a form that enables us to scrutinize it carefully. He is to be commended in publishing it without trying to hide any details and even pointing out himself some of the flaws of this kind of study.
I want to go straight to the bottom line. The median overall survival for the fludarabine group was 20 months; for the FC group it was 31 months; for the FCR group it was 49 months. The difference between F and FC had only a 1% risk of having occurred by chance, and the difference between FC and FCR had only a 5% risk of occurring by chance.
Although these figures are impressive and mean that the idea that FCR is better than FC which is better than F alone should certainly be listened to, the way that the data were collected means that the concept is not proven.
The first thing to say is that this was not a randomized trial. Randomization is a great device for ensuring that you are comparing like with like. Scientists who dissect physiological mechanisms prefer to work in inbred strains of mice. All the mice are identical, but humans have so many variables. It is usually impossible to find people who match each other. Even if you choose people with say, Stage 3 or 4, unmutated VH genes, high CD38 levels, high ZAP-70 levels, del 11q, high LDH, high serum CD23 levels, high beta-2 microglobulin levels; even then there will be considerable variation within that group. Patients are all different. Randomization means that the good and poor risk characteristics will tend to cancel each other out is the sample is large enough.
The second thing to say is that these drugs were not tested at the same time. The 241 patients given F were treated between 1984 and 1993; the 111 patients treated with FC were treated between 1995 and 1999; and the 143 patients treated with FCR began treatment between 1999 and 2001. Even if treatment of the disease had not improved in the past 20 years, there is little doubt that supportive care is better. It is admirable that the paper has given so much detail about this. For the F group, there was no routine use of prophylactic antibiotics or growth factors. In contrast some 35 of the FC group were given GM-CSF and 13 amifostine, but again prophylactic antibiotics were not used. For the FCR group Bactrim was given twice weekly and valacyclovir daily. Routine G-CSF was not given, but of course, as time passed G-CSF became much more routinely available in the community, as did better antibiotics and better antifungals.
The third thing is that these patients were not really matched. True, there were no significant differences between the FC and FCR groups in terms of them in terms of age, WBC, Hb, Platelets, beta-2M or albumin, but the F group were 2-3 years older, and had a slightly higher (10K) WBC, and lower HB (1g/dL), though the platelet count was higher (20K). Perhaps more important is what prior treatment the patients had received. 45% of the F group had received 3 or more types of treatment, compared with 31% of the FC group and 29% of the FCR group. In addition 75% of the F group were resistant to alkylating agents, compared to 46% of the FC group and 27% of the FCR group.
We know very well that outcome in CLL is chiefly determined by which ‘bucket’ the CLL falls in. The ‘gold standard’ for determining this is the VH gene mutational status. Unfortunately this was not done in any of these patients. The prognostic test that was done was beta-2M, and the median beta-2M was the same for all three groups. However, beta-2M is influenced not only by the rate of progression of the CLL, but also by the tumor mass, and simply having a lot of CLL is not life threatening.
So, what can we learn from this paper?
First, if you need treatment, and you have already had treatment previously and relapsed or failed to get even a partial remission, then FCR is better than FC in producing remissions, in producing complete remissions, in how long the remission will last and in overall survival. Although, the studies were not contemporaneous and not randomized, the two groups were very similar and the most likely explanation for the difference in outcome is that rituximab really does make a difference.
Second, the addition of cyclophosphamide to fludarabine is still an open question. The F group was more heavily treated than the other two groups, and were treated at a time when supportive care was not so good and significantly at a time when there were no rituximab-containing regimens to rescue them with. We know from another historical comparison (Byrd et al Blood 2005;105:49-53) that rituximab adds similar benefit to fludarabine alone. Although the recent German study shows that FC gives more and better remissions than F alone, so far there has been no difference in overall survival. Therefore the possibility remains that the less toxic FR is no worse than FCR in this situation.
Third, this paper does not address the question of what should be given as first line treatment for CLL. My considered view is still that treatment should be delayed as long as possible and begun with the treatment that does the least harm. But it is still possible that early effective treatment might be most beneficial in certain subgroups.