Lung cancer is the leading cause of cancer-related deaths for both men and women worldwide (1) and is a serious and growing health problem in China, with approximately 600,000 deaths each year. The cancer rate has been rising in China since the early 1980s because of the increasing smoking population (2). Thoracic radiotherapy (RT) remains the standard and widely utilized treatment modality for inoperable, locally advanced non-small cell lung cancer (LA-NSCLC).

Randomized studies have demonstrated that concurrent chemoradiotherapy (CCRT) yielded better survival rates than the sequential approach but with higher treatment toxicities (3, 4). Besides efficacy, treatment toxicity does influence the treatment of choice by oncologists and patients (3, 4). Comprehensive literature reviews have shown that toxicity profile is more favorable in low-dose CCRT, but low-dose CCRT has not been widely adapted worldwide (5). In addition, in the era of CCRT, the chest tumor control rate has been disappointing, averaging less than 50% by radiographic criteria. Uncontrolled chest tumors (including both the primary and hilar/mediastinal lymph nodes) represent a major barrier to further therapeutic gain, as surviving cancer cells in residual chest tumors can seed distant metastases. In this context, the significance and the impact of chest local–regional tumor control for LA-NSCLC on survival outcome has been analyzed in 1,390 patients treated in 7 legacy Radiation Therapy Oncology Group (RTOG) trials by CCRT for LA-NSCLC (6). This analysis revealed that the chest local–regional tumor control rate at 3 years was 38% at best and only 14% using a more strict definition (7). This analysis did find a powerful association between chest local–regional tumor control and overall survival (OS), which was statistically significant (p = 0.0001). Such finding highlights the importance of chest tumor control on the survival outcome of LA-NSCLC.

Optimal CCRT in maximizing chest tumor control and minimizing toxicities remains to be defined. Taxane-based CCRT has been favored for the treatment of LA-NSCLC by many oncologists worldwide. Taxanes are known to have antiangiogenic effects on tumor vasculature (8, 9). These are also cell cycle-specific chemotherapeutic agents that can cause cytokinetic stabilization of the spindle microtubules leading to apoptotic cell death (10, 11). The cell cycle effect of paclitaxel on the arrest in the G2/M phase, the most radiosensitive phase, makes it an ideal radiation sensitizer to enhance RT effects (12). In the U.S., weekly paclitaxel-based CCRT has been published in many phase 1, 2, and 3 clinical trials for CCRT, and has been shown to be better tolerated than cisplatin-based CCRT (13, 14).

Weekly low-dose carboplatin and paclitaxel with concurrent radiation combination is a widely accepted standard in the U.S. and Europe for LA-NSCLC. Despite common practice, the optimal dose-schedule of paclitaxel for CCRT treatment of LA-NSCLC remains to be defined. Previously, a phase 1 and 2 clinical study was conducted using pulsed, low-dose paclitaxel CCRT for LA-NSCLC (15, 16). Based on preclinical investigations, it was hypothesized that a regimen of schedule-dependent, pulsed, low-dose paclitaxel CCRT could yield effective radiosensitization to improve chest tumor control. This clinical trial has reported promising tumor response rates and a high rate of infield tumor control with low toxicities (15–17). We also conducted a phase 1 clinical study testing this hypothesis-based study design in China (18). Following up the phase 1 study, we conducted a multicenter randomized phase 3 study comparing two low-dose sensitizing paclitaxel schedules of CCRT for LA-NSCLC. We hypothesized that a schedule-dependent, pulsed low-dose paclitaxel (15 mg/m², three times per week) CCRT regimen would yield better recurrence-free survival (RFS) than the commonly applied weekly paclitaxel (45 mg/m²) radiation regimen due to more effective radiosensitization.

Low-dose chemotherapy CCRT for LA-NSCLC has previously shown promising results with low toxicities by different investigators (15–17, 21–23). Among these, the best known is the randomized phase 3 study by Radiotherapy and Lung Cancer Cooperative Groups of the European Organization for Research and Treatment of Cancer, which compared very low-dose (6 mg/m²) daily cisplatin CCRT vs. weekly cisplatin (30 mg/m²) CCRT vs. RT alone for LA-NSCLC (21). Results demonstrated better survival of patients receiving daily low-dose cisplatin and RT than patients who received RT alone or weekly cisplatin. Such low doses of daily cisplatin in this context were thought to be the radiosensitizing effect of low-dose cisplatin in improving RT chest tumor control, which resulted in better survival while avoiding the toxicities of high-dose chemotherapy.

However, low-dose CCRT for LA-NSCLC has not been widely adapted, while poor chest tumor control of LA-NSCLC remains a major challenge for oncologists. Several single institution phase 1 and 2 studies have attempted RT dose-escalation to 74 Gy to improve chest tumor control and showed promising chest tumor control and survival (24–27). However, results of a phase 3 study (RTOG 0617) testing RT dose-escalation to 74 Gy vs. standard 60 Gy was disappointing (28). The median OS was 28.7 months (95% CI 24.1–36.9) for the 60 Gy arm and 20.3 months (17.7–25.0) for the high-dose 74 Gy arm (HR 1.38, 95% CI 1.09–1.76; p = 0.004). The exact reasons for the worse survival outcome in the 74 Gy arm remain to be elucidated. Nevertheless, the outcome of RTOG 0617 highlights the unresolved challenge in improving chest tumor control, and the need for novel strategies to enhance RT efficacy in the combined modality treatment of LA-NSCLC.

Pulsed low-dose paclitaxel was previously reported to be an effective radiosensitizing strategy (15–17). Chen et al. (17) reported that the kinetics of tumor shrinkage after pulsed low-dose paclitaxel CCRT were different from the standard conventional CCRT schedules, in that rapid tumor shrinkage occurred within 1 month after completion of CCRT, which is in contrast to the gradual tumor regression over a 5- to 6-month interval commonly observed (29). In addition to the rapid initial tumor responses, the infield (radiation port) tumor control was above 97% at 3 years. The exact mechanism of such rapid and effective tumor response and durable infield tumor control of pulsed low-dose paclitaxel CCRT is unknown, but one can theoretically attribute to the combined effects of paclitaxel G2/M cell cycle effect (9), its apoptotic effect (10, 11), tumor reoxygenation effect (30), and/or the antiangiogenic effect of low-dose chemotherapy delivered through metronomic dose schedules (31, 32). In the same context, we acknowledge that because the tumor response of standard treatment schedules is a gradual process, choosing any time point after CCRT to assess response rate is somewhat arbitrary and may be in favor of Arm 1. We chose the 1-month time point with the understanding of this issue. However, the RFS analyzed by the Kaplan–Meier method provided complementary information that was clinically more important than response rates, and Kaplan–Meier analysis would have accounted for the changes over a long period of time for both local tumor control and distant tumor control.

Our study reported significantly better tumor response rate, better RFS, and lower rates of Grade 3 and 4 leukopenia and neutropenia in the pulsed low-dose paclitaxel arm. The low rate of toxicities is consistent with the literature review on low-dose CCRT (5). We neither observed a significant difference in the patterns of failure of the two arms nor did we observe a significant difference in esophagitis or pneumonitis or other radiation-related acute side effects in the two arms. One may wonder why the better RFS in Arm 1 did not result in better OS, and if treatment-related toxicities may contribute to the lack of trend toward improving OS of Arm 1. We reviewed the Grade 5 toxicities of both arms, and there was only one Grade 5 lung toxicity in Arm 1 and two in Arm 2, thus Grade 5 toxicity did not contribute to the lack of OS benefit. Much larger sample sizes would have been necessary to reach any statistical power to detect the difference in OS.

We acknowledge that there are potential weaknesses of our study. The random number method is an assignment process in which all patients were assigned to either Group A or B completely at random, without monitoring the size of the groups. Our study design had planned to screen 134 patients with 67 randomized in each arm and to allow a 10% drop out rate. This design would have resulted in an estimated 60 patients in each arm for the final data analyses. Due to the method of random number table for the assignment of randomization, our study enrolled 74 patients in Arm 1 and 60 patients in Arm 2. In retrospect, we acknowledge that it would have been better to have a built-in process to balance assignments between the arms. However, despite the minor imbalance in randomization, the study resulted in sufficient number of cases in each arm for the study endpoint analyses (71 in Arm 1 and 59 in Arm 2), and there was no statistically significant imbalance of patient characteristics in the two arms (Table 2).

We chose to analyze our data based on patients completing the treatment regimen in the assigned arm instead of basing on the intention to treat (ITT). We think that final analyses of patients completing the treatment regimen in each arm have served the goal of our study better. The benefit of analysis based on ITT is to avoid effects from dropout and/or crossover, which may break the random assignment to the treatment groups in a study. Our study had very low dropout rates in both arms and did not allow for crossover, thus basing ITT for the analysis would not have made much difference.

Current practice standard is CCRT upfront for patients with good performance status, while induction chemotherapy followed by RT has been reserved for patients who cannot tolerate CCRT (33). A randomized phase 3 study did not demonstrate any benefit of adding induction chemotherapy to CCRT (34), and there is the lack of prior studies on induction chemotherapy followed by very low-dose CCRT for LA-NSCLC. We acknowledge that our study design is different from the current standard CCRT treatment regimens for LA-NSCLC due to logistical reasons and pattern of referrals in our region. Patients enrolled in our study had received 2–4 cycles of induction chemotherapy prior to enrollment. Because our primary goal was to assess low-dose chemotherapy in sensitizing chest RT to improve chest tumor responses, we designed the enrollment of patients to include the 2–4 cycles of cisplatin-based induction chemotherapy upfront, with the assumption that as long as all patients had received 2–4 cycles of cisplatin containing induction chemotherapy, the potential systemic effect on distant micrometastasis from induction chemotherapy would balance both arms. We acknowledge that it would have been helpful if the study were designed to include the induction chemotherapy followed by CCRT and had assessed tumor response to induction chemotherapy, but this was not feasible due to local practice and referral pattern.

In conclusion, our study showed that three times weekly paclitaxel (15 mg/m²) yielded better RFS and tumor responses, with lower hematologic toxicity than the weekly paclitaxel (45 mg/m²) schedule for CCRT in patients with LA-NSCLC who have received 2–4 cycles of induction chemotherapy. Based on this study, the survival of both arms was equivalent, but the low-dose arm experienced less toxicity. Whether low-dose paclitaxel CCRT translates to a survival benefit needs to be tested in a larger randomized study. In an era when maintaining quality of life during cancer therapy is important to patients, a regimen with low toxicity and better tumor control will be appropriate for those who cannot tolerate intense CCRT for LA-NSCLC.

This study was carried out in accordance with the recommendations of the institutional research subject review boards of the participating hospitals, with written informed consent from all subjects. All subjects gave written informed consent in accordance with the Declaration of Helsinki.

HL: study design, patient enrollment, data collection and analysis, data interpretation, and drafting of manuscript; YC: study design, data examination, data analysis, major revision of the manuscript, and final critique of the manuscript; AS: patient enrollment, data collection, and data analysis; KP: provided input on chemotherapy, reviewed data, and review of manuscript; RY: patient enrollment and collection of data; YY: statistics; JL, HL, YW, TX, LF, HM, and JG: patient enrollment and collection of data; and GZ: study design, data analysis, and finalization of manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.