Survival StrategiesConnecting the Dots: Call offers perspective on GIST trends
By Jerry Call, LRG Science Coordinator
Note: This editorial attempts to convey and analyze some of the recent research material, with a focus on genotyping. It is written by a layperson and is for informational purposes only; it is not a substitute for talking to your doctor. Others, including GIST experts, may have a different point of view. This article should not be construed as representing the opinions of any other person.
At Life Fest 2006, Life Raft Group Executive Director Norman Scherzer and Science Coordinator Jerry Call gave a presentation, “Connecting the Dots” to try to connect much of the new GIST information. Knowledge about GIST continues to grow at a rapid rate. This knowledge comes from clinicians, researchers, drug manufacturers and from patient groups. With so much new information, it can be difficult to interpret it all. In the last 6 months quite a bit of new information about GIST has been published and presented at international conferences such as at ASCO (American Society of Clinical Oncology). Eventually the experts will meet and discuss this information and perhaps revise/ update GIST treatment guidelines. This editorial is about survival decision-making: connecting the dots to survive in the interim.
Much of the new information concerns genotyping. Genotyping is the genetic makeup of an individual, in the form of DNA. Typically, one refers to someone’s genotype with regard to a particular gene of interest. In GIST, it is usually used to describe the common mutations that occur in the KIT and PDGFRA genes usually to the level of the affected exon, e.g., KIT exon 11.
Genotyping can be used for:
1. Determining Gleevec dose levels (Heinrich et al., Debiec-Rychter et al.)
2. Predicting response to Gleevec (Heinrich et al.)
3. Predicting response to Sutent (Demetri, Heinrich, et al.)
4. Generating hypotheses about adjuvant treatment (author’s opinion)
5. Helping to evaluate new drugs
Items 1, 2 and 3 were covered in detail in the July and August 2006 editions of the Life Raft Group newsletter. You can review these by visiting our newsletter section.
The following table was developed by myself and is based on data presented by Dr. Debiec-Rychter, Dr. Verweij and colleagues.
As an aid in understanding this table, it is helpful to understand a few terms. Progression-free survival (PFS) is the length of time from the start of treatment (in this case Gleevec for unresectable or metastatic GIST) until the time treatment fails (significant new growth or new tumors). Median PFS is the time point at which half of the patients have failed treatment and half have not.
In normal English, “significant” means important, while in Statistics “significant” means probably true (not due to chance). A research finding may be true without being important. When statisticians say a result is “highly significant” they mean it is very probably true. They do not (necessarily) mean it is highly important.
P value is a statistical measurement. The logrank test computes a p value that answers this question: If the two populations actually have identical survival curves, what is the chance that random sampling of subjects would lead to as big a difference in survival (or bigger) as you observed? For Table 1, row 1, there is a 2.6 percent probability (p value = .026) of observing a difference this large (between the two curves) if the curves were drawn from the same population (i.e., having the same probability distribution generating curves). An observed difference that large or larger would be expected 2.6 percent of the time.
In the scientific and statistical world, a p value of .05 (5%) is generally the threshold that must be met to be “statistically significant.” The second row of Table 1 has a p value of .20 (20%), not statistically significant.
In Table 1, row 1, we see that the p value is .026. Thus, there is less than a 3 percent chance that the data from which the median PFS difference was calculated would have been generated from identical patient populations (for example if both groups had been taking the same dose of Gleevec); the data is considered to be statistically significant. Row 2 of this table is for 377 patients in the same trial that had mutational testing (genotyping). Row 1 and row 2 both contain patients with all of the different types of mutations (exon 11, exon 9, PDGFRA, wild-type, etc). Row 2 is simply a subset of row 1 that has mutational data. Note, however, that when looking at the same heterogeneous group of GIST patients, row 1 is statistically significant (p=.026) while row 2 is not statistically significant (p=.20). Because the PFS difference between dose groups is relatively small, statistical significance was lost when the sample size was reduced.
When the difference in survival curves is relatively small, it takes large numbers of patients to reach statistical significance. Also note that the method used to construct this table can introduce error. This is evident in the first row where the PFS estimated (by the author) from the survival curves (not shown) is 4 months; the actual PFS calculated from the data (by the EORTC) was 5 months.
From Table 1 we can see that the difference in PFS for (KIT) exon 9 patients is quite large; patients on high-dose Gleevec have about 15.5 months of additional (median) PFS vs. low-dose Gleevec. In addition, the p value (.0013 for the entire exon 9 PFS curve) is quite low. Because the difference is so large, it doesn’t take as many patients to reach statistical significance. Based on the data, we can see that:
1. The difference between dose groups is large. PFS is almost 5 times as long in the high-dose group.
2. High confidence that the data is accurate.
In addition to the PFS data, Dr. Michael Heinrich presented data at ASCO (2005) that exon 9 patients on highdose Gleevec had 8 times as much chance of responding to Gleevec (significant shrinkage) as those on lowdose Gleevec. Even though the data had a low p value (.03) it was not considered statistically significant at the time because it was an interim analysis.
From the data by Debiec-Rychter and Heinrich it is easy to conclude that exon 9 patients should be on highdose Gleevec. In fact, this was recommended in the paper by Debiec-Rychter et al. (European Journal of Cancer).
In fact, the response of exon 9 patients to low-dose Gleevec is so poor that one wonders if any exon 9 patient, including patients taking Gleevec as adjuvant treatment, should be on low-dose Gleevec. Patients that are not able to tolerate high-dose Gleevec (even with a dose-escalation period) have the option of taking Sutent, which has very good activity against exon 9 GIST tumors.
Early data from the phase II GIST trial suggested that wild-type GISTs did not do as well as other GIST patients. They did not respond as well to Gleevec and they did not live as long. This data was based on only 9 patients, however. Recent data from both phase III trials show that wild-type patients have much better survival times than previously thought. In fact, the survival times are similar to exon 9 patients. This raises some questions:
• Is the response of wild-type GIST to Gleevec better than we thought?
• Does wild-type GIST just have a slower natural course than other types of GIST?
It is difficult to interpret the dosage data on wild-type GIST. With a somewhat better initial response on low-dose Gleevec vs. high-dose Gleevec and 83 percent of patients getting significant benefit from crossover to high-dose Gleevec, it is difficult to justify starting patients at higher doses. Wild-type GISTs also have a relatively good response to Sutent with a median PFS of 20.9 months after failure on Gleevec.
In contrast to exon 9 and wild-type GIST, the data for exon 11 GISTs are not as clear-cut. If we were to only consider the EORTC data and we were to use strict scientific criteria, then we would conclude that no significant difference in PFS has been demonstrated. With a p value of 0.25, there is a 25 percent chance of getting as big (or larger) a difference as shown by the survival curves (from which the 4 months PFS advantage was calculated) even if both groups were exactly the same.
The exon 11 data is potentially subject to the same loss of significance (due to a reduction in numbers) as the overall data. Remember that the overall difference in PFS went from a statistically significant p value of .026 to a not statistically significant p value of .20 when the number of patients in the analysis was reduced from 946 to 377. Smaller differences in PFS require larger numbers of patients to prove that a real difference exists. The EORTC data is designed to be combined with the U.S./ Canadian phase III trial. When this occurs, the exon 11 picture should become clearer as the power of the study will increase.
When considering clinical results, it makes sense to evaluate all of the available information. The EORTC data is very valuable and will become even more valuable when it is combined with the other phase III data. But it is not the only information available. Would limited effective treatment options after failure of low-dose Gleevec lend more weight to the argument for higher-dose Gleevec? What are the options (and how effective are they) after Gleevec failure?
• Crossover to 800 mg after progression was only effective (as measured by the growth modulation index) in 7 percent of exon 11 patients in the EORTC study.
• With Sutent, the median PFS is about 5 months and only about 35 percent of exon 11 patients get 6 months of stability (significant shrinkage is fairly rare).
Other arguments favoring higher doses for exon 11 patients include:
• The dilution of data from dose reductions using intent-to-treat analysis. In the EORTC trial, 60 percent of patients assigned to 800 mg had permanent dose reductions (to 600 mg, 400 mg or below) but are still counted in the high-dose arm. This type of analysis (which is the gold standard) may tend to estimate the minimum potential benefit of the highdose arm (in this case).
• An internal LRG study found that when comparing actual dose to intent-to-treat dose, a greater difference in PFS was observed, with higher doses showing a greater benefit. This study has limitations including possible selection bias and subjective progression criteria (patient-reported data). The effect may be that this study estimates the maximum possible benefit of the high-dose arm.
• Gleevec levels may fall as much as 33 to 40 percent over time (several possible reasons have been cited). This may be riskier for patients who are on lower doses.
With exon 11 GISTs, an argument can be made for either low-dose Gleevec or high-dose Gleevec. Lowdose proponents can cite the data has failed to show statistical significance and the PFS difference is relatively small. High-dose proponents can cite the lack of efficacy of crossover to high-dose Gleevec and Sutent and that 4 to 5 months PFS benefit (which may exist) for high-dose Gleevec is equal to the benefit they might get from Sutent.
Gleevec produces side effects which can be significant. These side effects are worse at higher doses. The long-term effects of Gleevec are not known and could be worse at higher doses. With exon 11 patients, there seems to be a wider therapeutic range of Gleevec dosing especially compared to exon 9. Patients that might be more suitable for low-dose Gleevec (400 mg) might include:
• Those with more significant side effect issues.
• Those that have less confidence that the data shows a difference that is significant to them.
• Those that are willing to accept a little more risk of progression.
• Patients with good adherence to taking Gleevec (they don’t forget, reduce or skip Gleevec).
Patients that might be more inclined to want higher doses of Gleevec might include:
• Patients with less side effects.
• Patients that believe the overall data shows a difference between doses that is significant to them.
To summarize, exon 11 patients and their physicians appear to have a lot more flexibility. The wider therapeutic range appears to allow more fine-tuning to balance side effects against the possibility of longer PFS.