Introduction

Esophageal cancer is an increasingly common cancer with a poor prognosis [1]. In the East Asia, squamous cell carcinoma (SCC) is dominant, while in the Western world, adenocarcinoma is dominant in esophageal cancer [2, 3]. For both histological types, surgery including optimal lymph node dissection has shown the best results for cure, but surgical intervention alone is insufficient with regard to survival outcome in advanced ESCC [4–7].

The chemotherapeutic regimen including CF (cisplatin/5-FU) has been well established for drastic reduction of mass volume, and the clinical benefits of postoperative CF adjuvant in addition to surgery were for the first time proven to be superior to surgery alone (SA) in advanced ESCC in phase III JCOG9204 trial in Japan [8]. The effect of CF neoadjuvant chemotherapy prior to surgery (CF NAC) was subsequently compared with that of the postoperative CF adjuvant in clinical Stage II/III (cStage II/III) ESCC in the phase III JCOG9907 trial, and CF NAC showed even better outcomes for overall survival (OS) than the postoperative adjuvant chemotherapy [9]. Nevertheless, CF NAC remains considered underpowered with regard to early control of the diseases, where considerable patients with CF NAC showed either preoperative treatment failure or frequent early recurrences within 1 year after the initial treatments.

We had been therefore committed to develop a novel potent chemotherapeutic regimens to control such early treatment failure in clinical stage II/III ESCC by adding docetaxel to cisplatin and 5-FU (DCF) [10]. DCF chemotherapy as a candidate regimen of palliative first-line treatment in the metastatic esophageal cancer has a great potential to improve treatment efficacy in the Western world [11] and in the Easter world [12, 13]. Docetaxel, cisplatin, and 5-FU have different antitumor activity, which are involved in tubulin depolymerization inhibition, DNA cross linking, and DNA antimetabolite, respectively, and synergy was anticipated by concurrent administration to the tumor cells in an experimental animal model [14]. Reflected by this result, DCF NAC actually improved progression-free survival as compared to CF NAC in head and neck cancer [15], and is being challenged to be developed in the Western world [16] and the Eastern world [10, 17] in esophageal cancer.

We recently reported on short-term clinical outcomes including early survival results of DCF NAC in cStage II/III ESCC [10], in which concurrent administration of DCF caused frequent severe side effects, especially in terms of hematological adverse events, but exhibited excellent survival outcomes with regard to early disease control after R0 resection. In this study, we will report long-term survival outcome of DCF NAC in cStage II/III ESCC.

Patients and methods

Patients

Thirty eight cStage II/III (excluding cT4) ESCC patients entered DCF NAC between December 2009 and August 2012 [10], while 41 had been treated by CF NAC according to JCOG9907 phase III clinical trial [9] between September 2007 and December 2009 and used as a historical control for prognosis. All patients were intended for curative resection, however, the patients could also select definitive chemoradiation therapy (dCRT) with CF as a curative treatment [18] since December 2009, if the patients strongly wished it. In December 2009, the Institutional Review Board of the Kitasato University Hospital approved the use of DCF chemotherapy for the management of ESCC. We had referred to feasibility study of the National Cancer Center Hospital East in Japan [19].

All patients who had DCF NAC had given written informed consent, while all patients who had CF NAC were treated as a standard practice according to JCOG9907, and eligible criteria were previously published [9].

The study was performed according to the guidelines of the Declaration of Helsinki, as amended in Edinburgh, Scotland, in October 2000.

Survival outcomes were evaluated for progression-free survival (PFS) and overall survival (OS). Median follow-up period in DCF NAC (49 months ranged from 26 to 64), while that of CF NAC was 70 months, ranged from 6 to 87.

Borderline respectability was defined as patients with suspected clinical T4 by computed tomography (CT) in cancer board committee for esophageal cancer.

DCF and CF NAC in patients who were intended for surgical resection

Patients were scheduled to receive 3 cycles of DCF NAC. The regimen consisted of docetaxel 70–75 mg/m2 given as a 1-h intravenous infusion on day 1, cisplatin 70–75 mg/m2 as a 2-h intravenous infusion on day 1, and 5-fluorouracil 750 mg/m2 as a continuous 24-h peripheral infusion on days 1–5 of a 21-day cycle. From February 2011, the starting doses of docetaxel and cisplatin were increased to 75 mg/m2.

CF (JCOG9907 regimen) NAC with cisplatin plus 5-fluorouracil was repeated twice every 3 weeks. A dose of 80 mg/m2 cisplatin was given by intravenous drip infusion for 2-h on day 1; 5-fluorouracil was administered at a dose of 800 mg/m2 by continuous infusion on days 1 through 5.

In patients whose response to the first or second course of chemotherapy was progressive disease or whose side effect was not tolerated in DCF or CF NAC, the next course of chemotherapy was not given, in order to take advantage of the probability for curative resection. The non-responders representing stable disease received scheduled cycles of NAC except those who did not tolerate side effect.

Surgery

The surgery were scheduled 3–5 weeks after the completion of the last cycle of DCF or CF NAC. A standard esophagectomy was performed according to the McKeown method (right thoracotomy followed by laparotomy and neck incision with cervical anastomosis), and three-field (thoraco-abdominal and cervical) lymph node dissection was also performed if indicated. R0, R1, and R2 resections were defined as microscopically and curatively resected, microscopically remnant, and macroscopically remnant cancer cells, respectively, after surgery [20]. The post-operative complications were evaluated according to the Clavien–Dindo classification [21].

Follow-up

Patients were assessed before being registered, before chemotherapy, and every 3 months during postoperative follow-up term for the initial 2 years. During years 3–5, they were assessed every 6 months. The diagnosis of recurrence was made on basis of imaging and, if possible, cytological analysis or biopsy. Clinical response to NAC was assessed with gastroscopy after patients completed the each cycle of chemotherapy (no sooner than 2 weeks after completion of chemotherapy). Postoperative adjuvant chemotherapy was not recommended.

Statistical methods

Progression-free survival (PFS) and overall survival (OS) were calculated by the Kaplan–Meier method. The Kaplan–Meier method was used to estimate cumulative survival rates, and differences in survival rates were assessed by the log-rank test. Survival (OS and PFS) was measured from the initial date of NAC therapy to the date of events or the last follow-up. Progression events were defined as recurrence after operation (including thoracotomy), salvage surgery followed by definitive CRT, inoperable recurrences after dCRT, and all deaths were also considered as events. All statistical analyses including calculating propensity score (PS) were conducted with JMP 11.0 software (SAS Institute, Cary, NC, USA). Cox proportional hazards models were used to compute relative risks and 95 % confidential interval (CI). All p values are two-sided.

Results

Survival outcome of PFS and OS in patients with ESCC who underwent NAC

Eight patients actually selected dCRT as a curative treatment in DCF NAC group due to patients’ wishes, while 1 patient wished to undergo dCRT in CF NAC group as previously described [10]. As a result, R0 esophagectomy was performed in 24 (80 %) among the 30 patients who did not select dCRT for curative treatment after DCF NAC, while R0 esophagectomy was performed in 33 among the 38 patients (87 %) excluding the 3 patients who selected dCRT due to cT4 for curative treatment in CF NAC. DCF NAC is unlikely to improve the rate of R0 esophagectomy remarkably as compared to CF NAC [10].

As for PFS, in this current study, DCF NAC showed remarkably better prognosis than CF NAC (p = 0.012, Fig. 1a). 3 and 5 year PFS of DCF NAC was 66 %, while that of CF NAC was 38 %. Interestingly, all progression was recognized within 3 years after the initial treatment. On the other hand, as for OS, DCF NAC showed marginally significantly better prognosis than CF NAC (p = 0.093, Fig. 1b). Three year OS of DCF NAC was 79 %, while that of CF NAC was 65 %. Five year OS of DCF NAC was 75 %, while that of CF NAC was 57 %. Final survival rate of OS in DCF NAC was 75 %, while that of CF NAC was 45 %. Deaths due to disease progression were recognized beyond 5 years in CF NAC, while surviving patients who are being followed up beyond 3 years in DCF NAC were all cancer free (no recurrence) at present.

Fig. 1
figure 1

Prognosis of cStage II/III ESCC according to NAC. a Progression-free survival (PFS) and b overall survival (OS). Black circle deaths of other diseases. CF cisplatin/5-FU, DCF docetaxel/cisplatin/5-FU

Full size image

Clinical background of both DCF NAC and FP NAC in cStage II/III ESCC and calculation of propensity score to predict DCF NAC

We previously showed that clinical background was similarly distributed between both DCF NAC and FP NAC in cStage II/III ESCC [10]. Eligibility criteria were not totally same for CF and DCF, because DCF was based on clinical study, while CF was administered as practice. However, rate limiting step is cisplatin administration for NAC treatments. As a result, patient distribution did not show any significant difference [10]. We therefore made prediction model for DCF NAC (objective variable) by clinical factors as explanatory variables such as age, sex, clinical T (cT), clinical N, and borderline resectability in a multivariate logistic regression model (likelihood ratio test, sex, age, cT, borderline resectability, cN: p = 0.4399, p = 0.6262, p = 0.0969, p = 0.5766, p = 0.2105) (Fig. 2a). This model showed relatively good one for prediction of DCF NAC to have AUC = 0.67, and the most significant factor was cT, and odds ratio of cT1 against cT2 was 0.14 in DCF NAC as compared to CF NAC (p = 0.036). PS to predict DCF were significant different between DCF NAC and CF NAC (p = 0.019) (Fig. 2b). Density distribution of PS is also shown in Fig. 2c. If matching was done for OS events using JMP 11.0, logit exchange score of both DCF and CF were well correlated in the matching cases (54 cases), but not so in the non-matching cases (25 cases) (Fig. 2d, e), and there was no significant distribution of cT factor (data not shown).

Fig. 2
figure 2

Patient background between CF NAC and DCF NAC. a ROC curve according to risk model to predict NAC status. b Distribution of propensity score (PS) between CF NAC and DCF NAC. c Density distribution of PS of CF NAC and DCF NAC. d Matching of logit exchange score of CF NAC and DCF NAC. e Matching and non-matching cases by PS

Full size image

Multivariate Cox proportional hazards model in cStage II/III ESCC

Univariate prognostic analysis was first performed for clinical factors such as age, sex, cT factor, cN factor, and borderline resectability as well as NAC status (Table 1). As a result, univariate prognostic factors with significant difference were NAC status (p = 0.012) (Fig. 1a), or cT (p = 0.007) and borderline resectability (p = 0.0034) for PFS (Fig. S1), while they were NAC status (p = 0.093) (Fig. 1b), or cT (p = 0.019) and borderline resectability (p = 0.025) for OS (Fig. S2). Multivariate Cox proportional hazards model adjusted for the 3 significant prognostic factors (cT, borderline resectability, and NAC status) identified independent prognostic factors of NAC status for OS (p = 0.047), and NAC status (p = 0.0013) and cT (p = 0.0098) for PFS (Table 2). Both NAC status and PS were again applied to the multivariate Cox proportional hazards model, and DCF NAC was the candidate of the prognostic indicator for PFS (p = 0.0044) and OS (p = 0.063).

Table 1 Clinical variables and prognosis in esophageal squamous cell carcinoma with neoadjuvant chemotherapy
Full size table
Table 2 Multivariate Cox proportional hazards model in ESCC with NAC
Full size table

Multimodality treatment and outcomes after progression of CF NAC and DCF NAC

Fiver-year PFS of patients with CF NAC was 38 %, while 5-year OS of those patients was 57 % (Fig. 1a). While almost patients with the progression represented recurrence, multimodality treatments could cure some patients, where long-term survivors underwent dCRT with DCF, which had been developed originally in our institute [22]. Among 33 patients with R0 esophagectomy, 17 cases had recurrences. Six cases survived over 5 years, among whom 5 cases were alive with cancer free. Such cancer free patients were composed of 3 lymph node recurrences treated by dCRT with DCF and 2 distant metastasis (liver and lung) treated by operative resection.

On the other hand, 5-year PFS of patients with DCF NAC was 66 %, while 5-year OS of those patients was 75 %. Differently from CF NAC, 13 patients with DCF NAC had progression within 3 years, among whom 7 cases died due to disease progression. On the other hand, 6 cases were alive with cancer free at 41, 48, 49, 53, 54, 57 months after the initial curative treatments (surgery n = 4, dCRT n = 2). The 6 cases were composed of patients with thoracotomy due to surgical T4 (trachea) followed by dCRT with DCF (n = 1), anastomotic recurrence followed by dCRT (n = 1), lymph node recurrences after R0 esophagectomy followed by dCRT (n = 2), and dCRT followed by salvage esophagectomy (n = 2).

Discussion

In Japan, JCOG9907 phase III trial demonstrated that CF NAC improved OS, but that it did not improve PFS as compared to postoperative adjuvant CF therapy in cStage II/III [9]. This curious result could not have been fully explained by the reasonable mechanism. Oncologists would like to know why CF NAC could rescue some of the progressed patients, who were similarly distributed to both CF NAC and postoperative adjuvant group. Treatment of recurrent patients must have played a key role in improved OS of ESCC with CF NAC. These findings made us speculate that CF NAC could make a given effect for systemically disseminated cancer cells to become a localized disease.

On the other hand, in the western study to validate prognostic effect CF NAC failed to demonstrate that it improved OS in advanced esophageal cancer [23]. Although there have been few information with regard to treatments against recurrent cancers, the different results of the similar NAC in esophageal cancer between the western and eastern world may be derived from treatment efficacy against recurrence. Anyway, CF NAC is unlikely to be a confident NAC treatment in advanced esophageal cancer.

CF NAC has another critical issue that there were considerable number of patients who showed preoperative treatment failure and early recurrence within 1 year [9]. This may be due to the underpowered nature of the CF NAC regimen. On the contrary, DCF NAC is actually significantly stronger than CF NAC with regard to improving clinical response in ESCC with cStage II/III [10]. Docetaxel harbored mechanism to suppress tumor cells differently from cisplatin and 5-FU, thus synergy would be anticipated with its concurrent use in an animal experiment [14]. In fact, such combination is active in several types of cancer in term of improving patient survival as a preoperative adjuvant setting [15]. However, there have been no data which describe long-term prognostic effect in esophageal cancer. In our current study, we for the first time demonstrated that DCF NAC could have a promise to improve prognosis as compared to CF NAC in cStage II/III ESCC.

Since median follow-up period was only 49 months, this study should not conclude 5 year survival. The survival curve showed many censored cases after 40 months in DCF group. This may overestimate survival rate of DCF group. However, 3-year PFS may be reliable survival rate in this study. In our prognostic analysis for both DCF NAC and CF NAC, on the other hand, all events of disease progression occurred within 3 years, and follow-up of almost all patients were completed beyond 3 years after the initial treatments even among the DCF NAC patients, suggesting that we have almost all information with regard to disease progression. Although follow-up term remains short for OS, however importantly all the patients who have been followed up beyond 3 years were cancer free in cStage II/III ESCC with DCF NAC at present. In other meaning, our prognostic data could be called a relatively conclusive one in terms of long-term prognosis of DCF NAC, except other disease deaths.

Nevertheless, DCF NAC is likely to have weak point. In our early study, DCF NAC could not significantly improve R0 resection rate of esophageal cancer [10], suggesting that DCF remains insufficient in efficacy for the most aggressive cStage II/III ESCC such as surgical T4 disease. This finding is consistent with the previous press which describes the clinical efficacy of DCF in borderline resectable ESCC [17]. On the other hand, DCF NAC robustly improved prognosis in cStage II/III with R0 esophagectomy [10]. Such early observation is well being reflected by Kaplan–Meier curve of PFS; there was small difference for PFS within 1 year, while there was big difference for PFS beyond 1 year (see Fig. 1a). That is why DCF may robustly affect prognosis by reducing recurrence after R0 resection, indicating that DCF NAC can more potently affect systemic micro-metastasis of cancer cells rather than localized control than CF NAC in cStage II/III ESCC.

Limitations. In this study, we included relative small number of patients with DCF NAC (n = 38) and CF NAC (n = 41). If survival rates are compared after PS matching, balanced patient number was further reduced in number (n = 27, respectively), and no significant difference was recognized, especially for OS, while DCF NAC showed better prognosis as compared to CF NAC (Fig. S3). This modest difference for OS between CF and DCF may be affected by multimodality treatments for recurrences. PFS (38 % at 5 years) by CF NAC in our institute was inferior to the survival result of the JCOG9907 (42 % at 5 years), while OS (57 % at 5 years) by CF NAC was similar to that (57 % at 5 years), putatively due to recent adoption of rigorous treatments including dCRT for recurrent tumors.

Anyway, 5-year OS beyond 70 % may be very promising survival outcomes in cStage II/III ESCC. That is why we can believe that DCF NAC is a novel and precious therapeutic tool in cStage II/III ESCC. Phase III trial to compare prognosis of cStage II/III ESCC by multimodality treatments including DCF NAC is ongoing as a phase III NExT trial in Japan [24], and the final result including primary endpoint (OS) is highly anticipated to validate the efficacy of DCF NAC in esophageal cancer.