Intensity-modulated versus 3-dimensional conformal radiotherapy in the definitive treatment of esophageal cancer: comparison of outcomes and acute toxicity

Though the current standard of care for locally advanced esophageal cancer (EC) is neoadjuvant chemoradiotherapy (nCRT) followed by resection [1, 2], the latter is also associated with surgical morbidity and mortality. Two randomized trials have addressed nCRT versus definitive CRT (dCRT); despite local control (LC) benefits with surgery, no differences were discerned in distant metastasis or survival [3, 4]. Thus, dCRT is performed – and is indeed a viable recommendation – for select ECs [1, 5, 6].

However, virtually all seminal clinical trials for both nCRT and dCRT have utilized three-dimensional conformal radiotherapy (3DCRT) [2‐4, 7, 8]. The advanced and highly conformal technology of intensity-modulated radiotherapy (IMRT) has heretofore been studied only in limited capacities [9‐13]. It remains an attractive modality largely owing to the intimate anatomic relationship of the esophagus with both lungs and the heart. Indeed, several publications have posited a positive correlation between radiation dose to the lungs [14‐18] and heart [19‐22], with the risk of toxicities in these structures. Hence, the high conformality of IMRT could theoretically afford fewer toxicities to these organs-at-risk (OARs), especially in the setting of relative dose-escalation with dCRT, but supportive data are currently lacking. The aim of this study was to compare outcomes and acute toxicity of EC patients receiving definitive dCRT with either 3DCRT or IMRT.

This single-institutional study was approved by the university institutional review board and ethics committee, and patient/treatment/outcomes were retrospectively collected from an institutional database. Herein, we evaluated 93 patients (2000–2012) treated definitively for locally advanced EC. Specifically, inclusion criteria were histologically-proven esophageal squamous cell or adenocarcinoma, Karnofsky performance status (KPS) ≥70, curative treatment intent, systemic workup negative for distant metastases, and receipt of concomitant CRT. Patients were excluded if receiving <50 Gy definitively, as per prior studies [2]. The choice of radiation modality (IMRT or 3DCRT) was made at the radiation oncologist’s discretion, predominantly influenced by institutional guidelines and availability at the time of treatment planning.

All patients underwent systemic workup with referring providers prior to commencing CRT and a CT for radiation treatment planning. The gross tumor was first contoured, followed by a 4–5 cm craniocaudal margin (0.5-1 cm radially) and inclusion of locoregional lymphatic pathways to form the clinical target volume (CTV); another isotropic margin 0.5-1 cm was added to construct the planning target volume (PTV). If a boost was applied, the boost CTV was defined by adding a margin of 1.5-2 cm craniocaudally and 0.5-1 cm radially. The boost CTV-to-PTV margin was 0.5-1 cm in any direction. 3DCRT utilized a 3-5-field beam arrangement. IMRT was planned using either a step-and-shoot design or helical application mode. Boosts were either sequential or simultaneous. Image guidance was utilized in more contemporary time periods.

Chemotherapy was administered concomitantly with RT in all patients, and consisted of cisplatin 20 mg/m² + 5-FU 1000 mg/m² over 5 days at weeks 1, 5, 9, and 13. While receiving CRT, patients were monitored weekly on treatment with toxicity assessments, and at regular follow-up intervals with imaging.

Statistical analyses were performed using R (version 3.0.2, R Development Core Team, 2013, URL: http://​www.​r-project.​org/) combined with the packages ‘splines’, ‘survival’ (version 2.37-7, Therneau, 2014) and ‘xtable’ (version 1.7-1, Dahl, 2013). Tests were two-sided with p < 0.05 considered significant. Since this was an exploratory trial, all p-values are interpreted descriptively and no adjustment for multiple testing is applied. Chi-squared test assessed measures of association in frequency tables, and the t-test evaluated the equality of population distributions. Survival analysis was done using Kaplan-Meier methodology. A local relapse (LR) was defined as the locally recurrent disease including a biopsy-proven or radiologically-suspicious lesion in close proximity to the initial disease. Progression-free survival (PFS) was defined as the time from last oncologic treatment to death, local recurrence or distant metastasis, whichever occurred first. Overall survival (OS) referred to the time interval between initial diagnosis to death from any cause, with censorship based on particular follow-up times. The Cox proportional hazards model was used for univariate analysis (UVA) and multivariate analysis (MVA) to assess the effect of several clinical factors on LR, PFS and OS. The Wald test was used to assess the predictive value of the covariates in the model.

Toxicities were assigned at the time of initial occurrence by the treating physician according to the Common Terminology Criteria for Adverse Events v4.0, and retrospectively reviewed for purposes of this study.

Median and mean follow-up for all patients was 20.1 months and 34.7 months, respectively (range, 3 – 150.1 months). Table 1 displays clinical characteristics of the patient population. Of note, differences between groups included a more proximally localized upper tumor border (23.7 vs. 26.4 cm from the dental arcade, p = 0.031) and a higher mean cumulative radiation dose (57.2 vs. 55.4 Gy, p = 0.01) for the IMRT group.

Figure 1 illustrates results of local relapse rate (LR), OS and PFS for both groups. The 1- and 3-year LR was 20.4% and 28.6% in the 3DCRT group and 15.9% and 22.7% in the IMRT group, respectively (p = 0.620). Median PFS in the respective groups were 13.8 and 16.6 months, which corresponded to 3- and 5-year PFS of 32.4% and 21% compared to 37.4% and 31.6%, respectively (p = 0.448). Median OS in the respective cohorts were 18.4 and 42.0 months, which corresponded to 3- and 5-year OS of 35.5% and 23.9% compared to 50.9% and 38.9%, respectively (p = 0.198).

Univariate analysis (UVA) determined factors associated with LR, PFS, and OS (Table 2). None of the investigated factors influenced OS or PFS with statistically significant impact. However, there was a trend for worse OS and PFS for high-grade disease (p = 0.052 and p = 0.089, respectively) and for improved PFS with higher pre-treatment hemoglobin levels (p = 0.057). Higher cumulative RT doses beneficially impacted LR on UVA (p = 0.004). Multivariate analysis (MVA) included the factors RT technique, KPS, T-stage, N-stage, Grading, tumor histopathology and RT dose. Besides the confirmation of a beneficial impact of increased RT doses on LR (HR 0.686, 95%-CI 0.524 – 0.897; p = 0.006) and a borderline significant worse OS for high-grade tumors (HR 1.743, 95%-CI 1.003 – 3.029; p = 0.049) MVA did not reveal any further significant findings.

Acute, radiation-related toxicities experienced by both cohorts are displayed in Table 3. These predominantly included dysphagia, radiodermatitis and nausea/vomiting. Of note, dysphagia was observed in 70.5% for patients in the IMRT group vs. 42.9% in the 3DCRT group (p = 0.067). Owing to a short follow-up period and assumed issues with data integrity regarding late toxicity reporting in the cohorts, late toxicities were unable to be reliably presented herein.

In terms of dosimetric analysis, detailed treatment plan information was available in 76 patients. Basic dosimetric parameters of 3DCRT and IMRT plans are shown in Table 4. A trend towards larger PTVs was observed for IMRT (p = 0.060). There were no significant differences in normal tissue exposure between treatment techniques.

In our study cohort, the choice of radiation technique (3DCRT or IMRT) did not significantly influence survival in definitive chemoradiotherapy for esophageal cancer patients. These results correspond to previous findings [9, 10], yet contradictory results have been presented as well [13]. Nevertheless, the value of such an analysis is not diminished, because there is still a lack of data regarding the potential value of IMRT in the definitive, relatively dose-escalated setting. This will undoubtedly be an important issue in the coming years as the use of IMRT for EC increases [23].

The novelty and emphasis of our data were aimed at detecting differences in outcome endpoints in the setting of relative dose-escalation (in the definitive setting). Although median OS was 42.0 months with IMRT compared to 18.4 months with 3DCRT, this difference did not reach statistical significance. However, LR was significantly increased by higher cumulative radiation doses. Interestingly, radiation modality did not show a noticeable impact on LR though confounder analysis revealed a significantly higher mean cumulative RT dose in the IMRT group. We suppose that this difference between both groups, namely 1.8 Gy (3DCRT 55.4 Gy vs. IMRT 57.2 Gy, p = 0.01), is too small to translate into a considerable outcome difference of radiation techniques. However, our data reveal a dose-relationship resulting in an increased local tumor control with higher radiation doses. This finding is in line with other investigations that have noted outcome enhancements for dose-escalation [24]. In contrast, the RTOG 9405 trial did not discern advantages from dose-escalation [25], albeit with now-outdated techniques. In summary, our data support the hypothesis that increased radiation doses improve patient outcomes; however, the potential benefit of dose-escalation in EC remains inconclusive due to the heterogeneity of evidence [26].

Nevertheless, there is little debate that IMRT theoretically allows for safer dose-escalation. Dosimetric investigations have determined that advanced IMRT techniques provide numerical advantages over 3DCRT, but without outcome differences [27, 28]. This particularly applies to a reduction of high dose exposure to the OAR. For example, Münch et al. have demonstrated a significant reduction of the lung V_30Gy with VMAT (6.6%) compared to 3DCRT (11%) and of the heart V_30Gy (17.7% with VMAT vs. 50.4% with 3DCRT) [27]. In contrast, our study failed to demonstrate any significant differences in OAR dose exposure. However, slightly lower doses to the spinal cord and the heart were observed for IMRT and MLD was equal despite a larger PTV volume by trend of IMRT patients. Of note, the nature of IMRT and arc therapy results in increased low-dose areas to the heart and lungs, the ramifications of which are also unclear presently. In the study of Münch et al., the V_5Gy of the lungs and the heart were 90.1% and 100% with VMAT compared to 79.7% and 91% with 3DCRT, respectively [27]. Using rotational techniques, low-dose areas to the lung have been higher in some studies than the recommended limits in other (lung cancer) studies [13, 29]. It cannot go unmentioned, however, that inverse-planned therapies (e.g. IMRT, arc therapy, etc.) are heavily dependent on nuances of treatment planning, technique, delivery, and optimization, which can lead to a variety of potential results [30]. The use of helical IMRT herein has heretofore never been reported as a modality of IMRT delivery, and could be of further interest for future investigations.

Though there is a knowledge gap regarding the role of IMRT in EC, compelling data suggest that IMRT and other advanced techniques may decrease postoperative complications when used in the setting of nCRT, thus potentially decreasing surgical morbidity and mortality [17]. Notably, Lin et al. analyzed multiple cancer registries and performed inverse probability of treatment weighting-adjusted multivariate analysis [31]. Therein, radiation modality (3DCRT vs. IMRT) was independently associated with all-cause, other, and cardiac-specific mortality. It is thus important to be cognizant of cardiac doses in a neoplasm for which the heart was historically not a major priority but long-term survival is now increasing; this is similar to the findings of the RTOG 0617 trial in locally advanced non-small cell lung cancer (NSCLC) [32]. Because IMRT affords lower cardiac doses for both tumors, it has been postulated that IMRT should be the standard for NSCLC, despite a similar lack of comparative data in both NSCLC and EC [33].

There are several limitations of our study. In addition to the retrospective nature and smaller sample size, it is recognized that groups were somewhat imbalanced in terms of tumor location. The rather short follow-up time may also likely underestimate cardiac events, as these typically take at least 5–10 years to develop. Although the year of treatment was not a significant factor on outcome in UVA, advancements in diagnostics and treatment options after the first line during the inclusion period of twelve years may cause an undiscovered bias. Other newer supportive technologies also cannot be overlooked, such as staging PET and 4-dimensional CT planning. Moreover, patients with poor KPS were excluded, which also contribute a nontrivial proportion of EC patients. Also, limiting applicability were the low numbers of adenocarcinoma, as well as institutional/regional biases in referring patients for surgery versus definitive therapy. Taken together, however, our data should ideally signal other reports as to whether IMRT is indeed appropriate for several patient subgroups over 3DCRT.

If clinical benefits of IMRT in the definitive and/or neoadjuvant treatment of EC are eventually proven, it could signal further testing of an even more new modality known as proton beam therapy. Though issues with economic costs remain, this modality has been also associated with low perioperative complications, which could potentially offset these costs [34‐36]. A randomized trial of photons versus protons for EC is currently underway (NCT01512589).

These data showed that when comparing 3DCRT and IMRT, survival of patients with esophageal cancer is not affected. However, we found a lower rate of local relapses in the IMRT group potentially owing to dose-escalation. Further validation is needed from prospective studies.