Randomized comparison between chemotherapy plus best supportive care with best supportive care in advanced gastric cancer

Randomized comparison between chemotherapy plus best supportive care with best supportive care in advanced gastric cancer

Annals of Oncology 8: 163-168,1997. O 1997 Kluwer Academic Publishers. Printed in the Netherlands. Original article Randomized comparison between che...

2MB Sizes 0 Downloads 16 Views

Annals of Oncology 8: 163-168,1997. O 1997 Kluwer Academic Publishers. Printed in the Netherlands.

Original article Randomized comparison between chemotherapy plus best supportive care with best supportive care in advanced gastric cancer B. Glimelius,1 K. Ekstrom,1 K. Hoffman,1'3 W. Graf,2 P.-O. Sjoden,3 U. Haglund,2 C. Svensson,4 L.-K. Enander,5 T. Linne,6 H. Sellstrom7 and R. Heuman 8 Departments of 'Oncology and 2Surgery and l Centre for Caring Sciences, University of Uppsala, Uppsala; ^Department of Oncology, Huddinge University Hospital, Stockholm; Departments of i Surgery and 6Oncology, Centralsjukhuset, Karlstad, 7Department of Surgery, Sjukhuset, SSffle; 'Department of Surgery, Lasarettel, Mora, Sweden

Summary Background: The extent to which chemotherapy may relieve tumour-related symptoms, improve quality of life and prolong survival in patients with gastric cancer is not known in spite of the extensive use of this treatment modality. The aim of this study was to estimate any gain in the quantity and quality of life produced by chemotherapy in these patients. Patients and methods: Between January 1991 and February 1995, 61 patients with gastric cancer were randomized to either chemotherapy in addition to best supportive care or to best supportive care. Chemotherapy was allowed in the latter group if the supportive measures did not lead to palliation. Chemotherapy was the ELF-regimen consisting of 5-fluorouracil, leucovorin and etoposide, or, in elderly patients with poor performance, a 5-fluorouracil/leucovorin regimen (FLv). Quality of life was evaluated with the EORTC-QLQ-C30 instrument. Results: More patients in the chemotherapy group (45%, 14/ 31) had an improved or prolonged high quality of life for a minimum period of 4 months compared to those in the best

Introduction The use of chemotherapy in patients with gastric cancer has attracted great interest through the decades. In the 1970s, single agent activity with approximately 20% response rates was noticed, particularly using 5-fluorouracil (5-FU), mitomycin C, BCNU and doxorubicin [1]. Combinations of these drugs appeared to yield higher response rates, although this could not be confirmed in controlled trials [2-5]. No single trial had tested whether chemotherapy influenced survival. During the 1980s, single drug activity was also noted for cisplatin and etoposide, and 5-FU was increasingly biochemically modulated with methotrexate or leucovorin [6]. Various combinations including also these drugs appeared to give even higher response rates and also prolonged survival in a number of phase II trials (for a review, see [7]). In spite of the fact that chemotherapy has been extensively used in advanced gastric cancer, there were, however, no studies of its true palliative value, i.e., the extent to which it relieves tumour-related symptoms, improves

supportive care group (20%, 6/30, P < 0.05). A similar difference was seen in the treating physician's evaluation of whether the patient was subjectively improved or continued to do well for at least 4 months (17/31, 55% versus 6/30, 20%, P < 0.01). Overall survival was longer in the chemotherapy group (median 8 vs. 5 months) although the difference was not statistically significant (P = 0.12). After corrections for imbalances in pretreatment characteristics, chemotherapy treatment was, however, associated with a survival benefit (P = 0.003). Also, the quality-adjusted survival time and time to disease progression were longer for patients randomized to chemotherapy (median 5 vs. 2 months, P = 0.03). Conclusions: The results show that chemotherapy can add to both quantity and quality of life in advanced gastric cancer. The number of patients who benefit from treatment is, however, still rather limited. Key words: chemotherapy, gastric cancer, palliation, randomized study

quality of life and influences survival. A Swedish Consensus Conference in November 1990 stated that "the use of chemotherapy has no place in the routine care of patients with advanced gastric cancer" [8]. In view of this lack of knowledge, a controlled clinical trial was planned in 1990. The intention was to compare the effects of chemotherapy in addition to best supportive care against best supportive care alone, but the research ethics committee required the provision of chemotherapy upon request also in the control group. Therefore, the trial can not answer the question how much chemotherapy improves quality of life and prolongs survival but rather whether chemotherapy given to most patients is superior to a likely routine at many hospitals, namely that it is given only to selected patients, either immediately or later in the course when other measures are not sufficient to relieve tumour related symptoms. During the progress of the study, the knowledge about the effects of chemotherapy in gastric cancer has increased. Three trials have compared combination chemotherapy against a control group. Murad et al. [9] random-

164 ized 21 patients before they stopped randomization because of "unquestionable advantages in terms of the response rates and projected survival in the group receiving chemotherapy". Patient inclusion was continued in the treatment group and the results have been reported for 40 patients, 10 in the control group and 30 in the chemotherapy group. Median survival was 9 months in the treated group and 3 months in the control group (P = 0.001). Pyrhonen et al. [10] randomized 41 patients when the trial was stopped due to slow inclusion and apparently better results in the treated group. Median survival increased from 3 months to about 12 months (P = 0.001). Finally, Scheithauer et al. [11] found in a study including 37 patients that treatment improved median survival from 4 months in the control group to > 7.5 months (P < 0.05) in the treated group. Two trials have reported that one of the 'second generation' combinations is superior to the 'first generation' combinations [12, 13], although a survival difference was seen in only one trial [12]. Finally, two trials have compared various second generation combinations without finding any apparent superiority for one over another [14,15]. The study was performed in two phases - a pilot phase between January 1991 and May 1992 at one hospital followed by a multicentre phase between June 1992 and February 1995. Inclusion criteria, treatments and other routines were the same during the two phases. The present report gives the treatment results for all randomized patients. A preliminary report on the patients included in the pilot phase has been published [16].

best supportive care, given in both randomization groups, have been presented before [19].

Treatment effects Survival was calculated from the date of randomization to the date of death from any cause. At the end of the follow up on March 31 1996, 1 patient was alive in each group. Patient evaluations, performed every other month, or earlier if progressive disease was suspected, were made by the treating physician as previously described [19]. Objective response evaluations followed the International Union Against Cancer recommendations [20], and toxicity was classified according to the World Health Organization guidelines [21]. Quality of life assessments were performed using the EORTC-QLQC30 version 1.0 [22]. Besides presenting average scale scores at randomization and during follow-up, an overall categorization of whether a patient's quality of life had changed during treatment/follow-up or not was also performed by two raters (KE and KH) as described previously [19], In order to be classified as having a favourable quality of life outcome, the patient should have either an unchanged high quality of life at least to the second follow-up evaluation, i.e., for a minimum of four months or be substantially improved for the same time period. Finally, a quality of life adjustment of the survival time was made by the two raters [19].

Statistical methods All analyses were decided in advance and performed according to 'intention-to-treat'. In addition to the statistical methods detailed in a previous parallel trial in pancreatic and biliary cancer [19], a Cox proportional hazard regression model was used in order to compensate for imbalances in patient characteristics [23],

Results

Between January 1991 and February 1995, 61 patients below the age of 76 years with surgically non-curable, histologically verified gastric adenocarcinoma who did not fulfil exclusion criteria (serum creatinine level > 125 mmol/1, serum bilirubin level > 60 mmol/1, Karnofsky performance status (KPS) <50, previous chemotherapy, and other primary tumours) were mcluded after informed consent and were randomized at the Regional Oncological Centre in Uppsala to either immediate chemotherapy including best supportive care or to best supportive care where chemotherapy was allowed if the supportive measures did not result in palliation. Patients were stratified for tumour site (cardia vs. gastric) and study centre.

All pre-treatment characteristics of the randomized patients were not well balanced between the groups (Table 1). Performance status was significantly lower in the chemotherapy group, and more of those patients had multiple tumour site involvement and disease symptoms. Patients randomized to chemotherapy started treatment after a median of 7 days (Table 2). In the best supportive care group, 4 patients requested chemotherapy. In addition, 8 patients in this group ultimately obtained chemotherapy. Radiotherapy was given to slightly more patients in the best supportive care group and it was initiated earlier in that group (Table 2). Likewise, more patients had palliative surgery in the best supportive care group.

Treatment

Overall survival

In patients randomized to the chemotherapy group, treatment was to be started within two weeks of randomization. The chemotherapy regimen was the ELF-regimen [17] or in patients above the age of 60 with KPS 70 or less, sequential 5-FU and leucovorin (FLv). The ELF-regimen consisted of leucovorin 350 mg/m 2 , etoposide 120 mg/m2 and 5-FU 500 mg/ m2, given dairy on 3 consecutive days and repeated every third week. The FLv-regimen, 5-FU 500 mg/m 2 and leucovorin 60 mg/m2, was given on 2 consecutive days every second week [18]. Therapy continued until tumour progression, either objectively or subjectively, as long as toxicity was low. If a patient did not have any tumour-related symptoms at randomization, the planned number of treatments was 6 courses of ELF. The same regimens were recommended to patients in the best supportive care group, if they requested chemotherapy. The principles of

Overall survival tended to be longer in the group of patients randomized to chemotherapy (median 8 months) compared to those allocated to best supportive care (median 5 months, P - 0.12) (Figure 1). When the influence on survival was tested in a Cox model, also including sex and performance status, it was found that randomization group had an independent relation to survival (P = 0.937, SE(p) = 0.311, P = 0.003). Also a high performance status (P = -0.045, SE(P) = 0.011, P = 0.0001) and female gender (P = -0.748, SE(P) = 0.321, P = 0.02), were related to a longer survival in the Cox model. Other registered

Patients and methods Study population

165 Table 1. Patient characteristics at randomization.

Randomised Median age (range) years Sex (male/female) Site (cardia/gastric) Removed primary (yes/no) Interval" (median, range) KPS", mean (range) B-Hb g/L mean ±SD S-ALP" ukat/L ±SD S-ASAT ukat/1 ±SD S-ALAT ukat/1 +SD Tumor site Locoregional Liver Lymph nodes Peritoneal Lung Others 1 site/2 or more sites Type of symptoms Weight loss Pain Difficulty to swallow Tiredness Nausea/vomiting Others No symptoms 1 symptom only 2 or more symptoms

Table 2. Treatment of patients.

Chemotherapy1"

Best supportive careb

31 64 (45-75) 23/8 9/22 13/18 0(0-1000) 79 (50-100) 124 ± 11 5.7 ±4.3 0.68 ±0.59 0.53 ±0.38

30 63(40-74) 22/8 7/23 11/19 0(0-2855) 87(5O-10O)d 122±18 4.4 ±2.5 0.51 ±0.43 0.48 ±0.41

19 10 14 10 2 6 6/25

23 7 9

2Q

12 9 9 7

16 12 11 6 5 2 7 22

Chemotherapy* Randomised Refused allocation Chemotherapy (yes/no) Time to chemotherapy, median (range) days Type of chemotherapy (FLv/ ELF) Number of courses, median (range) Radiotherapy (yes/no) Time to radiotherapy, median (range) days Surgery (yes/no)

31 0 31/0 7 (0-27)

Best supportive care* 30 4 12/18 49(2-721)

8/23

8/4

6 (1-28)

5(1-15)

9/22 199 (87^23)

12/18 49(5-625)

4/27

9/21

° Number of patients unless otherwise indicated.

9 2 4 10/20

4

2 6 13 lld

" Interval - disease-free interval (days); KPS - Karnofsky performance status. c

The upper normal limits are: S-ALP 4.8 ukat/1 and S-ASAT and S-ALAT 0.70 ukat/1. d Refers to a statistically significant difference (P < 0.05) between the primary chemotherapy and best supportive care groups, respectively.

variables (see Table 1) were not significantly related to survival. Changes in quality of life Two patients in each group did not complete the questionnaire at randomization because of rapidly progressing disease. The average scores on most scales items and global health status did not differ between the groups at randomization although there were more problems with pain and nausea/vomiting in the chemotherapy group (Table 3). The number of patients who replied to the questionnaires declined during follow-up. This decline was more rapid in the best supportive care group (Table 3). The reasons for not replying to a questionnaire after 2 and 4 months were usually either death or that the patient was terminally ill. After the first interview, one patient in the best supportive care group refused further interviews. Table 3. Quality of life at randomization (R) and at the first two evaluations (1, 2) after 2 and 4 months, respectively, in the chemotherapy and best supportive care groups. Chemotherapy R(29)'

1(26)

Best suppportive care R(28) 2(19) 1(21) Average ± SD score

2(11)

Primary chemotherapy Best supportive cars 0.5 -

12

16 months

Figure 1. Overall survival in patients randomized to chemotherapy ( • n = 31) or to best supportive care ( Q n - 30). The difference is not statistically significant (P = 0.12).

Functional scales Physical 78 ±25 Role 67 ±34 Emotional 74 ±21 CogiuUve 81 ± 17 Social 82 ±23 Global health status 55 ± 18 Symptom scales/items Pain 33±29 b Fatigue 43 ±27 Nausea/vomiting 22 ±26" Appetite 33 ±35 Dyspnoea 18 ±23 Diarrhoea 11 ±22

74 ±30 69 ±39 73 ±23 87 ±26 79 ±31 62 ±26

70 ±28 58 ±41 73 ±21 84±21 71 ±29 63 ±22

81 +22 74 ±34 72 ±23 88 ±19 80 ±23 64 ±20

77 ±29 69 ±36 73 ±26 84 ±28 78 ±34 66 ±22

90± 16 79 ±37 76 ±25 79 + 34 82 ±27 63 ±22

24 + 22 44 ±27 18 ±22 31 ±32 19 ±23 8± 14

23 ±24 44 ±26 14 ± 19 33 ±34 26 ±28 6 + 13

18 ±20 34 ±26 9 ± 14 36 ±35 11 ± 19 14 ±22

18 ± 22 32 ±33 32 ±36 38 ±29 12 ±15 14 ±24

18 ±25 28 ±24 13 + 18 17± 14 20 ± 18 14± 17

* Thefigureswithin parentheses refer to the number of patients interviewed. Refers to a statistically significant difference (P < 0.05) between the primary chemotherapy and best supportive care groups, respectively.

166 Table 4. Worst toxicity after chemotherapy in patients randomised to chemotherapy. FLv

ELF

WHO grade

Leulcopenia Thrombocytopenia Nausea/ vomiting Diarrhoea Stomatitis Dermatitis Hair loss Infection Conjunctivitis

0

1

2

8 8

-' -

0 0

0 0

0 0

13 22

-

7

1(12)

0

0

0

13

6(26) 3(13)

0 1(13) 0 0 0

0 0 0 0 0 0 0 - 0 0 0 0

19 18 21 0 21 20

3(13) 1 (4) 0 0 1(4) 3(13) 1(4) 0 0 2 (9) 0 0 2(9) 21(91) 2 (9) 1 (4) 2 (9) 0 0

7 1(12) 5 2(25) 8 0 80 8 5 3 (38)

3 4 0 1 2 Number of patients (percent) 6(26) 0

3

4

prolonged stationary disease, and 10 had progressive disease. One of the patients in the best supportive care group treated with chemotherapy (ELF) had a PR. The median time to objective disease progression (or death in patients not objectively evaluated) was 6 (0-20+) months in the chemotherapy group and 2 (0-13) months (P 0.03) in the best supportive care group.

3(13) 1(4) 1(4) 0

Toxicity

1(4)

The overall toxicity of the FLv treatment was low (Table 4). In contrast, the toxicity of ELF was higher with grade 3 or 4 toxicities (except alopecia) in 5 (22%) patients. In addition, all ELF treated patients had more or less total alopecia. There were no toxic deaths.

0

* Not registered as toxicity.

Compared to the scores at randomization all mean scale scores remained at approximately the same level in both groups (Table 3). In the overall categorization, 14 (45%) patients were considered by the raters to have a favourable quality-oflife outcome in the chemotherapy group compared to 6 (20%) in the best supportive care group (P < 0.05). In the best supportive care group, one of these was attributed to chemotherapy, one to irradiation, three to 'a naturally prolonged asymptomatic period' and one to the effect of steroids and other palliative efforts. In the chemotherapy group, the treating physician considered 17 (55%) patients to have had either a prolonged symptom-free period or improved symptomatology in the absence of severe toxicity. This number was lower (6, 20%, P < 0.01) in the best supportive care group. The median time to subjective disease progression was 6 (020+) months in the chemotherapy group and 2 (0-14) months (P < 0.01) in the best supportive care group. When the treating physicians' and the raters' evaluations were compared, concordance was observed in all but 7 (11%) patients. The average quality-adjusted survival was longer in the group of patients randomized to chemotherapy than in the best supportive care group (median 6 vs. 2 months, P - 0.03). The average proportion of the total survival time with high or improving quality of life was also higher in the chemotherapy group (69% vs. 47%, P < 0.05). Objective responses There was no requirement for the patients to have measurable disease prior to entry, nor was there a requirement that measurements were to be performed during treatment. However, repeated measurements of tumour size were often done in order to monitor treatment. In these patients in the chemotherapy group, 1 patient had a complete remission, 4 had a partial remission (PR) (overall response rate 23%, ELF 4/16, FLv 1/6), 7 (32%) had a

Discussion This study strongly indicates that chemotherapy can prolong survival and improve quality of life in a proportion of the patients with gastric cancer, but the precise magnitude of the survival benefit is difficult to estimate. Due to imbalances in pre-treatment characteristics, the survival benefit was not statistically significant in the univariate analyses (P = 0.12), whereas a multivariate analysis revealed a highly statistically significant difference (P = 0.003). Further, patients facing a rapidly progressing fatal disease such as advanced gastrointestinal cancer often demand active treatment, even if potentially toxic [24] and of unproven value. Therefore, the requirement of the research ethics committee to provide chemotherapy upon request in the control group, will always preclude an untreated control group. Since 4 patients in the best supportive care group refused allocation and were treated immediately with chemotherapy, and a total of 40% of the patients in the control group received chemotherapy, it is possible that the survival benefit is underestimated. The survival in the control group, if no chemotherapy had been given at all, can of course only be speculated upon. In comparison to the five months seen here, the median survival of three or four months, reported in the three other trials including an untreated control group [9-11], gives some indication of its duration, although such comparisons are difficult due to unknown selection mechanisms. These 4 trials, although including a total of only 159 patients randomized between chemotherapy and no (initial) chemotherapy thus indicate that survival is prolonged by approximately 6 months (range 3-9 months). Stronger scientific proof of survival prolongation would have been desirable, particularly since a proper comparison was only made in 3 of the trials but in the light of the difficulties to include patients in a controlled clinical trial with a no active treatment group, this will be difficult to reach. In this trial, 17 (26%) of 66 patients in Uppsala in which informed consent prior to randomization was aimed at refused inclusion. Most (12, 71%) of these patients wanted active treatment outside the trial. The pro-

167 Table 5. Quality of life outcomes overall and in relation to objective and subjective responses in the two randomized groups.

Tumour site Gastric Objective response CR PR SD PD Not evaluated Subjective response1" Symptom-free/ improved Unchanged/ deteriorated

Chemotherapy group

Best supportive care group

Improvement/ prolonged high quality of life*

All others

Improvement/ prolonged high quality of life*

All others

12(39)

19(61)

6/20

24(80)

1 1 3 1(10) 6

0 3 4 9(90) 3

0 0 1 1(5) 4

0 1 0 20 (95) 3

12(71)

5(29)

4

0

14(100)

2(8)

2 22 (92)

* QoL-patient, minimum duration 4 months. As evaluated by the treating physician.

portion of patients unwilling to accept inclusion at the other participating hospitals is not known. There are no data in the literature favouring the use of a particular cytostatic drug/combination in gastric cancer. Based upon a presumed favourable toxicity profile and a demonstrated antitumoural activity in phase II trials, not apparently inferior to other 'second generation' combinations, we selected the ELF-regimen. This regimen was originally designed for patients above the age of 65 not considered to tolerate more intensive regimens [17]. After having piloted it as a palliative treatment prior to the start of the trial, we found it too toxic for the majority of the elderly patients. We therefore selected a 5-FU/leucovorin regimen, extensively used in colorectal cancer [18], in the elderly patients with poor performance status. We saw definite toxicity together with good responses, also in younger patients. A recent EORTC trial has shown that the antitumour activity of ELF in gastric cancer is similar to other, commonly used doxorubicin (FAMtx) and cisplatinum (FUP) containing regimens [15]. More recently, a regimen containing epirubicin, cisplatin and protracted venous infusion 5-FU (ECF) has been found to give a survival advantage compared to FAMtx in a preliminary analysis of a multicenter trial [25]. Whether it is superior as a palliative treatment is not yet known. In this, and in the parallel trial in pancreatic and biliary cancer [19], the quality of life assessments were done with the EORTC QLQ-C30. Even if this questionnaire has excellent psychometric properties [22], and is therefore used extensively, it has, however, not been used longitudinally to any major extent in prospective randomized trials. The expected difficulties with attrition [26, 27] in all longitudinal studies, particularly in the palliative situation where survival may be short, and how the increasing number of drop-outs with time complicates the analyses and interpretation of longitudinal data were ex-

tensively discussed in the report on the parallel trial [19]. In the previous publication, we also described and discussed the categorization used to estimate whether a patient had improved or not. Although, this categorization was usually straightforward with either clear improvements/stable high quality of life or clear deterioration/no change in a patient with a poor quality of life, greater difficulties were noted in this trial. The possibility to use the EORTC QLQ C-30 for this purpose will be the topic of a separate publication. The present study has shown that improved quality of life/symptomatic relief without too debilitating adverse effects occur in slightly less than half of the patients, i.e., more than occasionally and more often than what 'intensive' best supportive measures (as practised in this study) can achieve. A practical consequence of this result is that the reluctance felt by many physicians to inform their patients about this option should be reduced. On the other hand, we do not advocate the routine use of this treatment in the light of the yet limited effects and the costs [28], but rather to consider it to selected patients after realistic and adequate information. If the patients accept treatment, it is important to detect non-responding patients at an early stage and be prepared to terminate further treatment.

Acknowledgement This work was supported by the Swedish Cancer Society.

References 1. Macdonald JS, Schein PS, Woolley PV et al. 5-Fluorouracil, mitomycin C and adriamycin (FAM): A new combination chemotherapy program for advanced gastric carcinoma. Ann Intern Med 1980; 93: 533-6. 2. Kovach JS, Moertel CG, Schutt AJ et al. A controlled study of combined 1,3-bis (2 chloroethyl)-l-nitrosourea and 5-fluourouracil therapy for advanced gastric and pancreatic cancer. Cancer 1974; 33: 563-7. 3. Cocconi G, De Lisi V, Di Blasio B. Randomized comparison of 5-FU alone or combined with mitomycin C and citarabine (MFC) in the treatment of advanced gastric cancer. Cancer Treat Rep 1982; 66: 1263-6. 4. De Lisi V, Cocconi G, Tonato M et al. Randomized comparison of 5-FU alone or combined with carmustine, doxorubicin and mitomycin (BAMR) in the treatment of advanced gastric cancer: A phase III trial of the Italian Clinical Research Oncology Group (GOIRQ. Cancer Treat Rep 1986; 70:481-5. 5. Cullinan SA, Moertel CG, Wieand HS et al. Controlled evaluation of three drug combination regimens versus fluorouracil alone for the therapy of advanced gastric cancer. J Clin Oncol 1994; 12:412-6. 6. Preusser P, Achterrath W, Wilke H et al. Chemotherapy of gastric cancer. Cancer Treat Rev 1988; 15: 257-77. 7. Willce H, Preusser P, Fink U et al. New developments in the treatment of gastric carcinoma. Semin Oncol 1990; 1: 61-70. 8. Consensus Statement Cytostatic Drug Treatment in Advanced Cancer. Stockholm: MFR Spri 1990. 9. Murad A, Santiago F, Petroianu A et al. Modified therapy with 5-fluorouracil, doxorubicin, and methotrexate in advanced gastric cancer. Cancer 1993; 72: 37^»1. 10. Pyrhdnen S, Kuitunen T, Nyandoto P et al. Randomized compar-

168

11.

12.

13.

14.

15.

16.

17.

18.

19.

ison of fluorouracil, epidoxorubicin and raethotrexate (FEMTX) plus supportive care with supportive care alone in patients with nonresectable gastric cancer. Br J Cancer 1995; 71: 587-91 Scheithauer W, Kornek G, Zek B et aL Palliative chemotherapy versus supportive care in patients with metastatic gastric cancer: A randomized trial. Second Int Conf Biol, Prev, Treatm GI Malignancy, Koln, Germany 1995; 68 (Abstr). Wils JA, Klein HO, Wagener DJT et al. Sequential high-dose methotrexate and fluorouracil combined with doxorubicin - a step ahead in the treatment of advanced gastric cancer A trial of the EORTC Gastrointestinal Tract Cooperative Group. J Clin Oncol 1991; 9: 827-31. Cocconi G, Bella M, Zironi S et al. Fluorouracil, doxonibicin, and mitomycin combination versus PELF chemotherapy in advanced gastric cancer: A prospective randomized trial of the Italian Oncology Group for Clinical Research. J Clin Oncol 1994; 12: 2687-93. Kelsen D, Atiq O, Saltz L et al. FAMTX versus etoposide, doxonibicin, and cisplatin: A random assignment trial in gastric cancer. J Clm Oncol 1992; 4: 541-8. Rougier P, Wils J, Eilke et al. Advanced gastric cancer: Comparison of FAMTX (5-FU, adriamycin, methotrexate) versus ELF (etoposide, 5-FU, leucovorin) versus FUP (infusional 5-FU + cisplatin). Results from an EORTC trial of the GITCCG and the AIO. Proc Eur Conf Clin Oncol 1995; 8. 546 (Abstr). Glimelius B, Hoffman K, Haglund U et al. Initial or delayed chemotherapy with best supportive care in advanced gastric cancer. Ann Oncol 1994; 5: 189-90. Wilke H, Preusser P, Stahl M et al. Etoposide, folinic acid, and 5-fluorouracil in carboplatin-pretreated patients with advanced gastric cancer. Cancer Chemother Pharmacol 1991; 29: 83-4. Nordic Gastrointestinal Tumour Adjuvant Therapy Group. Biochemical modulation of 5-fluorouracil: A randomized comparison of sequential methotrexate, 5-fluorouracil and leucovorin versus sequential 5-fluorouracil and leucovorin in patients with advanced symptomatic colorectal cancer. Ann Oncol 1993; 4: 235-41. Glimelius B, Hoffman K, SjSden PO et al Chemotherapy improves survival and quality of life in advanced pancreatic and biliary cancer. Ann Oncol 1996; 7: 593-600.

20. Hayward JL, Rubens RD. Assessment of response to therapy in advanced breast cancer. Br J Cancer 1977; 35: 292-8. 21. Miller AB, Hoogstraten B, Staquet et al. Reporting results of cancer treatment. Cancer 1981; 47: 207-14. 22. Aaronson NK, Ahmedzai S, Bergman B et al. The European Organization for Research and Treatment of Cancer QLQ-C30: A quality-of-life instrument for use in clinical trials in oncology. J Natl Cancer Inst 1993; 85: 365-76. 23. Cox DR. Regression models and life tables. JR Stat Soc 1972; 34: 187-20. 24. Slevin ML, Stubbs L, Plant HJ et al. Attitudes to chemotherapy: Comparing views of patients with cancer with those of doctors, nurses, and general public. BMJ 1990; 300:1458—60. 25. Webb A and Cunningham D: Curing gastric cancer - hone the scalpel with magic? Br J Cancer 1996; 73: 418-9 (Editorial). 26. Hopwood P, Stephens RJ, Machin D for the MRC Lung Cancer Working Party. Approaches to the analysis of quality of life data: Experiences gained from a Medical Research Council Lung Cancer Working Party palliative chemotherapy trial. Qual Life Res 1994; 3: 339-52. 27. Sigurdardottir V, Bolund C, Sullivan M. Quality of life evaluation by the EORTC questionnaire technique in patients with generalized malignant melanoma on chemotherapy Acta Oncol 1996; 35: 149-59. 28. Glimelius B, Hoffman K, Graf W et al. Cost-effectiveness of palliative chemotherapy in advanced gastrointestinal cancer. Ann Oncol 1995; 6: 267-74. Received 23 September 1996; accepted 3 December 1996.

Correspondence to: Bengt Glimelius, MD Department of Oncology University of Uppsala S-751 85 Uppsala Sweden