INTRODUCTION

About one third of patients with non-small-cell lung cancer (NSCLC) are estimated to have locally advanced, unresectable disease without evidence of extrathoracic dissemination.^1,2 Clinical trials conducted in the 1970s established the efficacy of radiation therapy (RT) in patients with locally advanced NSCLC. The optimal dose of RT was determined to be 60 Gy administered in 2 Gy/fraction during 6 weeks.³ Despite undergoing definitive radiation, patients experienced a high incidence of local and distant relapse. Trials then focused on the effect of adding chemotherapy to RT. Chemotherapy was thought to act in multiple ways that included sensitization of cancer cells to radiation, and also exhibited a cytotoxic effect on micrometastatic disease.

The value of adding induction therapy to RT was established by the Cancer and Leukemia Group B (CALGB) 8433 trial.⁴ The median survival for patients treated with induction chemotherapy of cisplatin plus vinblastine followed by radiation was 13.7 months, compared with 9.6 months for patients treated with RT alone (cisplatin 100 mg/m² on days 1 and 29 and vinblastine 5 mg/m² as a weekly intravenous bolus for 5 weeks). The survival probability at 5 years was 2.8 times greater for patients in the combination arm. Subsequently, the Radiation Therapy Oncology Group (RTOG) 88-08 trial⁵ validated CALGB 8433 by also concluding that RT plus chemotherapy is superior to RT alone.

Recent phase III trials^6,7 included concurrent chemoradiotherapy and administration of accelerated hyperfractionated radiation. In phase I and II trials,^8,9 radiation dose-escalation treatment using three-dimensional conformal external-beam techniques have shown a promising effect on survival, but this has yet to be confirmed in controlled studies. Despite modest improvements, local treatment failure still exceeds 80% when assessed by bronchoscopic biopsy, and distant metastases ultimately develop in the majority of patients.

Clinical and laboratory studies suggest that most epithelial tumors have a doubling time of less than 5 days.^10–12 In such rapidly proliferating tumors, prolonging total duration is detrimental because it results in accelerated tumor repopulation.¹³ Repopulation is hypothesized to be a factor that limits the success of conventional dose-escalation approaches. A European phase III randomized trial¹⁴ for locally advanced NSCLC patients tested continuous (no weekend breaks: the continuous hyperfractionated accelerated radiation therapy [CHART] regimen), hyperfractionated accelerated radiation therapy (HART; three treatments of 1.4 to 1.5 Gy, each separated by 6 hours), and accelerated (36 fractions during 12 days) schedules of RT versus standard schedules of radiotherapy (60 Gy in 6 weeks). Compared with the standard 60-Gy schedule, the CHART regimen produced a statistically significant 10% survival benefit at 2 years. The odds ratio in favor of the CHART regimen was 0.75, suggesting a 25% decline in the risk of death; nevertheless, the CHART regimen has not met with widespread clinical acceptance in the United States or Europe because of acute mucosal side effects, as well as the need for inpatient hospitalization and weekend treatments. In addition, chemotherapy was not included as a part of the treatment with CHART.

The phase II Eastern Cooperative Oncology Group (ECOG) protocol 4593¹⁵ tested a modified CHART regimen with weekends off (HART) and three fractions per day delivered during an 8-hour workday in an outpatient context. Thirty patients were treated with HART, delivered to a total dose of 57.6 Gy during 17 days with weekend breaks and 4-hour interfraction intervals. All patients required planning based on computed tomography (CT), with specific avoidance of the spinal cord in at least one field daily, and minimization of esophagus volume irradiated in all fields. No patient required interruption or discontinuation of treatment because of toxicity. Treatment-related toxicities greater than grade 3 included esophagitis in six patients. All acute toxicities resolved after therapy. In 28 assessable patients, the response rate within the radiation field was 64%. With a minimum follow-up of 19 months in surviving patients, the median and 1-year survival was 13 months and 57%, respectively.

On the basis of favorable phase III CHART data, as well as the results of the pilot phase II ECOG trial 4593, we conducted a phase III trial, ECOG 2597, to compare once-daily radiation therapy (qdRT) with HART after induction chemotherapy in locally advanced stage III NSCLC.

PATIENTS AND METHODS

Eligibility criteria included age older than 18 years; ECOG performance status of 0 or 1; biopsy-proven unresectable stage IIIA or IIIB NSCLC; bidimensional measurable disease; no pleural effusion on chest x-ray; no collapse of an entire lung; no prior malignancies in the preceding 5 years, with the exception of nonmelanoma skin cancer or carcinoma-in-situ of the cervix; no prior radiation or chemotherapy; no active peptic ulcer disease, esophageal reflux, or hiatal hernia; and consent to abstain from smoking during RT. Patients were ineligible if their tumor location was such that 100% of the cardiac volume would not receive more than 45 Gy, or if 50% or more of the cardiac volume would receive no more than 50 Gy. An informed consent procedure approved by an institutional review board and the National Cancer Institute was a prerequisite to enrollment.

The chemotherapy regimen consisted of two cycles of carboplatin (days 1 and 22) area under the time-concentration curve 6 mg/mL/min and paclitaxel 225 mg/m² during a 3-hour period on day 1 administered 3 weeks apart and delivered before radiotherapy. (Fig 1). Dose reductions were permitted for both hematologic and nonhematologic effects. Growth factor support was not routinely used, but was permitted as secondary prophylaxis.

After the completion of two cycles of chemotherapy, patients were reassessed with chest CT to ensure the absence of metastatic progression. In the absence of metastatic progression, patients were randomly assigned to one of two different radiotherapeutic regimens, with treatment to begin between days 43 and 50. In the qdRT arm, the total dose was 64 Gy in 32 fractions of 2 Gy each, delivered 5 days per week. For most patients, an initial anteroposterior field arrangement was used for approximately 40 Gy; this covered the primary tumor and all enlarged lymph nodes. In addition, elective nodal radiation of selected stations was allowed, based on tumor location and nodal status. Subsequently, this region received a total dose of 50 Gy, using either lateral or oblique portals, and a final cone-down boost increased the dose to 64 Gy for the tumor and all enlarged lymph nodes, with a 1- to 1.5-cm margin. Postchemotherapy CT scans were used for tumor definition. Electrons were permitted for treating the supraclavicular fossae only, and all photon energies had to be a minimum of 4 MV. CT-based treatment planning was recommended, but lung density corrections were not used; the prescription was to the isocenter and not to an isodose surface, and dose heterogeneity within the tumor was limited to 10%. A system of rapid port review was used and provided immediate feedback to the treating physician for therapy modification as needed. Standard dose limitations were used for normal tissues. In the experimental (HART) arm, the total dose was 57.6 Gy on the tid fractionation schedule (Fig 2).

Simulation and CT-based treatment planning were used, and esophageal contrast was used at simulation to define the location of the esophagus. Corrections for lung transmission were not used for dosimetric calculations. The minimum interval between fractions was 4 hours. The first and third fraction of each day consisted of anteroposterior-posteroanterior fields encompassing the primary tumor and draining lymphatics with a 1- to 1.5-cm margin; the fraction size for these fields was 1.5 Gy. The second fraction of each day used lateral or oblique photon fields, encompassed all gross disease (primary tumor and involved nodes) with a 1-cm margin, and excluded the spinal cord. The fraction was interdigitated between fraction 1 and fraction 3, and the fraction size was 1.8 Gy. Attempts were made to design the fraction 2 field to minimize the volume of esophagus treated without compromising the margin around tumor or spinal cord. Treatment began on a Monday and finished on the third Tuesday, for a total of 12 planned treatment days during 15 elapsed days.

Acute and late toxicities were scored using the National Cancer Institute Common Toxicity Criteria, version 2. The primary statistical objective was to assess survival, with toxicity, response rates, duration of response, and patterns of failure as secondary objectives. For the purposes of sample size calculation, the expected median survival for the control arm was assumed to be 14 months, based on CALGB 8433 and RTOG 88-08; with a total accrual of 294 patients, the study had an 87% power with a two-sided α = .05 for detecting a 50% improvement in median survival from 14 to 21 months. With a 5% ineligibility rate and a 10% progression rate before radiotherapy, accrual was projected at 338 patients. Overall survival was calculated from the date of enrollment to the date of death, or was censored at the time of last follow-up. Relapse-free survival was calculated from the date of enrollment to the date of death, documented relapse, or death, whichever occurred first, or was censored at the time of last follow-up.

Response to treatment was assessed using CT scans after two chemotherapy cycles and then again at least 4 weeks after completion of RT. A complete response was scored as resolution of all disease. A partial response was scored as a more than 50% reduction in the cumulative bidimensional tumor products for all sites of disease; stable disease constituted less than 50% reduction or less than 25% increase in this parameter, and progressive disease constituted more than 25% increase in this parameter and/or development of new foci of disease.

RESULTS

One hundred forty-one patients (in contrast to the necessary 338, or 42% of target) from 18 ECOG institutions were enrolled between June 1998 and June 2001. On the basis of prior ECOG NSCLC protocol accrual rates, we estimated a monthly accrual of 6.5 patients. However, the actual accrual rate was lower than anticipated (5.64 patients per month), leading to concern about timely trial completion. Ultimately, low patient accrual, the logistics of HART, mucosal toxicity, and early data from other trials demonstrating superior results with concurrent rather than sequential chemoradiotherapy led the data and safety monitoring committee to close ECOG 2597 early. The following results are based on data available in September 2004.

Of the 141 patients enrolled, eight were ineligible for the study and were excluded, primarily because of the presence of metastatic disease or prior chemotherapy. Of the 133 patients assigned to and eligible for induction chemotherapy, 119 (83% of all enrolled and 89.5% of all eligible patients) were randomly assigned to the two radiotherapy arms (59 and 60 patients to the qdRT and HART arms, respectively), indicating a patient loss of 14 of 133 or 10.5% (which is close to the predicted loss of 10%) secondary to disease progression, death, or patient refusal. Of these 119 patients, a total of seven patients (three receiving qdRT and four receiving HART) were found to be ineligible during the radiotherapy quality assurance process and were excluded, yielding an effective cohort of 56 in each arm. Therefore, the overall rate of patients eligible for random assignment was 79% of all enrolled patients, or an effective loss rate of 21%—slightly higher than the initially projected 15% loss. Baseline characteristics of the 112 patients randomly assigned to the qdRT and HART arms were similar and are listed in Table 1.

The median time to completion of RT was 51 elapsed days (range, 27 to 76 days) for qdRT, compared with 17 elapsed days (range, 17 to 33 days) for HART. RT completion rates were 53 patients (94.6%) for qdRT compared with 54 patients (96.4%) for HART. The reasons for noncompletion of RT in the five patients are as follows: progression in two patients receiving qdRT, and unknown reasons in one patient receiving qdRT and two patients receiving HART.

DISCUSSION

In the European Organisation for Research and Treatment of Cancer (EORTC) trial, comparison of standard RT to CHART resulted in a significant survival improvement (20% to 29% at 2 years) with CHART. The survival effect was more pronounced in patients with squamous cell carcinoma (SCC); however, it must be noted that 80% of the patients on the EORTC study had SCC, which suggests a good match between this treatment modality and SCC histology. Although the survival advantage was impressive, it was gained at the expense of increased early and late toxicity. The EORTC trial demonstrated that accelerated treatment provided a 23% reduction in risk of local progression, which provides some support for the hypothesis that accelerated hyperfractionation helps overcome accelerated tumor-cell repopulation.

In small-cell lung cancer, evidence from a phase III Intergroup trial¹⁶ shows that improved local control translates to a survival gain. Four hundred seventeen patients were randomly assigned to four cycles of cisplatin and etoposide plus radiotherapy beginning with the first cycle of chemotherapy. A 5-year survival rate of 26% compared with 16% was achieved on the once-daily schedule (45 Gy; the once-daily regimen was 5 weeks in duration and the twice-daily regimen was 3 weeks in duration). Both the EORTC trial in NSCLC and the Intergroup trial in small-cell lung cancer strongly suggest that shortening the RT treatment time improves outcome.

Recent in vivo experiments and clinical studies show that rapidly proliferating tumors (as determined by DNA content, S-phase fraction, and potential doubling times) are associated with inferior outcomes when treated with qdRT schedules. Biologic modeling in NSCLC^17,18 has shown that an increased biologically equivalent dose can be achieved by shortening the radiation delivery schedule and increasing the dose per fraction. A reanalysis of previously published RTOG data¹⁹ showed that the rate of proliferation of lung cancer is rapid. Of 397 lung cancer patients reanalyzed, 70 experienced treatment delays of > 5 and consequently had a probability of inferior survival as follows: without treatment delay, probability of 1-year survival was 37% compared with 1% with delay. Without treatment delay, probability of 3-year survival was 56% compared with 17% with delay. Median loss in survival probability was calculated to be 1.6% per day of delay beyond 6 weeks, resulting in the clonogenic doubling time of 3 to 3.5 days.

Acceleration of radiation treatment involves the delivery of the target dose in less time, and is analogous to the concept of dose-intensity in the delivery of cytotoxic chemotherapy. To prevent the excessive late tissue toxicities that result from single large daily fractions, multiple smaller fractions are used each day (hyperfractionation). The interval between multiple daily fractions (typically 4 to 8 hours) provides normal tissues with time to repair sublethal radiation damage, the half-life of which is estimated to be approximately 1.5 hours. Repair of sublethal injury in CNS tissue (primarily the spinal cord) appears to be somewhat longer, indicating the need for intervals more than 6 hours between radiation treatments that include the spinal cord, or a reduction in the total dose delivered to the spinal cord.

The accelerated radiation regimen CHART was evaluated for NSCLC patients in Europe, but the regimen has not come into widespread use because of the need for weekend treatment, a minimum 12-hour treatment day, higher esophagitis rates, and the use of three on-cord fractions per day. The HART regimen addresses these limitations with some success, perhaps with the exception of a slightly higher esophagitis rate (25% for HART v 16% for qdRT).

In ECOG 2597, we noted that although there was a substantial but not statistically significant improvement in survival with HART, there was no apparent corresponding improvement in local response rates. In fact, the response rates (22% v 25%) were substantially lower than might be expected from our own pilot ECOG 4593 trial,¹⁵ in which the response rate was 64%. Possible explanations for improvement in survival without improvement in response in ECOG 2597 include that the definition of best response did not differentiate local from systemic disease, so extrathoracic failure patterns may have clouded the response analysis; there was no central review of imaging studies; and HART treatment results in a more pronounced postirradiation radiographic effect, confounding response/progression with superimposed postirradiation changes.

In the 1990s, two large randomized trials^6,7 established chemotherapy with concurrent radiation as superior to sequential administration of chemotherapy and radiation (Table 5) but at the expense of increased local toxicity. The West Japan Lung Cancer Group⁶ randomly assigned patients (age younger than 75 years; ECOG performance status 0, 1, or 2) to undergo either concurrent chemoradiotherapy (split course) or sequential chemotherapy followed by radiation. The concurrent arm had a superior response rate (84% v 66%; P = .0002) and median survival (16.5 v 13.3 months). Distant relapse occurred in a majority of the patients in both groups. The RTOG 9410 trial was a three-arm, randomized, phase III trial that compared sequential with concurrent chemoradiotherapy. Sequential therapy (arm 1) consisted of two cycles of cisplatin and vinblastine followed by 60 Gy external-beam radiation administered in once-daily fractions beginning on day 50. In arm 2, the same chemotherapy was administered, whereas radiation was administered concurrently starting on day 1 of the chemotherapy cycle. Arm 3 used hyperfractionated RT with 69.2 Gy administered as twice-daily fractions. Patients in arm 3 received concurrent chemotherapy of cisplatin and oral etoposide. A total of 611 patients with unresected stage II/III NSCLC were enrolled onto the study. The median survival was superior (17.0 v 14.6 months; P = .038) for patients with concurrent chemoradiotherapy (arm 2) compared with sequential therapy (arm 1). Again, the incidence of acute toxicity was higher in the concurrent arms. In ECOG 2597, the outcome with the sequential chemotherapy followed by HART appears to be comparable or possibly even higher than the concurrent approaches. The overall decrease in distant failure in the HART arm is intriguing: in addition to a survival benefit, HART also offers the convenience of shorter treatment duration (2.5 instead of 7 weeks). Subsequent trials of the HART regimen will need to address the issues that led to early closure of this trial, including slow accrual, logistics of HART, mucosal toxicity, and the fact that concurrent chemoradiotherapy appears to be more effective than sequential.

We hypothesize that the use of a brief chemotherapy schedule followed by accelerated radiotherapy works by altering tumor cell kinetics so that the confounding effect of accelerated repopulation is lessened, resulting in a survival gain. In fact, with recent data suggesting that epithelial neoplasms such as head and neck cancer be radiosensitized through epidermal growth factor receptor inhibitory strategies, the HART approach in combination with epidermal growth factor receptor inhibitors may prove to be even better. Although ECOG 2597 did not meet its accrual goals and could not provide statistical significance, the survival trend (20.3 months for HART v 14.9 months for qdRT) suggests that additional exploration of the HART treatment strategy is warranted.

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. Consultant/Advisory Role: Chandra P. Belani, Bristol-Myers Squibb; Henry Wagner, Eli Lilly. Honoraria: Chandra P. Belani, Bristol-Myers Squibb; Henry Wagner, Genentech. Research Funding: David Johnson, ECOG. For a detailed description of these categories, 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.