In relation to all patients presenting with esophageal cancer from November 1997 to November 2007, the proportion of inoperable patients was higher in SCC's (23.8%; n = 104/437) as compared to ADC's (10.5%; n = 46/437) (in addition 5 patients with UDC). Both tumor entities revealed different reasons of inoperability: In patients with SCC, local tumor spread predominated with 45.7%, whereas in the majority of patients with ADC (58.7%), inoperability was due to hematogeneous metastases. Cardiopulmonary diseases causing contraindication for surgery were equally distributed among both tumor entities (17.1% in patients with SCC and 17.4% in patients with ADC).
The course of the disease was ascertained in 152 out of 155 (98%) patients by December 31st 2010; no data documenting the course were available in 3 patients at that time. One hundred and forty-eight patients had died from their malignancy; of the remaining 7 patients, 4 were alive and 3 were lost from follow-up. All 4 patients alive at the time of last follow-up had initially presented with a locally inoperable SCC and had undergone radiochemotherapy.
Six patients who underwent mucosectomy (n = 2) and tracheostomy (n = 1) or were lost from follow-up (n = 3) were not included in the analysis. The respective numbers of patients for SCC, ADC and UDC were 104, 46 and 5. According to the diagnostic work-up, tumor stage was IIA in 17 patients, IIB/III in 76 patients and IV in 54 patients. Contraindications for surgery were cardiopulmonary status in 26 patients, metastasis and tumor spread in 119, surgery refusal by 5, and no chance of esophageal substitution in five patients.
Brachytherapy was administered to 2 patients in group C and to 16 patients in group D. Seven patients in group B, 10 in group C, and 37 patients in group D completed their allocated treatment modality.
UDC patients were excluded since there were missing data in 3 (tumor stage), while the remaining 2 patients were excluded during the selection process.
In applying the sequential filters for confounding variables, only 123 out of 155 patients were considered for the analysis. We took care to include only patients without missing data, in terms of contraindications for surgery and TNM-stage by applying sequential filters to the Microsoft Excel database (Figure ). Contraindications for surgery were divided into 3 categories: a) cardiopulmonary, b) metastasis/non-resectable and c) other. Patients were considered accordingly by an analogy 1/5/1 (paralleling the rates among the 155 patients) in each of the 3 patient classes of the TNM-stage (IIA, IIB/III, IV) which were considered respectively by an analogy of 2/9/7 (paralleling the rates among the 155 patients) (Figure ). These sequential steps reduced the total of 155 patients by 8 patients (missing data), 19 patients by "TNM-stage" filtering, 2 patients by "contraindications for surgery" filtering and 3 patients lost to follow up (Figure ). The remaining 123 patients were assigned to four groups: SEMS group (A) n = 26 (SCC:10/ADC:16), Chemotherapy group (B) n = 12 (SCC:10/ADC:2), Radiotherapy group (C) n = 23 (SCC:11/ADC:12), Chemoradiotherapy group (D) n = 62 (SCC:52/ADC:10). For every patient in group B, approximately 2 patients in groups A and C, and 5 in group D were analyzed in terms of survival.
Figure 1 Only patients without missing data (in terms of the confounding variables) were included by applying sequential filters to the Microsoft Excel database. Contraindications for surgery were divided into 3 categories: a) cardiopulmonary (CP), b) metastasis/non-resectable (more ...)
As concerns the allocation of clinico-pathological characteristics among the 4 groups, all 21 parameters were equally distributed in terms of age, tumor localization, type of metastasis, histology, and cardiopulmonary status (Table ).
Comparison of clinico-pathological characteristics among the 4 groups of treatment modality
Overall median and mean patient survival was 6 months and 10.89 ± 12.63 months, respectively. The respective numbers for the 4 groups were as follows: Group A: 3/6.92 ± 8.4 months; Group B: 7/7.75 ± 6.6 months; Group C: 4/8.56 ± 9.5 months and Group D: 8/13.53 ± 14.7 months. Overall survival for the 4 groups is depicted in Table .
Cumulative proportion surviving of the 4 groups
Twenty-one variables were analyzed in univariate analysis of survival: gender, age, BMI, vital capacity (VC), FEV1, cTNM-classification, ASA-classification, nutritional status, tobacco consumption, alcohol intake, diabetes, Child-Pugh score, cardiopulmonary status, existence of a second carcinoma, histopathology, tumor localization, contraindication for surgery, cause of inoperability, site of metastasis and type of metastasis. Significant differences in overall survival were associated with tumor histology (P = 0.027), tumor localization (P = 0.019), and type of therapy (P = 0.005), respectively (Table ).
Univariate analysis of survival
In multivariate Cox's Proportional Hazard regression analysis of survival, the model including the predictors in univariate analysis gained statistical significance (P < 0.001), but treatment modality (P = 0.043) was the only independent predictor of survival (Table ).
Multivariate analysis of survival
With regard to the 12-month hazard rate/standard error of hazard rate for the 4 treatment groups: Group A = 0.14/0.07, Group B = 0.16/0.10, Group C = 0.15/0.06, and Group D = 0.04/0.01.
The difference in overall survival between Group A and Group D was highly significant (P < 0.01), and in favor of Group D. Comparing Group D versus Group B and Group D versus Group C, the difference in overall survival was marginally significant, and in favor of Group D (P = 0.069 and P = 0.059, respectively). The other possible comparisons in overall survival between Groups did not reach statistical significance (Figure ).
Cumulative proportion surviving for the patients of the 4 groups.
Aiming to establish a 2-year difference in overall survival of 24.3% vs. 3.8% for groups D and A respectively, there was a need for 26 patients per treatment group. The respective number for group D vs. C was 23 patients (24.3% vs. 0%), and for group D vs. B, it was 61 patients (24.3% vs. 8.6%). All the aforementioned comparisons of overall survival between groups fulfilled the sample size calculation, with the exception of one (group D: 62 patients vs. group B: 12 patients).
In order to investigate the rationale of our multidisciplinary team strategy to assign more adenocarcinoma patients to chemotherapy alone (n = 12) than to chemoradiotherapy (n = 9), survival curves of these two groups were compared. No statistical significant difference in overall survival was observed (P = 0.589). The same applied to squamous cell carcinoma patients undergoing chemoradiotherapy (n = 55) versus radiotherapy alone (n = 10) (P = 0.405).