Abstract
Background/Aim: The optimal radiotherapy dose for localized esophageal squamous cell carcinoma (ESqCC) patients treated with definitive concurrent chemo-radiotherapy (CCRT) is debated. The aim of our study was to compare patient outcomes using either standard or high radiotherapy dose. Materials and Methods: Eligible patients diagnosed between 2011 and 2015 from the cancer registry of our Institute were identified and a propensity score (PS)-matched cohort (1:1 for high vs. standard dose) was constructed to balance observable potential confounders (including organ at risk dose). The hazard ratio (HR) of death between high and standard dose was compared. Results: Our study population included 73/36 patients before/after PS matching. The HR of death at the high dose compared to the standard dose was 0.554 (95% confidence interval (CI)=0.308-0.998, p=0.049). Conclusion: Definitive CCRT using a high radiotherapy dose showed improved survival outcomes for localized ESqCC patients compared to standard dose.
Esophageal cancer is a major cause of cancer-related mortality worldwide (1, 2). Squamous cell carcinoma is the predominant malignancy in Asia whereas adenocarcinoma is more common in the western countries (2, 3). For late-stage localized esophageal squamous cell carcinoma (ESqCC), definitive concurrent chemoradiotherapy (CCRT) has been suggested as one of the standards-of-care treatments for many years in the North American, European, and Asian guidelines (4-8).
The optimal radiotherapy dose is a hot debate since the publication of INT-0123, a randomized controlled trial (RCT) (9). For example, strictly 50-50.4 Gy were suggested in the North American guideline (5), whereas up to 60 Gy had been mentioned in the European or Asian guidelines (7, 8). Due to the lack of a new RCT as stated in a systematic review published in 2015 (10) as well as in a review paper published in 2018 (11), many retrospective studies have been performed to address this issue as cited in the above 2018 review (12-14).
In order to appropriately compare patients treated with standard dose with those treated with high dose, patients treated with the standard dose because of organ at risk (OAR) dose constraint have to be excluded (i.e., they were not given high dose due to already high OAR dose even at standard dose). However, these dosimetric criteria were not considered in the above retrospective studies (12-14). Therefore, our study aimed to compare the outcomes of localized ESqCC patients treated with definitive CCRT using either standard or high radiotherapy dose, while controlling for covariables including OAR dose.
STROBE study flowchart and number of individuals at each stage of the study. 1Seventh American Joint Committee on Cancer, T1N0 excluded; 250 Gy [standard RT dose] vs. 60 Gy [high RT dose] in 1.8-2 Gy/fraction, ±5% in dose; also, critical organ dose within inclusion criteria; 3Regarding overall survival (according to death registry).
Materials and Methods
Study population and study design. In this retrospective study, clinical stage I-III (excluding T1N0) (15) unresected esophageal squamous cell carcinoma adult (age ≥18 years old) patients found in the cancer registry of our institute between 2011–2015 were included. All patients received definitive concurrent chemoradiotherapy with either standard (50 Gy +/- 5%) or high (60 Gy +/- 5%) dose radiotherapy (only via external beam), both in conventional fractionation. In order to make these two groups comparable in the OAR dose, the expected mean lung dose and spinal cord maximal dose were estimated if the radiotherapy plan was delivered for either 50 Gy (eMLD50 & eSCD50) or 60 Gy (eMLD60 & eSCD60) for each patient via reviewing RT planning record. Patients with eMLD60 >20 Gy or eSCD60 >50 Gy were excluded. Patients with previous cancer(s) or positron emission tomography (PET) not used in staging were also excluded. The date of diagnosis was adopted as the index date. The explanatory variable of interest [standard or high radiotherapy (RT) dose] was determined based on the cancer registry and decided the outcomes [overall survival (OS), progression-free survival (PFS), and esophageal cancer specific survival (ECSS)] using the recording in the cancer registry and the linkage with death registration. Then, potential confounders (see next section) were considered and propensity-score (PS) matched samples were constructed using various PS estimation methods [logistic regression (LR) and machine learning methods, including neural network (NN) & random forest (RF)], to evaluate the effectiveness of high RT dose (vs. standard RT dose). This study was approved by the research ethics committee of our institute (CMUH104-REC3-087).
Other explanatory covariables. For adjustment of potential non-randomized treatment selection, other covariables were collected as confounders, including patient demographic factors [age, gender, weight, smoking status], disease characteristics [clinical T-stage, N-stage, overall stage, maximum standardized uptake value (SUVmax) in staging PET, gross target volume (GTV)], and treatment [RT delivery, the use of peri-CCRT systemic therapy, RT break, eMLD50, eMLD60, eSCD50, and eSCD60]. These covariables were selected and modified by our experience in clinical practice and Taiwan Cancer Registry/National Health Insurance related studies (16-20), and were defined as the followings. Clinical stage was classified as T1-T3 vs. T4 for T-stage, N0-N1 vs. N2-N3 for N-stage and I-II vs. III for the overall stage. RT delivery was classified as image-guided radiotherapy (IGRT) or not. Peri-CCRT systemic therapy was classified as yes or no. The RT break interval was calculated by the exact RT duration (week) minus the expected RT duration (by 5 fractions per week) and classified as >1 week or ≤1 week. As mentioned in the literature, SUVmax was classified as ≥5.6 or <5.6 (21) and GTV was categorized as ≥27 ml or not (22).
Standardized difference (SDif) via various methods (LR_SAS, LR_R, NN_SAS, and RF_SAS). LR_SAS: Propensity score (PS) estimated by logistic regression and PS matched by SAS software; LR_R: PS estimated by logistic regression and PS matched by R software; NN_SAS: PS estimated by neuralnet and PS matched by SAS software; RF_SAS: PS estimated by random forest and PS matched by SAS software; eMLD50: expected mean lung dose if radiotherapy delivered at 50 Gy; eMLD60: expected mean lung dose if radiotherapy delivered at 60 Gy; eSCD50: expected spinal cord maximal dose if radiotherapy delivered at 50 Gy; eSCD60: expected spinal cord maximal dose if radiotherapy delivered at 60 Gy; GTV: gross target volume; IGRT: image-guided radiotherapy; peri-CCRT: peri-concurrent chemoradiotherapy; RT: radiotherapy; SUVmax: maximum standardized uptake value.
Statistical analysis. The software SAS 9.4 (SAS Institute, Cary, NC, USA) and R (R Development Core Team, R Foundation for Statistical Computing, Vienna, Austria) version 3.5.1 were used for statistical analyses. The propensity score method was used as advocated in the literature to balance the measured potential confounders (23, 24). The above covariables were used in the PS model construction using various methods to estimate the possible PS value as suggested in the literature (25-28) and then PS matching was performed (PSM, 1:1 paired matching). Finally, the balance of covariate was assessed via standardized difference (SDif) as suggested in several review papers (25, 29, 30). The hazard ratio (HR) of death for OS, PFS and ECSS was compared using Cox proportional hazards model with a robust variance estimator (24). The survival rates were obtained from the death registry (follow-up until December 31, 2017 or death). The E-value was also calculated as suggested in the literature (31) to evaluate the potential impact of potentially unmeasured confounder(s).
Results
Identification of the study population. Our study flow chart as suggested by the STROBE guideline (32) is depicted in Figure 1. The identified initial study population consisted of 73 clinical stage II-III unresected ESqCC adult patients receiving definitive CCRT (all treated with intensity-modulated radiotherapy) using either standard or high RT dose. PS values were estimated using LR, NN, RF and then PSM using SAS or R was performed. Figure 2 shows the distributions of SDif for each of the covariates applying various methods. About half of the covariables could not be moderately balanced (i.e., SDif ≤0.25) (29) for PS estimated by NN or RF methods whereas most covariables (except age) could be moderately balanced for PS estimated by the LR method. PS estimated by the LR method and PSM estimated using R (optmatch package) could achieve better covariable balance when compared to PSM using SAS (SDif≤0.25 for all covariables in PSM using R, and in PSM using R out performed PSM using SAS in 6 covariables while SAS outperformed R in 3 covariables). Therefore, the best approach was adopted (PS by LR and PSM by R) for the analysis and included 36 patients as the final study population (Table I).
Characteristics of unmatched and matched study population.
Outcomes. After a median follow-up of ten months (range=2-82 months), 32 deaths were observed (15 & 17 for high RT dose & standard RT dose groups, respectively). There was statistically significant difference when high RT dose was compared to standard RT dose (HR for death 0.554, 95%CI=0.308-0.998, p=0.049). The observed HR of 0.554 could be explained away by an unmeasured confounder that was associated with both the selection of treatments and the live/death risk ratio of 2.187 (E-value) fold each, but weaker confounding could not do so. The Kaplan–Meier OS curve is shown in Figure 3. The results were not statistically significant for PFS (HR 0.648, 95%CI=0.346-1.214, p=0.176) and ECSS (HR=0.581, 95%CI=0.317-1.064, p=0.079).
Kaplan–Meier overall survival curve (in years).
Discussion
In this single institution retrospective study, the survival outcome of localized ESqCC patients treated with definitive CCRT using high radiotherapy dose was found to be better compared to that using standard dose, even after controlling for covariables including OAR dose.
Our results were comparable with the studies (12-14) mentioned in the 2018 review paper (11), as well as our previous population-based study (17) in that high dose was associated with better outcomes. However, the results of the current study might be more reliable because they were controlled for OAR dose which was not considered in the abovementioned previous studies. However, our current study was still a retrospective study, and a recent RCT (NCT01937208) published as a conference paper reported no statistically significant difference (33). Therefore, the full paper of this RCT is eagerly awaited.
There were several limitations to our study. Firstly, as a non-randomized study, was prone to potential unmeasured confounders although care was taken to include potential confounders like OAR dose which were not addressed by the studies mentioned above. Secondly, the impact of salvage treatment was not investigated (34). Finally, the sample size of our study was small.
Therefore, the interpretation of the results of our study is not definitive, but should rather be viewed as supplementary while waiting for ongoing phase 3 trials. In addition to the above RCT (NCT01937208), only one trial (NCT02556762) was found that compares simultaneous modulated accelerated boost versus standard dose by searching https://clinicaltrials.gov/ in Nov 2018 using keywords “Esophageal Cancer | concurrent chemoradiotherapy|Phase 3”.
In conclusion, definitive CCRT using a high radiotherapy dose showed improved survival outcomes for localized ESqCC patients compared to standard dose, even after controlling for covariables including OAR dose.
Acknowledgements
This work was partly supported by Ministry of Science and Technology (MOST 107-2314-B-039-026-) and China Medical University Hospital (DMR-108-054). The corresponding Author would like to thank Dr. Ya Chen Tina Shih for her mentoring in health services research.
Footnotes
↵* These Authors contributed equally to this study.
Conflicts of Interest
The Authors declare no conflicts of interest regarding this study.
- Received November 22, 2018.
- Revision received December 9, 2018.
- Accepted December 10, 2018.
- Copyright© 2019, International Institute of Anticancer Research (Dr. George J. Delinasios), All rights reserved
