Aims: Neoadjuvant chemotherapy is used to downstage locally advanced oesophagogastric cancer. This study assessed whether changes in dysphagia and weight correlated with radiological and pathological assessment of response and surgical decision-making.
Materials and methods: All patients with locally advanced carcinoma of the lower oesophagus or oesophagogastric junction treated with neoadjuvant ECF (epirubicin, cisplatin, and 5-fluorouracil) chemotherapy from January 2000 to January 2003 were included in this study. Patients were considered to be operable depending upon their chemotherapy response. Weight and swallowing were assessed before and after chemotherapy. Statistical analysis was carried out using ANOVA, unpaired t test and Fisher's exact test.
Results: Seventy-eight patients (male-female ratio: 6.8: 1; median age: 62.2 years; range: 44.1-78.0 years) underwent a median of three cycles (range: 1-7) of neoadjuvant ECF chemotherapy. Forty patients (51%) gained weight, and swallowing improved in 53 patients (68%). Radiological changes (based on computed tomography) were assessed according to WHO criteria: complete response (5%), partial response (27%), stable disease (46%) and progressive disease (15%). Patients whose swallowing improved gained significantly more weight (P < 0.0001). Swallowing (P = 0.0009) was significantly improved in radiological responders but not weight (P = 0.06); when radiological non-responders were separated into stable and progressive disease, patients with progressive disease were identified as failing to gain weight (P = 0.005). Both swallowing (P < 0.0001) and weight gain (P < 0.0001) were better in patients undergoing surgery. The use of changes of weight (P = 0.42) and swallowing (P = 0.61) failed to separate pathological responders from nonresponders in the subset of patients undergoing surgery.
Conclusions: Weight gain and improved swallowing are good but not absolute indicators of radiological response to chemotherapy and patient selection for surgery. However, changes in these variables are not sufficiently sensitive to identify pathological responders from non-responders.