Breast Cancer Research and Treatment 9: 87-100, 1987 © Martinus Nijhoff Publishers, Boston - Printed in the Netherlands
Viewpoints
Drug dosage intensity - a panel discussion* William L McGuire (moderator), James Goldie, William Hryniuk and Douglass C Tormey
University of Texas Health Science Center, San Antonio TX USA; Cancer Center Agency of British Columbia, Vancouver BC Canada," Ontario Cancer Foundation Hamilton Regional Cancer Center, Hamilton, Ontario, Canada; Wisconsin Clinical Cancer Center, Madison WI, USA
Key words: chemotherapy, dose intensity, drug resistance, drug scheduling, toxicity, trial design Summary
There is now widespread interest in the subject of drug dose intensity in cancer treatment. Recent retrospective analyses have shown a clear-cut benefit for those patients receiving either the intended drug dosage in a particular regimen or the highest dosage achievable through an escalation scheme. In this discussion, the participating physicians will review dose intensity in light of these retrospective studies, highlighting the particular tumor types in which benefits have been demonstrated. Other important issues will be discussed, such as scheduling of drug combinations and the importance of using toxicity as a biologic response monitor of drug efficacy. The panelists will also provide opinions as to the best design for aprospective clinical trial to test dose intensity.
Definition of dose intensity and its importance in treatment outcome DR. MCGUIRE: Let us begin the panel discussion by
asking Dr. Hryniuk to define dose intensity, discuss the evidence that it is important in treatment outcome, and indicate in which tumor types it is most effective. DR. HRYNIUK: Dose intensity is the amount of drug
administered per unit time. For a single drug regimen it may be expressed simply as mg/mVwk, regardless of the particular schedule used. For regimens containing only one drug, remission rate correlates with dose intensity in advanced colorectal cancer, in breast cancer, and in oat cell cancer. However, for regimens containing more than one
drug, dose intensity must first be calculated as mg/m2/wk for each drug and then expressed as a decimal fraction of the dose intensity of the same drug in an arbitrarily chosen reference regimen. These decimal fractions (called relative dose intensities) are averaged to give the average relative dose intensity. If the regimens being analyzed are missing one or more of the drugs of the standard regimen, a zero is assigned as the dose intensity for the missing drug(s). However, the aggregate of the relative dose intensities is still divided by the number of drugs in the standard regimen. When regimens analyzed in this manner are aligned in order of their average relative dose intensities, treatment outcome correlates with the alignment. Thus, remission rates in metastatic breast cancer, disease-free survival in stage II
*A series of teleconferences has been organized under the auspices of Bristol-Myers to address several major current questions in oncology.A panel of recognizedexperts with a moderator has been assembled to discuss each question, and we are reporting a number of these discussionsin Breast Cancer Research and Treatment. This is reprinted from 'OncologyViewpoints', courtesyof Bristol-Myers Oncology Division, EvansvilleIN 47721, USA.
88
W L M c G u i r e et al.
breast cancer, and remission rate and survival in advanced ovarian cancer correlate with average relative dose intensity. A recently completed review shows that remission rate during salvage chemotherapy of ovarian cancer also correlates with dose intensity. Others have shown dose intensity of MOPP (mechlorethamine, Oncovin [vincristine], procarbazine, prednisone) chemotherapy correlates with outcome in Hodgkin's disease and nonHodgkin's lymphoma. Dose intensity, calculated after taking into account dosage reductions and treatment delays for toxicity, is designated as received dose intensity. Received dose intensity correlates much more closely with outcome than does dose intensity calculated from the intended protocol. Analysis of dose intensity can also detect whether the individual drugs comprising a combination regimen are truly equivalent in activity. Perhaps the most important practical application may be in individual patients, e.g., those receiving CMF chemotherapy for advanced breast cancer. Here, received dose intensity predicts the probability of a response. Thus, in the early stages of treatment in each case a clinical decision may be taken to continue CMF chemotherapy, or switch to alternative treatments, or discontinue all chemotherapy. The concept of dose intensity must be separated from the concept of total dose. Thus, two regimens may be of identical dose intensity, but differ in total dose because of differences in duration of treatment. Alternatively, two regimens may deliver an identical total dose, but may differ greatly in dose intensity, with the abbreviated regimen being of higher dose intensity, and therefore more active. It remains to be seen whether, and to what extent, dose intensity correlates with outcome independently of total dose. Dr. Howard Skipper at the Southern Research Institute has concluded that, in retrospect, dose intensity is also a key determinant of treatment outcome and toxicity in previously performed animal experiments. However, the animal data were not designed to test dose intensity or distinguish it from total dose. DR. MCGUIRE"Dr. Tormey, do you have the same
perspective regarding the definition of dose intensity, and do you view the retrospective evidence mentioned as indicating its importance in actually achieving improved relapse-free survival and overall survival? DR. TORMEY: [ think we can all generally agree with Dr. Hryniuk's definition. Certainly the retrospective evidence using that definition supports the concept of dose intensity being an important predictor of treatment outcome. There are also a number of trials using different drug schedules, such as CMFVP (cyclophosphamide, methotrexate, fluorouracil, vincristine, and prednisone) in breast cancer, as well as some high-dose vs. low-dose trials suggesting that higher dose intensity regimens are better. However, it is not clear whether it is the dose intensity that improves the treatment regimen, or whether it is in fact related to tumor and host kinetics. DR. GOLDIE: I think that Dr. Hryniuk has come up with a very useful definition for dose intensity, primarily, as he says, because it allows one to look directly at clinical trial data, and to make comparisons across trials. It is simple, but elegant in its simplicity, It allows you to get a quantitative handle on what is actually happening in many chemotherapy protocols. It may not be the only way to define dose intensity, but I think it is a very useful, pragmatic one, In terms of the evidence for its value, in addition to what the other speakers have cited, another experiment in dose intensity is when very high doses of chemotherapy are given together with autologous bone marrow replacement. Under these conditions, tumors that are refractory to conventional doses of chemotherapy may sometimes show sustained responses, again consistent with there being a dose-response effect in chemotherapy. It is not such a straightforward matter to calculate the dose intensity in a single large dose, because you would get an almost infinite dose intensity since it is given at a discrete point in time. What we have done is to take into account what in statistics is called the edge effect for instantaneous events. For example, we would calculate the dose intensity of a
Drug dosage intensity single high dose by taking into account the normal recovery time that you would expect before you could administer a theoretical second dose, even if you do not give one. This type of analysis allows you to at least make a comparison between a single high dose and programs involving multiple dosages. I think Dr. Hryniuk's is a very useful definition, and is consistent with a large body of information suggesting a dose-response effect in chemotherapy.
Dose intensity vs. number of drugs
DR. MCGUIRE:Dr. Tormey, could you tell us about the protocols in the Eastern Cooperative Group (ECOG) study that you designed? A large number of different drugs were used in some of the regimens. Could dose intensity be related to, or be a function of, the actual number of different drugs in a particular regimen? DR. TORMEY: i think we are talking about several different issues. One is the difference between CMF (cyclophosphamide, methotrexate, and fluorouracil) and CMFP (cyclophosphamide, methotrexate, flouorouracil, and prednisone), for example, where CMFP was clinically superior to CMF. When one does the analysis of those trials one notices two things. First, there are higher doses of CMF delivered in the CMFP regimen, the delays are the same in both regimens, and the toxicity is actually less in the CMFP regimen. I do not think that prednisone is particularly adding to the therapeutic effect in terms of increasing cell kill, but rather in terms of allowing a higher dose of CMF to be delivered, and this would be consistent with the relative dose intensity analysis. In that sense, more drugs are contributing to the dose intensity, but indirectly. Second, if one adds increasing numbers of myelosuppressive drugs to a regimen, then one winds up reducing the doses of the drugs, and does not always increase the actual dose intensity of that given regimen. However, if one can add increasing nonmyelosuppressive drugs to a basic myelosup-
89
pressive regimen, than one does end up increasing the overall dose intensity, assuming those drugs are cytocidal. Thus, there are actually two effects. One is an indirect mechanism of increasing the number of drugs, which increases the dose that you can give of your cytocidal agents, which would then increase the dose intensity. The second is that, by increasing the number of cytocidal nonmyelosuppressive agents, one could also increase the dose intensity. DR. MCGUIRE:Dr. Goldie, you have always been a proponent of using multiple drugs in a sequential way to avoid the problem of drug resistance, and have always emphasized that these drugs must be given at effective doses. Could you comment on the question of dose intensity versus the number of types of different drugs - not scheduling, but specifically the number in a particular regimen? DR. GOLDIE: Even with considerations of dose intensity, virtually all of the effective chemotherapy protocols that we have involve the use of several drugs that are quite different in their mode of action, and at least to some degree, are non-crossresistant. This also indicates that in addition to the question of dosage, tumor heterogeneity is another problem that must be overcome. To visualize this from the point of view of combination chemotherapy and how it relates to dose intensity, imagine a very simple system that has only one class of cells in it and essentially no resistant cells, but is growing back at a particular rate. In this case, the ability to control the tumor will be a direct function of the dose intensity or the frequency of administration of a single agent. In real systems, however, it is a more complex problem due to the presence of subpopulations of resistant cells. In addition to controlling sensitive cells in the tumor, you have to also give adequate doses of other drugs that will each be effective in controlling the resistant subpopulations. To accomplish this, each one of these drugs has to be given at a minimally effective dose intensity. Another way to look at it is that an effective combination chemotherapy protocol essentially reduces the heterogeneous system to a homogeneous system.
90
W L McGuire et al.
In other words, there are no cells within the tumor that are resistant to all of the agents in the Combination. The combination may be considered as if it were a complex single drug that has to be given in an adequate dose intensity to kill off the cells faster than they are repopulating. So, I therefore believe there is a relationship between the number of drugs in a regimen and dose intensity. The fact that more than one drug is required to eradicate most tumors shows that both dose intensity and tumor heterogeneity must be considered. DR. MCGUIRE:Dr. Hryniuk, do you agree that dose intensity could be simply a function of the number of different drugs in each regimen or is that preempting the question of dose intensity? DR. HRYNIUK:The answer is both yes and no. The average relative dose intensity derives from the aggreg.ate of the relative dose intensities of all the drugs, divided by the number of drugs. If the aggregate is high, the average relative dose intensity will be high. For example, in a series of regimens which contain some of the drugs of a four-drug standard regimen, the aggregate for a particular test regimen can be high even if only one or two of the drugs of the standard regimen are represented. Alternatively, the aggregate can be low even if all four drugs are represented. In either case the aggregate is divided by four, and the resultant figure is the one that correlates best with treatment outcome. When drugs with non-overlapping toxicities are combined at maximally tolerated dose intensities, the aggregate will be higher the greater the number of drugs used. Thus, dose intensity is a simple arithmetic description of a fundamental principle which we all follow in designing and applying combination chemotherapy regimens. I should add that these considerations apply only if the drugs being considered are of approximately equal antitumor activity.
Dose scheduling: considerations
DR. MCGUIRE: Dr. Hryniuk, what are your views
concerning dose scheduling? How important is it? For example, is there any evidence favoring either intermittent high-dose therapy or the same total dose given continuously? DR. HRYNXUK:For purposes of calculating dose intensity, one must completely ignore scheduling. Otherwise it is not possible to calculate mg/mZ/wk for each drug. We therefore tacitly assume that scheduling is not a direct determinant of antitumor effect. This is a useful assumption but in some cases it may be incorrect. For example, we know that scheduling may be a critical determinant of antitumor effect of cytosine arabinoside in animal model systems and probably also in humans. Furthermore, all drugs demonstrate threshold effects. Below a certain dose nothing happens; above a certain dose, responses begin. Scheduling is obviously important in assuring that thresholds are exceeded. Finally, as a special example of the importance of threshold, high systemic pulses of some drugs must be given to get adequate amounts through the blood-brain barrier. However, most variations in scheduling occur at dose intensities above threshold, and tumor targets are almost always outside the CNS. In such situations we are probably safe in ignoring schedule as a direct determinant of antitumor effect. Nevertheless, I believe scheduling is a crucial determinant of toxicity in all situations, and in that sense, scheduling governs received dose intensity and antitumor activity. There is good evidence for this. Two randomized trials of combination chemotherapy in advanced breast cancer, one by the SWOG and one by the SECSG, showed that greater amounts of drugs could be tolerated if given continuously in low doses instead of intermittently in high doses. Paradoxically, the 'high-dose' intermittent schedules turned out to be the low dose intensity schedules, and produced inferior antitumor effects. Doxorubicin is better tolerated if given weekly instead of every three weeks, and the weekly schedule produces better antitumor effects. Another example of the importance of scheduling is the use of fluorouracil by infusion. Given in this way much more drug can be administered compared to the weekly push method, and the antitumor effects are greater.
Drug dosage intensity On the other hand, in animals with minimal burden leukemia and in children with leukemia in remission, methotrexate if given daily in small doses is quite toxic. If given in large doses separated by several days, much more drug can be tolerated, i.e., intermittent schedules deliver much higher dose intensities of methotrexate. These schedules also produce greater antitumor effects than do the daily dose schedules. However, in animals with a heavy burden of leukemia or children with leukemia in relapse, the tolerance for methotrexate is poor regardless of the schedule used. Only small amounts of drug (with a low dose intensity) can be administered, and tumor response is much reduced, regardless of schedule. Therefore, to predict the effect of a schedule one must consider the state of the host, the stage of the tumor, and the particular drug. DR. MCGUIRE:Dr. Goldie, what are your comments on this question of scheduling? DR. GOLDIE: I think there are at least two or three components to the whole concept of scheduling. There are the various pharmacologic considerations that determine the optimal method for administering individual drugs, which varies quite a bit. Some work better when given by infusion, others by IV push doses, and so on. I would like to point out that the structure of the protocol itself is a kind of scheduling. Consider, for example, MOPE in the treatment of Hodgkin's disease. Instead of giving M O P P in the usual way, with the drugs effectively bunched together on days i and 8, all of the nitrogen mustard is given in sequential doses first, until all of the mustard in the protocol is used; then we switch to vincristine, to procarbzine, and so on. Over a sixmonth period, one could end up with the same relative dose intensity as standard MOPP. Instead of giving the drugs concurrently, they can be staggered so that one can be given, and then switched to another, and so on. With what we know about how Hodgkin's disease responded to sequential single agents in the 1960s, not bunching them together would produce a distinctly inferior therapeutic result. The aggregation of agents, therefore, given at
91
least within a very short period, is clearly an important component of scheduling. This is where tumor heterogeneity must be considered. DR. HRYNIUK: I do not completely agree with you, Jim. Not until one can calculate whether it is dose intensity or scheduling can we be sure why MOPP is better than M--~O--~P---~P. Dose intensity could be the explanation for the bunching effect. DR. GOLDIE: There has never been a prospectively randomized study of this effect, but you do not have to give all of the mustard on days 1 and 8 once a month. Larger doses of mustard can be given every three weeks until the total dose intensity for the six-month protocol is exhausted. It would probably be exhausted in the first two or three months; then you could switch to procarbazine, and then to vincristine. When averaged over six months, the same value for a six-month period would be obtained as for the regular MOPP. If just a part of that time frame - one month - was observed, it would be different. But if looked at over six months, it would be the same. DR. HRYNIUK: It seems that the mustard would be used up very quickly, and then the next drug would have to be used, and the next drug, so that the same dose intensity would be obtained and perhaps the same therapeutic effect would result. DR. GOLDIE: I agree that there has not been a definitive experiment, but that was more or less what was done in the i960s with a number of tumors. We gave one drug for a period of time, often in very high individual dosage; when that would cease to work, we would switch to another one, and so on. In Hodgkin's disease, there was a series of remissions, but it was very rare to produce cures. I think it was the concatenation of all the active agents, so that not only was there a high-dose intensity, but increased diversity of agents over a short time frame, that began to produce cures. I think this was the big impact that MOPP had. DR. TORMEY: I tend to agree more with Dr. Goldie, not only because of the M O P P observation, but
92
W L McGuire et al.
also considering a similar experience in the 1960s and early 1970s in the treatment of testicular cancer. In reality, dose intensity is only one determinant of treatment results. One has to consider the schedule as well. I would also like to re-emphasize the point that continuous therapy may be better not only because of dose intensity, but also because of the schedule with respect to the tumor kinetics and the host-toxicity patterns. It would really be very difficult to sort out these effects. The continuous regimen, which appears to be therapeutically superior in breast cancer, is also a higher dose-intensity regimen, so I am not sure that the two can be separated that easily.
Toxicity and dose escalation DR. MCGUIRE: In the discussion of scheduling, I believe Dr. Hryniuk mentioned that scheduling certainly accounts for the observed toxicity, but he dismissed this as a nuisance. There are some people who view toxicity as a biological response to the drug, which perhaps could be used profitably. Dr. Tormey, could or should drug dosage be escalated to achieve a pre-set toxicity level? Or should we just administer drugs at a pre-set dosage regardless of whether a certain level of toxicity is achieved? DR. TORMEY: I expect that we are going to have some agreement on that question. My goal is to provide the maximum dose compatible with patient safety. This would imply that in some patients the doses would be escalated to achieve a pre-set toxicity level; in others, reductions will have to be undertaken. I think this approach is supported by the concept that we wish to achieve the maximum dose possible in a given patient in order to get as high up the dose-response curve for a given drug as we can. This brings up an interesting question. If one group of patients has a relatively low relative dose intensity because of reduction for toxicity, is the response rate really going to be less than that of a group of patients who get a fairly high relative dose intensity, but with no toxicity at all? I am not sure if that particular analysis has been done. I know this
is being examined prospectively in a couple of current E C O G adjuvant trials. DR. MCGUIRE:Dr. Goldie, what do you think about the question Dr. Tormey has just raised? DR. GOLDIE: I tend to agree with Dr. Tormey. I think there are some practical problems when you start to individualize the protocols for individual patients, escalating the doses for some and reducing them for others. From a biological point of view, however, it certainly makes sense, although if you are talking about large-scale clinical trials, fine tuning each patient's dose may be rather difficult to accomplish. Yet I agree that in individual patients, you want to push the chemotherapy to what is maximally tolerable. And I suspect that in many protocols over the years we have been significantly underdosing patients. I suspect that with the right schedule, you can probably use much larger doses of chemotherapy than we have been inclined to use in the past. Also, the information about the role that prednisone played in many protocols is illuminating. I think it does a lot more than just affect marginated pools of granulocytes. It seems to genuinely improve the marrow's tolerance to chemotherapy, and reduces the incidence of significant sepsis. There may be other similar maneuvers that we could employ in our protocols that would allow us to push the dosages up. DR. MCGUIRE:Dr. Hryniuk, do you think toxicity is something we can use as an end point of dose intensity, or is it just a nuisance resulting from inappropriate scheduling? DR. HRYNIUK: Dr. Tormey has brought up a very important point. We need instruments which accurately and reproducibly measure aggregate toxicity in each of the spheres where toxicity is important. I think the E C O G scale is a very good first approximation. Once developed, such instruments could provide guidelines to an achievement of a particular defined level of toxicity so that we can answer Dr. Tormey's question prospectively. Toxicity is not only a nuisance- once a certain dose intensity is reached, toxicity will probably rapidly increase to
Drug dosage intensity intolerable levels, no matter what schedule is used. Thus, toxicity limits the usefulness of each particular drug. However, when considering methods to increase dose intensity, I do not favor escalation. It is difficult to describe the precise conditions under which a decision can be made to escalate and how much can be allowed. For practical reasons, I would favor the use of a pre-set starting dose intensity, delivered in the least toxic schedule as predetermined by a formal phase I study of fully functional, previously untreated patients with minimal tumor burden. I would then allow de-escalation from that dose intensity. Besides, if the concept of dose intensity has any validity at all, it argues against dose escalation. It predicts that the best results will be obtained from maximum dose intensity delivered early, before resistant tumor cells have had a chance to grow while one is groping for the maximally tolerated dose. I think prednisone and prophylactic antibiotics are important additions. At the very least, their use allows the physician peace of mind after he/she has given high dose intensity therapy. In practical terms, I do not favor escalation because, having been involved in trials using doses with escalation clauses, I have had to learn what everybody else has learned: It is difficult to make decisions to escalate. DR. TORMEY: I disagree somewhat with Dr. Hryniuk. It does not seem practical to provide a protocol in which 80 or 90% of the patients will achieve significant toxicity in the first cicle. By doing that in a large-scale scenario such as a cooperative group, there may be a major problem in terms of toxicity, and perhaps, even death of patients. The result is that we end up providing regimens in which the median patient, in fact, will achieve a reasonable level of toxicity. I would rather build in a dose escalation clause and use it effectively. Within any cooperative group, there are more violations of dose escalation clauses than there are of dose reduction clauses. How much of that can be overcome with increasing education, I am not sure. I certainly agree that you want to get as much of a dose in on the first cycle as possible, but I think one
93
has to deal with the limits of practicality in the private practice sector and in the cooperative se{ting. DR. HRYNIUK: When I say I am not in favor of escalation, I mean the overall treatment should provide the highest dose intensity. We are conducting a pilot study of super-FAC in which the first two, three, or four doses are well tolerated, but the fifth, sixth, and seventh doses are not. We must fine tune the drug scheduling in the regimen to allow the last few doses to be administered at the highest possible dose intensity. I think Dr. Tormey and I may end up at the same place, but we need instruments measuring aggregate toxicity to tell us where that place is, so we can compare results between trials. DR. TORMEY: I would certainly agree with that within limits, but I am not sure where the aggregate line should be drawn. When one considers the heterogeneity of these tumors, one has to begin to wonder how many drug doses are actually useful. In some scenarios, perhaps only the first three courses are actually providing an antitumor effect in which the sensitive cells are still in a larger proportion in the tumor than the resistant cells. After three cycles, resistant cells may be dominant, so no further therapeutic effect will be achieved. In other scenarios, that may result after six cycles. We just do not have enough information with individual patients, or even aggregates of patients, so we can draw the line - whether it should be two, four, or six months of therapy, in terms of dose intensity.
Dose intensity and the development of drug resistance DR. MCGUIRE: I would like to get back to the question of dose intensity and the development of drug resistance. Dr. Goldie, could you comment on that? DR. GOLDIE: The clue to drug resistance still being an important factor, even when considering dose intensity, relates to the fact that you have to use
94
W L McGuire et al.
multiple agents in order to get the maximum therapeutic effect. If drug resistance was not a problem, then theoretically you could accomplish all objectives by giving a single drug at its maximum-tolerated dose intensity. We know that has not worked, at least in the vast majority of tumors. One might think of a tumor being composed of a group of smaller tumors, each with its own population of sensitive and resistant cells to a particular drug; and that in treatment, one matches each drug in the combination regimen to one of the subsets of cells within the tumor. You therefore have to do two things. You have to give an adequate dose intensity of the individual agent for it to deal effectively with the cells that it can deal with; at the same time, you have to build into the protocol structure a format that allows all of the agents that will be used to be administered over a relatively short period of time. In other words, you do not break up the protocol. You therefore give all of the agents of one type first, and then switch to the second, and so on. This has to do with the fact that, regardless of the drug dose being given, if there is a population of cells within the tumor that is not being affected by therapy for a protracted period, then that subpopulation is being allowed many opportunities to evolve to higher levels of resistance and to develop multiple levels of resistance, as in the pleiotropic resistance phenomenon. Therefore, you have to get at all of the subsets early, and the protocol structure has to reflect this. At the same time, all of the subclones have to be blanketed within the tumor effectively, and the agents being given have to do this in an appropriate dose intensity. DR. ~C~UIRE: Dr. Tormey, do you have anything to add regarding drug resistance? DR. TORMEY:I agree with what Dr. Goldie just said. To summarize, I would say one should give the maximum amount of drug possible in the shortest period of time. DR. GOLDIE: Yes, I think that is true, bearing in mind that there will always have to be compromises with the theoretical optimum because of considera-
tions of toxicity. If we did not have to worry so much about host toxicity, then we could follow the standard treatment regimens of antimicrobial infections by administering large doses of antibiotics over a relatively short, intense period of time. But we are unable to do that with cancer chemotherapy because of the toxicity considerations. DR. HRYNIUK:Actually the concepts of dose intensity and of rumor heterogeneity are complementary, and allow us to predict two kinds of treatment failure. One type is failure to reduce the bulk of disease because the patient has not been treated with high enough dose intensity for long enough to eradicate the fast-growing, sensitive tumor cells. In this situation average relative dose intensity applies for all the drugs if they are all active against the bulk of disease. The second type is failure to cotitrol a smaller core of the tumor - the resistant mutants. Here, the concept of average relative dose intensity is still applicable. Resistant mutants may exhibit much less kill (or even no kill) on exposure to changes in dose intensity of some of the drugs in a regimen. For practical purposes, such drugs should be assigned a dose intensity of zero in calculating the aggregate relative dose intensity. Yet we would all agree it is important to treat these resistant cells from the outset. One strategy might be to hold the resistant mutant cells 'in check' with a lower dose intensity of some of the drugs which are active against them, while clearing the bulk of the disease with the highest dose intensity of the drugs which are most active against the bulk but not the mutants. After this initial cytoreduction, one would then 'leapfrog' by increasing the doses of the drugs used to hold the resistant cells in check, and replacing the drug most active against the tumor bulk with an alternative drug active against the mutants. A major difficulty throughout all of this is that we do not know when to stop therapy. To determine the optimum duration of treatment we will have to conduct trials varying the time spans or total dose, fixing dose intensity at some constant level.
Drug dosage intensity Problems of retrospective analyses of dose intensity DR. MCGUIRE:We began by first defining dose intensity, and then reviewing some of the evidence that this is important and the various tumor types where retrospective analyses have shown this to be true. There have been some published criticisms of retrospective analyses and the methodology that has been used. I would like to review what some of these limitations are, and ask whether any new methodologies have been investigated that get around some of the previously published limitations. DR. TORMEY: A major problem that I have with these analyses concerns the fact that, at this point in time, one is doing a retrospective correlative type of study when one looks at dose intensity. Clearly, the studies are not all comparable with respect to everything from eligibility to dose modification schemes, and even to definitions of end points. Consequently, one has the same problem here that one has in any retrospective analysis. One of the best examples of this problem of retrospective analyses relates to dose delivery in the adjuvant breast cancer trials. Dr. Redmond's publication was elegant in showing that even a placebo has a marked effect at different doses. Another issue involves the difficulties in separating out drug effects within the regimens that are utilized. The regimens tend to reduce all drugs simultaneously for different toxicities, and it is not clear that one is getting a maximal dose of any given drug for a particular toxicity. We have reached the point where what we have with relative dose intensity and the whole concept is a simple and elegant way to look at dose intensity. We are now ready to begin to develop prospective tests of the concept proper. DR. MCGUIRE:Dr. Goldie, what are your comments about the limitations of retrospective studies of dose intensity? DR. GOLDIE: Rather than deal with the whole concept globally, I will pick out a couple of specific problems that I know that Dr. Hryniuk has had with his own studies.
95
One basic problem in retrospective analyses is simply the quality of the data. What you want, in fact, is to know what doses are actually given; even from your own institutional record, this information is sometimes hard to find. Also, if one looks at studies from another center or another country, then it may be very difficult to find out what the actual administered doses were for many of the patients. At the very least, it certainly involves a lot of work in terms of looking up old order sheets and so on, because we tended not to record this information as part of the published study; also investigators will talk about the projected doses given and not the actual ones. It is clearly the actual given doses that are important. Therefore, this is a technical difficulty in doing retrospective analyses, and it does limit the number of studies that can actually be subjected to this kind of analysis. There are broader problems, as with any retrospective analysis. My own view is that this has been overstated in the literature in recent years. I think there are many useful things that can be done retrospectively, provided attention is given to the appropriate details. Finally, I think we should remember that dose intensity is basically a scientific hypothesis that can be subjected to disproof under the appropriate conditions but can never be proven beyond a shadow of a doubt. You cannot do that with scientific theories. All you can ask is: are the results consistent with the hypothesis? I think that is what we are seeing so far in the retrospective analyses that have been done. They are consistent with the hypothesis that there is a dose-response effect. Certainly other explanations are possible, but that does not necessarily mean the other explanations are correct. We therefore have to remember the rules that govern the testing of scientific hypotheses. Given a legitimate theory, it can be subjected to unambiguous disproof, but it can never be proven beyond the shadow of a doubt because there will always be other competing explanations. Consequently, to satisfy a lot of people, there will have to be prospective studies done. DR. MC~UIRE: Dr. Hryniuk, what would be the major limitation or weakness of the retrospective ana-
96
WL McGuire et al.
lyses that have been performed by yourself and others to date? DR. HRYNIUK: Collecting the actual data. Some studies were not designed, conducted, or reported with dose intensity in mind, nor were the data readily available which allowed one to calculate actual received dose intensity. Furthermore, I think the comments of Drs. Goldie and Tormey are right to the point. The concept of dose intensity is a simple mathematical way of looking at our experience in chemotherapy. While the concept may give us insights which will allow treatment improvements to be devised, it should first be tested prospectively to determine its validity and its limitations.
Optimal design for a prospective trial of dose intensity DR. MCGUIRE:IS there any prospective evidence that the high-dose regimens are superior to low-dose regimens? And if there are, could we list it? But first, I would like to discuss what would be the best trial design for prospectively testing the concept of dose intensity. DR. HRYNIUK:First of all, we should say high dose intensity instead of high-dose regimen, and low dose intensity instead of low-dose therapy, because as discussed earlier, high-dose regimens may provoke inordinate delays due to toxicity and turn out to be low dose intensity regimens, while low-dose regimens may allow uninterrupted treatment and turn out to be high dose intensity regimens. Considerable data from prospective trials suggest that high dose intensity regimens are superior to low dose intensity regimens. There were two studies by ALGB in lymphoblastic leukemia, and one by Pinkel, and a study of Hodgkin's lymphoma by Brendley and Frye. There were six breast cancer studies - t w o by ECOG, one by the SWOG, one by the SECSG, and two studies reported at the 1986 meeting of the American Society of Clinical Oncology. In one of these latter studies the total dose was kept constant while high and low dose intensity
were compared. High dose intensity was a clear winner. Varying dose intensities of VePesid a (etoposide, VP-16) in small cell lung cancer were prospectively studied by Cavalli, and of cis-platihum in testicular cancer (maximal disease) by the SWOG. In both studies, the high dose intensity regimens were superior. While the bulk of evidence derived from randomized trials supports the significance of dose intensity, high dose intensity regimens may not always be superior. We recently completed a study in small cell lung cancer escalating cyclophosphamide and to a much lesser extent Adriamycin. Results in the escalated arm were not superior. The Ludwig group, in testing adjuvant CMF + prednisone in one to three node-positive premenopausal patients, noted that prednisone allowed a higher dose of C, M, and F to be given. However, these higher doses did not improve treatment outcome. Similarly, in the SWOG study of testicular cancer described earlier, patients with minimal disease receiving the high-dose regimen did not derive any additional benefits. I would suggest that randomized trials which failed to show superiority for high dose intensity regimens probably did not produce large enough increases in dose intensity to impact on a shallow dose-response relationship. In addition, treatment delays in the high-dose arms may have reduced received dose intensities to the same levels as those in the 'low' dose arms. Unfortunately treatment delays are never reported and yet when calculating received dose intensity, delays must be accorded equal weight with dose reductions. Failure to detect differences could also have been due to beta-type errors, due to small sample sizes. Finally, the whole hypothesis relating outcome to dose intensity could be erroneous. We will just have to wait and see what answers come from well designed randomized prospective trials testing dose intensity in a variety of circumstances. DR. MCGUIRE: Could you tell us how you would design the ideal prospective trial to study dose intensity? DR. HRYNIUK: We are currently working on this.
Drug dosage intensity The first thing confronting us is that not only does the study have to be scientifically valid, but it also has to be medically and ethically acceptable. That places constraints upon trial design. Initially, a careful phase I study is required to determine which schedule produces the least toxic, highest dose intensity regimen using drugs of equivalent activity. Furthermore, this study must be conducted in fully functional, previously untreated patients with minimal to moderate disease burden. We have taken two years to complete these necessary preliminaries and have reached the point where we could begin a definitive randomized trial testing dose intensity. Patients will be randomized to three levels of dose intensity with the highest dose intensity being that predicted as tolerable from the pilot studies. Thus, we can be sure that patients allocated to the highest dose intensity will be able to receive it. We will also separate the three levels of dose intensity sufficiently and accrue large enough numbers of patients at each level to be able to detect and define the differences in responses between the three groups. Finally, and above all, the treatment will be of such duration as to ensure that each group gets the same total dose of drug. Dose intensity will be the only variable tested. The end point will be response rate, since the trial will be in patients with metastatic breast cancer. When the trial is completed we hope the three points give us the 'straight' portion of the sigmoid curve relating response to dose intensity. Incidentally, a special example of such a trial would be one with the highest dose intensity point derived from patients receiving megadoses of chemotherapy followed by autologous marrow rescue. The lower points would be derived from groups receiving two, three, or four courses of the same drugs in one half, one third, or one quarter of the dose intensity of the megadose treatment. I must emphasize that if one is testing the basic concept of dose intensity using a combination regimen, the drugs must be of approximately equal activity and the dose intensities of all the drugs should be raised or lowered together. On the other hand, if one of the drugs were suspected of being more active and the aim
97
were to find the dose-response relationship of only the most active drug, then one would vary the dose intensity for only that drug. DR. MCGUIRE: If your phase I studies define the upper limit that, according to the retrospective studies, might perhaps be the best dose, has anyone on your institution review board or planning committee raised the ethical question of treating people? Who is going to get the lowest homeopathic equivalent? Do you see any problems in that part of the study design? DR. HRYNIUK: The answer is not too difficult. Firstly, one must choose to deal with a disease where, as far as is known, response rate does not impact on survival. Then one must ensure that each treatment group gets the same total dose of drug. For example, the high dose intensity patients could be finished in 16 weeks, whereas the lowest dose intensity patients could continue for 48 weeks (using the three-point design described above). One would argue that treating with a combination regimen for a longer period might be just as effective as treating at a higher dose intensity for a shorter time. Moreover, the group receiving chemotherapy at the lower dose intensity would probably experience fewer side effects. That would be important if one were only palliating with chemotherapy. Finally, one would introduce salvage chemotherapy for the low dose intensity group, if this group proved to have a lower response rate and one were concerned that this would jeopardize overall survival. DR. TORMEY:I think the trial Dr. Hryniuk designed is an interesting one, and I suppose there is a host of different trials that one could conceive of. Clearly the use of a standard dose, or a/high dose, versus a lower dose with fixed schedules would be one method. Intermittent therapy versus continuous treatment with the same per day density would be another type of trial. One could select a couple of standard regimens with different dose densities and test them directly, or take the same total dose, but use different schedules. The problem with all of these is the issue of
98
W L McGuire et al.
tumor heterogeneity and scheduling as it relates to toxicity patterns and tumor kinetics. One could design a host of trials that would support the hypothesis, but still not really be sure that one of the other explanations is not playing a part in creating that support. In the same manner, one or two of the trials might be negative because of one of those other explanations. It goes back to what Dr. Goldie was saying earlier: One is not going to be able to really prove it in a hard core sense. Similarly, it is also very difficult to prove the Goldie-Coldman hypothesis directly. One can come up with lots of support, both retrospectively and now prospectively, but it is still questionable as to whether one of the other explanations is not creating a false impression or contributing to the result. DR. MCGUIRE: Dr. Tormey, I think you are very critical of trial designs, but as an expert on trial design, could you come up with some scheme that avoids most of the pitfalls that you mentioned? How would you do your own trial? DR. TORMEY: I do not have a single, reasonably sized trial to examine this directly. There are too many other variables involved. I also think the only way one can approach it is to select the variables of dose intensity, dose rate, and schedule, and to do sequential trials looking at these things one after another until you complete them. These would be too many arms involved in one single trial to test the hypothesis. DR. MCGUIRE:Where would you begin? What is the first trial that you would design in this sequence of trials leading to the ultimate answer? What do you think is the most critical issue that might be addressed initially? DR. TORMEY: I would probably start with a standard or slightly higher than standard dose versus a lower dose. That is not too different from what Dr. Hryniuk is doing now, and it is identical to what A L G B is currently doing in one of their trials. My second choice would probably be to go to the same total dose, but on different schedules.
DR. MCGUIRE: Dr. Goldie, how would you design a trial to test the concept of dose intensity? DR. GOLDIE: First of all, as the other panelists have said, although it might seem simple, it is really a very complex problem. And I doubt that any one trial could ever be constructed that would adequately test all of these facets. One would probably have to mount a series of trials. Ideally, I would like to see the trials conducted in tumors in which there are really unambiguous end points such as a cure, so that there would be a very clear-cut measure of how effective a protocol is. That, of course, limits the tumor types that could be tested quite substantially. I think one of the problems is that empirically over the last decade, many of the curative protocols are quite high on the dose intensity scale, and one clearly would not want to deliberately underdose a curative protocol just to create a difference between the two treatment arms. But I might want to look at a few signal tumors in which there are measurable cure rates with an existing standard protocol. Then I would want to see whether you could improve the dose intensity of that protocol, maybe by a factor of 15 to 20%, into a range that should still produce acceptable toxicity; then I would mount a randomized sudy between the two arms. Again, I think that the other variables that have to be looked at may have to be the subject of other trials. If you try to control too many of these variables in the same study, it will be confusing. You would probably have to look at the outcome of a large number of studies to see whether overall they are consistent with the concept of a dose-response effect.
Concluding remarks DR. MCGUIRE: I would now like to hear concluding remarks from the participants. Is there anything that might affect an oncologist's practice after reading about dose intensity? DR. EIRYNIUK: I think prospective randomized trials will probably underscore the importance of dose
Drug dosage intensity intensity. In all likelihood, the greatest mistake in curative chemotherapy is to unnecessarily back away from using effective dose intensities. In this category I would include adjuvant chemotherapy of stage II breast cancer. On the other hand, the greatest mistake in palliative chemotherapy would be to push treatment to the limit of patient tolerance in the mistaken belief that one would thereby greatly improve results. Increased toxicity will undoubtedly result from such inappropriate treatment. These mishaps would probably retard acceptance of prospective trials testing the concept of dose intensity in more appropriate situations. In terms of where to go from here - I am a bit surprised that Drs. Goldie and Tormey don't agree completely with the test design I have proposed. However, I must admit that my co-investigators and I have had two years to come up with the trial designs I described earlier. Perhaps the tests seem simple to us because we don't think scheduling is important except as it governs toxicity. To us, the major question has become: is dose intensity important or is the total dose more important? This simple question will require a huge topography of trials before it can be completely answered. To expedite all of these studies, we need consensus on an instrument for measuring toxicity, and a clear consensus on how to report received dose intensity. DR. TORMEY:The comment regarding toxicity consensus is worthwhile. There is currently a document circulating among the cooperative groups that is attempting to achieve common toxicity guidelines among physicians doing clinical trials in the United States. If that happens, it will go a long way towards improving intertrial comparisons. It is clear we all feel that relative dose intensity is important. It fits with all of our animal data concepts and with basically everything we have learned about treating human disease. The way that it has been formulated at the beginning of this discussion is simple and elegant, and enables one to deal conceptually with it in a very straightforward fashion. In terms of how to translate that concept into the everyday management of patients, one really needs
99
to deliver the maximum amount of drug the patient can tolerate over the shortest time frame. This does not necessarily mean that one should pick a regimen that one is not currently using, even though in some cases it might be appropriate. If you are using a given regimen, I would build in dose escalation and achieve that toxicity in the first or second cycle; then I would maintain that toxicity pattern throughout treatment rather than waiting several cycles of therapy to achieve that toxicity. OR. 6OLD~E: The great weight of evidence - certainly from experimental systems and to the extent it has been studied systematically in clinical chemotherapy - is that there is indeed a dose-response effect in chemotherapy; therefore, relative dose intensity is important. In fact, it would be astonishing, given what we know about the basic biology of these processes, if this were not the case. The burden of proof should really be on those who claim there is no dose-response effect because that does not make a lot of sense scientifically. In practical terms, regarding what this means as a message for the clinical chemotherapist, we have to be prepared to push our regimens a bit higher than we have. Also psychologically, although we may not like it, we have to accept occasional, inadvertent serious side effects in chemotherapy. It is more difficult, psychologically, to deliberately plan this, to build it into the protocol, and to know that 10% of the patients entering this study may have major problems. For animal studies, however, we know we have to push chemotherapy up into this range before one starts to see really significant effects. Our psychological attitudes have to be changed. The climate under which chemotherapy is to be administered has to encourage the physician to have the moral courage to persist with the treatment; neither the physicians nor the health care system should be critical of the occasional incidences where there have been serious or even fatal reactions to chemotherapy, because that may be the price of getting better results with our existing agents. DR. I-IRYNIUK: The greatest use of the concept of dose intensity is that it stresses that treatment delay
100
W L M c G u i r e et al.
arithmetically equals no treatment. If I w e r e to e n t r e a t o n c o l o g i s t s to r e m e m b e r any p a r t of the c o n c e p t , t h a t w o u l d b e it.
Address for offprints: W.L. McGuire Medicine/Oncology University of Texas Health Science Center San Antonio TX 78284 USA