gms | German Medical Science

The promising safety profile and the high response rates demonstrated by imatinib make it difficult to perform a standard phase III trial in order to compare survival with the previous conservative standard treatments based on interferon-alpha (IFN), since most of the patients randomised to IFN cross over to imatinib [Ref. 1]. Nevertheless, patients, physicians, pharmaceutical companies, and health authorities would appreciate to be able to assess the expected survival advantage with imatinib.

In 1998, for patients treated with IFN the New CML Score was published, a prognostic system which is able to distinguish three risk group with statistically significantly different survival probabilities [Ref. 2]. As a consequence, IFN treatment is meanwhile given up in patients with high risk according to the New CML score. In intermediate-risk and low-risk patients, the achievement of major cytogenetic response (MCR) has a statistically significant impact on survival, independent of the New CML score. Now, the notion was to estimate the rise in survival probabilities for intermediate-risk and low-risk patients when treated with imatinib by developing simulation samples considering (i) the new medication's higher MCR rates and (ii) the survival probabilities for patients with and without MCR stratified for each risk group which were observed in an IFN-treated patient sample.

Our data base of the Collaborative CML Prognostic factors project [Ref. 2] comprised 680 IFN-treated intermediate-risk or low-risk patients with complete data on cytogenetic response and survival. Within 21 months of treatment, 83% of all first MCR had been achieved and the survival of 583 patients was still under observation. The 269 patients of the low-risk group (46% of 680) consisted of 88 patients (33% of 269, 16 of 88 died) with MCR within 21 months and 181 patients without (81 died). The 10-year survival probabilities of patients with MCR vs. patients without MCR were 0.74 [95% confidence interval (c.i.): 0.63; 0.85] vs. 0.20 [c.i.: 0.10; 0.31]. The median survival time of the non-major responders was 78 months. Within the 314 intermediate-risk patients, the MCR rate was 24% (n=75, 18 died) while 239 patients (157 died) remained without MCR during the first 21 months. The respective 10-year survival probabilities equalled 0.60 [c.i.: 0.45; 0.75] and 0.12 [c.i.: 0.05; 0.19], with 66 months median survival time for the non-major responders. To estimate survival probabilities under imatinib treatment, two assumptions were met: (i) Within each of the four groups determined by the specific combination of New CML score risk (low, intermediate) and MCR (yes, no), the survival probabilities under imatinib and IFN treatment were expected to be the same. (ii) For both risk groups, among imatinib-treated patients a 21 months MCR rate of 85% was presumed, a rather conservative estimate deducted from O'Brien et al. [Ref. 2]. For the estimation of survival probabilities under imatinib treatment, in accordance with (i) and (ii), a sample of low-risk patients was constituted by randomly drawing 85% of its patients and their survival data out of the 88 patients with MCR and 15% out of the 181 patients without MCR. In the same way, a patient sample representing imatinib treatment in the intermediate-risk group was gained. For each simulation sample 10,000 random drawings were performed.

The 8,500 drawings for the simulation sample of low-risk patients with MCR resulted in 1,545 deaths and a 10-year survival probability of 0.74. For the 1,500 simulated non-major responders, 690 deaths and a 10-year survival probability of 0.19 were observed, median survival time was 75 months. Of the 8,500 major responders of the simulated intermediate-risk group 1,969 had died and the 10-year survival probability was 0.61, whereas the simulation of the 1,500 non-major responders led to 1,002 deaths, a 10-year survival probability of 0,12, and a median survival time of 64 months. All results were expectedly and necessarily close to those of the corresponding IFN-treated risk group defined by the combination of New CML Score and MCR (yes or no).

Without stratification for MCR, the 269 IFN-treated low-risk patients had a median survival time of 95 months and a 10-year survival probability of 0.37 [c.i.: 0.27; 0.45], altogether. For the 10,000 cases representing imatinib-treated low-risk group, a 10-year survival probability of 0.66 was observed. The 314 IFN-treated intermediate-risk patients revealed 72 months median survival time and a 10-year survival probability of 0.22 [c.i.: 0.15; 0.29], whereas the corresponding simulation sample for imatinib treatment showed a 10-year survival probability of 0.53. Compared to the 10-year survival probabilities of the IFN-treated real patients samples, both samples simulating results under imatinib treatment prognosticated an increase of around 0.30.

In contrast to IFN-based treatment, the MCR rate of patients treated with imatinib is much higher. However, due to the relatively short observation period, median and long-term survival of imatinib-treated patients is not known. With IFN, New CML Score risk group and MCR (yes or no) are validated prognostic factors for survival time. Again, short observation time does not allow conclusions on the meaning of the New CML score nor whether MCR under imatinib induces the same survival advantage as observed with IFN, for sure. However, there is strong evidence that imatinib with its high response rates will lead to an increase in survival probabilities in comparison to IFN.

The idea was to combine the survival probabilities depending on the values of the prognostic factors for IFN treatment with the information on the higher MCR rates under imatinib and to create simulated patient samples prognosticating long-term survival under imatinib treatment. To assume the same survival probabilities for IFN- and imatinib-treated patients within the specific risk groups determined by the combination of New CML score risk (low, intermediate) and MCR (yes, no) was seen as a rather conservative than overoptimistic approach. Comparing the simulated patient samples with the IFN-treated patients samples, within each risk group, the higher MCR rates in the simulation samples led to an estimated increase in the 10-year survival probability of around 0.30, a larger improvement than seen with an earlier, more conservative approach where all risk groups were estimated together [Ref. 3]. Thus, prognosis by means of a simulation study offered the conclusion that increase in survival probabilities with imatinib treatment will be decisive, both in the low-risk as well as in the intermediate-risk group.