It has been roughly 20 years since chemotherapy dose escalation was proposed as a possible strategy for improving outcomes in patients with breast cancer.[1,2] This concept has sustained a series of remarkable rollercoaster-like controversies, with heated arguments at national meetings, substantial lay press coverage, patients suing their insurance companies seeking coverage, legislative fiats requiring third-party payment long before critical data were available, and a well-publicized episode of clinical scientific fraud that is nearly unprecedented in its audacity. How did we get here from there?
Clearly, there is a dose response for chemotherapy in breast cancer, both in laboratory models and in the clinic. Indeed, in both the metastatic and adjuvant settings, abundant data demonstrate that some therapy is better than no therapy. Prospective, randomized clinical trials have demonstrated an added benefit of increasing the dosage of regimens like CMF (cyclophosphamide [Cytoxan, Neosar], methotrexate(Drug information on methotrexate), fluorouracil(Drug information on fluorouracil) [5-FU]) and CAF (cyclophosphamide, Adriamycin, 5-FU) from levels of minimal toxicity to those that produce modest toxicity.[3,4] So, why hasn’t the promise of "more is better" rung true?
In this issue of Oncology, Drs. Armstrong and Davidson provide an elegant discussion of the preclinical and theoretical infrastructures that have supported an enormous amount of clinical research, as well as personal angst (for patients, physicians, and the public at large). They conclude, correctly, that "despite nearly 2 decades of study, the hypothesis that increasing chemotherapy doses above standard levels will positively affect clinical outcome in breast cancer remains unproven."
This statement, with which I absolutely agree, represents a colossal disappointment. Put simply, dose escalation for patients with breast cancer, especially to very high doses requiring bone marrow stem cell support (ie, high-dose chemotherapy/bone marrow transplantation, or HDC/BMT), does not represent the cure for cancer we had hoped for. Nonetheless, progress has been made, and there are lessons to be learned.
First, as has always been true in every scientific endeavor, a "good idea is not a fact." By 1980, very high-dose chemotherapy had been shown to cure a percentage of patients with leukemia who were destined to die of their disease after standard therapy. Indeed, this observation earned a well-deserved Nobel prize. However, in a second lesson learned, breast cancer is not leukemia. Early enthusiasm over high-dose therapy for breast cancer should have been tempered (and it was in some circles) by a thorough knowledge of the long and heterogenous natural history of breast cancer, even in the metastatic setting.
Barring nearly 100% cure rates in the metastatic setting, historical controls are of value only in the very early stages of clinical experience in this disease. There is no question that phase I trials of HDC/BMT were required and important, as treatment-related mortality dropped from an unacceptable 30% to the current level of 3% to 5% from the early 1980s to the late 1980s. Clearly, prospective, multi-institutional, randomized trialsespecially in patients with relatively favorable prognosescould not have been performed in the early 1980s.
However, the principal drop in mortality for patients undergoing high-dose chemotherapy was achieved by the early 1990s, and it is unconscionable that we are just now getting the results of the prospective, randomized trials a decade later. The heterogeneity of the natural history of this disease, coupled with the pitfalls of selection bias, seemed to have been forgotten by many. Biases are hard to overcome, whether in society at large or in medicine. Nonetheless, by the late 1980s the safety of HDC/BMT was sufficiently improved, so that only a few phase II trials demonstrating some promise for this approach were required to proceed with large prospective, randomized clinical trials, to definitively investigate its relative worth.
As with any new approach, everyone involved should have accepted the investigational nature of HDC/BMT and agreed to keep their enthusiasm sufficiently muffled to facilitate the rapid completion of these important trials. As with any investigational therapy, no center should have offered HDC/BMT outside of a clinical trial, and for the most part, those trials should have been concurrently controlled studies.
At the 1993 meeting of the American Society of Clinical Oncology (ASCO), more than 25 phase II studies of HDC/BMT for breast cancer were reported, each using strikingly similar regimens and each coming to the same conclusion: High-dose chemotherapy is feasible and promising for patients with breast cancer. Since at least three National Cancer Institute-sponsored, multi-institutional, prospective, randomized clinical trials addressing HDC/BMT were already open in the United States by this time, these conclusions seemed anticlimactic.
Those patients, and the other thousands who were treated completely outside of a clinical trial, should have been offered the opportunity to participate in the large prospective, randomized clinical trials being run at that time in the United States. The largest two of these, which both targeted women with 10 or more positive nodes and who did not have metastases, were performed by the Cancer and Leukemia Group B (CALGB) and the Eastern Cooperative Oncology Group (ECOG). Sadly, they took an inordinately long time to accrue. Indeed, the ECOG study, which opened in 1990 and required only approximately 700 patients, has still not been reported. The CALGB study, for which preliminary results have been reported, required nearly 6 years for accrual. During the time it took to enroll roughly 1,500 patients to these two studies (1990-1997), more than 12,000 node-positive patients were enrolled into other US Intergroup and National Surgical Adjuvant Breast and Bowel Project trials.
Admittedly, these rapidly accruing studies were more successful, in part, because the pool of patients was larger. (The HDC/BMT studies were restricted to patients with 10 or more nodes). However, it is also clear that the HDC/BMT studies were less successful because of the vitriolic debate, fueled by investigator and physician bias, which was played out in the public arena. In fact, during this time, the Autologous Blood and Marrow Transplant Registry (ABMTR) recorded nearly 1,500 stage II and III breast cancer patients who were treated with HDC/BMT outside of randomized trials.
Of note, the ABMTR estimates that it captured only 50% or fewer patients who were actually treated. I suspect that had these 3,000 or so patients been told that HDC/BMT was a "good idea" but far from proven, many would have participated in the clinical trials to determine whether the potential benefit was a "fact." The ultimately successful accrual of the CALGB and ECOG trials, driven by those who were willing to test their biases, speaks to this opinion.
Is the Concept of HDC/BMT Dead?
A third lesson to be learned is that a single study does not a whole story make. The series of Oxford meta-analyses clearly demonstrate that even a single well-designed clinical trial is not as meaningful as several highly powered studies addressing the same question.[7-9] Following the 1999 ASCO meeting, at which the preliminary results of several trials were released, it was widely reported by the lay press that the idea of HDC/BMT was effectively dead. It was remarkable how quickly the pendulum had swung from the overabundant enthusiasm of the late 1980s to dogmatic skepticism, based on heterogeneous and immature data.
At the time of this writing, at least five prospective, randomized clinical trials of HDC/BMT for high-risk adjuvant patients have been reported. Each of them has a different trial design with different follow-up. The results from some of them appear completely equivalent, while others suggest a small benefit. Currently, at the least, we can conclude that HDC/BMT is not the "home run" that had been anticipated. In other words, it clearly does not have the impact of penicillin for pneumonia or cisplatin(Drug information on cisplatin) (Platinol) for testicular cancer. However, there may be some small benefit for selected patients. The challenges now are to determine the magnitude of that benefit and to assess whether that benefit is worth the risks, toxicities, and costs for individual patients and society at large.
Making Use of Available Results
The HDC/BMT experience suggests that many patients with metastatic disease achieve a complete remission, and it is reasonable to consider HDC/BMT as a platform for other investigational therapies. In that regard, Drs. Armstrong and Davidson discuss the use of immunotherapies. Integrating HDC/BMT and immunotherapy is particularly appealing. One may wish to manipulate bone marrow stem cells ex vivo so that they serve as mediators of the immune response, for example, by inserting immune-regulating genes or treating with various immunologically active cytokines.[11,12] Indeed, one may not need the HDC aspect of HDC/BMT to achieve this goal. In this regard, the field of minitransplants, in which bone marrow stem cells are harvested, manipulated, and reimplanted without bone marrow ablation, is progressing rapidly.
Other novel therapies might be applied with or after HDC/BMT, such as the application of inhibitors of angiogenesis or tyrosine kinases. In theory, these might be more effective in patients with minimal residual disease than in those with large established metastases. Although it is anticipated that these novel therapies will be tested in the true adjuvant setting, HDC/BMT might reduce tumor burden in patients with more advanced disease to the point where they would have an improved chance of working.
Patients Who Might Achieve Greater Benefit
Finally, in the debate about HDC/BMT, we seem to have forgotten another lesson from our experience with breast cancer: Based on either prognostic or predictive factors, subgroups of patients may be identified who are more likely to benefit than others. Currently, patients have been selected as candidates for HDC/BMT based solely on prognostic features that principally delineate tumor burden (locally advanced disease, large number of positive nodes, or metastases). A growing body of literature has demonstrated that certain molecular features can serve as predictive factors for specific therapies.[14-16] The best example of such a predictive factor is the very tight correlation of estrogen-receptor content with benefit from endocrine therapy.
Recent studies have suggested that markers like p53 mutations and overexpression/amplification of erbB2 may predict the activity of specific chemotherapeutic agents, such as the anthracyclines and taxanes.[17,18] Given the reliance on alkylating agents in most HDC/BMT regimens, it seems possible that measures of alkylating agent resistance, such as glutathione synthetase or O-methyltransferase levels and/or activity, might identify subgroups in whom dose escalation is more effective.
In summary, it is time for us to get beyond the debate about whether HDC/BMT "works." Currently, we can conclude that there is a small window of patients for whom it is not inappropriate (an intentional double-negative) and for whom participation in studies that address the next generation of questions is appropriate. As with all new and investigational therapies, with time we will be able to determine the relative worth of HDC/BMT. If an additional benefit does emerge, society must decide whether that benefit outweighs the costs.
In the meantime, our focus needs to shift from questions regarding dose to very different approaches, which may or may not encompass HDC. The exciting decline in breast cancer mortality during the past decade is tempered by the unacceptable loss of 40,000 women this year from the disease. Clearly, we must redouble our efforts to bring exciting new therapies into the arena. In doing so, we must also remember lessons from the past and remain cautiously optimistic but scientifically skeptical.