Lactic Effectiveness of Locoregional Radiation Therapy in Breast Cancer

Transcription

1 Does Locoregional Radiation Therapy Improve Survival in Breast Cancer? A Meta-Analysis By Timothy J. Whelan, Jim Julian, Jim Wright, Alejandro R. Jadad, and Mark L. Levine From the Departments of Medicine, Clinical Epidemiology, and Biostatistics, McMaster University, and Cancer Care Ontario, Hamilton Regional Cancer Centre, Hamilton, ON, Canada. Submitted July 13, 1999; accepted November 29, Address reprint requests to Timothy J. Whelan, Hamilton Regional Cancer Centre, 699 Concession St, Hamilton, ON L8V 5C2, Canada; tim.whelan@hrcc.on.ca by American Society of Clinical Oncology X/00/ Purpose: Recent randomized trials in women with node-positive breast cancer who received systemic treatment report that locoregional radiation therapy improves survival. Previous trials failed to detect a difference in survival that results from its use. A systematic review of randomized trials that examine the effectiveness of locoregional radiation therapy in patients treated by definitive surgery and adjuvant systemic therapy was conducted. Methods: Randomized trials published between 1967 and 1999 were identified through MEDLINE database, CancerLit database, and reference lists of relevant articles. Relevant data was abstracted. The results of randomized trials were pooled using meta-analyses to estimate the effect of treatment on any recurrence, locoregional recurrence, and mortality. Results: Eighteen trials that involved a total of 6,367 patients were identified. Most trials included both preand postmenopausal women with node-positive breast cancer treated with modified radical mastectomy. The type of systemic therapy received, sites irradiated, techniques used, and doses of radiation delivered varied between trials. Data on toxicity were infrequently reported. Radiation was shown to reduce the risk of any recurrence (odds ratio, 0.69; 95% confidence interval [CI], 0.58 to 0.83), local recurrence (odds ratio, 0.25; 95% CI, 0.19 to 0.34), and mortality (odds ratio, 0.83; 95% CI, 0.74 to 0.94). Conclusion: Locoregional radiation after surgery in patients treated with systemic therapy reduced mortality. Several questions remain on how these results should be translated into current-day clinical practice. J Clin Oncol 18: by American Society of Clinical Oncology. DURING THE LAST 50 years, the efficacy of postoperative locoregional radiation (to the chest wall or breast and regional lymph nodes) in women who undergo surgery for breast cancer has been examined in a number of randomized clinical trials. 1,2 Results demonstrate a reduction in breast cancer locoregional recurrence but no difference in overall survival. Many of these studies were of small sample size. Hence, the role of postoperative locoregional radiation has remained unclear, and as a result, there is variation in its use in clinical practice. The results of recent randomized trials demonstrate that, after mastectomy, postoperative locoregional therapy improves survival in women with node-positive breast cancer who also received adjuvant systemic therapy. 3-5 The results of these studies differ from those of previous studies and support the hypothesis that when systemic therapy is given to reduce the burden of micrometastatic disease, locoregional radiation may impact on overall survival. These studies have stimulated much discussion concerning the use of locoregional radiation therapy in routine clinical practice. 6,7 There remain a number of unanswered questions regarding the generalizability of these findings to patients with node-negative disease and to those treated with breastconserving surgery. There is the concern that the use of locoregional radiation in patients treated with more doseintensive or anthracycline-containing chemotherapy may be less effective and will be associated with an increase in cardiac toxicity 8,9 and acute leukemia. 10 The rate of significant arm lymphoedema that impacts on a patient s quality of life may also increase We wanted to review all trials of locoregional radiation therapy in women treated with systemic therapy to determine if the mortality effects observed in recently published studies were consistent with those in other trials and to assess the generalizability of these findings to current practice. Previous meta-analyses have either not included trials in which patients were treated with adjuvant systemic therapy or have not focused on this group of studies. 1,2 Our specific objectives were to conduct a systematic review of randomized trials that examined the effectiveness and toxicity of locoregional radiation therapy in patients with breast cancer treated by definitive surgery and adjuvant systemic therapy, to perform a meta-analysis of the results of these trials, and to consider possible factors (patient- and treatment-related) that could influence the treatment effect. Study Identification METHODS A structured search was conducted to identify randomized controlled trials of locoregional radiation therapy after definitive surgery in 1220 Journal of Clinical Oncology, Vol 18, No 6 (March), 2000: pp

2 LOCOREGIONAL RADIATION THERAPY women with breast cancer treated with systemic therapy, which was defined as adjuvant chemotherapy or hormonal therapy. A trial was suitable for inclusion if it fulfilled the following criteria: was published in a peer-reviewed journal in any language; all patients were treated by definitive surgery, ie, either radical/ modified radical mastectomy or lumpectomy plus an axillary dissection; patients in both treatment arms received the same systemic therapy; allocation of locoregional radiation treatment was said to be randomized; radiation therapy was delivered to the regional lymph nodes and chest wall or breast; and median follow-up was 5 years or more. Abstracts, as well as published papers, were acceptable. If the same trial had been published more than once, the most recently published data were used. Potentially eligible studies were identified by use of the following strategy: A MEDLINE and CancerLit search was completed for the period from 1966 to July Search terms included the following combined subject headings: breast neoplasms, systemic treatment, radiation therapy, randomization, and meta-analysis. The citation lists of all retrieved articles were examined to identify other potentially relevant reports. In addition, we manually reviewed relevant journals published in the first 6 months of A citation identified by any of the search strategies was reviewed by at least two of the investigators. The decision to select an article was based on information available in the published report and was reached by consensus. Three trials were considered inappropriate for inclusion in our study: one trial that included patients with locally advanced disease not treated by definitive surgery, 14 one that compared monochemotherapy plus radiation with polychemotherapy alone, 15 and one that had less than 5 years of follow-up. 16 Potentially eligible studies were randomly sorted into two groups, and each group was assigned to an investigator for independent review and data abstraction. After completing the review, each of the reviewers assessed the other reviewer s studies. Any disagreement in abstracted data was resolved by referral to the hard copy of the article or by review by another investigator. We assumed that the randomization was adequately implemented and that follow-up was complete in the studies that met our inclusion criteria, unless otherwise stated in the published reports. The methodologic quality of the randomized controlled trials 17 was considered by the use of a validated instrument 18 that independently assesses the method of randomization, the use of double blinding (not applicable), and the description of withdrawal and dropouts. Scores range from 0 (low quality) to 5 (high quality). Data Collection and Statistical Analysis The following information was gathered from each report: name of the first author, year of publication, number of patients randomized to locoregional radiation therapy or no locoregional radiation therapy, stage of disease, type of surgery, type of systemic therapy, locoregional sites irradiated, technique used, dose and fractionation schedule of radiation, sequencing of chemotherapy and radiation therapy, median follow-up, and the number of patients who experienced treatment toxicity, whose disease recurred at any site, whose disease recurred locally, and who died. The analysis was performed on published data; no attempt was made to obtain data on individual patients. The reported follow-up times for any recurrence, locoregional recurrence, and death varied between studies. The maximum published follow-up interval that was available was used. Whenever possible, the raw number of events was used, which was the case for the majority of the trials. In two trials, the number of events was estimated from published survival curves. For this approach, the follow-up time was restricted to a point at which approximately 50% of the patients had been observed (median followup). The number of events was obtained by applying a set square to the survival curve at the time of median follow-up, reading off the percentage surviving, and multiplying by the total number of patients randomized to the group to estimate the absolute number of survivors. The validity of the abstracted data was assessed by repeated crosschecking. This approach is approximate but reasonably accurate in the context of constant hazard and was considered sufficiently robust for the purposes of this analysis. In view of the limited description of toxicity in the reports, such data were not summarized. To estimate the effect of treatment on any recurrence, locoregional recurrence, and mortality, the results of randomized trials were pooled using meta-analysis. The primary analysis combined the study-specific odds ratios by use of precisionbased (or inverse variance) weights (alternatively called logit or Woolf 19 estimators) under the assumptions of both fixed and random effects as described by Laird and Mosteller. 20 In the fixed effect model, the weighting of two of the largest studies 3,5 was 50.4% of the total. In the random effects model, the weighting for the two Danish studies was 37.8%. Results were similar, and those of the random effects model are shown because they are the most conservative. The pooled treatment effect is expressed as an odds ratio ( 95% confidence interval [CI]) such that estimates more than 1.0 favor control and those less than 1.0 favor radiation therapy. All P values and 95% CIs were two-sided. Assessments of homogeneity and overall association were undertaken by use of 2 tests. Exploratory Analysis of Factors That Influence the Treatment Effect 1221 An exploratory analysis was performed for patient and technical factors that might influence the effect of treatment on mortality. The consistency of the treatment effect (odds ratio) was compared between studies grouped by the following variables: extent of disease (advanced, defined as stage III disease or 50% of patients with three nodes positive, v early), degree of axillary dissection (extensive, defined as an axillary evacuation, complete removal of contents, or a minimum number of nodes [ six] removed, v less extensive, defined as level II dissection with no minimum number of nodes specified), anthracycline-based chemotherapy (yes v no), radiation technique (mega- v orthovoltage), extent of radiation (all locoregional sites, defined as chest wall or breast and supraclavicular, axillary, and internal mammary nodes, v not), dose of radiation therapy ( 45 Gy v 45 Gy), timing of radiation therapy ( 6 months since initiation of systemic therapy v 6 months), rate of locoregional failure in the control arm ( median v median), and methodologic quality score of study ( 2 v 2). A random effects regression model was applied to the data according to the methods of Berkey et al. 21 All factors were coded as 1 or 0, and each was assessed individually and in the context of a multivariate model that contained the other factors. The factors, degree of axillary dissection, and rate of locoregional recurrence were not included in the latter analysis because of the limited number of studies in which this

3 1222 WHELAN ET AL Table 1. RCTs of Locoregional Radiation Therapy: Patient Characteristics Trial (first author, year initiated) No. of Patients Stage Surgery Extent of AX Dissection Chemotherapy DeBoer, NS M NS CMFP BCG Foroglou, NS M NS Chemo-endocrine Grohn, 24 Klefstrom, III MRM AX fat removed VAC levamisole Tramprisch, NS M NS LMF Blomqvist, II MRM AX evacuation CAFt tamoxifen Hayat, II MRM NS CMF Amparo, 29 Gervasio, II MRM NS AC Cooper, 31 Muss, II RM/MRM 10 nodes removed CMF, L-PAM Schmoor, II MRM six nodes dissected CMF Griem, II & III MRM NS CMF, MF, AC McArdle, 35, II MRM Clearance of AX contents CMF Vélez-Garcia, II & III RM/MRM Complete dissection 10 nodes CMF Ahmann, 41 Martinez, II & III MRM Complete removal of AX CFP contents Olson, III MRM eight nodes (median, 17) CAFTH Ragaz, II MRM Level II CMF Arwidi, 44 Ryden, 45 Tennvall-Nittby, I & II MRM Dissection to AX vein Cyclophosphamide, tamoxifen Overgaard, 5, ,375 II & III TM AD Level I and part of II Tamoxifen Overgaard, 3 Mouridsen, ,708 II & III TM AD Level I and part of II CMF Abbreviations: AC, doxorubicin, cyclophosphamide; AD, axillary dissection; AX, axillary; BCG, bacille Calmette-Guérin; CAFt, cyclophosphamide, doxorubicin, ftorafur; CAFTH, cyclophosphamide, doxorubicin, fluorouracil, tamoxifen, fluoxymesterone; CFP, cyclophosphamide, fluorouracil, prednisone; CMF, cyclophosphamide, methotrexate, fluorouracil; CMFP, cyclophosphamide, methotrexate, fluorouracil, prednisone; LMF, Leukeran, methotrexate, fluorouracil; L-PAM, melphalan; M, mastectomy, not otherwise specified; MF, methotrexate, fluorouracil; MRM, modified radical mastectomy; NS, not specified; RCT, randomized controlled trial; RM, radical mastectomy; TM, total mastectomy; VAC, vincristine, doxorubicin, cyclophosphamide. information was available. Only main effects (no interactions) were considered. All testing was performed treating the ratio of the estimated coefficient to its standard error as a t statistic, with degrees of freedom equal to the number of studies minus the number of estimated coefficients minus three. All P values and 95% CIs were two-sided. RESULTS Eighteen randomized trials that met our inclusion criteria were identified and reviewed in detail (Tables 1 and 2). Fifteen trials were published independently, 3-5,24,25,27-48 and three trials were published only as part of a metaanalysis. 22,23,26 The studies, which comprised a total of 6,367 patients, were initiated between 1973 and One half of the trials (n 9) involved fewer than 200 patients, and only two trials involved more than 1,000 patients (median, 209 patients; range, 50 to 1,708 patients). Median follow-ups ranged from 7.5 to 14.5 years. Patient Characteristics Most trials included both pre- and postmenopausal women. Two trials included only premenopausal patients, 3,4 and one trial included only postmenopausal patients. 5 The majority of the trials limited eligibility to patients who were node-positive. One trial was limited to patients with more than four positive nodes, 40 and two trials were limited to patients with stage III disease. 25,43 Several trials included patients with node-negative breast cancer, stage III disease, with either primary tumors greater than 5 cm or involvement of the skin or muscle. 3,5,25,34,43 Only one trial included patients with node-negative breast cancer with primary tumors 2 to 5 cm. 46 In the majority of trials, patients were treated with modified radical mastectomies. No trials were identified that treated patients with lumpectomies plus axillary dissections. The extent of axillary dissection was reported in 12 trials. In the majority of these trials, patients were treated with extensive axillary dissections. 25,27,32,33,36,40,42,43,46 In three trials, patients were treated with less extensive dissections. 3-5 Systemic Therapy By definition, all trials included patients treated with systemic therapy, and in three trials, different systemic therapy was used for different strata. 32,34,46 Cyclophosphamide, methotrexate, and fluorouracil (CMF) chemotherapy was used in nine trials, 3,4,22,28,32-34,36,40 an anthracyclinebased regimen in five trials, 25,27,30,34,43 other polychemotherapy in three trials, 26,34,42 and monochemotherapy in two trials. 32,46 Combined chemo-endocrine therapy was used in three trials, 23,27,43 and tamoxifen alone was used in two

4 LOCOREGIONAL RADIATION THERAPY 1223 Table 2. RCTs of Locoregional RT: Interventions RT Radiation Regimen Trial (first author, year initiated) Site Dose (Gy) Fraction Time (weeks) Energy Timing of RT DeBoer, CW, SC, AX Megavoltage NS Foroglou, CW, SC, AX, IMN Megavoltage/orthovoltage NS Grohn, 24 Klefstrom, CW, SC, AX, IMN Megavoltage Prechemo Tramprisch, CW, SC, AX Megavoltage NS Blomqvist, CW, SC, AX, IMN Megavoltage Sandwich, 2/3 cycles Hayat, CW, SC, AX, IMN Megavoltage Sandwich, 6/7 cycles Amparo, 29 Gervasio, CW, SC, AX, IMN Megavoltage/orthovoltage Prechemo Cooper, 31 Muss, CW, SC, AX, IMN Megavoltage Prechemo Schmoor, CW, SC, AX, IMN Megavoltage Sandwich, 2/3 cycles Griem, CW, SC, AX Megavoltage Postchemo McArdle, 35, CW, SC, AX, IMN Orthovoltage Prechemo Vélez-Garcia, CW, SC, AX, IMN Megavoltage Prechemo Ahmann, 41 Martinez, CW, SC, AX, IMN Megavoltage/orthovoltage/ Concurrent electrons Olson, CW, SC, AX, IMN Megavoltage Postchemo Ragaz, CW, SC, AX, IMN Megavoltage Sandwich, 4/5 cycles Arwidi, 44 Ryden, 45 Tennvall-Nittby, CW, SC, AX, IMN Megavoltage/orthovoltage/ electrons Concurrent Overgaard, 5, CW, SC, AX, IMN Megavoltage/electrons Concurrent Overgaard, 3 Mouridsen, CW, SC, AX, IMN Megavoltage/electrons Sandwich, 1/2 cycles Abbreviations: RT, radiation therapy; CW, chest wall; SC, supraclavicular lymph nodes; AX, axillary lymph nodes; IMN, internal mammary nodes; Prechemo, before chemotherapy; Postchemo, after chemotherapy. trials. 5,46 Two trials used immunotherapy in addition to chemotherapy. 22,25 Radiation Therapy In the majority of trials, radiation was delivered to the chest wall, supraclavicular, axilla, and internal mammary nodal areas. In three trials, the internal mammary nodes were not irradiated 22,26,34 ; in two other trials, radiation to the chest wall was optional, depending on the size of the primary tumor 32 or whether the tumor involved the skin. 42 Field arrangements or techniques varied from trial to trial and within trials. The chest wall was irradiated with two tangent fields 4,32-34,43 or with a single direct electron or photon field. 3,5,25,27,42,46 The supraclavicular and axillary nodes were irradiated either with an anterior field with a posterior patch 3-5,25,27,34,42,43,46 or with an anterior field alone. 3,32-34,43,46 The internal mammary nodes were irradiated with a single anterior field 4,25,43,46 or were included in chest wall irradiation, either by the wide tangents 32,37 or by an electron field to the chest wall. 3,5 An inverted L field (or hockey stick) was used to treat the supraclavicular, axillary, and internal mammary nodes in four trials. 27,32,33,40 Techniques to avoid substantial cardiac irradiation by the use of electrons alone 3,5,46 or mixed electron photon beam 33,43 were used in five trials. The technique was not described in seven trials. 22,23,26,28,30,36,40 Radiation was delivered primarily with megavoltage linear accelerators. Orthovoltage was used either solely 36 or in combination with megavoltage machines 23,30,42,46 in five trials. The dose of radiation ranged from 35 to 60 Gy given in 12 to 30 fractions. Radiation was delivered in to 7 weeks. The most common fractionation schedule was 50 Gy in 25 fractions over a 5-week period. 3,5,23,25,28,32,33,40,42,43 Compliance with radiation therapy was reported in seven trials and ranged from 68% to 100% (median, 96%). 3-5,27,34,36,43 In two trials, fewer than 85% of the patients who were randomized to radiation therapy received the intervention. 34,43 The scheduling of radiation therapy and chemotherapy was described in 15 trials. Radiation was given before chemotherapy in five trials, 25,30,32,36,40 sandwiched between cycles in five trials, 3,4,27,28,33 concurrent with systemic therapy in three trials, 5,42,46 and after chemotherapy in two trials. 34,43 Methodologic Quality of Studies The methodologic quality scores along with the number of trials that achieved them were as follows: 1 (seven trials), 2 (seven trials), and 3 (four trials). The four trials with a score of 3 provided adequate description of the randomization procedure and the handling of withdrawals and dropouts. No trial involved blinded allocation to treatment, thus higher scores were not obtained.

5 1224 WHELAN ET AL Fig 1. Meta-analysis of locoregional radiation therapy randomized trials: any recurrence. Abbreviations: N, number; OR, odds ratio; CI, confidence interval; TAM, tamoxifen; CMF, cyclophosphamide, methotrexate, fluorouracil. Toxicity Data on toxicity of therapy were variably reported. Information was available from eight trials. 3-5,27,30,34,42,43 Acute toxicity was infrequently reported, occurring in the trials as follows: severe skin toxicity, 2.7% 43 and 5% 42 ; myelosuppression attributed to radiation therapy, 2% 43 and 32% 27 ; and radiation pneumonitis, 1%, 4 15%, 42 and 23%. 27 Radiation esophagitis occurred in 17% of patients 42 in one trial. This last study had particularly high rates of acute and long-term toxicity and was the only trial in which radiation was given concurrently with chemotherapy. With respect to late toxicity, no cases of brachial plexus neuropathy were reported. Arm edema was reported in three trials. The incidence ranged from 0% to 25% (median, 3%) in nonirradiated patients and from 10% to 54% (median, 12%) in irradiated patients. 4,42,43 Cardiac toxicity, primarily congestive heart failure, was reported in six trials. 4,27,30,34,42,43 In trials using CMF, no cardiac complications were reported in patients treated with chemotherapy alone. 4,42 One case of pericarditis was reported in a patient treated with CMF and radiation therapy. 42 In trials in which patients were treated with anthracycline-containing chemotherapy, 27,30,34,38 the incidence of congestive heart failure in nonirradiated patients ranged from 0% to 19.2% (median, 2.6%). The incidence of cardiac failure in irradiated patients ranged from 1.9% to 23.6% (median, 3.2%). In two other studies, 3,5 no increase in 12-year cumulative morbidity or mortality from ischemic heart disease was observed in irradiated patients. 49 The incidence of secondary cancers was reported in only two trials 4,32 ; no increase was noted in irradiated patients. One case of acute myelogenous leukemia was reported in a patient treated with CMF and radiotherapy. 4 Recurrence and Mortality Data for recurrence were available for 13 trials. Radiation was shown to reduce the risk of any recurrence, with an odds ratio of 0.69 (95% CI, 0.58 to 0.83; P.00004) (Fig 1). This seemed to be largely a result of a reduction in local regional recurrence, with an odds ratio of 0.25 (95% CI, 0.19 to 0.34; P ) (Fig 2). Data for mortality were available for all trials. Radiation was shown to reduce mortality with an odds ratio of 0.83 (95% CI, 0.74 to 0.94; P.004). The test for heterogeneity was negative (P.26) ( Fig 3). A positive treatment effect was seen in six of nine trials that contained more than 200 patients. In two of the three trials with negative treatment effects, compliance with radiation therapy was poor. The meta-analysis was performed excluding the two large Danish trials. 3,5 The resulting odds ratio for mortality was 0.89 (95% CI, 0.76 to 1.05; P.17). We compared these two studies with the remaining studies in a regression model; the difference between the study data sets was not significant (P.15). Exploratory Analysis of Factors That Influenced the Treatment Effect On univariate analysis, only timing of radiation therapy ( 6 months v 6 months) was statistically significant (Table 3). The odds ratio for mortality in the trials in which radiation was administered within 6 months of starting chemotherapy was 0.78, compared with 1.14 in trials in

6 LOCOREGIONAL RADIATION THERAPY 1225 Fig 2. Meta-analysis of locoregional radiation therapy randomized trials: locoregional recurrence. which radiation was delayed. On multivariate analysis, timing of radiation continued to demonstrate an effect on treatment (P.03), and radiation technique (megavoltage v orthovoltage therapy) was also shown to be predictive of treatment effect (P.05). DISCUSSION The role of locoregional radiation therapy after definitive surgery in the management of breast cancer has been evaluated extensively since The majority of early trials focused on patients who did not receive adjuvant systemic therapy. These studies showed that radiation decreases locoregional recurrence, but an effect on overall survival has not been detected. Many of these trials were of relatively small sample size. An update reported by Cuzick et al 1 of a meta-analysis of trials that were initiated before the era of chemotherapy or hormonal therapy suggests that locoregional radiation after mastectomy decreased deaths caused by breast cancer but that this decrease was offset by an increase in deaths caused by cardiac disease. The excess risk for cardiac mortality seems to be greatest for trials that used older radiation techniques in which a high dose was delivered to the myocardium, eg, with the use of orthovoltage therapy to treat the chest wall 50 or wide tangents to treat the internal mammary nodes. 51 The results of recently published randomized trials of radiation therapy after mas- Fig 3. Meta-analysis of locoregional radiation therapy randomized trials: mortality.

7 1226 WHELAN ET AL Table 3. Factor Effect of Radiation on Mortality: Influence of Patient or Treatment Factors* No. of Studies Odds Ratio 95% CI P (random effects) Extent of disease Early Advanced AX dissection Less Extensive Extensive Anthracycline use No Yes Radiation technique Megavoltage Orthovoltage Extent of radiation All sites Not all sites Dose of radiation 45 Gy Gy Timing of radiation 6 months months Locoregional recurrence 24% % Methodologic quality * Univariate analysis in random effects regression model. Median value. tectomy in patients treated with systemic therapy demonstrate that radiation not only reduces the risk of locoregional failure, but improves survival. 3-5 These trials have stimulated much discussion about the role of locoregional radiation after surgery in present-day clinical practice. 6,7 To gain a better understanding of the use of this modality, we performed a systematic review of all randomized trials of locoregional radiation therapy after definitive surgery in patients treated with systemic treatment. The results of the meta-analysis are consistent with the three recently published trials, ie, locoregional therapy not only reduced local failure, but improved disease-free and overall survival. 3-5 Why are the results of our meta-analysis different from those of previous randomized trials and meta-analyses? 52 First, this meta-analysis focused only on patients who were treated with systemic therapy. The meta-analysis by Cuzick et al 1 did not include these trials. The overview by the Early Breast Cancer Trialists Collaborative Group did not specifically focus on this group of patients. 2 We wanted to test another hypothesis that could potentially explain why previous meta-analyses failed to detect an impact of radiation therapy on mortality. This hypothesis was that patients required adjuvant systemic therapy to allow locoregional radiation to manifest its effect. Systemic therapy, particularly chemotherapy, though effective in preventing distant metastases, is likely to be less effective in preventing locoregional recurrence in which the tumor burden is large. In patients who have distant failures reduced with chemotherapy, the effect of radiation therapy on preventing locoregional recurrence and resulting secondary systemic recurrence may be more evident. We performed a systematic review to determine which trials were appropriate to include. Each trial was scrutinized to determine eligibility and to extract appropriate data. One trial 14 included in the previous overview by the Early Breast Cancer Trialists Collaborative Group was excluded in our meta-analysis because it contained patients with locally advanced disease not treated by definitive surgery. One published trial 25 that was not included in the overview was included in our meta-analysis. Second, a number of the trials in our meta-analysis had longer follow-up than was available in trials from previous published meta-analyses. 2 Third, many trials included in our meta-analysis used relatively modern radiotherapy techniques, which delivered a more uniform dose and avoided excessive cardiac irradiation that could have resulted in increased tumor-cell kill and decreased cardiac mortality. The two Danish studies are noteworthy in this regard for their use of a technique that substantially reduced cardiac irradiation. The limited reports of long-term toxicity support this association, 49 but further follow-up will be necessary to confirm this effect. We recognize that two of the largest trials in the metaanalysis were positive. However, it is important in any meta-analysis to include all studies. We used the random effects model to weight all the trials that were included in our analysis. This is conservative because it gives less weight to larger trials than does the fixed effects model. 3,5 In addition, although there are limitations to such an approach, we performed the meta-analysis excluding the two Danish trials. The results, excluding these trials, were not inconsistent with the overall results. Even though our results demonstrate an impact of locoregional radiation on survival, the issue is whether the results are generalizable to current clinical practice. The reviewed studies were initiated 15 to 25 years ago and, for the most part, represent patients with node-positive breast cancer treated with modified radical mastectomy, CMF chemotherapy, or tamoxifen who received radiation to the chest wall and all nodal areas at risk either before or within several months after starting systemic therapy. Breast cancer treatment has changed dramatically since the inception of these

8 LOCOREGIONAL RADIATION THERAPY studies in terms of surgical management, systemic treatment, and radiation therapy. Presently, many women with node-positive breast cancer are treated with lumpectomies and level I and II axillary dissections. Most are also treated with breast irradiation consistent with the goal of breastconserving therapy. It remains unclear what additional benefit further locoregional radiation would have in this situation. The technique for providing breast irradiation often results in a substantial dose of radiation not only to the chest wall, but to the dissected lower axilla and a proportion of the internal mammary nodes. Systemic treatment has also changed. Presently, many women are treated with more effective anthracycline-based regimens followed by longterm hormonal therapy. Again, it remains unclear what additional benefit in absolute terms locoregional radiation would have in this context, and concern remains about the potential for increased cardiac toxicity when radiation therapy is delivered in addition to anthracycline-based chemotherapy. 9 Finally, the timing and extent of radiation therapy has also changed. Radiation is now commonly delivered after chemotherapy to avoid acute interactions with chemotherapy, and the internal mammary nodes are infrequently treated because of the low risk of recurrence and to avoid excessive cardiac irradiation. The results of the exploratory analyses showed that when radiation is delivered with older techniques, such as the use of orthovoltage, or is given 6 months after the initiation of systemic therapy, it may be less effective. It is important that these results are interpreted cautiously, because they involve indirect nonrandomized comparisons between trials and because the relatively small number of trials evaluated leads to the risk of false-positive and false-negative conclusions. The use of orthovoltage is associated with increased intrathoracic, including cardiac, irradiation 53 and has been associated previously with increased long-term cardiac morbidity. 50,54 Delay in radiation therapy has been associated with decreased locoregional control. 55,56 Radiation therapy was delivered after six months of systemic therapy in only three trials, and it is unclear whether the effect on treatment had more to do with poor compliance in these studies or with a true biologic effect. Locoregional radiation therapy also seemed to be less effective when all regional sites were not irradiated and in patients treated with anthracycline-based chemotherapy, but these associations were not statistically significant. These results raise many interesting questions about the incorporation of locoregional radiation therapy in modern practice, in which anthracycline-based chemotherapy is commonly used and radiation therapy is often given after completion of systemic therapy. These associations need to be evaluated further in future trials that assess the role of locoregional radiation therapy A potential limitation of our meta-analysis is that it was based on trial-specific rather than patient-specific data. These different types of meta-analyses have been referred to as meta-analysis of the literature and meta-analysis of individual patient data (MAP), respectively. 57 A number of concerns have been discussed in the literature regarding meta-analyses that are based on trial-specific or aggregate data. 57,58 These include lack of incorporation of unpublished trials, inclusion of trials with relatively short followup, the use of estimated event rates obtained from published reports, and analysis based on end-of-trial event rates instead of on the measurement of the effect of treatment over time. The systematic review in this study included both published and unpublished trials that were published as part of a previous meta-analysis. One trial with a relatively short follow-up was excluded from the analysis, and in only two of 18 trials were event rates estimated. The meta-analysis performed on these studies was based on end-of-trial event rates. Concern that such analyses that use odds ratios may be less accurate than those that incorporate hazard ratios has not always been substantiated. In a previous study of patients with ovarian cancer, the results of these analyses were not markedly different. 57 The median follow-up of trials incorporated in our meta-analysis was more than 10 years. It is likely that any treatment effect obtained from radiation given 10 years previously would be less at the time of follow-up. However, a 10-year follow-up may not provide sufficient time for any negative effect on mortality, such as cardiac toxicity, to become evident. 59 This represents a potential limitation of the analysis, but the data support a benefit of locoregional radiation for up to 10 years. Through a systematic review, we had the opportunity to extract data not found in previous meta-analyses regarding the methodologic quality of studies, the influence of treatment factors (eg, timing of radiation) on outcome, and toxicity. 1,2 Another potential benefit of an analysis based on published data is the inclusion of trials for which only published data is available, as was the case for one particular study in our analysis. 25 We believe a MAP would be an important step in the assessment of locoregional radiation therapy and may provide a better quantitative estimate of the treatment effect. However, given the large number of trials involved in our meta-analysis, it is unlikely that the results of the meta-analysis would change qualitatively. Individual patient data would also afford investigators more power to consider potential predictive factors of the treatment effect. It is important to note, though, that a MAP may not overcome the problems of lack of generalizability of results caused by changes in practice over time. The results of this meta-analysis support an evolution in oncologists thinking concerning the biology of breast

9 1228 WHELAN ET AL cancer. One half century ago, the prevailing theory was that breast cancer spread by stepwise local extension, which resulted in more extensive surgery. By the 1970s and 1980s, this view had been replaced by that of breast cancer as a systemic disease. The recent thinking is that both hypotheses are valid. 60 Our results support the notion that, in the presence of adjuvant systemic therapy, local regional control is important and that reduction in locoregional recurrence may prevent secondary systemic spread from regional sites and, thus, prolong survival. Since the inception of these trials, many changes have occurred in the management of breast cancer, and it remains unclear how locoregional radiation should be incorporated into current practice. Several randomized trials that evaluate the integration of locoregional radiation into the current management of breast cancer are in progress or in the planning stages. On the basis of our review, it is vital that these studies have sufficient power to detect important clinical differences and to consider the potential impact of treatment variables such as radiation technique and timing of radiotherapy on mortality and toxicity. These studies will help clarify the role of this treatment modality in the current multidisciplinary management of breast cancer. 1. Cuzick J, Stewart H, Rutqvist L, et al: Cause-specific mortality in long-term survivors of breast cancer who participated in trials of radiotherapy. J Clin Oncol 12: , Early Breast Cancer Trialists Collaborative Group: Effects of radiotherapy and surgery in early breast cancer: An overview of the randomized trials. N Engl J Med 333: , Overgaard M, Hansen PS, Overgaard J, et al: Postoperative radiotherapy in high-risk premenopausal women with breast cancer who receive adjuvant chemotherapy: Danish Breast Cancer Cooperative Group 82b Trial. N Engl J Med 337: , Ragaz J, Jackson SM, Le N, et al: Adjuvant radiotherapy and chemotherapy in node-positive premenopausal women with breast cancer. N Engl J Med 337: , Overgaard M, Jensen MB, Overgaard J, et al: Postoperative radiotherapy in high-risk postmenopausal breast-cancer patients given adjuvant tamoxifen: Danish Breast Cancer Cooperative Group DBCG 82c randomised trial. Lancet 353: , Hellman S: Stopping metastases at their source. N Engl J Med 337: , Recht A, Bartelink H, Fourquet A, et al: Postmastectomy radiotherapy: Questions for the twenty-first century. J Clin Oncol 16: , Valagussa P, Zambetti M, Biasi A, et al: Cardiac effects following adjuvant chemotherapy and breast irradiation in operable breast cancer. Ann Oncol 5: , Shapiro CL, Hardenbergh PH, Gelma R, et al: Cardiac effects of adjuvant doxorubicin and radiation therapy in breast cancer patients. J Clin Oncol 16: , Curtis RE, Boice JD Jr, Stoval M, et al: Risk of leukemia after chemotherapy and radiation treatment for breast cancer. N Engl J Med 326: , Swedborg I, Wallgren A: The effect of pre- and postmastectomy radiotherapy on the degree of edema, shoulder-joint mobility, and gripping force. Cancer 47: , Larson D, Weinstein M, Goldberg I, et al: Edema of the arm as a function of the extent of axillary surgery in patients with stage I-II carcinoma of the breast treated with primary radiotherapy. Int J Radiat Oncol Biol Phys 12: , Tobin MB, Lacey HJ, Meyer L, et al: The psychological morbidity of breast cancer-related arm swelling: Psychological morbidity of lymphoedema. Cancer 72: , Buzdar AU, Hortobagyi GN, Kau S-W, et al: Breast cancer adjuvant therapy trials of M. D. Anderson Hospital: Results of three REFERENCES studies, in Salmon SE (ed): Adjuvant Therapy of Cancer V. Orlando, FL, Grune & Stratton, 1987, pp Spangenberg JP, Nel CJ, Anderson JD, et al: A prospective study of the treatment of stage III breast cancer. S Afr J Surg 24:57-60, Schulz K-D, Reusch K, Schmidt-Rhode P, et al: Consecutive radiation and chemotherapy in the adjuvant treatment of operable breast cancer, in Salmon SE, Jones SE (eds): Adjuvant Therapy of Cancer III. New York, Grune & Stratton, 1981, pp Liberati A, Himel HN, Chalmers TC: A quality assessment of randomized control trials of primary treatment of breast cancer. J Clin Oncol 4: , Jadad AR, Moore RA, Carrol D, et al: Assessing the quality of reports of randomized clinical trials: Is blinding necessary? Control Clin Trials 17:1-12, Woolf B: On estimating the relation between blood group and disease. Ann Hum Genet 19: , Laird NM, Mosteller F: Some statistical methods for combining experimental results. Int J Technol Assess Health Care 6:5-30, Berkey CS, Hoaglin DC, Mosteller F, et al: A random-effects regression model for meta-analysis. Stat Med 14: , DeBoer G, Paterson AHG, Pritchard KI: [unpublished study], in Early Breast Cancer Trialists Collaborative Group: Effects of radiotherapy and surgery in early breast cancer: An overview of the randomized trials. N Engl J Med 333: , Foroglou P, Giokas G, Lissaios B: [unpublished study], in Early Breast Cancer Trialists Collaborative Group: Effects of radiotherapy and surgery in early breast cancer: An overview of the randomized trials. N Engl J Med 333: , Grohn P, Heinonen E, Klefstrom P, et al: Adjuvant postoperative radiotherapy, chemotherapy, and immunotherapy in stage III breast cancer. Cancer 54: , Klefstrom P, Grohn P, Heinonen E, et al: Adjuvant postoperative radiotherapy, chemotherapy, and immunotherapy in stage III breast cancer: II. 5-year results and influence of levamisole. Cancer 60: , Tramprisch HJ: [unpublished study], in Early Breast Cancer Trialists Collaborative Group: Effects of radiotherapy and surgery in early breast cancer: An overview of the randomized trials. N Engl J Med 333: , Blomqvist C, Tiusanen K, Elomaa I, et al: The combination of radiotherapy, adjuvant chemotherapy (cyclophosphamide-doxorubicinfluorouracil) and tamoxifen in stage II breast cancer: Long-term

10 LOCOREGIONAL RADIATION THERAPY follow-up results of a randomized trial. Br J Cancer 66: , Hayat H, Brufman G, Borovik R, et al: Adjuvant chemotherapy and radiation therapy vs chemotherapy alone for stage II breast cancer patients: A national randomized trial. Ann Oncol 1:21, 1990 (suppl, abstr) 29. Amparo M, Gervasio H, Albano J, et al: Adjuvant chemotherapy versus radiotherapy and chemotherapy in operable breast cancer with positive nodes: Results at 14 years of a randomized trial. Ann Oncol 5:16, 1994 (suppl 8, abstr) 30. Gervasio H, Alves H, Rito A, et al: Phase III study: Adjuvant chemotherapy versus adjuvant radiotherapy plus chemotherapy in women with node-positive breast cancer. Breast J 4:S88, 1998 (suppl 1, abstr) 31. Cooper MR, Rhyne AL, Muss HB, et al: A randomized comparative trial of chemotherapy and irradiation therapy for stage II breast cancer. Cancer 47: , Muss HB, Cooper MR, Brockschmidt JK, et al: A randomized trial of chemotherapy (L-PAM vs CMF) and irradiation for node positive breast cancer: Eleven year follow-up of a Piedmont Oncology Association trial. Breast Cancer Res Treat 19:77-84, Schmoor C, Bastert G, Bojar H, et al: Effect of radiotherapy in addition to 6 cycles CMF in node positive breast cancer patients: German Breast Cancer Study Group (GBSG). Eur J Cancer 34:S59, 1998 (suppl 5, abstr) 34. Griem KL, Henderson IC, Gelman R, et al: The 5-year results of a randomized trial of adjuvant radiation therapy after chemotherapy in breast cancer patients treated with mastectomy. J Clin Oncol 5: , McArdle CS, Calman KC, Cooper AF, et al: The social, emotional and financial implications of adjuvant chemotherapy in breast cancer. Br J Surg 68: , McArdle CS, Crawford D, Dykes EH, et al: Adjuvant radiotherapy and chemotherapy in breast cancer. Br J Surg 73: , Vélez-Garcia E, Carpenter JT, Moore M, et al: Postsurgical adjuvant chemotherapy with or without radiotherapy in women with breast cancer and positive axillary nodes: The South-Eastern Cancer Study Group (SECSG) Experience, in Jones SE, Salmon SE (eds), Adjuvant Therapy of Cancer IV. Orlando, FL, Grune & Stratton, 1984, pp Vélez-Garcia E, Moore M, Vogel CL, et al: Postmastectomy adjuvant chemotherapy with or without radiation therapy in women with operable breast cancer and positive axillary lymph nodes: The Southeastern Cancer Study Group experience. Breast Cancer Res Treat, 3:49-60, 1983 (suppl 1) 39. Vélez-Garcia E, Carpenter JT, Moore M, et al: Postsurgical adjuvant chemotherapy with or without radiotherapy in women with breast cancer and positive axillary nodes: Progress report of a South- Eastern Cancer Study Group (SEG) Trial, in Salmon SE (eds), Adjuvant Therapy of Cancer V. Orlando, FL, Grune & Stratton, 1987, pp Vélez-Garcia E, Carpenter JT, Moore M, et al: Postsurgical adjuvant chemotherapy with or without radiotherapy in women with breast cancer and positive axillary nodes: A South-Eastern Cancer Study Group (SEG) Trial. Eur J Cancer 28A: , Ahmann DL, O Fallon JR, Scanlon PW, et al: A preliminary assessment of factors associated with recurrent disease in a surgical adjuvant clinical trial for patients with breast cancer with special emphasis on the aggressiveness of therapy. J Clin Oncol 5: , Martinez A, Ahmann D, O Fallon J, et al: An interim analysis of the randomized surgical adjuvant trial for patients with unfavorable breast cancer. Int J Radiat Oncol Biol Phys 10:106, 1984 (abstr) 43. Olson JE, Neuberg D, Pandya KJ, et al: The role of radiotherapy in the management of operable locally advanced breast carcinoma: Results of a randomized trial by the Eastern Cooperative Oncology Group. Cancer 79: , Arwidi A, Aspegren K, Augustsson N-E, et al: Postoperative radiation therapy in mammary carcinoma stage II: Target volume, organs at risk, absorbed dose, time-dose schedule, and dose to organs at risk in a prospective investigation.acta Radiol Oncol Radiat Phys Biol 18: , Ryden S, Ferno M, Moller T, et al: Long-term effects of adjuvant tamoxifen and/or radiotherapy: The South Sweden Breast Cancer Trial. Acta Oncol 31: , Tennvall-Nittby L, Tengrup I, Landberg T: The total incidence of loco-regional recurrence in a randomized trial of breast cancer TNM stage II. Acta Oncol 32: , Overgaard M, Christensen JJ, Johansen H: Evaluation of radiotherapy in high-risk breast cancer patients: Report from the Danish Breast Cancer Cooperative Group (DBCG 82) trial. Int J Radiat Oncol Biol Phys 19: , Mouridsen HT, Rose C, Overgaard M, et al: Adjuvant treatment of postmenopausal patients with high risk primary breast cancer: Results from the Danish adjuvant trials DBCG 77 C and DBCG 82 C. Acta Oncol 27: , Højris I, Overgaard M, Christensen JJ, et al: Morbidity and mortality of ischemic heart disease in 3083 high-risk breast cancer patients given adjuvant systemic treatment with or without postmastectomy irradiation. Radiother Oncol 48:S120, 1998 (suppl 1, abstr) 50. Haybittle JL, Brinkley D, Houghton J, et al: Postoperative radiotherapy and late mortality: Evidence from the Cancer Research Campaign trial for early breast cancer. BMJ 298: , Rutqvist LE, Lax I, Fornander MD, et al: Cardiovascular mortality in a randomized trial of adjuvant radiation therapy versus surgery alone in primary breast cancer. Int J Radiat Oncol Biol Phys 22: , Jadad AR, Cook DJ, Browman GP: A guide to interpreting discordant systematic reviews. CMAJ 156: , Fuller SA, Haybittle JL, Smith REA, et al: Cardiac doses in post-operative breast irradiation. Radiother Oncol 25:19-24, Jones JM, Ribeiro GG: Mortality patterns over 34 years of breast cancer patients in a clinical trial of post-operative radiotherapy. Clin Radiol 40: , Recht A, Come SE, Gelman RS, et al: Integration of conservative surgery, radiotherapy, and chemotherapy for the treatment of early-stage node-positive breast cancer: Sequencing, timing, and outcome. J Clin Oncol 9: , Buchholz TA, Austin-Seymour MM, Moe RE, et al: Effect of delay in radiation in the combined modality treatment of breast cancer. Int J Radiat Oncol Biol Phys 26:23-35, Stewart LA, Parmar MK: Meta-analysis of the literature or of individual patient data: Is there a difference? Lancet 341; , Oxman AD, Clarke MJ, Stewart LA: From science to practice: Meta-analyses using individual patient data are needed. JAMA 274: , Cuzick, Stewart H, Peto R, et al: Overview of randomized trials of postoperative adjuvant radiotherapy in breast cancer. Cancer Treat Rep 71:15-29, Hellman S: Natural history of small breast cancers. J Clin Oncol 12: , 1994