Meta-Analysis to Evaluate the Role of Interferon in Follicular Lymphoma
http://www.100md.com
《临床肿瘤学》
St Bartholomew's Hospital, London, UK
University of Minnesota, Minneapolis, MN
Taussig Cancer Center, Cleveland, OH
Centre Jean Bernard, Le Mans, France
University Medical Center, Utrecht, The Netherlands
Wilmot Cancer Center, University of Rochester, Rochester, NY
Georg-August-University, Gottingen, Germany
Hospital Universitario de la Princesa, Madrid, Spain
Ospedale Civile, Venice, Italy
Oncology Hospital, IMSS, Mexico City, Mexico
ABSTRACT
PURPOSE: To determine whether interferon (IFN) -2, when given with or following chemotherapy, influences response rate, remission duration, and survival in newly diagnosed patients with follicular lymphoma.
PATIENTS AND METHODS: Ten phase III studies evaluating the role of IFN-2 in 1,922 newly diagnosed patients with follicular lymphoma were analyzed. Updated individual patient data were used to perform meta-analyses for response, survival, and remission duration.
RESULTS: The addition of IFN-2 to initial chemotherapy did not significantly influence response rate. An overall meta-analysis for survival showed a significant difference in favor of IFN-2, but also showed significant heterogeneity between studies. Further analyses were carried out in order to explain this heterogeneity, and to define the circumstances in which IFN-2 prolonged survival. The survival advantage was seen when IFN-2 was given: (1) in conjunction with relatively intensive initial chemotherapy (2P = .00005), (2) at a dose 5 million units (2P = .000002), (3) at a cumulative dose 36 million units per month (2P = .000008), and (4) with chemotherapy rather than as maintenance therapy (P = .004). With regard to remission duration, there was also a significant difference in favor of IFN-2, irrespective of the intensity of chemotherapy used, IFN dose, or whether IFN was given as a maintenance strategy or with chemotherapy.
CONCLUSION: When given in the context of relatively intensive initial chemotherapy, and at a dose 5 million units ( 36 x 106 units per month), IFN-2 prolongs survival and remission duration in patients with follicular lymphoma.
INTRODUCTION
Interferon (IFN) was introduced into the treatment of follicular lymphoma following observation of activity in L1210 leukemia1 and AKR lymphoma,2 but with limited understanding of its mode of action. Phase II studies (using an empirical dose and schedule derived from the original Scandinavian osteogenic sarcoma trial3) in low-grade lymphoma, resulted in response rates of 30% to 50%, regardless of the source of IFN-2.4-11 Such responses were, however, almost always incomplete and, in general, short lived. Activity having been demonstrated, the concept of synergy between IFN and conventional cytotoxic drugs was tested and confirmed in murine models of leukemia12 and lymphoma,13 in human breast cancer xenografts in nude mice,14 and in patients with follicular lymphoma known to be refractory to chemotherapy.15,16
A series of randomized studies followed, in newly diagnosed patients with a spectrum of histologic subtypes, evaluating the use of IFN-2 given in conjunction with chemotherapy.17-29 In some studies, IFN-2 was given concurrently with chemotherapy, in others, it was used as "maintenance" therapy, while in still others, it was used throughout. The chemotherapy regimens varied in intensity, from relatively low doses of alkylating agents (chlorambucil or cyclophosphamide), to doxorubicin or mitoxantrone-containing regimens. The dose and schedule of IFN-2 also varied, though more than half of the studies embraced the standard, albeit empirical, dose of 3 x 106 units, given three times weekly, for 6 months or 1 or 2 years (or in the case of one study,26 until recurrence). Further variability was introduced by the fact that some studies permitted (or advocated) the use of radiotherapy on completion of chemotherapy to sites of bulky disease at presentation, or to residual disease.25,27,28
In addition to the variation in histologic subtype engendered by the inclusion of patients with any type of low-grade lymphoma (though the majority of patients in fact had follicular lymphoma), there were also differences in eligibility criteria. Some studies were very permissive, allowing enrollment of not only patients who had a specific indication for treatment, but also of those who did not. Conversely, others, for example the GELA (Groupe d'Etudes des Lymphomes de l’Adulte) study,19,20 focused on patients deemed to have an adverse prognosis (by GELA criteria), while one study included only patients in whom a complete remission (CR) had been achieved with prior chemotherapy.27
As a consequence of this heterogeneity (in both patient population and entry criteria), there are major discrepancies in the results, even when only studies with a similar design are considered. Therefore, for this meta-analysis, only patients with follicular lymphoma were included. The aim was to determine whether IFN-2 given at any point improves response rate, remission duration, and survival.
PATIENTS AND METHODS
Studies
At the inception of the project, review of the published literature (and of completed analyses presented in abstract form) led to identification of eight phase III studies of newly diagnosed patients with "low-grade" or, specifically, follicular lymphoma. A protocol was written and distributed to individual authors and groups, who were then asked to provide updated (ie, updated from the original publication or analysis for that study) individual patient information, which was transmitted electronically to us (W.G.). Data were checked and only patients described as having follicular lymphoma were selected for inclusion in the analysis. Central review of pathology was not undertaken. On this basis, an initial meta-analysis was conducted and presented.30 Data from two studies that had not been published at that time were subsequently made available.24,28 The present meta-analysis, therefore, comprises the results for patients included in ten phase III studies.
Patients
The analysis relates to 1,922 patients whose clinical characteristics are listed in Table 1. The majority of patients had stage IV disease by virtue of bone marrow involvement.
Treatment
Treatment in terms of chemotherapy regimen and the dose, schedule, and timing of IFN-2 is summarized in Table 2, by Study Group (or country of origin when a formal group was not involved). In six of the studies, IFN-2 was given as part of initial therapy17-24,29; in four studies, it was given only as maintenance therapy25-28; and in four studies, patients could be randomly assigned to receive it throughout.17,18,24,29
Chemotherapy. Chemotherapy regimens were categorized by the intensity of chemotherapy as follows: studies utilizing relatively "less intensive" chemotherapy are defined as those that used chlorambucil, cyclophosphamide, or cyclophosphamide/vincristine/prednisolone (CVP) as initial therapy (Table 2). Those studies that used an anthracycline or mitoxantrone-based combination regimen were deemed "more intensive" (Table 2).
Interferon. The dose of IFN-2 was similarly categorized by individual dose and cumulative dose administered over a period of 4 weeks. Individual doses were grouped by those doses equal to or less than 5 x 106 (ie, 2 x 106 units/m2 or 3 x 106 units per dose). The dose per month follows closely; ie, the latter doses, given 3x weekly, translate into a monthly dose of 36 x 106 units. Hence, for the meta-analysis, the studies were divided into those studies using more than 36 x 106 units per month or that amount (Table 2).
Meta-Analyses: Methodology
Response data (CR/partial response v the rest), survival times, and durations of remission were calculated for each individual patient, and then the overview methodology described in detail by Peto31 was applied. For the response meta-analysis, data from the six studies that evaluated the effect of adding IFN-2 to initial chemotherapy were used.17-24,29 For the survival and remission duration meta-analyses, data from all ten studies were included.
In brief, for response, the expected number of responses was calculated for each trial arm, assuming no difference in response rate between the two arms. This was then compared with the observed number of responses, and the differences (observed-to-expected ratio [O-E]) calculated for each trial. For survival and remission duration, the expected number of events (deaths or recurrences) in each arm, assuming no difference between the treatments, was derived for each trial. This was then compared with the observed number of events and the differences (O-Es) calculated. For response, survival, and remission duration, these O-E figures were then summed over the whole set of trials to form a grand total. The variance of the individual estimates was also derived and summed, giving a variance and a standard deviation (SD) for the grand total. The number of SDs (z) by which the grand total differs from zero gives an estimate of the significance of any effect. The O-Es were also tested for heterogeneity to see whether the scatter of results was unexpected. (The sum of [O-E]2/variance should be distributed as 2n – 1 if the scatter is random, where n is the number of studies.)
For response, survival, and remission duration, the size of any treatment effect, specifically the typical odds ratio (TOR), was simply estimated as exp(z/SD), with approximate 95% CI expressed as exp (z/SD ± 1.96/SD). CIs for the individual trials were calculated in the same fashion. The TORs and their CIs for each trial, together with the grand total, are shown in Figures 1-7. The size of the solid squares is proportional to the amount of information each trial contains. Because many individual trials are displayed, 99% CIs are shown for the individual trials. For the overall result, 95% CIs are given (represented by the open diamonds in the figures). These odds ratios were converted into 5- and 10-year differences in survival and remission duration as follows: the odds ratio, Or, is related to the hazard ratio, , by the equation = 1/(1 – [Or/100]). If the survival proportions at any time, t, are P1 and P2, then = loge(P1)/loge(P2), so, if the overall proportion surviving at time t is P, then P = (n1P1+n2P2)/(n1+n2) where n1 and n2 are the sample sizes at time t (which can be calculated as described by Simon32). It follows that P1 +(n2/n1)P1 – P(n1+n2)/n1 = 0, and this can be easily solved numerically to give P1, P2 and the difference P1 – P2. For response, the calculations are identical using the proportions responding instead of the survival proportions.
Four of the 10 studies had a double randomization, as listed in Table 2. Clearly, for each of these studies, only a proportion of randomly assigned patients initially reached the second randomization (for a variety of reasons including toxicity, patient refusal, and death). Those failing to reach the second randomization had a worse survival than those who did, because of this selection process. Furthermore, the numbers not reaching the second randomization were not necessarily the same as in the two initial randomization groups. For instance, in the UK study, 47 (45%) of 104 patients initially randomly assigned to receive IFN-2 failed to reach the point of second randomization, compared with 30 (30%) of 100 of those patients initially randomly assigned to the chlorambucil-only arm. Therefore, in these studies, to analyze just those patients who received "any IFN-2" versus the rest of the patients would bias the result, almost certainly in favor of the IFN-2 group. To avoid this problem and to make it possible to perform a meta-analysis of IFN-2 given at any point versus no IFN-2, such studies have been regarded as having two separate strata, ie, comprising patients who reached the second randomization and those who did not. (For the Non-Hodgin's Lymphoma Study Group [NHLSG] study, information on the second randomization was unavailable, and so for this study only the initial randomization was used in the meta-analysis graphs). The remaining three studies are therefore each represented twice on the meta-analysis graphs, reflecting the two different strata.
The survival of the IFN-2 and no IFN-2 groups was plotted using the method described by Gregory33 (see Fig 8). This enables the survival percentages to be calculated stratified by center without assuming proportional hazards, and produces a survival comparison that represents the meta-analysis result and an adjusted 2 statistic that matches the P value from the meta-analysis. To check the proportional hazards assumption and to evaluate possible changes in the hazard ratio over time, a smoothed (instantaneous) hazard was plotted over time (stratified by center and thus matching the meta-analysis findings) by calculating the hazard ratio for each death time as the hazard for the 100 deaths occurring both before and after that time. This limited the time scale of the graph somewhat because the first 100 deaths occurred during the first year, whereas the last 100 deaths occurred during years 7 to 11, when follow-up is relatively short. However, this methodology made it possible to gain an approximation of the pattern of changing hazard with time. The number 100 was chosen after experimentation to dampen out large fluctuations in the hazard that would obscure the pattern. It was then possible to test for any observed trend in the hazard by calculating hazards over adjacent intervals (eg, each year) and then calculating a weighted correlation coefficient (where the weights are the SEs of the hazard estimates) for these hazards with time. The meta-analysis methodology assumes proportional hazards (so the hazard ratio would be constant over time), thus these techniques make it possible to discern significant departure from this assumption.
Because significant heterogeneity was observed between studies (see Results), further analyses were carried out in an attempt to explain this heterogeneity. A number of hypotheses were generated. The following appeared the most likely to explain the observed differences; that IFN-2 might increase survival when given: (1) concurrently with, or following, relatively intensive chemotherapy; (2) at a dose equal to or greater than 5 x 106 units; (3) at a cumulative dose in excess of 36 x 106 units over a period of 4 weeks; or (4) with initial chemotherapy rather than as maintenance therapy.
RESULTS
Response
The addition of IFN-2 to initial chemotherapy did not significantly influence response rate overall (2P = .41). The estimated difference in percentage response rate was –2% (ie, in favor of the no IFN group; 95% CI, –6% to 3%; Fig 1), though there was considerable heterogeneity between studies (2P = .005), and a highly significant benefit from IFN-2 was observed in the GELA study (2P = .006).
Survival
Considering all eligible patients, the use of IFN-2 at any point in treatment appeared to prolong survival (2P = .004; Fig 2). The corresponding survival curve is shown in Figure 8, and the estimated 5- and 10-year survival differences (see Meta-Analysis: Methodology) were 5.5% (95% CI, 2% to 9%) and 8% (95% CI, 3% to 13%), respectively. However, there was significant heterogeneity between studies (2P = .003). Furthermore, the Mexican study's result is clearly an outlier, because there is a much greater survival effect in this trial than in any of the other studies. When the data from the Mexican study were excluded, the results were still significant (2P =.03), with 5- and 10-year survival differences of 4% (95% CI, 0.4% to 8%) and 6% (95% CI, 0.6% to 11%), respectively, and the heterogeneity effect was of borderline significance (2P = .07). However, there are no a priori grounds for excluding the Mexican study. The main difference between the latter and the other studies is the fact that only patients in CR were eligible for randomization. Their survival overall may thus be superior to that of patients in the other studies which included those patients in whom a partial response had been achieved. For patients failing to respond, the use of IFN-2 at any point in treatment did not prolong survival (Z = –0.81; 2P = .42; data not shown), with any trend being in favor of the no IFN group.
Further inspection of the data raised the four hypotheses listed in the Patients and Methods section; these hypotheses were investigated further. Analyzing each of the four hypotheses separately, for patients receiving relatively intensive initial therapy, IFN-2 used at any point was associated with prolongation of survival (2P = .00005; Fig 3; 5-and 10-year survival differences being 11% [95% CI, 6% to 16%] and 15% [95% CI, 8% to 23%], respectively. In contrast, for patients receiving less intensive initial therapy (ie, alkylating agents or CVP), there was no significant improvement in survival (2P = .93; Fig 4).
Considering the individual dose of IFN-2, for patients receiving 5 x 106 units or more per dose, there was a highly significant increase in survival (2P = .000002; Fig 5; 5- and 10-year survival differences being 15% [95% CI, 9% to 21%] and 20% [95% CI, 12% to 28%], respectively). Conversely, for patients receiving less than 5 x 106 units per dose, there was no effect on survival (2P = .75; data not shown).
Similarly, in terms of the monthly dose of IFN-2 administered, for patients receiving more than 36 x 106 units over 4 weeks, there was a significant improvement in survival (2P = .000008; Fig 6; 5- and 10-year survival differences being 18% [95% CI, 10% to 25%] and 23% [95% CI, 14% to 33%], respectively), compared with no significant effect when IFN-2 was used at the typical dose of 3 x 106 units 3x weekly, (equivalent to 36 x 106 units per 4 weeks; 2P = .43; data not shown).
For studies in which IFN-2 was given initially together with chemotherapy, there was a significant improvement in survival (2P = .004; data not shown; 5- and 10-year survival differences being 7% [95% CI, 2% to 12%] and 9% [95% CI, 3% to 15%], respectively), compared with no significant effect in studies in which it was given only as maintenance (2P = .20; data not shown). In the latter group, the only study that demonstrated an effect was the Mexican study, which was, again, an outlier. Excluding this study reduced the significance to 2P = .98.
It should be noted that none of the latter four results are significantly altered by the exclusion of the data from the Mexican study. For instance, for patients receiving 5 x 106 units or more per dose, there was still a highly significant increase in survival (2P = .0002; data not shown). The odds reduction was 37% ± 10% (hardly different than with the inclusion of the Mexican study's data), and 5- and 10-year survival differences were also largely unaffected, being 13% (95% CI, 6% to 20%) and 17% (95% CI, 8% to 25%), respectively. The other three comparisons show similar small differences from the equivalent analyses that included the Mexican data (data not shown). This is because, though the Mexican study shows a relatively large survival difference, it is a relatively small study, and the significance and magnitude of the meta-analysis differences are influenced more by the studies with larger numbers of patients, such as those conducted by the Eastern Cooperative Oncology Group (ECOG) and GELA.
Plotting the hazards as they change over time, as described in Patients and Methods, showed a pattern of increasing hazard over time (Fig 9). There is a big rise in risk of death from not having received IFN during the 3- to 4-year period after entry, and though the risk reduces slightly thereafter, it is maintained at least up to 7 years from entry, and probably beyond. Although there may be some random element to this pattern, it is very clear that the risk increases over time from entry. It therefore appears that IFN produces an effect that is maintained, or even increases, over time. This was confirmed by a weighted correlation of adjacent yearly hazards that gave a correlation coefficient of r = 0.60 (2P = .04). Thus the 10-year survival estimates are probably slight underestimates of the true difference (by approximately 25%, so the overall reported 10-year survival difference of 8% is probably closer to 10%), and this is confirmed by inspection of the adjusted survival graph (Fig 8) and survival estimates taken from it, which show this slightly larger difference (the adjusted survival curve methodology does not assume proportional hazards33 and can therefore produce these estimates). The 5-year survival estimates appear to be much more accurate, being in the middle of the survival range for this analysis. It should be noted that the departure from proportional hazards is not sufficient to invalidate the findings of the study; its main effect is to create a slight underestimate of the 10-year survival figures.
To confirm this finding, the meta-analyses were repeated using only data from the patients who had already survived 3 years (because the hazard is close to zero for the first 3 years, which is different from its later values), and the basic findings were unaltered; if anything the results tended to be more significant, and with less heterogeneity between studies.
Remission Duration
Overall, the addition of IFN-2 at any point significantly prolonged remission duration when all ten studies were considered (2P < .000001; Fig 7; 5- and 10-year remission duration differences being 11% [95% CI, 7% to 15%] and 10% [95% CI, 6% to 13%], respectively); the corresponding remission duration curve is shown in Figure 10. Evaluating the effect of maintenance IFN-2 alone (ie, considering only the four studies in which it was not used as part of initial therapy), there was still a highly significant difference (2P = .0002; 5- and 10-year remission duration differences being 13% [95% CI, 6% to 19%] and 12% [95% CI, 6% to 19%], respectively), in contrast to the survival comparisons.
DISCUSSION
These results demonstrate that the addition of IFN-2 to conventional chemotherapy as part of initial therapy for newly diagnosed patients with follicular lymphoma does not result in any improvement in response rate. However, the meta-analysis confirms that IFN-2 does prolong remission duration and overall survival, and also prolongs survival and remission duration under specific circumstances that relate to the intensity of the chemotherapy used, the dose of IFN-2, or both. The point at which IFN-2 is given, ie, as part of initial therapy rather than as maintenance therapy, also appears to influence outcome. There was no survival difference following only maintenance IFN-2 therapy (except in the Mexican study).
To summarize, overall, there was an effect of IFN-2 on survival; it seems likely that this relates to some or all of the following factors: intensity of the initial chemotherapy, dose or dose-intensity of IFN-2, and whether the IFN-2 was given as part of initial therapy or as maintenance therapy. However, it is difficult (if not impossible) to determine which of these hypotheses is correct. The statistical differences between them are marginal. For each hypothesis there is at least one counter-example, or one study, that is the exception. It thus seems probable that more than one factor is at work; for instance, that IFN-2 is effective at higher doses, and in studies in which the chemotherapy is more intensive, though this theory is impossible to prove from these data.
The difference in survival (Figs 8 and 9) increases over time. The survival effect appears to be a late effect, and the difference between the curves may therefore widen further with longer follow-up. This is in keeping with the difference in remission duration, which shows a consistent effect from about 1 year onwards, with a difference of at least 10% at 10 years. This might be expected to translate eventually into a similar size of survival difference, as is becoming apparent in the survival curve.
It is important to recognize that the data on which the meta-analysis is based reflect protocol dose and not actual dose administered. Originally, it had been the intention to analyze the results using the dose given (of both chemotherapy and of IFN-2), but not all of the data sets included this information. Thus it was not possible to do this, because individual patients' medical records could not be reviewed owing to the logistical difficulties of doing so for ten studies worldwide.
The prolongation in survival in the context of relatively more intensive chemotherapy is unlikely to be simply a reflection of an increase in the intensity of the chemotherapy. Intensification of treatment in follicular lymphoma has increased remission duration34 but not survival.35-43
An earlier meta-analysis based on data from 8 studies (as published in manuscript or abstract form) also showed significantly better 5-year survival and longer progression-free survival at 3 and 5 years for patients with low-grade lymphoma who received IFN in randomized trials.44 Furthermore, the advantage was most marked in studies using anthracycline-containing induction chemotherapy. It is of interest that despite the methodology being very different (in that the present study used individualized updated patient data), the main conclusions are in fact the same.
Thus, the meta-analysis results support the use of IFN-2 as part of treatment for follicular lymphoma, at least when given in the context of some of the chemotherapy regimens that have been in use during the last 20 years. The difficulty with applying this conclusion now is the clinical toxicity of the treatment that is not negligible. A further problem is the fact that the studies from which the data are derived were conducted at a time when standard initial treatment for follicular lymphoma in the United Kingdom was chlorambucil, in Europe, CVP, and in the United States, cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP). With the development of alternative treatment modalities such as fludarabine-based regimens, anti-CD20–containing treatments, radioimmunotherapy, and high-dose treatment, it is difficult to know how best to incorporate this information into the algorithm of therapy. With the advent of pegylated IFN-2, which is associated with fewer side effects, it is possible to envisage the use of a better tolerated preparation as part of the control arm in future phase III studies, against which experimental options could be evaluated. Finally, the heterogeneity of the findings draws attention to the potential importance of variations in dose and dose-intensity of both chemotherapy and of biologic agents when these are combined in clinical trials.
Authors' Disclosures of Potential Conflicts of Interest
The following authors or their immediate family members have indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. Research funding: A.Z.S. Rohatiner, Schering-Plough; P. Solal-Celigny, Schering-Plough. For a detailed description of this category, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and Disclosures of Potential Conflicts of Interest found in Information for Contributors in the front of each issue.
Acknowledgment
We are very grateful to all the data managers at individual hospitals and of the cooperative groups who contributed data, and thank Margaret Cresswell for preparing the manuscript. Andrew Lister and Ama Rohatiner are supported by Cancer Research UK.
The authors are affiliated with the following organizations: Cancer and Leukemia Group B (CALGB), B. Peterson; Eastern Cooperative Oncology Group (ECOG), B. Peterson, E. Borden; Groupe d’Etudes des Lymphomes de l’Adulte (GELA), P. Solal-Celigny; European Organisation for Research and Treatment of Cancer (EORTC), A. Hagenbeek; Southwest Oncology Group (SWOG), R.I. Fisher; German Low Grade Lymphoma Study Group (GLSG), M. Unterhalt; LNH-PRO, R. Arranz.
NOTES
Supported in part by a grant from Schering-Plough.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Gresser I, Brouty-Boye D, Thomas M-T, et al: Interferon and cell division, I: Inhibition of the multiplication of mouse leukaemia L1210 cells in vitro by an interferon preparation. Proc Natl Acad Sci U S A 66:1052-1058, 1970
Gresser I, Maury C, Tovey MG: Interferon and murine leukaemia VII: Therapeutic effect of interferon preparations after diagnosis of lymphoma in AKR mice. Int J Cancer 17:647-651, 1976
Strander H, Adamson U, Aparisi T, et al: Adjuvant interferon treatment of human osteosarcoma. Recent Results Cancer Res 68:40-44, 1978
Gutterman JU, Blumenschein GR, Alexanian R: Leukocyte interferon-induced tumour regression in human metastatic breast cancer, multiple myeloma and malignant lymphoma. Ann Intern Med 93:399-406, 1980
Horning SJ, Merigan TC, Krown SE: Human interferon alpha in malignant lymphoma and Hodgkin's disease. Cancer 56:1305-1310, 1985
Louie AC, Gallagher JC, Sikora K, et al: Follow up observations on the effect of human leukocyte interferon in non-Hodgkin's lymphoma. Blood 58:712-718, 1981
Quesada JR, Hawkins M, Horning S, et al: Collaborative Phase I-II study of recombinant DNA-produced leukocyte Interferon (Clone A) in metastatic breast cancer, malignant lymphoma and multiple myeloma. Am J Med 77:427-432, 1984
O'Connell MJ, Colgan JP, Oken MM, et al: Clinical trial of recombinant leukocyte A interferon as initial therapy for favourable histology non-Hodgkin's lymphomas and chronic lymphocytic leukaemia. J Clin Oncol 4:128-136, 1986
Wagstaff J, Loynds P, Crowther D: A phase II study of human recombinant DNA-a2 interferon in patients with low-grade non-Hodgkin's lymphoma. Cancer Chemother Pharmacol 18:54-58, 1986
Leavitt J, Ratanathathorn V, Ozer H, et al: Alfa-2b Interferon in the treatment of Hodgkin's disease and non-Hodgkin's lymphoma. Semin Oncol 14:18-23, 1987 (suppl 2)
Foon KA, Sherwin SA, Abrams PG, et al: Treatment of advanced non-Hodgkin's lymphoma with recombinant leukocyte A interferon. N Engl J Med 311:1148-1152, 1984
Chirigos MA, Pearson JW: Cure of murine leukaemia with drugs and interferon treatment. J Natl Cancer Inst 51:1367-1368, 1973
Gresser I, Maury C, Tovey M: Efficacy of combined interferon cyclophosphamide therapy after diagnosis of lymphoma in AKR mice. Eur J Cancer 14:97-99, 1978
Balkwill FR, Moodie EM: Positive interactions between human interferon and cyclophosphamide or adriamycin in a human tumor model system. Cancer Res 44:904-908, 1984
Rohatiner AZS, Richards MA, Barnett MJ, et al: Chlorambucil and interferon for low grade non-Hodgkin's lymphoma. Br J Cancer 55:225-226, 1987
Chisesi T, Capnist G, Vespignani M, et al: Interferon alfa-2b and chlorambucil in the treatment of non-Hodgkin's lymphoma. J Investig New Drugs 5:35-40, 1987
Chisesi T, Congiu M, Contu A, et al: Randomized study of Chlorambucil (CB) compared to Interferon (Alfa-2B) combined with CB in low-grade non-Hodgkin's lymphoma: An interim report of a randomized study—Non-Hodgkin's Lymphoma Co-Operative Study Group. Eur J Cancer 27:S31-S33, 1991 (suppl 4)
Peterson BA, Petroni GR, Oken MM, et al: Cyclophosphamide versus cyclophosphamide plus interferon alfa-2b in follicular low-grade lymphomas: An intergroup phase III trial (CALGB 8691 and EST 7486). Proc Am Soc Clin Oncol 16:14a, 1997 (abstr 48)
Solal-Celigny P, Lepage E, Brousse N, et al: Recombinant interferon alfa-2b combined with a regimen containing doxorubicin in patients with advanced follicular lymphoma. N Engl J Med 329:1608-1614, 1993
Solal-Celigny P, Leage E, Brousse N, et al: Doxorubicin-containing regimen with or without interferon alfa-2b for advanced follicular lymphomas: Final analysis of survival and toxicity in the Groupe d'Etude des Lymphomes Folliculaires 86 Trial. J Clin Oncol 16:2332-2338, 1998
Smalley RV, Andersen JW, Hawkins MJ, et al: Interferon alpha combined with cytotoxic chemotherapy for patients with non-Hodgkin's lymphoma. N Engl J Med 327:1336-1341, 1992
Andersen JW, Smalley RV: Interferon Alfa plus chemotherapy for non-Hodgkin's lymphoma five-year follow-up. N Engl J Med 329:1821-1822, 1993
Smalley RV, Weller E, Hawkins MJ, et al: Final analysis of the ECOG I-COPA trial (E6484) in patients with non-Hodgkin's lymphomas treated with interferon alfa (IFN-alpha2a) plus an anthracycline-based induction regimen. Leukemia 15:1118-1122, 2001
Arranz R, Garcia-Alfonso P, Sobrino P, et al: Role of interferon alfa-2b in the induction and maintenance treatment of low-grade non-Hodgkin's lymphoma: Results from a prospective, multicenter trial with double randomization. J Clin Oncol 16:1538-1546, 1998
Hagenbeek A, Carde P, Meerwaldt JH, et al: Maintenance of remission with human recombitant interferon alfa-2a in patients with stages III and IV low-grade malignant non-Hodgkin's lymphoma: European Organisation for Research and Treatment of Cancer, Lymphoma Co-operative Group. J Clin Oncol 16:41-47, 1998
Unterhalt M, Hermann R, Koch P, et al: Long term interferon alpha Maintenance prolongs remission duration in advanced low grade lymphomas and is related to the efficacy of initial cytoreductive chemotherapy. Blood 88, 1996 (suppl 1, abstr 1801)
Aviles A, Duque G, Talavera A, et al: Interferon alpha 2b as maintenance therapy in low grade malignant lymphoma improves duration of remission and survival. Leuk Lymphoma 20:495-499, 1996
Fisher RI, Dana BW, LeBlanc M, et al: Interferon alpha consolidation after intensive chemotherapy does not prolong the progression-free survival of patients with low-grade non-Hodgkin's lymphoma: Results of the Southwest Oncology Group randomized phase III study 8809. J Clin Oncol 18:2010-2016, 2000
Rohatiner AZ, Radford J, Deakin D, et al: A randomised controlled trial to evaluate the role of interferon as initial and Maintenance therapy in patients with follicular lymphoma. Br J Cancer 85:29, 2001
Rohatiner AZ, Gregory W, Peterson B, et al: A meta-analysis (MA) of randomised trials evaluating the role of interferon (IFN) as treatment for follicular lymphoma. Proc Am Soc Clin Oncol 17:4a, 1998 (abstr 11)
Peto R: Why do we need systematic overviews of randomised trials Stat Med 6:233-240, 1987
Simon RJ: Confidence intervals for reporting results of clinical trials. Ann Intern Med 105:429-435, 1986
Gregory WM: Adjusting survival curves for imbalances in prognostic factors. Br J Cancer 58:202-204, 1988
Steward WP, Crowther D, McWilliam LJ, et al: Maintenance chlorambucil after CVP in the management of advanced stage, low-grade histologic type non-Hodgkin's lymphoma: A randomized prospective study with an assessment of prognostic factors. Cancer 61:441-447, 1988
Bagley CM, De Vita VT, Berrard CW, et al: Advanced lymphosarcoma: Intensive cyclical combination chemotherapy with cyclophosphamide, vincristine and prednisolone. Ann Intern Med 76:227-234, 1972
Portlock CS, Rosenberg SA, Glatstein E, et al: Treatment of advanced non-Hodgkin's lymphomas with favourable histology: Preliminary results of a prospective trial. Blood 47:747-756, 1976
Kennedy BJ, Bloomfield CD, Kiang DT, et al: Combination versus successive single agent chemotherapy in lymphocytic lymphoma. Cancer 41:23-28, 1978
Lister TA, Cullen MH, Beard MEJ, et al: Comparison of combined and single agent chemotherapy in non-Hodgkin's lymphoma of favourable histological subtype. Br Med J 1:533-537, 1978
Ezdinli E, Anderson J, Melvin F, et al: Moderate versus aggressive chemotherapy of nodular poorly differentiated lymphoma. J Clin Oncol 3:769-775, 1985
Longo DL, Young RC, Hubbard SM, et al: Prolonged initial remission in patients with nodular mixed lymphomas. Ann Intern Med 100:651-656, 1984
Anderson T, Bender RA, Fisher RI, et al: Combination chemotherapy in non-Hodgkin's lymphoma: Results of long-term follow up. Cancer Treat Rep 61:1057-1066, 1977
Bitran JC, Golomb HM, Ultmann JE, et al: Non-Hodgkin's lymphoma, poorly differentiated and mixed cell types: Results of sequential staging procedures, response to therapy and survey of 100 patients. Cancer 42:88-95, 1978
Glick JH, Barnes JM, Ezdinli EZ, et al: Nodular mixed lymphoma: Results of a randomised trial failing to confirm prolonged disease-free survival with COPP chemotherapy. Blood 58:920-925, 1981
Allen IE, Ross SD, Borden SP, et al: Meta-analysis to assess the efficacy of interferon-alpha in patients with follicular non-Hodgkin's lymphoma. J Immunother 24:58-65, 2001(A.Z.S. Rohatiner, W.M. Gr)
University of Minnesota, Minneapolis, MN
Taussig Cancer Center, Cleveland, OH
Centre Jean Bernard, Le Mans, France
University Medical Center, Utrecht, The Netherlands
Wilmot Cancer Center, University of Rochester, Rochester, NY
Georg-August-University, Gottingen, Germany
Hospital Universitario de la Princesa, Madrid, Spain
Ospedale Civile, Venice, Italy
Oncology Hospital, IMSS, Mexico City, Mexico
ABSTRACT
PURPOSE: To determine whether interferon (IFN) -2, when given with or following chemotherapy, influences response rate, remission duration, and survival in newly diagnosed patients with follicular lymphoma.
PATIENTS AND METHODS: Ten phase III studies evaluating the role of IFN-2 in 1,922 newly diagnosed patients with follicular lymphoma were analyzed. Updated individual patient data were used to perform meta-analyses for response, survival, and remission duration.
RESULTS: The addition of IFN-2 to initial chemotherapy did not significantly influence response rate. An overall meta-analysis for survival showed a significant difference in favor of IFN-2, but also showed significant heterogeneity between studies. Further analyses were carried out in order to explain this heterogeneity, and to define the circumstances in which IFN-2 prolonged survival. The survival advantage was seen when IFN-2 was given: (1) in conjunction with relatively intensive initial chemotherapy (2P = .00005), (2) at a dose 5 million units (2P = .000002), (3) at a cumulative dose 36 million units per month (2P = .000008), and (4) with chemotherapy rather than as maintenance therapy (P = .004). With regard to remission duration, there was also a significant difference in favor of IFN-2, irrespective of the intensity of chemotherapy used, IFN dose, or whether IFN was given as a maintenance strategy or with chemotherapy.
CONCLUSION: When given in the context of relatively intensive initial chemotherapy, and at a dose 5 million units ( 36 x 106 units per month), IFN-2 prolongs survival and remission duration in patients with follicular lymphoma.
INTRODUCTION
Interferon (IFN) was introduced into the treatment of follicular lymphoma following observation of activity in L1210 leukemia1 and AKR lymphoma,2 but with limited understanding of its mode of action. Phase II studies (using an empirical dose and schedule derived from the original Scandinavian osteogenic sarcoma trial3) in low-grade lymphoma, resulted in response rates of 30% to 50%, regardless of the source of IFN-2.4-11 Such responses were, however, almost always incomplete and, in general, short lived. Activity having been demonstrated, the concept of synergy between IFN and conventional cytotoxic drugs was tested and confirmed in murine models of leukemia12 and lymphoma,13 in human breast cancer xenografts in nude mice,14 and in patients with follicular lymphoma known to be refractory to chemotherapy.15,16
A series of randomized studies followed, in newly diagnosed patients with a spectrum of histologic subtypes, evaluating the use of IFN-2 given in conjunction with chemotherapy.17-29 In some studies, IFN-2 was given concurrently with chemotherapy, in others, it was used as "maintenance" therapy, while in still others, it was used throughout. The chemotherapy regimens varied in intensity, from relatively low doses of alkylating agents (chlorambucil or cyclophosphamide), to doxorubicin or mitoxantrone-containing regimens. The dose and schedule of IFN-2 also varied, though more than half of the studies embraced the standard, albeit empirical, dose of 3 x 106 units, given three times weekly, for 6 months or 1 or 2 years (or in the case of one study,26 until recurrence). Further variability was introduced by the fact that some studies permitted (or advocated) the use of radiotherapy on completion of chemotherapy to sites of bulky disease at presentation, or to residual disease.25,27,28
In addition to the variation in histologic subtype engendered by the inclusion of patients with any type of low-grade lymphoma (though the majority of patients in fact had follicular lymphoma), there were also differences in eligibility criteria. Some studies were very permissive, allowing enrollment of not only patients who had a specific indication for treatment, but also of those who did not. Conversely, others, for example the GELA (Groupe d'Etudes des Lymphomes de l’Adulte) study,19,20 focused on patients deemed to have an adverse prognosis (by GELA criteria), while one study included only patients in whom a complete remission (CR) had been achieved with prior chemotherapy.27
As a consequence of this heterogeneity (in both patient population and entry criteria), there are major discrepancies in the results, even when only studies with a similar design are considered. Therefore, for this meta-analysis, only patients with follicular lymphoma were included. The aim was to determine whether IFN-2 given at any point improves response rate, remission duration, and survival.
PATIENTS AND METHODS
Studies
At the inception of the project, review of the published literature (and of completed analyses presented in abstract form) led to identification of eight phase III studies of newly diagnosed patients with "low-grade" or, specifically, follicular lymphoma. A protocol was written and distributed to individual authors and groups, who were then asked to provide updated (ie, updated from the original publication or analysis for that study) individual patient information, which was transmitted electronically to us (W.G.). Data were checked and only patients described as having follicular lymphoma were selected for inclusion in the analysis. Central review of pathology was not undertaken. On this basis, an initial meta-analysis was conducted and presented.30 Data from two studies that had not been published at that time were subsequently made available.24,28 The present meta-analysis, therefore, comprises the results for patients included in ten phase III studies.
Patients
The analysis relates to 1,922 patients whose clinical characteristics are listed in Table 1. The majority of patients had stage IV disease by virtue of bone marrow involvement.
Treatment
Treatment in terms of chemotherapy regimen and the dose, schedule, and timing of IFN-2 is summarized in Table 2, by Study Group (or country of origin when a formal group was not involved). In six of the studies, IFN-2 was given as part of initial therapy17-24,29; in four studies, it was given only as maintenance therapy25-28; and in four studies, patients could be randomly assigned to receive it throughout.17,18,24,29
Chemotherapy. Chemotherapy regimens were categorized by the intensity of chemotherapy as follows: studies utilizing relatively "less intensive" chemotherapy are defined as those that used chlorambucil, cyclophosphamide, or cyclophosphamide/vincristine/prednisolone (CVP) as initial therapy (Table 2). Those studies that used an anthracycline or mitoxantrone-based combination regimen were deemed "more intensive" (Table 2).
Interferon. The dose of IFN-2 was similarly categorized by individual dose and cumulative dose administered over a period of 4 weeks. Individual doses were grouped by those doses equal to or less than 5 x 106 (ie, 2 x 106 units/m2 or 3 x 106 units per dose). The dose per month follows closely; ie, the latter doses, given 3x weekly, translate into a monthly dose of 36 x 106 units. Hence, for the meta-analysis, the studies were divided into those studies using more than 36 x 106 units per month or that amount (Table 2).
Meta-Analyses: Methodology
Response data (CR/partial response v the rest), survival times, and durations of remission were calculated for each individual patient, and then the overview methodology described in detail by Peto31 was applied. For the response meta-analysis, data from the six studies that evaluated the effect of adding IFN-2 to initial chemotherapy were used.17-24,29 For the survival and remission duration meta-analyses, data from all ten studies were included.
In brief, for response, the expected number of responses was calculated for each trial arm, assuming no difference in response rate between the two arms. This was then compared with the observed number of responses, and the differences (observed-to-expected ratio [O-E]) calculated for each trial. For survival and remission duration, the expected number of events (deaths or recurrences) in each arm, assuming no difference between the treatments, was derived for each trial. This was then compared with the observed number of events and the differences (O-Es) calculated. For response, survival, and remission duration, these O-E figures were then summed over the whole set of trials to form a grand total. The variance of the individual estimates was also derived and summed, giving a variance and a standard deviation (SD) for the grand total. The number of SDs (z) by which the grand total differs from zero gives an estimate of the significance of any effect. The O-Es were also tested for heterogeneity to see whether the scatter of results was unexpected. (The sum of [O-E]2/variance should be distributed as 2n – 1 if the scatter is random, where n is the number of studies.)
For response, survival, and remission duration, the size of any treatment effect, specifically the typical odds ratio (TOR), was simply estimated as exp(z/SD), with approximate 95% CI expressed as exp (z/SD ± 1.96/SD). CIs for the individual trials were calculated in the same fashion. The TORs and their CIs for each trial, together with the grand total, are shown in Figures 1-7. The size of the solid squares is proportional to the amount of information each trial contains. Because many individual trials are displayed, 99% CIs are shown for the individual trials. For the overall result, 95% CIs are given (represented by the open diamonds in the figures). These odds ratios were converted into 5- and 10-year differences in survival and remission duration as follows: the odds ratio, Or, is related to the hazard ratio, , by the equation = 1/(1 – [Or/100]). If the survival proportions at any time, t, are P1 and P2, then = loge(P1)/loge(P2), so, if the overall proportion surviving at time t is P, then P = (n1P1+n2P2)/(n1+n2) where n1 and n2 are the sample sizes at time t (which can be calculated as described by Simon32). It follows that P1 +(n2/n1)P1 – P(n1+n2)/n1 = 0, and this can be easily solved numerically to give P1, P2 and the difference P1 – P2. For response, the calculations are identical using the proportions responding instead of the survival proportions.
Four of the 10 studies had a double randomization, as listed in Table 2. Clearly, for each of these studies, only a proportion of randomly assigned patients initially reached the second randomization (for a variety of reasons including toxicity, patient refusal, and death). Those failing to reach the second randomization had a worse survival than those who did, because of this selection process. Furthermore, the numbers not reaching the second randomization were not necessarily the same as in the two initial randomization groups. For instance, in the UK study, 47 (45%) of 104 patients initially randomly assigned to receive IFN-2 failed to reach the point of second randomization, compared with 30 (30%) of 100 of those patients initially randomly assigned to the chlorambucil-only arm. Therefore, in these studies, to analyze just those patients who received "any IFN-2" versus the rest of the patients would bias the result, almost certainly in favor of the IFN-2 group. To avoid this problem and to make it possible to perform a meta-analysis of IFN-2 given at any point versus no IFN-2, such studies have been regarded as having two separate strata, ie, comprising patients who reached the second randomization and those who did not. (For the Non-Hodgin's Lymphoma Study Group [NHLSG] study, information on the second randomization was unavailable, and so for this study only the initial randomization was used in the meta-analysis graphs). The remaining three studies are therefore each represented twice on the meta-analysis graphs, reflecting the two different strata.
The survival of the IFN-2 and no IFN-2 groups was plotted using the method described by Gregory33 (see Fig 8). This enables the survival percentages to be calculated stratified by center without assuming proportional hazards, and produces a survival comparison that represents the meta-analysis result and an adjusted 2 statistic that matches the P value from the meta-analysis. To check the proportional hazards assumption and to evaluate possible changes in the hazard ratio over time, a smoothed (instantaneous) hazard was plotted over time (stratified by center and thus matching the meta-analysis findings) by calculating the hazard ratio for each death time as the hazard for the 100 deaths occurring both before and after that time. This limited the time scale of the graph somewhat because the first 100 deaths occurred during the first year, whereas the last 100 deaths occurred during years 7 to 11, when follow-up is relatively short. However, this methodology made it possible to gain an approximation of the pattern of changing hazard with time. The number 100 was chosen after experimentation to dampen out large fluctuations in the hazard that would obscure the pattern. It was then possible to test for any observed trend in the hazard by calculating hazards over adjacent intervals (eg, each year) and then calculating a weighted correlation coefficient (where the weights are the SEs of the hazard estimates) for these hazards with time. The meta-analysis methodology assumes proportional hazards (so the hazard ratio would be constant over time), thus these techniques make it possible to discern significant departure from this assumption.
Because significant heterogeneity was observed between studies (see Results), further analyses were carried out in an attempt to explain this heterogeneity. A number of hypotheses were generated. The following appeared the most likely to explain the observed differences; that IFN-2 might increase survival when given: (1) concurrently with, or following, relatively intensive chemotherapy; (2) at a dose equal to or greater than 5 x 106 units; (3) at a cumulative dose in excess of 36 x 106 units over a period of 4 weeks; or (4) with initial chemotherapy rather than as maintenance therapy.
RESULTS
Response
The addition of IFN-2 to initial chemotherapy did not significantly influence response rate overall (2P = .41). The estimated difference in percentage response rate was –2% (ie, in favor of the no IFN group; 95% CI, –6% to 3%; Fig 1), though there was considerable heterogeneity between studies (2P = .005), and a highly significant benefit from IFN-2 was observed in the GELA study (2P = .006).
Survival
Considering all eligible patients, the use of IFN-2 at any point in treatment appeared to prolong survival (2P = .004; Fig 2). The corresponding survival curve is shown in Figure 8, and the estimated 5- and 10-year survival differences (see Meta-Analysis: Methodology) were 5.5% (95% CI, 2% to 9%) and 8% (95% CI, 3% to 13%), respectively. However, there was significant heterogeneity between studies (2P = .003). Furthermore, the Mexican study's result is clearly an outlier, because there is a much greater survival effect in this trial than in any of the other studies. When the data from the Mexican study were excluded, the results were still significant (2P =.03), with 5- and 10-year survival differences of 4% (95% CI, 0.4% to 8%) and 6% (95% CI, 0.6% to 11%), respectively, and the heterogeneity effect was of borderline significance (2P = .07). However, there are no a priori grounds for excluding the Mexican study. The main difference between the latter and the other studies is the fact that only patients in CR were eligible for randomization. Their survival overall may thus be superior to that of patients in the other studies which included those patients in whom a partial response had been achieved. For patients failing to respond, the use of IFN-2 at any point in treatment did not prolong survival (Z = –0.81; 2P = .42; data not shown), with any trend being in favor of the no IFN group.
Further inspection of the data raised the four hypotheses listed in the Patients and Methods section; these hypotheses were investigated further. Analyzing each of the four hypotheses separately, for patients receiving relatively intensive initial therapy, IFN-2 used at any point was associated with prolongation of survival (2P = .00005; Fig 3; 5-and 10-year survival differences being 11% [95% CI, 6% to 16%] and 15% [95% CI, 8% to 23%], respectively. In contrast, for patients receiving less intensive initial therapy (ie, alkylating agents or CVP), there was no significant improvement in survival (2P = .93; Fig 4).
Considering the individual dose of IFN-2, for patients receiving 5 x 106 units or more per dose, there was a highly significant increase in survival (2P = .000002; Fig 5; 5- and 10-year survival differences being 15% [95% CI, 9% to 21%] and 20% [95% CI, 12% to 28%], respectively). Conversely, for patients receiving less than 5 x 106 units per dose, there was no effect on survival (2P = .75; data not shown).
Similarly, in terms of the monthly dose of IFN-2 administered, for patients receiving more than 36 x 106 units over 4 weeks, there was a significant improvement in survival (2P = .000008; Fig 6; 5- and 10-year survival differences being 18% [95% CI, 10% to 25%] and 23% [95% CI, 14% to 33%], respectively), compared with no significant effect when IFN-2 was used at the typical dose of 3 x 106 units 3x weekly, (equivalent to 36 x 106 units per 4 weeks; 2P = .43; data not shown).
For studies in which IFN-2 was given initially together with chemotherapy, there was a significant improvement in survival (2P = .004; data not shown; 5- and 10-year survival differences being 7% [95% CI, 2% to 12%] and 9% [95% CI, 3% to 15%], respectively), compared with no significant effect in studies in which it was given only as maintenance (2P = .20; data not shown). In the latter group, the only study that demonstrated an effect was the Mexican study, which was, again, an outlier. Excluding this study reduced the significance to 2P = .98.
It should be noted that none of the latter four results are significantly altered by the exclusion of the data from the Mexican study. For instance, for patients receiving 5 x 106 units or more per dose, there was still a highly significant increase in survival (2P = .0002; data not shown). The odds reduction was 37% ± 10% (hardly different than with the inclusion of the Mexican study's data), and 5- and 10-year survival differences were also largely unaffected, being 13% (95% CI, 6% to 20%) and 17% (95% CI, 8% to 25%), respectively. The other three comparisons show similar small differences from the equivalent analyses that included the Mexican data (data not shown). This is because, though the Mexican study shows a relatively large survival difference, it is a relatively small study, and the significance and magnitude of the meta-analysis differences are influenced more by the studies with larger numbers of patients, such as those conducted by the Eastern Cooperative Oncology Group (ECOG) and GELA.
Plotting the hazards as they change over time, as described in Patients and Methods, showed a pattern of increasing hazard over time (Fig 9). There is a big rise in risk of death from not having received IFN during the 3- to 4-year period after entry, and though the risk reduces slightly thereafter, it is maintained at least up to 7 years from entry, and probably beyond. Although there may be some random element to this pattern, it is very clear that the risk increases over time from entry. It therefore appears that IFN produces an effect that is maintained, or even increases, over time. This was confirmed by a weighted correlation of adjacent yearly hazards that gave a correlation coefficient of r = 0.60 (2P = .04). Thus the 10-year survival estimates are probably slight underestimates of the true difference (by approximately 25%, so the overall reported 10-year survival difference of 8% is probably closer to 10%), and this is confirmed by inspection of the adjusted survival graph (Fig 8) and survival estimates taken from it, which show this slightly larger difference (the adjusted survival curve methodology does not assume proportional hazards33 and can therefore produce these estimates). The 5-year survival estimates appear to be much more accurate, being in the middle of the survival range for this analysis. It should be noted that the departure from proportional hazards is not sufficient to invalidate the findings of the study; its main effect is to create a slight underestimate of the 10-year survival figures.
To confirm this finding, the meta-analyses were repeated using only data from the patients who had already survived 3 years (because the hazard is close to zero for the first 3 years, which is different from its later values), and the basic findings were unaltered; if anything the results tended to be more significant, and with less heterogeneity between studies.
Remission Duration
Overall, the addition of IFN-2 at any point significantly prolonged remission duration when all ten studies were considered (2P < .000001; Fig 7; 5- and 10-year remission duration differences being 11% [95% CI, 7% to 15%] and 10% [95% CI, 6% to 13%], respectively); the corresponding remission duration curve is shown in Figure 10. Evaluating the effect of maintenance IFN-2 alone (ie, considering only the four studies in which it was not used as part of initial therapy), there was still a highly significant difference (2P = .0002; 5- and 10-year remission duration differences being 13% [95% CI, 6% to 19%] and 12% [95% CI, 6% to 19%], respectively), in contrast to the survival comparisons.
DISCUSSION
These results demonstrate that the addition of IFN-2 to conventional chemotherapy as part of initial therapy for newly diagnosed patients with follicular lymphoma does not result in any improvement in response rate. However, the meta-analysis confirms that IFN-2 does prolong remission duration and overall survival, and also prolongs survival and remission duration under specific circumstances that relate to the intensity of the chemotherapy used, the dose of IFN-2, or both. The point at which IFN-2 is given, ie, as part of initial therapy rather than as maintenance therapy, also appears to influence outcome. There was no survival difference following only maintenance IFN-2 therapy (except in the Mexican study).
To summarize, overall, there was an effect of IFN-2 on survival; it seems likely that this relates to some or all of the following factors: intensity of the initial chemotherapy, dose or dose-intensity of IFN-2, and whether the IFN-2 was given as part of initial therapy or as maintenance therapy. However, it is difficult (if not impossible) to determine which of these hypotheses is correct. The statistical differences between them are marginal. For each hypothesis there is at least one counter-example, or one study, that is the exception. It thus seems probable that more than one factor is at work; for instance, that IFN-2 is effective at higher doses, and in studies in which the chemotherapy is more intensive, though this theory is impossible to prove from these data.
The difference in survival (Figs 8 and 9) increases over time. The survival effect appears to be a late effect, and the difference between the curves may therefore widen further with longer follow-up. This is in keeping with the difference in remission duration, which shows a consistent effect from about 1 year onwards, with a difference of at least 10% at 10 years. This might be expected to translate eventually into a similar size of survival difference, as is becoming apparent in the survival curve.
It is important to recognize that the data on which the meta-analysis is based reflect protocol dose and not actual dose administered. Originally, it had been the intention to analyze the results using the dose given (of both chemotherapy and of IFN-2), but not all of the data sets included this information. Thus it was not possible to do this, because individual patients' medical records could not be reviewed owing to the logistical difficulties of doing so for ten studies worldwide.
The prolongation in survival in the context of relatively more intensive chemotherapy is unlikely to be simply a reflection of an increase in the intensity of the chemotherapy. Intensification of treatment in follicular lymphoma has increased remission duration34 but not survival.35-43
An earlier meta-analysis based on data from 8 studies (as published in manuscript or abstract form) also showed significantly better 5-year survival and longer progression-free survival at 3 and 5 years for patients with low-grade lymphoma who received IFN in randomized trials.44 Furthermore, the advantage was most marked in studies using anthracycline-containing induction chemotherapy. It is of interest that despite the methodology being very different (in that the present study used individualized updated patient data), the main conclusions are in fact the same.
Thus, the meta-analysis results support the use of IFN-2 as part of treatment for follicular lymphoma, at least when given in the context of some of the chemotherapy regimens that have been in use during the last 20 years. The difficulty with applying this conclusion now is the clinical toxicity of the treatment that is not negligible. A further problem is the fact that the studies from which the data are derived were conducted at a time when standard initial treatment for follicular lymphoma in the United Kingdom was chlorambucil, in Europe, CVP, and in the United States, cyclophosphamide, doxorubicin, vincristine, and prednisone (CHOP). With the development of alternative treatment modalities such as fludarabine-based regimens, anti-CD20–containing treatments, radioimmunotherapy, and high-dose treatment, it is difficult to know how best to incorporate this information into the algorithm of therapy. With the advent of pegylated IFN-2, which is associated with fewer side effects, it is possible to envisage the use of a better tolerated preparation as part of the control arm in future phase III studies, against which experimental options could be evaluated. Finally, the heterogeneity of the findings draws attention to the potential importance of variations in dose and dose-intensity of both chemotherapy and of biologic agents when these are combined in clinical trials.
Authors' Disclosures of Potential Conflicts of Interest
The following authors or their immediate family members have indicated a financial interest. No conflict exists for drugs or devices used in a study if they are not being evaluated as part of the investigation. Research funding: A.Z.S. Rohatiner, Schering-Plough; P. Solal-Celigny, Schering-Plough. For a detailed description of this category, or for more information about ASCO's conflict of interest policy, please refer to the Author Disclosure Declaration and Disclosures of Potential Conflicts of Interest found in Information for Contributors in the front of each issue.
Acknowledgment
We are very grateful to all the data managers at individual hospitals and of the cooperative groups who contributed data, and thank Margaret Cresswell for preparing the manuscript. Andrew Lister and Ama Rohatiner are supported by Cancer Research UK.
The authors are affiliated with the following organizations: Cancer and Leukemia Group B (CALGB), B. Peterson; Eastern Cooperative Oncology Group (ECOG), B. Peterson, E. Borden; Groupe d’Etudes des Lymphomes de l’Adulte (GELA), P. Solal-Celigny; European Organisation for Research and Treatment of Cancer (EORTC), A. Hagenbeek; Southwest Oncology Group (SWOG), R.I. Fisher; German Low Grade Lymphoma Study Group (GLSG), M. Unterhalt; LNH-PRO, R. Arranz.
NOTES
Supported in part by a grant from Schering-Plough.
Authors' disclosures of potential conflicts of interest are found at the end of this article.
REFERENCES
Gresser I, Brouty-Boye D, Thomas M-T, et al: Interferon and cell division, I: Inhibition of the multiplication of mouse leukaemia L1210 cells in vitro by an interferon preparation. Proc Natl Acad Sci U S A 66:1052-1058, 1970
Gresser I, Maury C, Tovey MG: Interferon and murine leukaemia VII: Therapeutic effect of interferon preparations after diagnosis of lymphoma in AKR mice. Int J Cancer 17:647-651, 1976
Strander H, Adamson U, Aparisi T, et al: Adjuvant interferon treatment of human osteosarcoma. Recent Results Cancer Res 68:40-44, 1978
Gutterman JU, Blumenschein GR, Alexanian R: Leukocyte interferon-induced tumour regression in human metastatic breast cancer, multiple myeloma and malignant lymphoma. Ann Intern Med 93:399-406, 1980
Horning SJ, Merigan TC, Krown SE: Human interferon alpha in malignant lymphoma and Hodgkin's disease. Cancer 56:1305-1310, 1985
Louie AC, Gallagher JC, Sikora K, et al: Follow up observations on the effect of human leukocyte interferon in non-Hodgkin's lymphoma. Blood 58:712-718, 1981
Quesada JR, Hawkins M, Horning S, et al: Collaborative Phase I-II study of recombinant DNA-produced leukocyte Interferon (Clone A) in metastatic breast cancer, malignant lymphoma and multiple myeloma. Am J Med 77:427-432, 1984
O'Connell MJ, Colgan JP, Oken MM, et al: Clinical trial of recombinant leukocyte A interferon as initial therapy for favourable histology non-Hodgkin's lymphomas and chronic lymphocytic leukaemia. J Clin Oncol 4:128-136, 1986
Wagstaff J, Loynds P, Crowther D: A phase II study of human recombinant DNA-a2 interferon in patients with low-grade non-Hodgkin's lymphoma. Cancer Chemother Pharmacol 18:54-58, 1986
Leavitt J, Ratanathathorn V, Ozer H, et al: Alfa-2b Interferon in the treatment of Hodgkin's disease and non-Hodgkin's lymphoma. Semin Oncol 14:18-23, 1987 (suppl 2)
Foon KA, Sherwin SA, Abrams PG, et al: Treatment of advanced non-Hodgkin's lymphoma with recombinant leukocyte A interferon. N Engl J Med 311:1148-1152, 1984
Chirigos MA, Pearson JW: Cure of murine leukaemia with drugs and interferon treatment. J Natl Cancer Inst 51:1367-1368, 1973
Gresser I, Maury C, Tovey M: Efficacy of combined interferon cyclophosphamide therapy after diagnosis of lymphoma in AKR mice. Eur J Cancer 14:97-99, 1978
Balkwill FR, Moodie EM: Positive interactions between human interferon and cyclophosphamide or adriamycin in a human tumor model system. Cancer Res 44:904-908, 1984
Rohatiner AZS, Richards MA, Barnett MJ, et al: Chlorambucil and interferon for low grade non-Hodgkin's lymphoma. Br J Cancer 55:225-226, 1987
Chisesi T, Capnist G, Vespignani M, et al: Interferon alfa-2b and chlorambucil in the treatment of non-Hodgkin's lymphoma. J Investig New Drugs 5:35-40, 1987
Chisesi T, Congiu M, Contu A, et al: Randomized study of Chlorambucil (CB) compared to Interferon (Alfa-2B) combined with CB in low-grade non-Hodgkin's lymphoma: An interim report of a randomized study—Non-Hodgkin's Lymphoma Co-Operative Study Group. Eur J Cancer 27:S31-S33, 1991 (suppl 4)
Peterson BA, Petroni GR, Oken MM, et al: Cyclophosphamide versus cyclophosphamide plus interferon alfa-2b in follicular low-grade lymphomas: An intergroup phase III trial (CALGB 8691 and EST 7486). Proc Am Soc Clin Oncol 16:14a, 1997 (abstr 48)
Solal-Celigny P, Lepage E, Brousse N, et al: Recombinant interferon alfa-2b combined with a regimen containing doxorubicin in patients with advanced follicular lymphoma. N Engl J Med 329:1608-1614, 1993
Solal-Celigny P, Leage E, Brousse N, et al: Doxorubicin-containing regimen with or without interferon alfa-2b for advanced follicular lymphomas: Final analysis of survival and toxicity in the Groupe d'Etude des Lymphomes Folliculaires 86 Trial. J Clin Oncol 16:2332-2338, 1998
Smalley RV, Andersen JW, Hawkins MJ, et al: Interferon alpha combined with cytotoxic chemotherapy for patients with non-Hodgkin's lymphoma. N Engl J Med 327:1336-1341, 1992
Andersen JW, Smalley RV: Interferon Alfa plus chemotherapy for non-Hodgkin's lymphoma five-year follow-up. N Engl J Med 329:1821-1822, 1993
Smalley RV, Weller E, Hawkins MJ, et al: Final analysis of the ECOG I-COPA trial (E6484) in patients with non-Hodgkin's lymphomas treated with interferon alfa (IFN-alpha2a) plus an anthracycline-based induction regimen. Leukemia 15:1118-1122, 2001
Arranz R, Garcia-Alfonso P, Sobrino P, et al: Role of interferon alfa-2b in the induction and maintenance treatment of low-grade non-Hodgkin's lymphoma: Results from a prospective, multicenter trial with double randomization. J Clin Oncol 16:1538-1546, 1998
Hagenbeek A, Carde P, Meerwaldt JH, et al: Maintenance of remission with human recombitant interferon alfa-2a in patients with stages III and IV low-grade malignant non-Hodgkin's lymphoma: European Organisation for Research and Treatment of Cancer, Lymphoma Co-operative Group. J Clin Oncol 16:41-47, 1998
Unterhalt M, Hermann R, Koch P, et al: Long term interferon alpha Maintenance prolongs remission duration in advanced low grade lymphomas and is related to the efficacy of initial cytoreductive chemotherapy. Blood 88, 1996 (suppl 1, abstr 1801)
Aviles A, Duque G, Talavera A, et al: Interferon alpha 2b as maintenance therapy in low grade malignant lymphoma improves duration of remission and survival. Leuk Lymphoma 20:495-499, 1996
Fisher RI, Dana BW, LeBlanc M, et al: Interferon alpha consolidation after intensive chemotherapy does not prolong the progression-free survival of patients with low-grade non-Hodgkin's lymphoma: Results of the Southwest Oncology Group randomized phase III study 8809. J Clin Oncol 18:2010-2016, 2000
Rohatiner AZ, Radford J, Deakin D, et al: A randomised controlled trial to evaluate the role of interferon as initial and Maintenance therapy in patients with follicular lymphoma. Br J Cancer 85:29, 2001
Rohatiner AZ, Gregory W, Peterson B, et al: A meta-analysis (MA) of randomised trials evaluating the role of interferon (IFN) as treatment for follicular lymphoma. Proc Am Soc Clin Oncol 17:4a, 1998 (abstr 11)
Peto R: Why do we need systematic overviews of randomised trials Stat Med 6:233-240, 1987
Simon RJ: Confidence intervals for reporting results of clinical trials. Ann Intern Med 105:429-435, 1986
Gregory WM: Adjusting survival curves for imbalances in prognostic factors. Br J Cancer 58:202-204, 1988
Steward WP, Crowther D, McWilliam LJ, et al: Maintenance chlorambucil after CVP in the management of advanced stage, low-grade histologic type non-Hodgkin's lymphoma: A randomized prospective study with an assessment of prognostic factors. Cancer 61:441-447, 1988
Bagley CM, De Vita VT, Berrard CW, et al: Advanced lymphosarcoma: Intensive cyclical combination chemotherapy with cyclophosphamide, vincristine and prednisolone. Ann Intern Med 76:227-234, 1972
Portlock CS, Rosenberg SA, Glatstein E, et al: Treatment of advanced non-Hodgkin's lymphomas with favourable histology: Preliminary results of a prospective trial. Blood 47:747-756, 1976
Kennedy BJ, Bloomfield CD, Kiang DT, et al: Combination versus successive single agent chemotherapy in lymphocytic lymphoma. Cancer 41:23-28, 1978
Lister TA, Cullen MH, Beard MEJ, et al: Comparison of combined and single agent chemotherapy in non-Hodgkin's lymphoma of favourable histological subtype. Br Med J 1:533-537, 1978
Ezdinli E, Anderson J, Melvin F, et al: Moderate versus aggressive chemotherapy of nodular poorly differentiated lymphoma. J Clin Oncol 3:769-775, 1985
Longo DL, Young RC, Hubbard SM, et al: Prolonged initial remission in patients with nodular mixed lymphomas. Ann Intern Med 100:651-656, 1984
Anderson T, Bender RA, Fisher RI, et al: Combination chemotherapy in non-Hodgkin's lymphoma: Results of long-term follow up. Cancer Treat Rep 61:1057-1066, 1977
Bitran JC, Golomb HM, Ultmann JE, et al: Non-Hodgkin's lymphoma, poorly differentiated and mixed cell types: Results of sequential staging procedures, response to therapy and survey of 100 patients. Cancer 42:88-95, 1978
Glick JH, Barnes JM, Ezdinli EZ, et al: Nodular mixed lymphoma: Results of a randomised trial failing to confirm prolonged disease-free survival with COPP chemotherapy. Blood 58:920-925, 1981
Allen IE, Ross SD, Borden SP, et al: Meta-analysis to assess the efficacy of interferon-alpha in patients with follicular non-Hodgkin's lymphoma. J Immunother 24:58-65, 2001(A.Z.S. Rohatiner, W.M. Gr)