Soft Tissue Sarcoma Trials: One Size No Longer Fits All

Soft tissue sarcomas are a heterogeneous group of diseases that share a mesenchymal origin. Despite this heterogeneity, these tumors have always been pooled together in clinical studies and treated as if they were all the same. This issue of Journal of Clinical Oncology features five articles that each, in their own way, indicate that soft tissue sarcoma trial design should considerably be changed in future studies. Recent molecular biology research has identified several subtypespecific oncogenes and their protein products that could serve as treatment targets. Such cancer cell–specific treatment will obviously depend on the molecular characteristic that, in the case of soft tissue sarcomas, may not be shared by all of the subtypes and would thus require subtype-specific trials and treatments. GI stromal tumors (GISTs) may serve as the model for this. Until the year 2000, GIST used to be treated similarly to all other soft tissue sarcoma subtypes, with minimal effects from chemotherapy. Overexpression of the oncogene product receptor c-KIT (CD117), often harboring activating mutations, turned out to be a frequent—albeit not uniform—characteristic of GIST. The activity of the KIT tyrosine kinase (TK) inhibitor imatinib was clearly shown from the year 2000 onward. Subsequent detailed analysis has revealed that exon-9 mutations in KIT are less sensitive to imatinib than others. And the majority of KIT-negative GISTs are still sensitive to imatinib because they express mutated platelet-derived growth factor, one of the other targets of the agent. The data indicate not only that the disease may be drug-sensitive as a result of specific molecular changes, but also that the required drug dose may be dependent on specific changes within these molecular changes. This in itself is a historic finding. In this issue of JCO, Demetri et al add an additional layer to dose selection. In their randomized phase II study that was the basis of marketing approval for imatinib in GIST, they sampled the first 73 of the total 147 patients to assess imatinib pharmacokinetics (PK). By analyzing the PK data, they found that despite a high interpatient variability, clinical outcomes seemed to correlate with imatinib trough plasma levels at steady-state. Patients with a minimum plasma concentration of imatinib less than 1,100 ng/mL showed a shorter time to progression of 11.3 months, compared with more than 30 months for patients with higher minimum plasma concentration. In patients harboring KIT exon 11 mutations in the tumor (n 39), which are now known to be the most sensitive to imatinib, the clinical benefit rate (complete response partial response stable disease) was 67% for patients with trough levels less than 1,100 ng/mL, compared with 100% for all others (P .001). It is important to realize that the clinical benefit was assessed by RECIST (Response Evaluation Criteria in Solid Tumors Group). We now know that RECIST does not appropriately discriminate between complete response, partial response, and stable disease in GIST but is appropriate for assessing progressive disease. Therefore, by using the clinical benefit rate (as opposed to progression) and time to progression end points, the reported observations likely have major relevance, although the sample size for the analysis is fairly small, which limits the power of the observation. We can thus not yet recommend using the data for routine monitoring purposes. As stated by the authors, phase III studies should additionally test whether monitoring imatinib plasma levels might optimize clinical outcomes for GIST patients. The study allowed dose escalation from 400 mg to 600 mg in the case of progressive disease. By PK quartiles, the frequency of dose escalation was nine of 18 (50%) in Q1, 15 of 36 (41.7%) in Q2-Q3, and four of 19 (21%) in Q4 (P .165, 2 test). So the higher the exposure, the lower the number of processed dose escalations. Although this could lend additional support to the suggested dependence of antitumor effect on drug exposure levels, it is not known if dose escalation was performed in a consistent way. The measured day 29 area under the curve (AUC) in this study seems to be consistently lower than the day 1 AUC, in line with previous observations. This could point to an increase of drug clearance over time and raises the possibility of AUC-guided intrapatient dose escalation over time. Again, much larger data sets will be needed to explore this option. The most important conclusion from this study for trial design is that for targeted agents such as imatinib, dose selection may be based on specific mutations in the given target as well as on individual plasma exposure. For phase III studies with such agents, it may thus be important to include simple PK assessments by measuring trough levels, as well as to include detailed analysis of the molecular abnormality in the tumor tissue of the patients in the study. As indicated above, soft tissue sarcomas perhaps should not be pooled together in clinical studies. In an effort to assess this problem, Chugh et al performed a phase II study in 190 soft tissue sarcoma patients treated with imatinib, published in this issue of JCO, using a Bayesian hierarchical statistical model. This is an interesting model that borrows information across disease subtypes and allows for evaluation of each subtype individually. The model used both cohort expansion and early stopping rules, defined in advance, that allowed the outcomes in one subtype to influence the decision making in other subtypes. This was done under the assumption that the various subtypes are interrelated. When the study was started, it was not yet known that, in particular, the latter assumption for the molecular targets (KIT and platelet-derived growth factor) involved and the JOURNAL OF CLINICAL ONCOLOGY E D I T O R I A L VOLUME 27 NUMBER 19 JULY 1 2009