SMALL-CELL CARCINOMA of the lung contributes

Transcription

1 Does Thoracic Irradiation Improve Survival and Local Control in Limited-Stage Small-Cell Carcinoma of the Lung? A Meta-Analysis By P. Warde and D. Payne Purpose: Our main purpose was to determine whether the addition of thoracic radiation therapy to systemic improves 2-year survival, improves local (intrathoracic) tumor control, and affects treatmentrelated mortality in patients with limited-stage small-cell carcinoma of the lung. Design: Eleven randomized trials addressing this issue were identified using a computerized literature search (Medline and Cancerline) and by polling senior investigators in the field. A meta-analysis was then performed and the results of the trials were analyzed in two ways, the odds ratio (OR) (Peto) method and the risk difference method (Dersimonian and Laird). Results: The overall OR for benefit of thoracic radiation on 2-year survival (ie, the odds of surviving 2 years among patients allocated to radiation compared with the odds of surviving 2 years among patients allocated to control) is 1.53 (95% confidence interval [CI], 1.30 to 1.76; X2 = 12.76; P <.001). The risk difference method showed SMALL-CELL CARCINOMA of the lung contributes to approximately 20% of all cases of lung cancer. The staging classification suggested by the Veterans Administration Lung Cancer Group (VALG) is widely used, and divides patients into those with limitedstage disease and those with extensive-stage disease.' Approximately 30% of patients have limited-stage disease, which is defined as tumor confined to one hemithorax and the regional lymph nodes, which include the mediastinal, ipsilateral supraclavicular, and the contralateral hilar nodes. The majority of these patients have occult metastatic disease and systemic is now the accepted primary modality of treatment. There is still uncertainty about whether it is necessary to give thoracic radiation therapy in addition to systemic, and this question has been addressed in a number of randomized trials with apparently conflicting results. Radiation therapy (RT) directed at the primary tumor might be anticipated to improve local control because the disease is responsive to irradiation, and the treatment can be directed at the region of gross disease From the Department of Radiation Oncology, University of Toronto and Princess Margaret Hospital, Toronto, Canada. Submitted August 26, 1991; accepted January 15, Address reprint requests to P. Warde, MB, MRCPI, FRCPC, 500 Sherbourne St, Toronto, Ontario, M4X 1K9 Canada by American Society of Clinical Oncology X/ $3. 00/0 that radiation therapy improved 2-year survival by 5.4% (95% Cl, 1.1% to 9.7%). Local control results were available for only nine studies, the OR for treatment benefit is 3.02 (95% Cl, 2.80 to 3.24; x 2 = ; P <.0001), and intrathoracic tumor control was improved by 25.3% (95% CI, 16.5% to 34.1%). The OR for excess treatment-related deaths in the thoracic radiation-treated patients was 2.54 (95% CI, 1.90 to 3.18;? = 8.24; P <.01). The risk difference for treatment-related deaths was 1.2% (95% CI, - 0.6% to 3.0%). Conclusions: This meta-analysis shows a small but significant improvement in survival and a major improvement in tumor control in the thorax in patients receiving thoracic radiation therapy. However, this is achieved at the cost of a small increase in treatment-related mortality. J Clin Oncol 10: o 1992 by American Society of Clinical Oncology. without systemic toxicity. Improved local control in turn might be expected to improve the survival rate, even in a disease like limited-stage small-cell lung cancer, which is essentially systemic. This question has been debated since the introduction of for disease treatment, and now there are at least 11 randomized trials with mature data addressing this issue. Not all of these trials conclude that thoracic irradiation is beneficial. Indeed, at a consensus meeting on the management of small-cell cancer in 1988,2 it was concluded that "a definitive recommendation for thoracic RT in combined modality regimens cannot yet be made." This statement may be considered conservative because the oncology community seems to have widely adopted the practice. Indeed thoracic irradiation is a component of virtually all North American limited-stage small-cell cancer trials; the principal questions address dose, fractionation, timing, or volume. Nevertheless, controversy remains in some circles in which it is noted that some randomized trials have been negative and some positive, and it is concluded that thoracic irradiation is not indicated. However, when the apparent benefits of treatment, although significant, are of modest magnitude, the method of meta-analysis of a group of trials may provide the opportunity to form a conclusion with greater confidence than would be the case when one simply contemplates the same group of trials with various outcomes. We have summarized studies of this issue and performed a meta-analysis of published randomized trials. Our object is to evaluate 890 Journal of Clinical Oncology, Vol 10, No 6 (June), 1992: pp

2 META-ANALYSIS OF XRT IN SMALL-CELL LUNG CANCER 891 the effect of thoracic RT on survival, intrathoracic control of disease, and treatment-related toxicity in patients with limited-stage small-cell carcinoma of the lung treated with. General Principles DESIGN The meta-analysis method used in this overview was designed to meet the criteria suggested by Sacks et al. 3 The first step was to develop a research protocol that specified the primary and secondary research objectives of the analysis, how studies were to be identified, the inclusion and exclusion criteria for studies, how data were to be abstracted, and how the statistical analyses were to be performed. Research Objectives The primary research objective was to assess the effect of the addition of thoracic irradiation to on 2-year survival in patients with limited-stage small-cell carcinoma of the lung. The secondary research objectives were to assess the toxicity of the addition of thoracic irradiation and to assess the effect of thoracic irradiation on local (intrathoracic) control of disease at 2 years. Literature Search We decided to confine the analysis to randomized trials of thoracic irradiation in patients with limited-stage small-cell carcinoma of the lung. The studies used in the analysis were identified from a computerized literature search using Medline and Cancerline and by polling senior investigators in the field. A manual search was also performed using references derived from review articles and trial reports. Data Abstraction To avoid bias in the data abstraction process, the data from the studies was abstracted independently by two observers (P.W. and D.P.), who subsequently compared results. When the 2-year actuarial survival/local control results were not specified in the text of the report, these data were obtained from actuarial curves. We then multiplied the survival/local control percentages by the number of patients in the appropriate arms of the studies to generate the proportion of patients surviving (or maintaining local control) at 2 years with and without thoracic radiation therapy. Local control data at 2 years was not available in all studies; in some studies it was available only at the time of death, in others it was available only at the time of reporting. We decided to include these data in our analyses but have indicated when this has occurred. In two studies detailed information on patterns of failure was available, including reasons for excluding patients from the analysis.' In both studies, we elected to score all excluded patients as having local failure in both the thoracic radiotherapy and control arms. When results were available in abstract form, only the principal investigator was contacted for a more complete description of the results (R. Creech, personal communication, February 1990).4 Statistical Methods We combined the results for 2-year survival, local control at 2 years, and toxic deaths in two ways. The first method used was the classical Peto method that adapts the Mantel and Haenszel method for combining data from contingency tables. 5 This method produces an estimate of the odds ratio (OR) for treatment effect (the odds of success among patients allocated to treatment compared with the odds of patients allocated to control), the SE of this estimate, and a statistical test of the null hypothesis of no treatment effect. This is the most commonly used statistical method for pooling data and is well described by Yusuf et al. 6 The second method we used was the DerSimonian and Laird method; in this approach the parameter of interest in each study is the difference between event rates in the treated and untreated groups. 7 This method is more conservative than the Peto approach if there is substantial heterogeneity between the studies and it gives an estimate of the overall rate difference, its SE, and a statistical test of the null hypothesis of no treatment effect (eg, if the 2-year survival in the treated group was 20% and 15% in the untreated group, the rate difference is 5%). The DerSimonian and Laird method is more intuitively appealing to clinicians because it estimates actual gains that can be expected in terms of percentages of patients treated. To assess the extent to which our meta-analysis results were potentially distorted through publication bias, we performed sensitivity analyses by adding the results of a single negative randomized trial to the analysis and by varying the size of the treatment groups to raise the P value above.05 (so that there would not be any statistically significant difference when looking at the X 2 for overall treatment effect). RESULTS The trials that were identified are listed in Table 1.4,8-17 The definition of limited-stage disease was somewhat different in the various studies. In some studies, patients with pleural effusion were excluded from entry, whereas in other trials these patients were excluded only if the pleural fluid cytology was positive. Also, in some studies patients were excluded based on tumor size and pulmonary function, whereas in other studies these were not stated to be exclusion criteria. Some studies randomized only patients showing a clinical complete response, presumably because of a belief that thoracic RT was required if grossly evident disease remained after. Although it is apparent that selection criteria for these studies are not uniformly identical, the population is one in which the treating physicians believed they were treating regional disease and that a regional treatment might make an important difference. In a general sense, this population is reasonably homogeneous. Furthermore, the meta-analysis method does not "pool" studies for comparison but rather studies a collection of properly randomized comparisons. The 11 trials considered here thus constitute an appropriate data set. The agents used, whether prophylactic cranial irradiation (PCI) was given, the thoracic irradiation dose/fractionation details, and the timing of the thoracic irradiation are also outlined in Table 1.

3 892 WARDE AND PAYNE Table 1. Randomized Trials of Thoracic Radiation Therapy With Chemotherapy in Limited-Stage Small-Cell Lung Cancer Author Chemotherapy Prophylactic Cranial Irradiation Dose/Fraction Thoracic Radiation Timing Bunn et a19 Osterlind et al 1 3 Kies et a1 10 Birch et al8 (LUN 328) Birch et a18 (LUN 81343) Ohnoshi 12 Souhami et a1 17 Perry et all4 Creech et a1 4 Nou et al15 Carlson et a116 CML/VAP CMVL CMVEA CAV CML CMVP/EAN AV/MC CVEA CMLAE CMAV/CMLV CLVP/EAM No NS No No 40 Gy/15F/3 wks 40 Gy/10 OF/4 wks 48 Gy/22F/6.5 wks 40 Gy/14F/7 wks 45 Gy/15F/8 wks 40 Gy/20F/4 wks 40 Gy/20F/4 wks 50 Gy/25F/5 wks 50 Gy/25F/5 wks 40 Gy/20F/4 wks 55 Gy/30/5-7 wks Concurrent with first dose of Days and days Started day 85 for 2 wks, rest then compared over 2.5 wks Given on wks 5, 8, and 11 Given wks 1,2, and 7 Between Ist + 2nd course After 4 courses of to patients with CR or PR Either with Ist or 4th cycle of Starting day 43 after completion of induction After 3 courses of After 6-9 month of induction Abbreviations: A, Adriamycin; C, cyclophosphamide; E, etoposide; L, lomustine (CCNU); M, methotrexate; N, mustine; P, procarbazine; V, vincristine; NS, not stated. Survival The 2-year survival results from the individual studies are shown in Table 2. In column 4, the observed number of events (2-year survivors) in the treated group in each study is given (denoted by O). Column 5 lists the expected number of events (2-year survivors) in each study if treatment were wholly without effect (denoted by E). This is calculated by multiplying the total number of patients in the treated group by the total number of events (2-year survivors) in both the treated and untreated groups and then dividing this by the total number of patients. The difference between the observed and expected (O-E) is a measure of the treatment effect and this is shown in column 6. The variance of (O-E) is shown in column 7. The OR for treatment effect (the odds of surviving 2 years among patients allocated to thoracic irradiation compared with the odds of patients allocated to control) is shown in column 8. This can be estimated using the formula: OR = exponential (O-E)/V. Column 9 contains the X 2 for treatment effect of the individual studies (this is significant at the.05 level if > 3.84, [1 degree of freedom (df)]). The overall OR, which is the benefit of thoracic irradiation on 2-year survival, is 1.53 (95% confidence interval [CI], 1.30 to 1.76). The Mantel-Haentzel X 2 for overall treatment benefit is highly significant at (1 df). The results for the DerSimonian-Laird method are also shown in Table 2. Column 10 has the event rate difference (the difference in 2-year survival between the thoracic radiotherapy-treated patients and the control patients for each study). The overall rate difference, which is the overall benefit of adjuvant thoracic irradiation on 2-year survival, is 5.4% (P <.05) with a 95% CI of 1.1% to 9.7%. Local Control Local (intrathoracic) control results were available for only nine of 11 studies, and these are shown in Table 3. The overall OR for treatment benefit was 3.02, (95% CI, 2.80 to 3.24), the x 2 for treatment effect was (P <.0001). The DerSimonian-Laird method showed that the effect of thoracic radiation to improve local control by 25.3% (95% CI, 16.5% to 34.1%). With both of the Peto and DerSimonian-Laird methods, there is a clear benefit of thoracic irradiation on intrathoracic relapse.

4 META-ANALYSIS OF XRT IN SMALL-CELL LUNG CANCER 893 Table 2. 2-Year Survival in Trials of Thoracic RT Year Survival x 2 for % Event Odds XRT Rate Author Thoracic RT Control O E O-E V Ratio Effect Difference Bunn et al 9 13/47 6/ Osterlind et al 1 3 3/69 9/ Kies et al1' 10/40 13/ Birch et al8 (LUN 328) 36/149 23/ Birch et a18 (LUN 81343) 43/147 47/ Ohnoshi et al1,1' 2 5/25 3/ Souhami et al 7 6/57 8/ Perryet all' 68/270 19/ Creech et al 4 22/113 14/ Nou et 11' 6/25 4/ Carlson et a116 10/24 10/ Overall 222/ / > NOTE. Overall OR, 1.53 (95% CI, 1.30 to 1.76). Mantel-Haentzel and Peto Methods-- for treatment effect, (30.19)2/71.4 = (P <.005). Overall event rate difference, 5.4% (95% CI, 1.1% to 9.7%). Abbreviation: O, observed number of events; E, expected number of events if treatment was wholly without effect; O-E, difference between observed and effected; V, variance of O-E. Toxic Deaths Data on toxic deaths was available from nine studies and these are shown in Table 4. The overall OR for treatment effect (the odds of treatment-related deaths in the thoracic irradiation group compared with the odds of treatment-related deaths in the control group) was 2.54 (95% CI, 1.90 to 3.18). The x 2 for treatment effect was 8.24 (P <.01). The overall rate difference using the DerSimonian-Laird Method was 1.2% (95% CI, -0.6% to 3.0%). By using both methods, there seemed to be an increased chance of death related to therapy if thoracic RT was given. Table 3. 2-Year Local Control in Trials of Thoracic RT Author Thoracic RT Control Bunn et a19 30/47 13/49 Perry et al" 4 146/270 17/129 Osterlind et al t3 * 22/57 9/60 Kies et ap'* 20/40 15/53 Birch et all (LUN 328)t 65/149 38/142 Birch et al8 (LUN 81343)t 59/147 59/222 Ohnoshi et 0 11 ",12 7/25 6/27 Nou et al 5 s 10/25 5/31 Carlson et al1 6 17/24 10/24 Total 376/ /737 NOTE. Overall OR, 3.02 (95% CI, 2.80 to 3.24). x 2 for treatment effect, (P <.0001). Overall rate difference, 25.3% (95% Cl, 16.5% to 34.1%). *Local control at time of reporting, not 2 years. tlocal control at 1 year. #Local control based on first site of recurrence. Sensitivity Analyses We added a hypothetical randomized trial of thoracic irradiation versus control to the primary analyses of survival, local control, and toxic deaths. The purpose of this was to see what size a single negative randomized trial would have to be to raise the P value for overall treatment effect above.05. For 2-year survival the trial would have to include more than 5,900 patients (assuming equal numbers of patients randomized to thoracic RT and control, and a 20% 2-year survival in both arms). The trial would have to have more than 35,000 patients to include local control data (again assuming equal numbers of patients randomized to thoracic RT and control, and a 55% local control in both arms). The trial size would need to have more than 8,000 patients to include toxic deaths data (again assuming equal numbers of patients in each trial arm and 1% toxic death rate in each arm). Homogeneity An underlying assumption in combining individual trial results to arrive at one summary measure is that the different results obtained in individual studies are due to chance alone. Therefore, all trial results are homogeneous and are estimates of one "true" effect. A x 2 test for heterogeneity (Mantel-Haentzel) was performed for the survival, local control, and toxic death data and was not statistically significant for any end point.1 5 Therefore, it is reasonable to assume that the differences in

5 894 Table 4. Treatment-Related Mortality in Trials of Thoracic RT Author Thoracic RT Control Bunn et ai9 7/47 3/49 Perry et al14 5/270 2/129 Osterlind et a1l3 2/69 0/76 Kies et alo 2/40 0/53 Ohnoshi et all1,12 2/25 0/27 Creech et al4 1/113 1/119 Birch et al8 (LUN 328) 0/149 2/142 Birch et al8 (LUN 81343) 8/147 4/222 Carlson et /24 0/24 Total 29/884 12/841 NOTE. Overall OR, 2.54 (95% CI, 1.90 to 3.18). x 2 for treatment effect, 8.24 (P <.05). Overall rate difference, 1.2% (95% CI, -0.6% to 3.0%). results between the different trials are caused by sampling variation only; recognizing the power of tests for heterogeneity/homogeneity is often low. DISCUSSION Meta-analysis is a relatively new scientific discipline that is being increasingly used in medical research. It is the process of critically reviewing and statistically combining the results of previous research. Sacks et a1 3 have suggested that the purposes of meta-analysis include the need (1) to increase statistical power for primary end points and for subgroups; (2) to resolve uncertainty when reports disagree; (3) to improve estimates of effect size; and (4) to answer questions not posed at the start of individual trials. Because individual randomized trials such as the ones used in this analysis are often too small to detect clinically important effects, it seems particularly important to apply this process of meta-analysis to the question of the benefit of thoracic RT in patients with limited-stage small-cell carcinoma of the lung. The methodologic issues of meta-analysis have been discussed extensively elsewhere, but it is important to point out that patients in one trial are not compared with those in another; instead a stratified analysis is performed in which a treatment effect is estimated within each study and then individual effects are combined to estimate the overall effect. In this study we have attempted to combine systematically the results of all published randomized controlled trials of thoracic RT for limited-stage small-cell carcinoma of the lung. Our results indicate that there is a modest improvement in overall survival and a large improvement in intrathoracic tumor control with the use of thoracic RT. However, these improvements in survival and local control are achieved at the cost of a small, but significant, increase in treatment-related mortality. WARDE AND PAYNE The small improvement in 2-year survival is consistent with the data from various retrospective reviews of nonrandomized data.is The results from retrospective studies have been suspect because of the concern that poor-risk/poor-performance status patients may not have been given combined modality treatment because of concerns regarding toxicity. This potential bias is overcome in prospective randomized trials such as the ones used in this meta-analysis. Our analysis dealt with the survival at 2 years only, and it is possible that the benefit of thoracic RT may not be present if survival beyond 2 years is examined. We were unable to do this because long-term follow-up data are not yet available in the literature. We have not addressed the issue of whether differences in how and when thoracic RT is given in relation to systemic affect survival. This analysis has also shown a substantial improvement in intrathoracic tumor control with RT. At first glance this would seem to be a persuasive argument for administering thoracic RT to all patients with limitedstage disease as improved local control in the chest is a prerequisite for improved survival. However, we observed from the analysis of the survival data that combining and thoracic RT has only a modest impact on 2-year survival. The use of combined modality therapy has also been shown in this analysis to increase treatment-related mortality. In four of nine studies in which the data were available, the OR (the odds of suffering a treatment-related death in the thoracic RT group compared with the odds of suffering a treatment-related death in the control group) for treatment effect was more than 3, indicating a considerable influence of thoracic RT on treatment-related deaths. The influence of combined modality therapy on the quality of life of patients with limited-stage small-cell lung cancer has not been looked at in any of the 11 studies. It is likely that with improvement in disease control in the chest the patients' quality of life has been improved, but this needs to be examined in a prospective manner. A recent Canadian trial demonstrated that thoracic RT should be administered concurrently with the earlier cycles of and not delayed until after all is given.' 9 This result further suggests that the magnitude of the survival benefit of RT may in fact be underestimated if treatment is given by the generally current practice of delaying RT until after all courses have been given. We believe that patients with small-cell lung cancer with limited-stage disease should receive thoracic RT as

6 META-ANALYSIS OF XRT IN SMALL-CELL LUNG CANCER 895 part of routine management, (especially if they are responding to and preferably early in the program), recognizing that in some patients, particularly those with poor performance status and/or impaired pulmonary function, treatment should not be considered mandatory. The survival benefit detected in this study is small, equivalent to, for example, a survival rate at 2 years of 15% without RT compared with 20% if RT is given. These results may be conservative; however, insofar as the RT technique, quality assurance, etc, may have improved over the 2 decades spanned by the studies. Indeed, the two large, well-done American studies11, 14 published in 1987 from the National Cancer Institute and the Cancer and Leukemia Group B support the notion that the advantages may be underestimated in this analysis. Further studies are needed to define the optimal dose and scheduling of thoracic radiation therapy, and it is essential that these studies incorporate quality-of-life assessments. Although this report suggests an important incremental contribution of chest radiotherapy, the high rates of local relapse emphasize the need for improved methods of local treatment. However, it also must be recognized that the development of distant metastases is an important cause of treatment failure and that development of more effective systemic should be the major thrust for research in this field. 1. Zelen M: Keynote address on biostatistics and data retrieval. Cancer Chemother Rep 4:31-42, Payne D, Arriagada R, et al: The role of thoracic radiation therapy in small-cell carcinoma of the lung: A consensus report. Lung Cancer 5: , Sacks HS, Berrier J, Reitman D, et al: Meta-analyses of randomized controlled trials. N Engl J Med 316: , Creech R, Richter M, Finkelstein D: Combination with or without consolidation radiation therapy for regional small-cell carcinoma of the lung. Proc Am Soc Clin Oncol 6:66A, 1987 (abstr) 5. Mantel N, Haenszel W: Statistical aspects of the analysis of data from retrospective studies of disease. J Natl Cancer Inst 22: , Yusuf S, Peto R, Lewis J, et al: Beta-blockade during and after myocardial infarction: An overview of randomised trials. Prog Cardiovasc Dis 27: , DerSimonian R, Laird N: Meta-analysis in clinical trials. Controlled Clin Trials 7: , Birch R, Omura G, Greco FM, et al: Patterns of failure in combined and radiotherapy for limited small-cell lung cancer: South Eastern Chemotherapy Group experience. Monogr Natl Cancer Inst 6: , Bunn PA, Lichter AS, Makuch RW, et al: Chemotherapy alone or with chest radiation therapy in limited stage small-cell lung cancer. Ann Int Med 106: , Kies MS, Mira JG, Crowley JJ, et al: Multimodal therapy for limited small cell lung cancer: A randomized study of induction combination with or without thoracic radiation in complete responders; and with wide-field versus reduced field radiation in partial responders; A South West Oncology Group study. J Clin Oncol 5: , 1987 REFERENCES 11. Ohnoshi T, Hiraki J, Kawahara S, et al: Trial comparing alone and plus chest irradiation in limited-stage small-cell lung cancer: A preliminary report. Jpn J Clin Oncol 16: , Ohnoshi T, Hiraki S, Kimura I: Randomised trial of alone or with chest irradiation in limited-stage small-cell lung cancer. Int Congr Ser 776: , Osterlind K, Hansen HH, Hansen HS, et al: Chemotherapy versus plus irradiation in limited small cell lung cancer: Results of a controlled trial with 5 years follow-up. Br J Cancer 54:7-17, Perry MC, Eaton WL, Propert KJ, et al: Chemotherapy with or without radiation therapy in limited small-cell carcinoma of the lung. N Engl J Med 316: , Nou E, Brodin O, Bergh J: A Randomised study of radiation treatment in small cell broncial carcinoma treated with two types of four drug regimens. Cancer 62: , Carlson RW, Sikicbi, et al: Late consolidative radiation therapy in the treatment of limited stage small cell lung cancer. Cancer 68: , Souhami RI, Geddes DM, Spiro SG, et al: Radiotherapy in small cell cancer of the lung treated with combination : A controlled trial. Br Med J 288: , Byhardt RW, Cox JD, Holoye PY, et al: The role of consolidation irradiation in combined modality therapy of small cell carcinoma of the lung. Int J Radiat Oncol Biol Phys 8: , Murray N, Coy P, Pater J, et al: National Cancer Institute of Canada. The importance of timing for thoracic irradiation in the combined modality treatment of limited stage small cell lung cancer. Proc Am Soc Clin Oncol 10:243, 1991 (abstr 831)