- © 1999 by American Society of Clinical Oncology
Safety and Efficacy of Using a Single Agent or a Phase II Agent Before Instituting Standard Combination Chemotherapy in Previously Untreated Metastatic Breast Cancer Patients: Report of a Randomized Study—Cancer and Leukemia Group B 8642
- Mary E. Costanza,
- Raymond B. Weiss,
- I. Craig Henderson,
- Larry Norton,
- Donald A. Berry,
- Constance Cirrincione,
- Eric Winer,
- William C. Wood,
- Emil Frei III,
- O. Ross McIntyre,
- Richard L. Schilsky and
- Cancer and Leukemia Group B
- From the Department of Medicine, University of Massachusetts Medical School, Worcester, MA; Walter Reed Army Medical Center, Washington, DC; Uniformed Services University of the Health Sciences, Bethesda, MD; University of California, San Francisco, CA; Memorial Sloan-Kettering Cancer Center, New York, NY; Duke University Medical Center, Durham, NC; Dana-Farber Cancer Institute, Boston, MA; Emory University School of Medicine, Atlanta, GA; Dartmouth Medical School, Cotton Cancer Center, Lebanon, NH; and Cancer and Leukemia Group B, Chicago, IL.
- Address reprint requests to Mary E. Costanza, MD, Department of Medicine, Division of Hematology Oncology, University of Massachusetts Medical Center, 55 Lake Ave North, Worcester, MA 01655; email mary.costanza{at}banyan.ummed.edu
Abstract
PURPOSE: We undertook a prospective, randomized phase III trial to evaluate the safety and efficacy of using a phase II agent before initiating therapy with standard combination chemotherapy in metastatic breast cancer patients.
PATIENTS AND METHODS: A total of 365 women with measurable metastatic breast cancer, previously untreated with chemotherapy for their metastatic disease, were randomized to receive either immediate chemotherapy with cyclophosphamide, doxorubicin, and fluorouracil (CAF) or up to four cycles of one of five sequential cohorts of single-agent drugs: trimetrexate, melphalan, amonafide, carboplatin, or elsamitrucin, followed by CAF.
RESULTS: The toxicity of each single agent followed by CAF was comparable to that of CAF alone. The cumulative response rates for the single agent followed by CAF were not statistically different from those of CAF alone (44% v 52%; P = .24). However, in the multivariate analysis, patients with visceral disease had a trend toward lower response rates on the phase II agent plus CAF arm (P = .078). Although survival and response duration also were not statistically significantly different between the two study arms (P = .074 and P = .069, respectively), there was a suggestion of benefit for the CAF-only arm.
CONCLUSION: The brief use of a phase II agent, regardless of its efficacy, followed by CAF resulted in response rates, toxicities, durations of response, and survival statistically equivalent to those seen with the use of CAF alone. These findings support the use of a new paradigm for the evaluation of phase II agents in the treatment of patients with metastatic breast cancer.
THE TREATMENT OF metastatic breast cancer has had limited success, with the goals of most physicians being symptom control, objective reduction in tumor burden, and possibly increased survival time, but generally not cure. With conventional chemotherapy regimens, approximately half of the patients with metastatic disease will obtain a response (usually a partial response), with a median duration of remission of less than 1 year. Many experienced practicing physicians believe that the meticulous use of sequential treatments, including hormone therapy, will result in superior survival rates. When chemotherapy is instituted in the metastatic setting, a frequently used guiding principle is to attempt to maximize response and response duration by first using the best standard chemotherapy, followed at progression with second- and third-line drugs. This strategy is diametrically opposed to that of some oncologists,1 who are convinced that studying phase II agents in chemotherapy-naive patients will alter little in the overall picture, that metastatic disease is virtually incurable at present, and that survival time is essentially uninfluenced by treatment regimens. This latter position has been challenged by two recent reports showing a modest but significant survival advantage of one treatment over another.2,3 These results emphasize the importance of studying the safety of using unproven or ineffective agents in metastatic breast cancer.
The expected benefit of studying phase II agents “up front” is the more rapid discovery of the true activity of previously untested new drugs. Until recently, phase II agents were first tested in patients whose tumors were no longer responding to most established drugs. Because these tumors were likely resistant to chemotherapy, not surprisingly, many phase II agents did not seem to have activity. One of the most effective drugs in the current treatment of breast cancer is doxorubicin. When initially tested in the phase II setting, doxorubicin was almost discarded as inactive because of its poor performance.4 The results of similar trials with heavily pretreated patients amplify this point.5 Used in a chemotherapy-naive group of patients, this drug may yield response rates that are three and four times as great as those observed in chemotherapy-refractory patients. Continuing attempts to find active drugs through the usual phase II mechanism are imperfect and perhaps self-defeating by design.6,7 Providing a more rapid and a fairer evaluation of potentially active agents is an important and pressing goal because progress during the last decade in improving survival in metastatic breast cancer has been poor.8 In addition, the discovery of better agents would most likely translate into an improvement in outcomes after adjuvant chemotherapy.
Although there are theoretical concerns that delaying the initiation of combination chemotherapy might adversely affect survival time and that the use of an ineffective or less effective single agent might permit the emergence of drug-resistant clones, previous studies comparing established single agents versus combination chemotherapy have failed to show such negative outcomes. Of seven studies comparing established single agents against combination therapy in advanced breast cancer, none has shown a significantly increased survival for patients receiving the combination therapy.9-15
There are no data regarding the potential impact of ineffective single agents before standard therapy. Neither is there evidence that briefly delaying the start of effective chemotherapy in patients with advanced disease might shorten their survival. This situation is not uncommon because experienced clinicians may intentionally defer treatment of patients with recurrent breast cancer for several months while they assess its rate of progression. Such a brief delay might be thought of as equivalent to the use of an ineffective phase II agent.
When this study was begun in 1986, the general oncologic community believed that it was imperative to conduct a definitive study assessing the safety of such a strategy, because a few academic centers had already committed themselves to using up-front phase II chemotherapy in metastatic breast cancer patients even though this had not been demonstrated to be a safe strategy.16 In subsequent years, many academic centers have used this strategy.17-23 The study reported here presents the final results of a large randomized trial, Cancer and Leukemia Group B (CALGB) 8642, designed to address the question of whether the use of a phase II agent before the use of a standard combination chemotherapy regimen would result in a significant difference (an increase or decrease) in toxicity, response rates, response duration, or survival times compared with the use of the standard chemotherapy regimen alone. To provide data about different classes of agents (those with different mechanisms of drug cytotoxicity), five single agents from four different drug classes were chosen for use as candidates for the single phase II drug tested in the experimental arm.
PATIENTS AND METHODS
Patient Selection and Eligibility Criteria
Women with histologically documented advanced breast cancer (stage IV, or inoperable disease) who had measurable disease and a performance status of 0 or 1 (Eastern Cooperative Oncology Group scale) and who had not received chemotherapy for their advanced disease were eligible for this study. Additional eligibility criteria included age more than 16 years and physiologic age less than 70 years; more than 12 months elapsed since completion of adjuvant chemotherapy; a total doxorubicin dose of 250 mg/m2 or less in the adjuvant setting; more than 4 weeks elapsed since hormone therapy cessation; more than 2 weeks elapsed since radiotherapy or major surgery; less than 50% of the pelvis previously irradiated; expected survival of more than 4 months; and no visceral crises as defined by lymphangitic spread to the lungs, bone marrow replacement, carcinomatous meningitis, or significant liver disease. In addition, the initial laboratory profile had to show adequate organ function, ie, granulocyte count greater than 1,800/μL, platelet count greater than 100,000/μL, hemoglobin level greater than 10 g/μL, creatinine level less than 1.8 mg/μL, blood urea nitrogen level less than 1.5 times normal, bilirubin level normal, AST and ALT levels less than 1.5 times normal, and alkaline phosphatase level less than 1.5 times normal. Patients could not have bilateral breast cancer, previous or concomitant malignancy, or other serious medical or psychiatric illnesses. Written informed consent was required.
Drug Dose and Schedule
Standard combination chemotherapy arm.
Cyclophosphamide, doxorubicin, and fluorouracil (CAF) was chosen as the standard combination therapy because it was a regimen familiar and acceptable to CALGB member investigators. The doses were cyclophosphamide 600 mg/m2 intravenously (IV) on day 1, doxorubicin 45 mg/m2 IV on day 1, and fluorouracil 500 mg/m2 IV on days 1 and 8 repeated every 4 weeks. After a total doxorubicin dose of 540 mg/m2 (including any adjuvant doxorubicin), methotrexate was to be substituted at 40 mg/m2 IV on days 1 and 8 (30 mg/m2 for patients > 60 years old). Treatment was to continue unless there was intolerable toxicity or until progression of disease.
Trimetrexate.
Trimetrexate (TMTX; NSC 352122) is a 2,4-diaminoquinazolone lipophilic derivative of methotrexate. As such, it is an effective folic acid antagonist, inhibiting dihydrofolate reductase. The TMTX concentration is not affected by the presence of leucovorin or methotrexate, which suggests a different transport mechanism.24-26 The dose of TMTX was 8 mg/m2/d IV for 5 days, repeated at 21-day intervals.27-30
Melphalan.
Melphalan (MEL) acts via an alkylating process that involves the transfer of alkyl groups to susceptible macromolecules such as DNA. MEL was chosen for this study because it is an alkylator and had not been studied either as an intravenous agent or as a single agent in patients with previously untreated metastatic breast cancer. The dose and schedule was 30 mg/m2 IV on day 1, repeated every 21 days.
Amonafide.
Amonafide (AMON; benzisoquinolinedione, NSC 308847) is a synthesized imide derivative of naphthalic acid31 and a site-specific intercalating agent and a topoisomerase II inhibitor.32-34 AMON was chosen because it was a previously untested phase II agent. The AMON dose was 300 mg/m2 IV four times a day for 5 days to be repeated every 21 days, with 100-mg/m2 escalations, blood counts permitting.35-37
Carboplatin.
Carboplatin (CBDCA; NSC 241240), or cis-diamine-1,1-cyclobutane dicarboxylatoplatinum II, is a cisplatin analog in which the carbon-1 groups are replaced by a heterocyclic organic structure. The initial dose was 400 mg/m2 given as an IV bolus.38-41 This was to be escalated by 50 mg/m2 depending on day-1 and nadir counts. The cycle was to be repeated every 28 days. CBDCA was chosen as a phase II agent because of its mechanism of action and because at the time of this study it was untested in patients with previously untreated metastatic breast cancer.
Elsamitrucin.
Elsamitrucin (ELSA; BMY-28090) is a novel actinomycete fermentation product42,43 that induces single-strand breaks in DNA and also inhibits topoisomerase I and II. The drug was to be given weekly for 16 weeks, which is equivalent to the length of four cycles of CAF. The ELSA dose was 25 mg/m2 IV bolus.44-46 ELSA was chosen for study because of its purported mechanism of action and because it was a previously unevaluated phase II agent.
Treatment Guidelines
Treatment duration.
Patients on the phase II arm were to be treated with no more than four cycles of the phase II agent before going on to standard CAF. If during this time the patient's disease progressed, the phase II agent was to be stopped and the patient was to go on to standard CAF therapy when the next cycle was due. Thus, patients whose disease progressed on treatment could have received as little as one cycle of phase II agent before beginning standard CAF. Stable or responding patients were to receive four cycles of therapy. At the end of four cycles, all patients, regardless of response status, were to begin standard CAF therapy.
Dose changes.
Doses were reduced according to standard accepted hematologic and chemical parameters. Dose escalation was mandated for two of the phase II agents: AMON and CBDCA. In the first instance, the escalation was required because of the variability in toxicity observed in individual patients during phase I testing. We now know from differences in patient acetylator status and the subsequent rate of plasma clearance and the generation of a toxic metabolite that this variation is related to alterations in AMON metabolism.47,48 In the case of CBDCA, escalations were mandated based on lack of bone marrow toxicity. At the time of the study, dosing by area under the curve was not a general practice. Dose adjustments for AMON and for CBDCA were to be made as follows. If day-15 granulocytes were equal to or more than 1,800/μL and platelets were equal to or more than 100,000/μL, the dose of AMON was to be increased by 100 mg/m2/d for 5 days. Under the same conditions, the CBDCA dose was to be increased by 50 mg/m2 on the next cycle of treatment.
Patient Evaluation
Pretreatment evaluation.
Pretreatment evaluation consisted of the following: a complete history and physical examination; performance status assessment; tumor measurements; and assessment of principal site of metastatic disease (visceral, soft tissue, or bone). Laboratory tests included the usual hematologic and chemical parameters, urinalysis, electrocardiogram, chest x-ray, and bone scan. Computed tomography of the abdomen and liver or bone marrow aspiration and biopsy were required only if there was clinical suspicion of disease.
On-treatment evaluation.
Evaluation during treatment included complete history and physical examination, tumor measurements, complete blood counts, and blood chemistries and urinalysis on day 1 of each cycle. Weekly complete blood counts were also required. Drug toxicity was to be evaluated on day 1 of each cycle. Liver scans and chest radiographs, if positive at baseline, were to be repeated every two cycles in the first 4 months, and then once every 4 months.
Criteria for evaluation.
Patients were stratified by estrogen receptor protein status, dominant site of metastatic disease, and previous adjuvant chemotherapy status. Toxicities were assessed using the CALGB expanded common toxicity criteria. Complete response (CR) was defined as the disappearance of all signs and symptoms of disease, without the appearance of new lesions, for a period of at least 4 weeks. Lytic lesions on radiographs had to recalcify to be scored as a CR. Partial response (PR) was defined as a reduction of 50% or greater in the sum of the products of the perpendicular diameters of all measurable lesions, without the appearance of new lesions or an increase in the size of existing lesions, for 4 or more weeks. Stable disease was defined as less than a 50% reduction in the sum of the products of the perpendicular diameters of all measured lesions, without the appearance of new lesions, for a period of more than 8 weeks. Progressive disease was defined as an increase in the product of the two perpendicular diameters of any measured lesion present at entry on study by 25% or more or the appearance of new lesions.
Statistical Analyses
Randomization.
Patients were randomly assigned to one of the two arms of the study. To avoid overaccrual to an inactive phase II agent, patient entry was temporarily suspended after 20 patients had been assigned to the phase II arm. If there was a least one response in that patient cohort, accrual to that agent was continued to a total of 50 patients. If there were no responses to a phase II agent among the first 20 patients, accrual to that agent was halted. Beginning in 1990, two patients were randomized to the phase II agent for each patient randomized to the CAF-only arm to complete phase II accrual more rapidly.
Confidence interval and power calculations.
With the sample size of 326 patients, there is an 81% probability of rejecting the null hypotheses of no difference in survival, given a true relative difference of 50% at 3 years of follow-up (ie, 25% v 37.5%, or a death ratio of 1.41). These calculations assume exponential survival distributions, comparisons using a two-sided log-rank test at the .05 significance level, 5 years of accrual at 64 patients per year, and 3 years of additional follow-up (when approximately three quarters of the patients will have died).
Statistical methods.
Survival time and response duration curves were drawn according to the Kaplan-Meier product-limit method. Response duration was the interval from first response (PR or CR) until disease progression or date last known to be disease-free. We used the log-rank test to compare two or more survival curves. Univariate and multivariate Cox proportional hazards models related various prognostic variables with survival-type end points, whereas logistic regression related different variables with tumor response. We used Pearson's correlation to assess the association of pairs of explanatory variables. Various demographic and pretreatment variables (Table 1) were included in the Cox model. We included treatment assignment, ie, CAF only or phase II agent followed by CAF. The primary objectives of the study were to estimate response rates to single-agent therapy and to compare survival time between arm I (CAF) and the pooled arm II (IIA, IIB, IIC, etc.). Secondary objectives were to compare overall response rates and overall response durations between the two arms.
Patient Characteristics
RESULTS
Accrual
Between December 23, 1986, and December 31, 1993, 365 patients were enrolled onto this study. Five patients never started treatment. Thirty-eight patients were considered ineligible for the study. Reasons for ineligibility included bilateral breast cancer, visceral crises, less than 4 weeks from cessation of hormone therapy, and no measurable disease. Median follow-up was 8.5 years. Data from 322 eligible patients are included in this report. Only 16 patients were accrued to the ELSA arm because at that point we had neared the accrual goal, and the study was to be terminated at the end of 1993.
Patient Characteristics
Table 1 lists patient characteristics by assigned treatment. The pretreatment characteristics were well balanced across the two arms of the study. About one third of all patients had received previous chemotherapy (adjuvant only), hormone therapy, and/or radiotherapy. Nearly two thirds of all patients had visceral disease and/or soft tissue metastases. About 40% of patients had bone involvement, and 40% of patients had three or more involved sites of metastases.
Toxicity
Toxicity was graded according to the CALGB expanded common toxicity criteria. Maximum toxicity is reported for each arm, including the five components of arm II. There were four toxic deaths: three patients on arm I (the CAF-only treatment, each from sepsis) and one patient on arm II (from cardiac failure). The death on the AMON + CAF arm actually occurred during the treatment with CAF at 12.9 months after entry onto study. Table 2 lists the maximum toxicities by treatment arm. For arm II, toxicities were those suffered at any time during the treatment with phase II agent + CAF. A striking result is the high incidence (34%) of grade 4 toxicity seen in patients on the MEL + CAF arm, which occurred only while they were receiving MEL alone. There was no indication that patients who experienced toxicities during the phase II agent administration had more severe toxicities when they went on to receive CAF than did patients who received CAF only.
Maximum Toxicities
Tumor Response
Response data are given for each arm. For arm II, the response rate is the highest degree of overall response achieved on the arm II regimen, ie, on phase II agent and subsequent CAF therapy. Table 3 lists the results for arms I and II and also for the component phase II agent + CAF arms. For response calculations, only patients assessable for response assessment are included. Three of the CAF-only patients were not assessable and four of the arm II patients were not assessable. The combined response rate (CR + PR) for the CAF-only arm was 52%, and that for the phase II agents + CAF arm was 44%. As Table 3 shows, the 95% confidence intervals overlap (43% to 60% and 36% to 51%). There was no significant difference between the overall response rates (P = .15).
Response Rates and 95% Confidence Intervals (n = 315)
The demographic and pretreatment variables listed in Table 1 were analyzed univariately and revealed several significant differences. Patients who had received previous adjuvant chemotherapy were less likely to have a response (37% v 52%; P = .018); patients with visceral metastases responded less well to chemotherapy than patients who did not have visceral disease (43% v 57%; P = .023); patients with soft tissue disease were more likely to respond than patients without soft tissue disease (52% v 39%; P = .031); and patients who were estrogen receptor–positive were more likely to respond than patients who were estrogen receptor–negative (56% v 42%; P = .029).
Multivariate logistic regression was performed using the previously described variables. The results are listed in Table 4. Only previous adjuvant chemotherapy and visceral involvement correlated with tumor response. Patients who had not received adjuvant chemotherapy before entering the trial were 1.7 times more likely to respond than patients who had received adjuvant chemotherapy (P = .045). Patients without visceral involvement were 1.6 times more likely to respond than patients with visceral metastases (P = .058). After adjusting for these two variables, treatment arm did not correlate with tumor response (P = .19). However, there was a suggestion of an interaction between treatment arm, visceral involvement, and tumor response. Patients treated on the CAF-only arm had similar response regardless of whether they had visceral disease (54% with v 56% without). However, on the phase II agent + CAF arm, there was a 60% response rate among patients without visceral metastases but only a 37% response rate among patients with visceral disease. After adjusting for the effect of previous adjuvant chemotherapy, visceral metastases, and treatment arm, this interaction was of marginal significance (P = .078).
Tumor Response: Multivariate Analysis (n = 301)
Response Duration
One hundred forty-nine patients achieved CR and/or PR; 73 were treated with CAF alone and 76 with a phase II agent followed by CAF. The duration of response by arm is shown in Fig 1 and Table 5. Patients on the CAF-only arm had a median response of 21 months, compared with patients on the phase II agent + CAF arm, whose median response duration was 15 months. This difference, which is in favor of the CAF-only arm, is not statistically significant (P = .069).
Response duration by treatment arm. (—) CAF-only arm (n = 73; 65 treatment failures; median response, 21.4 months). (– – – –) Phase II agent + CAF arm (n = 76; 69 treatment failures; median response, 15.0 months). Log-rank χ2 = 3.30, 1 df, P = .069.
Time to Clinical Event
Survival
Overall survival probability between the two arms of the study is shown in Fig 2 and Table 5. There was no statistically significant difference between the median survival durations of 20 versus 17 months (P = .074), although survival benefit slightly favored the CAF-only arm. Variables that were univariately of statistical significance (P ≤ .05) included previous adjuvant chemotherapy, previous radiation therapy, previous hormone therapy, number of previous treatment modalities, visceral disease, number of metastatic sites, and performance score. The survival advantage correlated with no previous adjuvant chemotherapy, radiation, or hormone therapy, fewer previous treatment modalities, absence of visceral metastases, fewer metastatic sites at study entry, and lower performance score. In the multivariate analysis shown in Table 6, the following variables were significantly associated with enhanced survival: fewer previous treatment modalities (P = .0028), lower performance score (P = .0002), and lack of visceral metastases (P = .010). Treatment arm did not correlate with survival. There was no interaction between treatment arm, visceral metastases, and survival.
Overall survival by treatment arm. (—) CAF-only arm (n = 144; 133 deaths; median survival, 19.6 months). (– – – –) Phase II agent + CAF arm (n = 178; 170 deaths; median survival, 16.6.months). Log-rank χ2 = 3.20, 1 df, P = .074.
Overall Survival: Multivariate Analysis (n = 308)
DISCUSSION
The results of this large randomized phase III trial in patients with previously untreated metastatic breast cancer demonstrate that pretreatment with a short course of a single agent before standard combination chemotherapy did not significantly alter the duration of survival or the overall response rate compared with the standard chemotherapy arm.
The drugs selected for use as phase II agents were chosen as surrogates for different mechanisms of cytotoxicity: an antimetabolite (TMTX), two alkylators (MEL and CBDCA), a topoisomerase II inhibitor (AMON), and a topoisomerase I inhibitor (ELSA). The agents were also selected because of the varied mechanisms of resistance to their activity: dihydrofolate reductase (TMTX), p-glycoprotein or multidrug resistance (TMTX, AMON, and ELSA), and glutathione S-transferase (MEL and CBDCA). The data from the individual phase II agent plus CAF cohorts suggest that there is no difference between CAF alone and the different agents studied. Although the overall response rate for the phase II agent + CAF arm was not significantly different from that of the CAF-only arm, individual phase II agent + CAF cohorts had response rates that varied nonsignificantly from less than to more than the response rate seen in the standard CAF arm. The same is true of the survival estimates. Patients who responded to a phase II agent continued to respond when institution of CAF was required, and some of them converted from a PR to a CR. Initial responses to a phase II agent ranged from 5% to 24%.49-51 With additional response rates to CAF after a phase II agent administration, cumulative response rates varied from 33% to 60%, with an overall response rate of 44%, a result not significantly different from the response rate to CAF alone (52%). These results imply that the peculiarities of each drug, whatever its mechanism of action or of cellular resistance, do not diminish the ability of patients to achieve benefit from the subsequent use of standard chemotherapy. Another conclusion is that there is no evidence that single-agent induction therapy with a moderately effective drug such as MEL or CBDCA provides an improved overall survival or overall response rate. However, this result does not predict what might have been the result if the induction agent had been given at a higher dose or had been more effective.
There were two disturbing findings in this study. First, we found an interaction, although weak, between treatment and visceral metastases with tumor response. Patients with visceral disease showed a trend toward a poorer response rate on the phase II agent + CAF arm: a difference of 23% (P = .087). Second, patients on the phase II agent + CAF arm had a median of 6 months less of response duration, although this is not a significant difference. These nonsignificant differences in response rate and response duration between patients on the phase II agent + CAF arm and patients on the CAF-only arm may be attributed to the fact that patients may have been delayed in getting a response for up to 4 months while receiving a relatively ineffective phase II agent. Any subsequent achievement of response when receiving CAF may have taken longer and resulted in a reduction of the duration of response. Given that patients were essentially asymptomatic (performance score of 0 or 1) to be eligible for this study, the decrease in response rate or response duration may not be clinically important, even if it had been statistically significant.
There are some limitations to our study. By design, we could not investigate every potential phase II agent. Thus, it is possible that a phase II agent yet to be discovered could penalize or benefit patients if used before standard chemotherapy. For this study, we chose five agents typical of their class or agents that were novel. Because we limited accrual for an inactive phase II agent to 20 patients and accrual for an active agent to 50 patients, we do not have the statistical power to compare individual components of the phase II arm. For such power, a minimum of 180 patients would have to be accrued for each phase II agent. Accordingly, by design, this study is limited in its ability to fully address issues of individual phase II agents. Again by design, this study does not address the durability of response to phase II agents. Because of the general concerns of clinicians to enter their patients onto the experimental arm, and the concern that prolonged exposure to a single agent might increase cross-resistance to the subsequent standard therapy, we limited the use of the phase II agent to four cycles, regardless of whether a patient was responding to the drug at that point. Therefore, for responding patients on phase II agents, we do not know the maximum duration of those responses. Although this study supports the general safety of using a single phase II agent before standard chemotherapy, it does not support using such agents beyond four cycles or 4 months.
As with any study, accrual of patients presents a potential for selection bias. Although randomization, including a two-to-one randomization, eliminates much of the bias between the two arms of the study, there remains the bias of selection of patients who are offered and who accept enrollment. We note that for the standard chemotherapy arm, median survival for patients with metastatic disease previously untreated with chemotherapy was 20 months. This interval is shorter than might have been expected and suggests selection bias. Such bias is important with regard to whether or not one can generalize our conclusions as appropriate for the population of women with previously untreated metastatic breast cancer. The group of women entered onto this study was characterized by the following: two thirds were 50 years old or older, two thirds had visceral disease, two fifths had received previous hormone therapy for their metastatic disease, and two fifths had estrogen receptor–negative cancers. Yet, because of the study eligibility requirements (performance score of 0 or 1), all of the women in this study were clinically “well.” They could not have any signs or symptoms of visceral crises, and their liver, renal, and hematologic blood values had to be in or close to the normal range. Therefore, this study cannot directly address the safety of using single agents up front in women with clinically symptomatic visceral disease or disease so advanced that their blood studies are grossly abnormal (>1.5 times normal). Patients with known visceral disease at the start of treatment were to have bimonthly imaging of the viscera during phase II therapy to detect early progression and to stop phase II therapy immediately. The suggestion of a harmful effect of phase II agents in this group of carefully monitored women leads us to urge caution in placing women with significant visceral disease on up-front phase II protocols. It should also be emphasized that all patients on this study had measurable disease. Therefore, we cannot directly address the issue of safety in women in whom disease is not measurable and in whom disease progression could occur unknowingly. One way to detect early progression of disease in such patients would be to perform routine imaging studies of viscera likely to develop detectable or measurable disease, but this would be cost-ineffective.
In conclusion, this study supports the overall safety of using phase II agents for up to four cycles before beginning standard chemotherapy for patients with previously untreated metastatic breast cancer who have good performance status and normal organ function. Response rates and survival times were not significantly increased or decreased between the two arms of the study. The safety of this treatment is based on halting the phase II agent when progression is noted and on beginning standard therapy no matter what the response after a maximum of four cycles of phase II agent therapy. Because subset analysis suggests the possibility of adverse outcomes in patients with visceral disease sites, we recommend caution in enrolling patients with clinically significant visceral involvement in future up-front phase II studies. However, in general, patients and their physicians wishing to enter phase II trials before beginning standard chemotherapy for metastatic disease can be assured that they are not jeopardizing response rates or overall survival times. These findings support the use of a new paradigm for the evaluation of phase II drugs in breast cancer patients. Such a strategy may not be safe in all types of cancer. For example, in tumors such as small-cell carcinoma of the lung, which grow rapidly or may become drug resistant rapidly, the use of an ineffective chemotherapeutic agent may compromise response rates and/or survival times. Finally, although this study did not result in the discovery of a new and highly effective chemotherapeutic agent for breast cancer treatment, the study design permits such discovery should a promising phase II agent emerge.
APPENDIX
Participants in the CALGB 8642 Study
The following individuals and institutions participated in this study: Robert Diasio, MD, University of Alabama, Birmingham, AL (supported by grant no. A47545); Robert Cooper, MD, Bowman Gray School of Medicine, Winston-Salem, NC (supported by grant no. CA03927); Thomas Shea, MD, University of North Carolina, Chapel Hill, NC (supported by grant no. CA47559); N.J. Vogelzang, MD, University of Chicago Medical Center, Chicago, IL (supported by grant no. CA04326); Herbert Zmaurer, MD, Dartmouth-Hitchcock Medical Center, Hanover, NH (supported by grant no. CA04326); Jeffrey Crawford, MD, Duke University Medical Center, Durham, NC (supported by grant no. CA47577); George P. Canellos, MD, Dana-Farber Cancer Institute, Boston, MA (supported by grant no. CA47642); Gerald Clamon, MD, University of Iowa Hospitals, Iowa City, IA (supported by grant no. CA47642); Marc Citron, MD, Long Island Jewish Medical Center, New Hyde Park, NY (supported by grant no. CA11028); Ernest Borden, MD, University of Maryland Cancer Center, Baltimore, MD (supported by grant no. CA31983); Michael Grossbard, MD, Massachusetts General Hospital, Boston, MA (supported by grant no. CA12449); F. Marc Stewart, MD, University of Massachusetts Medical Center, Worcester, MA (supported by grant no. CA37135); Michael C. Perry, MD, University of Missouri/Ellis Fischel Cancer Center, Columbia, MO (supported by grant no. CA12046); James Holland, MD, Mount Sinai Hospital, New York, NY (supported by grant no. CA04457); Daniel R. Busman, MD, North Shore University Hospital, Manhassat, NY (supported by grant no. CA35279); Ted Szatrowski, MD, New York Hospital–Cornell Medical Center, New York, NY (supported by grant no. CA07968); Louis A. Leone, MD, Rhode Island Hospital, Providence, RI (supported by grant no. CA08025); Ellis Levine, MD, Roswell Park Memorial Institute, Buffalo, NY (supported by grant no. CA02599); Stephen Graziano, MD, State University of New York Health Science Center, Syracuse, NY (supported by grant no. CA21060); Alvin Mauer, MD, University of Tennessee, Memphis, TN (supported by grant no. CA47555); Stephen Seagren, MD, University of California, San Diego, CA (supported by grant no. CA11789); Alan Venook, MD, University of California, San Francisco, CA (supported by grant no. CA60138); Nancy Dawson, MD, Walter Reed Army Medical Center, Washington, DC (supported by grant no. CA26806); and Daniel Ihde, MD, Washington University–Barnes Hospital, St. Louis, MO (supported by grant no. CA47546).
Acknowledgments
Supported in part by National Institutes of Health grants no. CA37135 (M.E.C.), CA26506 (R.B.W.), CA60138 (I.C.H.), CA33601-128 (D.A.B. and C.C.), CA47577 (E.W.), CA12449 (W.C.W.), CA32291 (E.F.), CA04326 (O.R.M.), and CA31946 (R.L.S.). The research for Cancer and Leukemia Group B 8642 was supported in part by grant no. CA31946 from the National Institutes of Health to the Cancer and Leukemia Group B.
Footnotes
-
The contents of this article are solely the responsibility of the authors and do not necessarily reflect the official views of the National Cancer Institute.
- Received September 21, 1998.
- Accepted January 21, 1999.
