The relationship between quality of life (EORTC QLQ-C30) and survival in patients with gastro-oesophageal cancer

It remains unclear whether any aspect of quality of life has a role in predicting survival in an unselected cohort of patients with gastro-oesophageal cancer. Therefore the aim of the present study was to examine the relationship between quality of life (EORTC QLQ-C30), clinico-pathological characteristics and survival in patients with gastro-oesophageal cancer. Patients presenting with gastric or oesophageal cancer, staged using the UICC tumour node metastasis (TNM) classification and who received either potentially curative surgery or palliative treatment between November 1997 and December 2002 (n=152) participated in a quality of life study, using the EORTC QLQ-C30 core questionnaire. On univariate analysis, age (P<0.01), tumour length (P<0.0001), TNM stage (P<0.0001), weight loss (P<0.0001), dysphagia score (P<0.001), performance status (P<0.1) and treatment (P<0.0001) were significantly associated with cancer-specific survival. EORTC QLQ-C30, physical functioning (P<0.0001), role functioning (P<0.001), cognitive functioning (P<0.01), social functioning (P<0.0001), global quality of life (P<0.0001), fatigue (P<0.0001), nausea/vomiting (P<0.01), pain (P<0.001), dyspnoea (P<0.0001), appetite loss (P<0.0001) and constipation (P<0.05) were also significantly associated with cancer-specific survival. On multivariate survival analysis, tumour stage (P<0.0001), treatment (P<0.001) and appetite loss (P<0.0001) were significant independent predictors of cancer-specific survival. The present study highlights the importance of quality of life (EORTC QLQ-C30) measures, in particular appetite loss, as a prognostic factor in these patients.

Gastro-oesophageal cancer is the third commonest cause of cancer death in the UK. Each year, there are approximately 16 500 new cases and over 13 000 deaths attributable to the disease. Overall survival is poor with the majority of patients presenting with advanced, inoperable disease and less than 15% surviving 5 years (Cancer Research UK, Information Resource Centre, 2004). Although there have been improvements in survival following surgery (Ando et al, 2000;Hundahl et al, 2000;Hofstetter et al, 2002;von Rahden and Stein, 2004), for the majority of patients current treatment offers little in terms of improved survival. As a result quality of life in these patients is likely to be of considerable importance (Aaronson, Bullinger and Ahmedzai, 1988;Aaronson et al, 1993).
The European Organisation for Research and Treatment of Cancer have developed and validated the EORTC-QLQ-C30 questionnaire designed to assess the quality of life of cancer patients (Aaronson et al, 1993). Disease-specific aspects of the questionnaire provide detailed information about the patients' perception of their health. Moreover, it has been reported that, in a few studies, the EORTC QLQ-C30 measurement of quality of life may have prognostic value in patients with gastro-oesophageal cancer (Conroy et al, 2006). Blazeby et al, (2001) reported that, in addition to age and TNM stage, physical function or emotional function had independent prognostic value in 92 patients with oesophageal cancer. However, treatment (whether or not the patient underwent surgery) was not included in the model (Blazeby et al, 2001). Fang et al (2004) studied 110 patients with squamous oesophageal cancer and concluded that there was evidence to support the correlation of patient-reported quality of life scores with survival; therefore, pretreatment physical functioning might be a surrogate marker of an unrecognised biological prognostic factor. Although performance status was significant on univariate analysis, it was not significant on multivariate analysis; whereas physical functioning was significant (Fang et al, 2004).
In contrast, in a study of more than 1000 patients with inoperable gastro-oesophageal cancer, entering three randomised clinical trials, Chau et al (2004) reported that no aspect of the QLQ-C30 had independent prognostic value when performance status was considered. However, physical function, role function and global quality of life were associated with survival on univariate analysis. There were no survival differences among patients with oesophageal or gastric cancer (Chau et al, 2004). However, this study was retrospective and included selected cohorts of patients.
Therefore, from the above it remains unclear whether any aspect of quality of life other than physical function has a role in predicting survival in an unselected cohort of patients with gastrooesophageal cancer. The aim of the present study was to examine the relationship between quality of life (EORTC QLQ-C30), clinico-pathological characteristics and survival in patients with gastro-oesophageal cancer.

Patients
Patients presenting with adenocarcinoma or squamous carcinoma of the gastric or oesophageal tract at the Royal Infirmary andSouthern General Hospital, Glasgow between November 1997 andDecember 2002 (n ¼ 152) participated in a quality of life study, using the EORTC QLQ-C30 core questionnaire.
The extent of tumour spread was recorded using the TNM 5th edition classification (Sobin and Wittekind, 1997). Tumours around the gastro-oesophageal junction were further classified according to tumour site, using the Siewert system; type 1 and 2 lesions of the gastro-oesophageal junction were designated as cancers of the oesophagus. Type 3 tumours of the cardia were designated as gastric cancers (Siewert and Stein, 1998).
For gastric cancers, TNM stage I -III tumours were considered to be potentially amenable to curative surgical resection. For oesophageal cancers, TNM stage I -III tumours, excluding T4, were deemed to be potentially amenable to curative surgical resection. Patients who had stage 1 and 2 disease but whose performance status was poor or who had significant comorbidity were deemed not suitable for surgery and went forward for active palliative treatment or supportive care. There were 152 patients included in the study, 69 patients underwent surgery and 83 patients received active palliative treatment or supportive care.
The study was approved by the Research Ethics Committee of the Royal Infirmary and Southern General Hospital, Glasgow.

Methods
Clinical and demographic variables were recorded at the patient's initial presentation and included age, sex, tumour type, site and length, TNM stage, ECOG performance status, weight loss and dysphagia.
Following diagnosis but prior to treatment the lead clinician approached patients as to whether they would participate in a study to examine their quality of life. If they gave informed consent they were given the EORTC QLQ-C30 questionnaire to complete.
Different aspects of quality of life were assessed using this cancer-specific 30-item questionnaire, which has six functional scales (physical, role, emotional, cognitive, social, global health status) and several questions relating to a range of physical symptoms (Aaronson et al, 1993). Patients marked to what extent each statement applied to them. A number of patients were excluded because they were unlikely to understand the questionnaire either due to language, brain metastases, delirium or confusion. Neither age nor performance status were considered when offering the patient questionnaire. Few subjects were excluded (less than 10 patients) and therefore in those patients offered the questionnaire the bias was likely to be small.

Statistics
Scoring algorithms have been produced by the EORTC Quality of Life Study Group. The sum of items in each category is added and the total divided by the number of questions in the category. A linear transformation is then undertaken to convert this to a percentage scale with a higher score representing a higher response level. Thus a high score for functional scale represents a high/healthy level of functioning. A high score for the global health status/quality of life represents a high quality of life. In contrast, a high score for the symptom scale represents a higher level of symptoms/problems (Aaronson et al, 1993).
Data are presented as the median and range. Survival was determined from the time of biopsy proven diagnosis, and the end point for survival analysis was cancer-specific death. Patients were followed up at their clinic or endoscopy appointments and information on date and cause of death was checked with that received by the cancer registration system through the Registrar General (Scotland). Deaths up to the end of April 2007 were included in the analysis.
Univariate and multivariate survival analysis and calculation of hazard ratios (HR) were performed using a Cox regression model. For simplicity of presentation, a single hazard ratio was calculated for each ordered categorical variable, corresponding to the relative risk between adjacent categories. Hazard ratios for EORTC quality of life and symptom scores relate to a one percentage point increase in the score. Owing to the large number of covariates examined, only those that were significant on univariate analysis were included in the multivariate analysis, and only main effects were considered. The analysis was performed using a backward stepwise procedure to derive a final model of the variables that had a significant relationship with survival. To remove a variable from the model, the corresponding P-value had to be greater than 0.05. The proportional hazards assumption was checked using log minus log plots.
Comparison of the association between tumour site, TNM stage, treatment and the functional (physical, role, emotional, cognitive, social, global health status) and physical symptoms (fatigue, pain and appetite loss) scales of the EORTC-QLQ-C30 quality of life questionnaire was carried out using the w 2 -test or Mann -Whitney U-test where appropriate. Analysis was performed using SPSS software (SPSS Inc. Chicago, IL, USA).

RESULTS
Patient characteristics and cancer specific survival analysis of patients with gastro-oesophageal cancer (n ¼ 152) are shown in Table 1. The minimum follow-up period was 54 months or until date of death, the median follow-up for survivors was 81 months, one patient was lost to follow up and one patient withdrew from the study. During this period 106 (70%) patients died from their disease and 14 (9%) died from comorbid disease.
On multivariate analysis, tumour stage (Po0.001), treatment (Po0.0001) and appetite loss (Po0.0001) were significantly independent predictors of cancer-specific survival. The relationship between appetite loss and cancer-specific survival in patients with gastro-oesophageal cancer is shown in Figure 1.
The relationship between tumour site, clinico-pathological characteristics and quality of life in patients with gastrooesophageal cancer is shown in Table 2. Compared with the gastric cancer patients, oesophageal cancer patients were older (Po0.01), had more dysphagia (Po0.001) and a poorer ECOG-ps (Po0.05). In terms of quality of life, compared with the gastric cancer patients, oesophageal cancer patients had higher emotional functioning (Po0.01), cognitive functioning (Po0.05), less nausea and vomiting (Po0.05).

DISCUSSION
In the present study tumour site was not associated with major differences in EORTC QLQ-C30 quality of life function or symptom scores. However, there were major differences in quality of life and symptom scores with increasing stage of disease. In particular, social functioning, fatigue, appetite loss and global quality of life were all impaired with increasing tumour stage.
As might be expected in view of these associations with tumour stage, the majority of quality of life and symptom scores predicted survival on univariate analysis. It was of interest, however, that   appetite loss remained an independently significant prognostic factor even after adjustment for TNM stage and treatment. Furthermore, the predictive value of appetite loss was maintained even after adjustment for all other clinico-pathological variables and quality of life and symptom scores. Taken together the results of the present study highlight the importance of appetite loss as a presenting symptom in patients with gastro-oesophageal cancer. Few studies have examined the relationship between aspects of quality of life and survival in patients with gastro-oesophageal cancer. The results of the present study are consistent with the report of Fang et al (2004) who reported that appetite loss was associated with poorer survival in 110 patients with oesophageal cancer. However, the association was much weaker than that of the present study and was not significant in multivariate analysis. Furthermore, the follow-up period and the number of patients who died of their disease were not defined. Blazeby et al (1995), in a smaller study of 59 patients with oesophageal cancer, also reported that appetite loss was associated with poorer survival.
The basis of the relationship between appetite loss and poorer cancer-specific survival cannot be determined by the present cross sectional study. However, it was of interest that appetite loss was closely associated with nausea and vomiting, dysphagia and weight loss and therefore it may be that these symptoms result in appetite loss and the consequent loss of weight, which has long been recognised to impact on outcome (Dewys et al, 1980).
A number of workers have implicated the systemic inflammatory response in this process (Kotler, 2000;MacDonald, 2007).
O 'Gorman et al (1988), in a cross sectional study, showed that in addition to appetite loss and weight loss, the systemic inflammatory response was an important factor in determining patients' quality of life (EORTC QLQ-C30) in gastro-intestinal cancer patients (O'Gorman et al, 1988). Therefore, it is of interest that two recent studies have shown that the presence of a systemic inflammatory response, as evidenced by an elevated C-reactive protein, predicts survival in both operable (Crumley et al, 2006a) and inoperable (Crumley et al, 2006b) gastro-oesophageal cancer patients. In the present study C-reactive protein concentrations, at the time of quality of life assessment, were available in 94 (62%) patients. Consistent with previous work an elevated C-reactive protein concentration was associated with increased appetite loss and when included in the multivariate analysis, an elevated C-reactive protein concentration was independently associated with poorer cancer-specific survival. However, even those patients without an elevated C-reactive protein concentration reported some appetite loss and the independent prognostic value of appetite loss remained, thus confirming the importance of appetite loss in the multifactorial nature of weight loss and poor outcome in these patients (MacDonald, 2007).
In summary, in patients with gastro-oesophageal cancer, routinely used prognostic factors are based predominantly on clinical and pathological findings. The present study highlights the importance of quality of life (EORTC QLQ-C30) measures, in particular appetite loss, as prognostic factors in these patients.