Quality of life after curative gastrectomy for gastric cancer in a randomised controlled trial

Quality of life (QOL) was studied in gastric cancer patients treated on a randomised, controlled trial comparing D1 (level 1) with D3 (levels 1, 2 and 3) lymphadenectomy. A total of 221 patients were randomly assigned to D1 (n=110) and D3 (n=111) surgery. Quality-of-life assessments included functional outcomes (a 14-item survey about treatment-specific symptoms) and health perception (Spitzer QOL Index) was performed before and after surgery at disease-free status. Patients suffered from irrelative events such as loss of partners was excluded thereafter. Main analyses were done by intention-to-treat. Thus, 214 D1 (106/110=96.4%) and D3 (108/111=97.3%) R0 patients were assessed. Longitudinal analysis showed that functional outcomes decreased at 6 months after surgery and increased over time thereafter, while health perceptions increased over time in general. On the basis of linear mixed model analyses, patients having total gastrectomy, advanced cancer and hemipancreaticosplenectomy, but not complications had poorer QOL than those without. D1 and D3 patients showed no significant difference in QOL. The results suggest that changes of QOL were largely due to scope of gastric resection, disease status and distal pancreaticosplenectomy, rather than the extent of lymph node dissection. This indicates that nodal dissection can be performed for a potentially curable gastric cancer.

Gastric cancer remains the second most common cause of cancerrelated deaths worldwide, and a major clinical challenge because of its poor prognosis and limited treatment options. The efficacy of nodal dissection is controversial (Cuschieri et al, 1999;Hartgrink et al, 2004). However, we have conducted a single institutional trial and demonstrated that D3 surgery has survival benefit (Wu et al, 2006b) but acceptable morbidity (Wu et al, 2004).
Quality of life (QOL) is looked upon as a multidimensional entity comprising physical, psychological, social and medical parameters. Cross-sectional studies (Koster et al, 1987;Kusche et al, 1987;Buhl et al, 1990;Korenaga et al, 1992;Jentschura et al, 1997;Wu et al, 1997a;Zieren et al, 1998;Svedlund et al, 1999;Thybush-Bernhardt et al, 1999) of QOL after gastric cancer surgery provided important background for design of this study, delineating a number of specific hypotheses and research questions related to surgery and survivorship. This report examines QOL of gastric cancer patients receiving either D1 or D3 surgery.

Gastric cancer surgical trial summary
From October 1993 to August 1999, 221 patients were randomised at laparotomy to compare D1 (level 1) and D3 (level 1, 2 and 3) lymphadenectomy surgery for gastric cancer (Wu et al, 2004(Wu et al, , 2006b. The primary surgical treatment report contains the full details. D3 surgery had more complications than D1 surgery, but both had no deaths (Wu et al, 2004(Wu et al, , 2006a. No patients received preoperative or postoperative adjuvant radiotherapy or chemotherapy. D3 surgery offers a survival benefit when done by welltrained, experienced surgeons (Wu et al, 2006b).

Eligibility for the QOL study
The QOL study, a companion protocol, was opened simultaneously with surgical trial. Patients were eligible if they were registered to the National Health Research Institutes and had fitted the requirement of operation criteria and randomisation. The exclusion criteria included patients not curatively resected and not treated according to protocol. All patients were at disease-free status when the survey was carried out, because tumour recurrence was the decisive exclusion factor. The Ethics Committee approved the study protocol and an informed consent was obtained from each patient.

Instruments, assessment strategy and procedures
The Spitzer QOL Index (Spitzer et al, 1981), a global health assessment with valid questionnaire, includes five items rated on a three-point scale: activity, daily living, health, support of family and friends, and outlook. Functional outcome was assessed using a 14-item survey designed to assess treatment-specific symptoms, largely gastrointestinal function. Low scores reflect more symptoms (Korenaga et al, 1992;Wu et al, 1997a). Its validity was checked by the Cronbach's a-scores.
The assessment strategy focused on QOL at a disease-free status not to be confounded by irrelevant events such as loss of partners and bony fractures after falling. We hypothesised that patients receiving D3 resection would have more symptoms and greater fatigue than patients receiving D1 surgery, with treatment arm differences most prominent at the 6-month assessment, and possibly continuing up to 1 year after random assignment. Questionnaires were administered at registration, 6 months, 1 year and annually thereafter until recurrence of tumour, if any. A research nurse (Hsueh-Pin Yu) briefly described and explained the procedure for responding to each question. Each interview took approximately 20 min.

Statistical considerations and analytic plan
All analyses used the intention-to-treat philosophy, whereby patients were grouped according to randomly assigned surgery treatment arms and under the assumption of a missing-at-random mechanism. The lymph node dissection effects on the functional outcome, and Spitzer QOL Index were examined by the linear mixed models (Diggle et al, 2002), adjusting for fixed effects for each time point, treatment, baseline covariate, age, sex, marital status, scope of gastric resection and gross appearance of tumour, and with random effects for intercept and each time point. Quadratic times as well as interaction terms were considered in the models to adjust for nonlinear temporal trend. A general unstructured covariance was assumed, that is, no specific forms of the random-effects covariance matrix were assumed. The lymph node dissection effects were assessed based on P-values of the fitted coefficients, adjusting for significant covariates in the model. The time trends of the functional outcome and Spitzer Index were assessed by significance tests of adjacent time periods in the mixed models, assisted by the longitudinal plots classified by the D1 and D3 treatment groups. These models and analyses were implemented by SAS PROC Mixed procedure.

Patient characteristics
In all, 214 R0 patients (mean age 63.9 ± 11.1 years) were recruited. Comparison of the QOL study sample with the final surgical trial sample demonstrated no statistical difference between D1 (106/ 110 ¼ 96.4%) and D3 (108/111 ¼ 97.3%) patients in the QOL study (P ¼ 0.772, w 2 test). Male patients were older (65.43 ± 10.3 vs 58.8 ± 12.2) and underwent more distal pancreaticosplenectomy procedures. Female patients as compared with male patients were more likely to be married (100 vs 90.8%). The two groups did not differ significantly according to age, gender, marital status, type of gastric resection, reconstructive procedures (Table 1), comorbidity and location of tumour. But D3 patients had more cholecystectomy, distal hemipancreaticosplenectomy procedures (Table 1) and complications. During follow-up, five patients encountered irrelevant events; two spouse suicides (fifth year), three bony fractures (at second, second and sixth years, respectively). Four patients immigrated to Mainland China (at second, second, third and third years, respectively) and therefore could not be assessed at regular time points thereafter. Quality of life was not assessed at scheduled across assessment points after surgery in 3.2% of the subjects due to prolonged absence overseas. The Cronbach's a-scores for treatment-specific symptoms were 0.67.

Longitudinal analysis of QOL
Nodal dissection effect showed transient discrepancy of food volume at baseline, diarrhoea and heartburn at first and second years after surgery ( Table 2). The linear mixed model analysis for functional outcome showed that there were no significant differences in scores between two surgical treatment groups (P ¼ 0.5338, 0.6423 and 0.3941 for nodal dissection effect and its interaction effects with time and time 2 , respectively), after adjusting for significant covariate effects (Table 3). These significant covariate effects included time trend, baseline score, age, gender, gross appearance of tumour, scope of gastric resection and hemipancreaticosplenectomy. Surgical complications, depth of cancer invasion and marital status did not affect functional outcomes, and therefore were dropped from the model. These results coincide with univariate analyses (which were not shown here) based on linear mixed effects models when a single covariate effect is considered, adjusted for time trends and baseline scores. The overall time trends (Figure 1) indicated that the functional outcomes were significantly lower at the 6th month after surgical treatment than at the baseline (Po0.0001), and improved from the sixth month to the first year after surgery (Po0.0001), continuously improving up to the second year after surgery (P ¼ 0.006). Thereafter, the differences in functional outcomes between adjacent years were not statistically significant. The time trends of functional outcomes had to be adjusted by significant interaction effects with time, including scope of gastric resection, gross appearance of tumour, age and baseline score.
For the Sptizer Index scores, the linear mixed model analysis revealed similar results that there were no significant differences in scores between two surgical treatment groups (P ¼ 0.5652, 0.2956 and 0.0784 for node dissection effect and its interaction effects with time and time 2 , respectively), after adjusting for significant covariate effects (Table 3). The significant covariate effects are similar to those on the functional outcomes. An interesting finding is that the marital status is a significant effect on the Sptizer Index scores, but not on the functional outcomes. The overall health perceptions (Spitzer Index) ( Figure 1) were significantly lower at the baseline than at the 6th month after surgical treatment (Po0.0001), and improved from the 6th month to the first year after surgery (P ¼ 0.0022). Thereafter, there were no significant  Table 3. Further analysis based on overall averaged scores showed that patients who underwent a distal pancreatic resection had poorer appetite (P ¼ 0.001 based on t-test), decreased consistency (Po0.0001) and volume of food (P ¼ 0.0003) and frequency of eating (Po0.0001), lost of more body weight (P ¼ 0.012), could not readily engage in full-time work or study (Po0.0001), lacked energy (Po0.0001) and reported less psychosocial support (P ¼ 0.021) than those without pancreatic resection.

DISCUSSION
The results of the longitudinal QOL measurements indicate that D1 and D3 gastric cancer patients have similar health perception and functional outcomes after surgery. When we designed QOL study the Functional Assessment of Cancer Therapy -General (FACT-G) (Cella et al, 1993) and the European Organization for Research and Treatment of Cancer Core Quality of Life Questionnaire (EORTC QLQ-C30) (Aaronson et al, 1993) were not available, and no Chinese version was available. The Spitzer QOL Index is a valid but suboptional questionnaire due to the limited number of items and ceiling effect. However, it is easily applied and can be summed into a single score while retaining its domain-specific properties. It has been used worldwide in various cancers (Greimel et al, 2002;Erridge et al, 2005) including gastric cancer (Kusche et al, 1987; Disease-specific symptom   Buhl et al, 1990). Our treatment-specific instrument, which was modified from Korenaga's study for Japanese gastric cancer patients, is a short questionnaire, which has proved to be reliable in implementation (Wu et al, 1997a(Wu et al, , 2000 and in current study (Cronbach's a-scores 0.67). Taken all questionnaires together, data acquisition for 19 items assisted by a research nurse had shown its validity throughout the entire study.
Patients had anxiety, depression, loss of appetite and body weight at diagnosed of gastric cancer, but returned to normal gradually after surgery. Similar to other observations (Jentschura et al, 1997;Wu et al, 1997a;Davies et al, 1998), patients after a total gastrectomy tended to suffer poorer tolerance of normal food, need for more frequent eating and loss of more body weight than those after a subtotal gastrectomy. And younger patients had better QOL than older (Koster et al, 1987;Wu et al, 1997a). We noted that gender plays an important role in facing gastric cancer. Male patients usually have more morbidity after gastric cancer surgery (Viste et al, 1988). It is noteworthy that female patients were younger and mostly married, but male patients tended to feel better and calmer, suffer less insomnia and have a more positive outlook. This may not be simply explained by different culture roles attributes for men and women, since Koster et al (1987) also recorded the same phenomenon in German gastric cancer patients. Lymph node dissection along the upper border of the pancreas is essential for radical gastric cancer surgery. Resection of distal pancreas to facilitate lymph node dissection had a high morbidity and mortality. (Bonenkamp et al, 1995;Cuschieri et al, 1996;Wu et al, 2006a) Our data revealed that diarrhoea, except at first year, was similarly experienced in both groups. This study further disclosed its worse functional outcome and health perception, but no negative effect of QOL among D3 patients was drawn as the number of affected patients (1% in D1 vs 12% in D3 patients) was small. However, a technique developed by Maruyama et al (1995) to preserve the pancreas is feasible and safe, and should be considered in gastric cancer surgery.
Although the morbidity rate was higher in D3 patients than in D1 patients, our analysis indicates that lymph node dissection did not adversely influence QOL. The possible explanations are that the anastomotic leaks were minor in extent and abscess was detected early. These patients had no fever and did not need intensive care. They consumed regular diet while treating by a closed continuous irrigation system for abscess, and leaks were managed nonoperatively with nutritional support (Wu et al, 2004) Nevertheless, one should be very cautious in interpreting these results because the postoperative morbidity rate was very low, as compared with other reported studies (Bonenkamp et al, 1995;Cuschieri et al, 1996). Our previous study and current trial showed that operative morbidity did not influence survival (Wu et al, 1997b(Wu et al, , 2006b. The findings we observed are apparently at variance with those of a previous study (Siewert et al, 1998). A similar disparity also exists in QOL.