Correspondence
Bone Marrow Transplantation versus Chemotherapy in Non-Hodgkin's Lymphoma
N Engl J Med 1995; 333:727-730September 14, 1995
- Article
To the Editor:
In their study of the efficacy of high-dose chemotherapy and autologous bone marrow transplantation in patients with non-Hodgkin's lymphoma who respond slowly to chemotherapy, Verdonck et al. (April 20 issue)1 used response criteria that may have influenced the outcome. A variety of criteria have been proposed to assess response in patients with lymphoma, but the most widely accepted are those ratified at the Cotswolds meeting,2 which define a partial remission as “a decrease by at least 50 percent in the sum of the products of the largest perpendicular diameters of all measurable lesions.” In contrast, Verdonck et al. considered patients with a reduction in tumor diameters of 25 percent or more to have had partial remissions. Thus, some of the randomized patients may not have fulfilled widely accepted criteria for an objective response. Did the outcome of patients with 25 to 50 percent reductions in tumor diameters differ from that of patients with a greater degree of cytoreduction?
It is possible that the patients most likely to benefit from early intensified therapy are those who have not had even a partial response after three cycles of standard therapy. Do the authors have any data on patients considered to be nonresponsive after three cycles of cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP)? Did any of these patients proceed directly to dose-intensive therapies, and what was their ultimate outcome?
2 ReferencesJohn F. Seymour, M.B., B.S.
Royal Melbourne Hospital, Parkville, Victoria 3050, Australia1
Verdonck LF ,van Putten WLJ ,Hagenbeek A , et al. Comparison of CHOP chemotherapy with autologous bone marrow transplantation for slowly responding patients with aggressive non-Hodgkin's lymphoma. N Engl J Med 1995;332:1045-1051
Full Text | Web of Science | Medline2
Lister TA ,Crowther D ,Sutcliffe SB , et al. Report of a committee convened to discuss the evaluation and staging of patients with Hodgkin's disease: Cotsw olds meeting. J Clin Oncol 1989;7:1630-1636
Web of Science | Medline
To the Editor:
The article from the Netherlands concerning autologous bone marrow transplantation has excited a lot of comment because of its conclusions about the place of marrow transplantation in the management of non-Hodgkin's lymphoma.
The authors are guilty of an error that a first-year statistics student would recognize: sophisticated statistical analysis has been applied to an inhomogeneous set of data. Immunoblastic lymphoma is as similar to diffuse, small-cleaved-cell lymphoma as is an apple to an avocado. Any attempt to draw therapeutic conclusions from large studies of such dissimilar biologic entities will support the null hypothesis every time.
What a pity that the authors chose to classify and recruit patients on the basis of the biologically compromised International Working Formulation. The modified Kiel classification allows a much more precise and biologically relevant separation of non-Hodgkin's lymphoma.1 Its use in their study would have resulted in a clear-cut distinction between entities and therefore a much more discerning choice of therapy.
Groups D, E, and F of the Working Formulation (follicular, large-cell; diffuse, small-cleaved-cell; and diffuse mixed-cell types, respectively) are all equivalent to the category of centroblastic–centrocytic lymphoma in the modified Kiel classification.2 With the use of this classification, three groups stand out distinctly, as shown in Table 1Table 1
Classification of Non-Hodgkin's Lymphoma in the Patients Studied by Verdonck et Al.. A slow response to therapy, incomplete but prolonged remissions, and a high frequency of marrow involvement are quite characteristic of centroblastic–centrocytic lymphoma. Most oncologists used to working with the modified Kiel classification would not consider autologous bone marrow transplantation a first-line treatment for this lymphoma.3 Precise taxonomy should be a prerequisite for all clinical trials involving non-Hodgkin's lymphoma. The study by Verdonck et al. does little to advance knowledge on the subject.
3 ReferencesSteve Flecknoe-Brown, F.R.A.C.P.
Sydney Adventist Hospital, Wahroonga, NSW 2067, Australia1
Lennert K, Feller AC. Histopathology of non-Hodgkin's lymphomas. New York: Springer-Verlag, 1992.
2
Lukes RJ, Collins RD. Tumors of the hematopoietic system. Armed Forces Institute of Pathology fascicle no. 28. Bethesda, Md.: Armed Forces Institute of Pathology, 1992:35.
3
Hiddemann W ,Unterhalt M . Current status and future perspectives in the treatment of low-grade non-Hodgkin's lymphomas. Blood Rev 1994;8:225-233
CrossRef | Web of Science | Medline
To the Editor:
Verdonck and colleagues showed that patients with aggressive non-Hodgkin's lymphoma do better with three plus five cycles of CHOP than with four cycles of CHOP plus high-dose chemoradiotherapy if they responded slowly to three cycles of CHOP, reaching only a partial remission. This result may have been contrary to the authors' expectations, but when calculated over the entire treatment time the intensity of chemotherapy is higher in the group treated with eight cycles of CHOP (total doses for eight cycles: 6000 mg of cyclophosphamide per square meter of body-surface area, 400 mg of doxorubicin per square meter, 16 mg of vincristine and 4000 mg of prednisone) than in the group given four cycles of CHOP plus high-dose cyclophosphamide and total-body irradiation. The total amount of drug in the latter group is half that in the former except for cyclophosphamide, which was given in a dose of 60 mg per kilogram of body weight for two days (approximately 4800 mg per square meter) and therefore reached a higher total dose (7800 mg per square meter). The total-body irradiation also has to be considered. The 8-Gy dose was administered in one fraction; when given together with high-dose cyclophosphamide this dose can ablate the bone marrow but is not high enough to eradicate aggressive lymphoma cells. Fractionated irradiation with a total dose of 40 Gy is necessary for that. The crucial point is that we want to treat the non-Hodgkin's lymphoma and not the bone marrow if it is not involved; the combination of total-body irradiation and high-dose cyclophosphamide is a good conditioning regimen for bone marrow ablation but possibly not the best regimen for the eradication of lymphoma cells.
Klaus Mross, M.D.
University Hospital Eppendorf, D-20251 Hamburg, GermanyTo the Editor:
The study by Verdonck et al. can be interpreted in several ways. The authors suggest that they compared the efficacy of CHOP chemotherapy with that of autologous bone marrow transplantation in patients with non-Hodgkin's lymphoma who did not have a complete remission with primary therapy and who had lymphoma-negative bone marrow (a slow response). Another interpretation of this study is that it examined whether a slow response to primary therapy is a prognostic factor that can predict the response to either of two second-line therapies: additional CHOP chemotherapy or autologous bone marrow transplantation. The study demonstrates that a slow response is not predictive of a difference in outcome, as compared with a fast response, and it does not predict which patients, if any, will respond to additional therapy with CHOP or autologous bone marrow transplantation therapy. There is no evidence that a slow response to primary therapy is a valid prognostic factor. The enthusiasm of the editorialist (April 20 issue)1 for the conclusion that autologous bone marrow transplantation does not provide an important benefit in non-Hodgkin's lymphoma must be tempered with the knowledge that this study shows only that a slow response cannot be used to identify a group of patients who will have a better response to autologous bone marrow transplantation than to additional CHOP chemotherapy.
1 ReferencesHarry B. Burke, M.D., Ph.D.
New York Medical College, Valhalla, NY 105951
Walker AM . Low power and striking results -- a surprise but not a paradox. N Engl J Med 1995;332:1091-1092
Full Text | Web of Science | Medline
To the Editor:
The report by Verdonck and colleagues must be interpreted with caution. Their conclusion, and that echoed in the accompanying editorial, was that autologous bone marrow transplantation did not improve the survival of patients with aggressive non-Hodgkin's lymphoma and slow, partial responses to CHOP chemotherapy. Some readers may therefore conclude that consolidation therapy with autologous bone marrow transplantation has no role in the initial care of patients with aggressive lymphoma and a poor prognosis.
However, the authors did not really treat patients with a poor prognosis. More than half the patients had a low risk or a low-to-intermediate risk, according to the International Index defined by the International Non-Hodgkin's Lymphoma Prognostic Factors Project,1 and they would be expected to do well with standard combination chemotherapy. In fact, only one patient in each arm of the randomized study was in the high-risk group. In addition, although survival was similar in slowly responding patients and in those with rapid responses, the relapse rate in patients who responded slowly was nearly half that of patients who responded rapidly (23 percent vs. 40 percent).
A significant number of slowly responding patients with good partial responses and a residual mass of unknown importance may have already been entering a complete remission with CHOP alone. A residual mass is common after treatment, particularly in patients with bulky mediastinal disease; scanning with gallium-67 would have been useful to evaluate patients with equivocal radiographic results, as would biopsies in selected patients. Because such additional tests were not used, it is much more difficult to be confident that there is a group with a poor prognosis among the patients who responded slowly.
Another caveat in interpreting this study relates to the efficacy of the conditioning regimen used in the high-dose arm of the trial. Combination therapy with a single fraction of total-body irradiation and cyclophosphamide at a dose of 120 mg per kilogram is a relatively modest treatment regimen for aggressive lymphomas and is probably inferior to regimens currently used at most autologous-transplantation centers.2
Although this study demonstrated that a slow response alone is not an appropriate criterion for early intervention with autologous bone marrow transplantation, this result is not related to the use of autologous transplantation in properly selected patients with relapsed disease or as possible initial therapy for those with poor prognostic features according to the International Index. In this regard, data from our institution demonstrated disease-free survival of 79 percent in poor-risk patients who underwent transplantation as part of their initial therapy.3 Although the criteria for selecting patients who will benefit most from immediate marrow-ablative therapy remain to be clarified, the role of autologous transplantation in patients with relapsed non-Hodgkin's lymphoma is established.
3 ReferencesStephen D. Nimer, M.D.
Andrew Zelenetz, M.D., Ph.D.
Carol Portlock, M.D.
Memorial Sloan-Kettering Cancer Center, New York, NY 100211
The International Non-Hodgkin's Lymphoma Prognostic Factors Project
Full Text | Web of Science | Medline2
Gulati S ,Yahalom J ,Acaba L , et al. Treatment of patients with relapsed and resistant non-Hodgkin's lymphoma using total body irradiation, etoposide, and cyclophosphamide and autologous bone marrow transplantation. J Clin Oncol 1992;10:936-941
Web of Science | Medline3
Gulati SC ,Shank B ,Black P , et al. Autologous bone marrow transplantation for patients with poor-prognosis lymphoma. J Clin Oncol 1988;6:1303-1313
Web of Science | Medline
To the Editor:
I was surprised to read in the abstract of the article by Verdonck et al. that disease-free survival was higher than overall survival in the transplantation group (60 percent vs. 56 percent). The same values for overall survival and disease-free survival were given in the text under the heading “Survival of the Randomized Patients” and in Figure 1, whereas the legend for Figure 1 suggests an overall survival rate of 65 percent (12 deaths among 34 patients). Is this just a typographic error, and if not, do these differences in the values affect the statistical analysis?
Waltraud Finckh, M.D.
Wendelinusstr. 3, D-97688 Bad Kissingen, GermanyTo the Editor:
We take issue with the editorial by Walker,1 which suggests that the study by Verdonck et al. yielded a clinically useful result. In the search for a role for high-dose therapy early in the natural history of non-Hodgkin's lymphoma, the choice of a suitable risk group that might benefit from such treatment is paramount. It is clear from the study that a slow response to CHOP was, through no fault of the investigators, not the correct choice on which to base randomization. The investigators themselves confirmed that the rapidity with which a complete response is achieved appears to have no prognostic importance. Data from the British National Lymphoma Investigation2 corroborate this; interestingly, patients with a slow response who nevertheless entered complete remission were shown to have an actuarial five-year survival rate in excess of 90 percent. In addition, of all the randomized patients in the Verdonck study, more than 50 percent were at low or low-to-intermediate risk, as defined by the International Index.3 So, although such a study cannot fail to yield an interesting result, it has done nothing to support or refute the prevailing hope that high-dose chemotherapy may improve survival for patients with non-Hodgkin's lymphoma who are truly at high risk. Such patients, who have a poor prognosis, must continue to be enrolled in randomized clinical trials.
3 ReferencesAdele K. Fielding, M.R.C.P.
Rachel M. Pearce, M.Sc.
Anthony H. Goldstone, F.R.C.Path.
London Hospitals, London WC1E 6AU, United Kingdom1
Walker AM . Low power and striking results -- a surprise but not a paradox. N Engl J Med 1995;332:1091-1092
Full Text | Web of Science | Medline2
Gribben JG ,Vaughan Hudson B ,Linch DC . The potential value of very intensive therapy with autologous bone marrow rescue in the treatment of malignant lymphomas. Hematol Oncol 1987;5:281-293
CrossRef | Web of Science | Medline3
The International Non-Hodgkin's Lymphoma Prognostic Factors Project
Full Text | Web of Science | Medline
To the Editor
: As a consulting statistician with approximately 25 years of practical experience, I take strong exception to the editorial remarks by Dr. Walker regarding statistical practices. I believe that the views expressed by Dr. Walker are quite distorted and far from the mainstream of statistical thinking.
Let me first point out that the techniques used for inferential statistics can be broadly categorized as follows: point estimation, interval estimation, and hypothesis testing. Each category has specific areas of usefulness, and techniques from several categories are often used jointly.
Walker seems to imply that only confidence-interval techniques should be used and that the use of hypothesis-testing methods is often inappropriate. In contrast, my experience has been that in most experiments, especially clinical trials, the paramount methods of data analysis are tests of hypotheses, rather than confidence-interval methods. This is also probably true for the biologic and medical literature. Moreover, there is a well-known duality between confidence intervals and hypothesis testing. One can interpret a confidence interval as the set of parameter values that would not be rejected in a test of hypothesis.
Walker further indicates that when hypothesis testing is performed, such tests should always be two-sided. Again, this opinion is misleading and inappropriate. If one has prior information about the probable direction of an effect, such information should be incorporated into the statistical procedure. Suppose that one is interested in pursuing a new medical therapy only if it can be shown to be superior to a standard therapy. Why use critical resources to test for the inferiority of the new therapy? When the direction of a test is specified in advance of the data analysis (i.e., in the study protocol), then no increase is introduced into the false positive rate of the procedure. Not to specify the direction of a test would cause important information to be ignored and would unnecessarily increase the size of a study. If it is the standard policy of the Journal to require all statistical tests to be two-sided, then I suggest that this policy should be critically reviewed.
In his criticism of Verdonck et al., Walker concludes that the authors found a clinically useful result despite having a study with insufficient power. In contrast, I support the methodologic approaches of Verdonck et al. and believe that they dealt with the design and analysis of the study in the proper way.
Murray R. Selwyn, Ph.D.
Statistics Unlimited, Wellesley Hills, MA 02181Author/Editor Response
The authors reply:
To the Editor: We agree with Drs. Burke and Nimer et al. that a slow response to the initial cycles of chemotherapy can no longer be used to predict the outcome of treatment in non-Hodgkin's lymphoma. We also agree — and discussed the possibility — that other subgroups of patients with non-Hodgkin's lymphoma, identified by different prognostic factors, could benefit from autologous bone marrow transplantation, such as high-risk patients identified with use of the International Index. At the moment, however, there are no scientific data to support this notion.
In response to the issue of the intensity of the treatment schedule raised by Nimer et al. and Mross: we know of no evidence that hyperfractionated total-body irradiation is more effective against non-Hodgkin's lymphoma than a single fraction of a high dose of total-body irradiation. In fact, it has been extremely difficult to compare schedules of total-body irradiation in different centers because of the many variables, some of which are not fixed (e.g., the dose rate). We do not agree with Dr. Mross that concentrated high-dose cytotoxic therapy, as is used in the context of autologous bone marrow transplantation, can be compared with prolonged chemotherapy by simply summing up the doses of chemotherapy.
There has been and continues to be much debate about pathological systems of classification for clinical studies of non-Hodgkin's lymphoma. Dr. Flecknoe-Brown suggests the use of the Kiel classification. We elected to apply the Working Formulation because it has been widely used to evaluate the outcome after new treatment approaches in non-Hodgkin's lymphoma. A new method of classification, recently designed by the International Lymphoma Study Group,1 may be an improvement, but this scheme has already raised a lot of controversy, and there is currently no agreement about its clinical usefulness.2
Dr. Seymour points to our use of a definition of partial response as a reduction in tumor diameters of 25 percent or more instead of a reduction of 50 percent or more. For the evaluation of early autologous bone marrow transplantation, however, we included all patients with a reduction in tumor diameters of 25 percent or more in the group with slowly responsive lymphoma. Because this analysis was done earlier than tumor response is usually evaluated, we used a less strict cut-off level for partial response.
Finally, there is no reason for Dr. Finckh to be concerned. The actuarial analysis corrects for the variable duration of follow-up of the patients. Therefore, the actuarial rate of overall survival of 56 percent is lower than the crude rate of 65 percent (22/34). The analysis of disease-free survival was restricted to patients who had a complete response, whereas all patients were included in the analysis of overall survival. Therefore, event-free survival is a better end point than disease-free survival. As can be seen in Figure 1 and Figure 2 of our article, event-free survival is lower than overall survival.
2 ReferencesLeo F. Verdonck, M.D., Ph.D.
University Hospital, Utrecht, 3508 GA Utrecht, the NetherlandsWim L.J. van Putten, M.Sc.
Anton Hagenbeek, M.D., Ph.D.
Dr. Daniel den Hoed Cancer Center, 3075 EA Rotterdam, the Netherlands1
Harris NL ,Jaffe ES ,Stein H , et al. A revised European-American classification of lymphoid neoplasms: a proposal from the International Lymphoma Study Group. Blood 1994;84:1361-1392
Web of Science | Medline2
Rosenberg SA . Classification of lymphoid neoplasms. Blood 1994;84:1359-1360
Web of Science | Medline
Author/Editor Response
Dr. Selwyn takes exception to the claims that hypothesis testing may be inappropriate for some problems and that, when performed, tests should always be two-sided. The study by Verdonck et al. provides a counterexample to Dr. Selwyn on both points. The one-sided test of hypothesis originally planned would have had a P value on the order of 0.8, and the outcome of autologous bone marrow transplantation would have been declared not significantly different from that of conventional therapy. It may be, as Fielding et al. suggest, that the patients enrolled in the trial were not the ones who should have been used for an evaluation of autologous bone marrow transplantation, but no one questions the empirical observation. Among the people who were enrolled, those assigned to autologous bone marrow transplantation did much worse than those assigned to CHOP. The confidence interval tells us that it is very unlikely that the transplantation protocol implemented by Verdonck et al. is much better than CHOP in such patients. This is clinically important. Dr. Selwyn's one-sided P value would cavalierly write the result off as not significant, so that similar patients would in the future be enrolled in trials of therapies from which they would be unlikely to benefit.
I see no objection to presenting P values. However, the usefulness of one-sided tests is in the eye of the beholder. There are few important studies for which Journal readers would be indifferent to a catastrophic and surprising outcome. If readers are interested in both positive and negative outcomes of a study, it hardly matters that the investigator can put on blinders and profess interest only in the positive result.
Alexander M. Walker, M.D., Dr.P.H.
Harvard School of Public Health, Boston, MA 02115
