5 Year Follow-up of First British Stomach CancerGroup Trial
WILLIAM H. ALLUM MICHAEL T. HALLISSEYKRYSTYNA A. KELLY
FOR THE BRITISH STOMACH CANCER GROUP*
Department of Surgery and West Midlands CRC Clinical TrialsUnit, Queen Elizabeth Hospital, Birmingham
Summary 411 patients were entered into a
prospective, randomised controlled trialof adjuvant chemotherapy after gastrectomy foradenocarcinoma. After a follow-up of at least 5 1/2 years therehas been no survival advantage for those receiving adjuvant5-fluorouracil and mitomycin C with or without an
induction course of 5-fluorouracil, vincristine, cyclo-phosphamide, and methotrexate compared with those
undergoing surgery only. There have been 366 deaths,including 22 from treatment-related conditions. Amultivariate analysis of prognostic factors demonstratedthat stage of disease, nodal and resection margininvolvement, and the presence of residual disease are
significant determinants of survival. Weight loss beforesurgery had a significant independent influence on survival.The combination of preoperative symptoms and
intraoperative findings may be used to select patients forradical or palliative procedures.
*Committee of the British Stomach Cancer Group and research registrars:Sid Arnott, Victor S. Brookes, John L. Craven, David J. Ellis, John W. L.Fielding, Sarah L. Fagg, Walter Gregory, Michael S. Hockey, Bruce G.Jones, M. C. Mason, Abraham Minawa, Adrian Timothy, Linda Ward, JohnA. H. Waterhouse, Stuart Winsey, and Peter F. M. Wrigley.
IN 1976 the British Stomach Cancer Group began amulticentre, prospective, randomised controlled trial of tworegimens of adjuvant chemotherapy in operable gastriccancer. The regimens combined 5-fluorouracil and
mitomycin C, which at that time were the two most activeagents when used alone or in combination. The controlsreceived placebo. Recruitment ended in May, 1981. Aninterim report was published with follow-up complete toAug 15, 1981.1 We now describe follow-up of at least 5tyears for all evaluable patients and we report prognosticfactors.
Patients and Methods
The protocol has been described.’ In summary, this was amulticentre trial in England and Wales. Patients aged 15-74 whohad undergone resection for histologically proven adenocarcinomaof the stomach were eligible. Residual microscopic or macroscopicdisease did not exclude patients. Exclusion criteria were previousmalignancy, previous chemotherapy or radiotherapy, stage 1 or
stage 4 disease, intestinal or biliary obstruction that could not berelieved surgically, or a blood urea of 12 mmol/1 or higher. Patientshad to be entered within 12 weeks of surgery. The patients werestratified by centre, age, sex, duration of symptoms, and stage ofdisease (see table 1 in ref 1), and randomised sequentially into one ofthree intravenous treatment groups after giving informed verbalconsent. Group A (n = 130) was the control patients, who receivedsaline every 3 weeks. Groups B (n= 140) and C (n =141 ) received5-fluorouracil 15 mg/kg and mitomycin C 150 g/kg every 3 weeks;group B received in addition a 5-day induction course of
cyclophosphamide, 5-fluorouracil, vincristine, and methotrexate.1Other clinical details, including surgical and histopathological data,were recorded. The induction course in group B was given as aninpatient; all other treatments were given on an outpatient basis andcontinued for 2 years unless withdrawn at the patient’s request orbecause of unacceptable side-effects or until death. Before everycourse of treatment, haematological and biochemical values,
TABLE I-PATIENTS’ CHARACTERISTICS
*Median; t no of patients.
including urinalysis, were measured and dosages were adjustedaccordingly. At every visit, clinical progress, laboratory data, drugdosages, and side-effects were recorded.The duration of survival was the criterion of "response".
Kaplan-Meier survival curves were drawn and the significance ofthe differences between curves was assessed with the log rank test.For the comparison between treatment groups, an overall X2 valueand a X2 value adjusted for stratification were calculated. In ananalysis of prognostic factors both univariate log rank analyses andmultivariate analyses based on the Cox proportional hazard modelwere done. For the Cox multiple regression, only cases withcomplete data for all the variables considered could be used (ie, 400patients). A stepwise procedure was used. Only factors whoseregression coefficients were significant at the 1% level wereconsidered to predict prognosis independently.
411 of the 455 patients originally randomised have beenevaluated. Details of the 44 exclusions have been described.1The median follow-up for all patients is 100 months with aminimum of 66 months. The patient characteristics arelisted in table 1. There was no major imbalance between thegroups.
The overall probability of survival is shown in fig 1. Themedian survival was 15-5 months (95 % confidence interval[CI] 14-17). At the time of analysis 45 patients were aliveand 366 had died. The longest duration of survival was 112months.
Fig l--Overall survival (366/411 deaths).
Fig 2-Survival by treatment.
Group A, 113/130 deaths; B, 127/140; and C, 126/141 (p=0’98).
Effect of Treatment
There was no significant difference between the threetreatment groups, with or without stratification (fig 2). Themedian survival was 43 days longer for the treated patientsthan for controls, with a 95% CI for the true difference of- 3 to 89 days. Treatment is unlikely to prolong survival bymore than 3 months.
Cause of Death
The cause of death in 366 patients (table II) was obtainedfrom the general practitioner or consultant in charge of thecase. 41 cases (11%) had necropsies. Malignant diseasecaused 322 deaths, 318 from recurrent disease and 4 fromnew primary tumours. These 4 included 2 patients who hadcarcinomas of the bronchus, 1 with carcinoma of the colon,and 1 patient in group C who had acute myeloblasticleukaemia induced by chemotherapy. There were 22treatment-related deaths, 4 during the induction course ingroup B, 1 from late marrow failure, and 1 from fibrosingalveolitis. 16 deaths were due to the haemolytic uraemicsyndrome; 15 occurred after the first year of treatment,which reflects the hypothesis that this condition is associatedwith prolonged exposure to mitomycin C.2,3 There were 13patients without residual disease at necropsy who weredocumented with residual disease at initial surgery.
ToxicityThe clinical, haematological, and biochemical side-effects
were as described.1 The frequency of all side-effects wassignificantly higher in the treatment groups than in theplacebo group. In addition to the deaths from the
haemolytic uraemic syndrome, there were a further 14patients in whom the condition was detected. All the
patients who had this syndrome were in the active treatmentgroups.
The following clinical features were not prognostic: sex,age (under or over 60), preoperative weight, and normal
weight. Contrary to the initial findings,’ the duration ofsymptoms was no longer significant (p =0-07). Defined asthe proportion of normal weight lost before treatment,weight loss was significant (p < 0-0001).The surgical features of most significance were those
describing the macroscopic extent of disease at initial
exploration and after resection: tumour size (p=0’02),residual tumour (p < 0-0001), metastases (p < 0-0001), livermetastases (p < 0-001), peritoneal metastases (p < 0-0001),and other metastases (p < 0 O1 ). The number of tumour siteswas not significant (p=0-08). Gastrectomy, classified astotal, proximal partial, and distal partial, was significant(p < 0-001). Distal partial gastrectomy had a better prognosisthan proximal partial gastrectomy, with total gastrectomyhaving the worst prognosis. Splenectomy was significantlyrelated to survival (p<0’01), although removal of otherorgans was not (oesophagus, pancreas, colon, mesocolon, oromentum).As we found before,’ clincopathological staging was a
deternlinant of survival (p < 00001; fig 3). Of the otherpathological fmdings, tumour size (p < 001), number ofsites (p<0-02), nodal involvement (p < 0-0001), andresection line involvement (p < 0-0001) were significant.The fact that serosal involvement was not significant(p=0-012) reflects the entry criteria rather than a lack ofclinical significance. Stage I cases were not eligible and only11 % of the patients had no serosal penetration.The univariate analysis assesses factors individually
without allowing for their inter-relations. For example,tumour site and size are closely related to the type ofgastrectomy. Therefore the Cox proportional hazardsmodel was used to investigate the independent prognosticsignificance of the variables in the univariate analysis. Theproportionality assumption was tested by stratifying thedata by each of the variables in turn: there was no evidencefor non-proportionality. The optimum scale ofmeasurement was chosen for each variable. The type of
gastrectomy was ranked according to independent evidencedemonstrating that total gastrectomy has a poorer prognosisthan proximal partial gastrectomy which has a poorerprognosis than distal partial gastrectomy.4When only the clinical features were used, weight loss was
the one factor significantly related to prognosis (p < 0-001),confirming the log rank analyses. For clinical and surgicalfindings, residual tumour (p<0-001) and type of
gastrectomy (p < 0-01) were significantly related to survival,and weight loss remained a significant independent
predictor of survival (p < 0-001). However, after wecontrolled for these three variables, none of the remainingvariables was significant. When clinical, surgical, andpathological findings were analysed, residual tumour, nodalinvolvement, resection line involvement, and weight losswere significant variables (p < 0-001). Serosal involvementwas not significant, confirming the log rank analysis, andmetastases were of borderline significance. That resectionline involvement is an independent prognostic factorconfirmed a previous analysis.s This factor has been
incorporated into the staging system being used in thesecond British Stomach Cancer Group trial (unpublished).This is vindicated because the use of resection lines gave a
significant improvement in the ability to predict survivalcompared with stage as defined in this trial (p = 0-0005). Inaddition, weight loss combined with stage of disease, asdefined in the second BSCG trial, showed a significantimprovement in the ability to predict survival comparedwith stage alone (p = 0-0001), which suggests that weightloss could also be usefully incorporated into the stagingsystem. No other variable was able to improve significantlyupon the staging system used in the second British StomachCancer Group trial.
Our trial failed to reject the null hypothesis that
chemotherapy confers no survival advantage when used asan adjuvant to resection in gastric adenocarcinoma. The95% CI suggests that chemotherapy is unlikely to prolongsurvival by more than 3 months, which would be of littleclinical importance even if such a difference could bedemonstrated to be of statistical significance in a largerstudy. The negative finding of this trial, however, should notdeter future developments to improve treatment of thiscondition.The failure of chemotherapy to improve survival in this
study and others6.7 questions the role of such treatment inthis disease. Many cytotoxic agents have limited activity ingastric cancer and perhaps it should be accepted that drugswhich produce only partial responses are unlikely to confersurvival benefits and should not be assessed further. Butcurrent treatments fail to control this disease and new
therapies must be explored. In addition there are reports ofsignificant benefits from chemotherapy in gastric8 andbreast9 cancer. The feature common to both these studieswas the use of a single cytotoxic agent started early in theperioperative period. In gastric cancer, mitomycin C startedwithin 48 h of surgery produced a 10% survival benefit inthe treated group. We are now evaluating early treatmentwith a single agent in a prospective randomised trial.We also showed the value of multivariate methods in
analysing factors that are independent indicators of disease.Stage of disease and the factors contributing to it wereconfirmed as important prognostic indicators. In addition,the multivariate analysis confirmed an earlier reports aboutthe importance of resection line involvement for prognosis,demonstrating its independent effect. The British StomachCancer Group recommends that patients with residualdisease should now be considered to have had a palliativeresection (stage 4A). In our study preoperative weight loss,which is not specifically related to the disease, was identifiedas an independent prognostic factor. It may be important toincorporate non-specific factors into a staging system, suchas in Hodgkin’s disease. Additionally multivariate analysishelps to clarify the relevance of factors found to be
significant by univariate analysis-splenectomy reducedsurvival significantly when analysed in isolation but themultivariate analysis showed that it is the extent of tumourwhich required a splenectomy that is of importance, not thesplenectomy itself.
Gastric cancer remains a common problem. Identifiableamong factors influencing survival are (1) stage of disease,(2) perioperative mortality, (3) extent of surgery, and (4)failure of adjuvant therapies for advanced disease. Screeningprogrammes for gastric cancer reduce mortality by 25% as aresult of the detection and surgical treatment of "early"disease.10 Much effort is directed in the UK to screening forcervical cancer (2000 deaths per annum) and breast cancer(14 000 deaths per annum), but little attention is paid togastric cancer (10 000 deaths per annum). However, pilotearly-detection programmes have resulted in an increase instage 1 disease." This is an area that should be further
developed. Whilst screening programmes are effective, mostcases will still be diagnosed at an advanced stage andmanagement remains a high priority. For the surgicallyresectable lesions there are two areas worthy of exploration:perioperative mortality and the role of surgeons in localcontrol of disease. Data from the West Midlands RegionalCancer Registry reveals a perioperative (30 day) mortality of13% for radical partial gastrectomies and 30% for radicaltotal gastrectomies. The result for total gastrectomies is
significantly related to the number of resections done peryear by a surgeon. These results compare unfavourably withthe operative mortality of cardiac or vascular surgery.However, an operative mortality of 2-5 % can be achieved inspecialised units. Whilst many consider that gastrectomyshould be a "general" surgical operation, the facts dictatethat specialised units must be developed. The role of surgeryin the local control of disease is awaiting evaluation in theMedical Research Council Surgical Trial. The rationale isthe high incidence of local failure after a radical gastrectomyin contrast to the reduced incidence of local recurrence andimproved survival following a more extended radical
gastrectomy (R2).We thank the Cancer Research Campaign, Medical Research Council, and
United Birmingham Hospitals Endowment Fund for support. Trialcoordination and data collection and analysis was done by the West MidlandsRegional Cancer Registry and the West Midlands Cancer Research
Campaign Clinical Trials Unit. We thank all members of and pathologistsassociated with (listed in ref 1) the British Stomach Cancer Group.
Correspondence should be addressed to M. T. H., Department of Surgery,Queen Elizabeth Hospital, Birmingham B15 2TH.
1. Fielding JWL, Fagg SL, Jones BG, et al. An interim report of a prospective,randomized, controlled study of adjuvant chemotherapy in operable gastric cancer:British Stomach Cancer Group. World J Surg 1983; 7: 390-09.
2 Jones BG, Fielding JW, Newman CE, Howell A, Brookes VS. Intravascular
haemolysis and renal impairment after blood transfusion in two patients onlong-term 5-fluorouracil and mitomycin C. Lancet 1980, i: 1275-57.
3. Rumpf KW, Reiger J, Lankisch PG, von Heyden HW, Nagel GA, Scheler FMitomycin C-induced haemolysis and renal failure Lancet 1980, ii 1037-38
4. Adashek K, Saugel J, Longmire WP Cancer of the stomach review of ten yearintervals. Ann Surg 1979; 189: 6-10.
5. Hockey MS, Fielding JWL, Kelly KA, et al. Resection line disease in stomach cancer.Br Med J 1984; 289: 601-03.
6 Longmire WP Jr, Kuzuma JW, Dixon WJ The use of triethylenthiophosphoramideas an adjuvant to the surgical treatment of gastric carcinoma. Ann Surg 1968, 167:293
7. VA Cooperative Surgical Adjuvant Study Group. Use of thio-TEPA as an adjuvant tothe surgical management of carcinoma of the stomach. Cancer 1965, 18: 291.
8. Imanaga H, Nakazato H. Results of surgery for gastric cancer and effect of adjuvantmitomycin C on cancer recurrence. World J Surg 1977; 1: 213-21.
9. Nissen-Meyer R, Kjellgren K, Malmino K, Mansson B, Norris T. Surgical adjuvantchemotherapy: results with one short course with cyclophosphamide after
mastectomy for breast cancer Cancer 1978, 41: 2088-9810. Miwa K Cancer of the stomach in Japan. Jpn J Cancer Res 1979; 22: 61-7511. Allum WH, Hallisey MT, Dorrell A, Lowe J, Fielding JWL Programme for early
detection of gastric cancer. Br Med J 1986, 293: 541.
COMPARISON OF RIFAMPICIN WITHPHENOBARBITONE FOR TREATMENT OF
PRURITUS IN BILIARY CIRRHOSIS
LAURA BACHS1MONTSERRAT ELENA2
ALBERT PARÉS1CARLOS PIERA3
Liver Unit1 and Clinical Biochemistry2 and Nuclear Laboratories,3Hospital Clinic i Provincial, University of Barcelona, Villarroel,
170-08036 Barcelona, Spain
Summary The anti-pruritic effects of rifampicin(10 mg/kg) and phenobarbitone (3 mg/kg)
were assessed in 22 patients with primary biliary cirrhosis ina crossover randomised clinical trial. Each agent was givenfor 14 days, with a 30-day washout period betweentreatments. 21 patients completed the course of rifampicinand 18 that of phenobarbitone; rifampicin was withdrawnfrom 1 patient when anaemia and renal failure developed,whereas 3 patients stopped taking phenobarbitone becauseof a rash and the 4th merely refused the drug. Rifampicinhad a greater anti-pruritic effect than phenobarbitone. Thesymptom improved in 19 patients taking rifampicin and in 8taking phenobarbitone, the degree of improvement beinggreater with rifampicin than with phenobarbitone. Pruritusdisappeared in 9 patients receiving rifampicin, and three ofthem were free of itch when switching over to
phenobarbitone. Both drugs were equally effective in
inducing hepatic microsomal function but rifampicin hasthe additional effect of reducing cholestasis. Its anti-pruriticeffect should be tested in long-term clinical trials.
PRURITUS is a distressing symptom of chronic
cholestasis, particularly in patients with primary biliarycirrhosis (PBC).1.2 Treatments used for relieving pruritusin PBC include cholestyramine,3 3 phenobarbitone,4-5androgenic steroids,6,7 plasmapheresis,8.9 phototherapy,"histamine antagonists," and rutosides. 12 Recent studies
suggest that rifampicin, an antibiotic that enhances hepaticmicrosomal function, could also relieve pruritus in patientswith PBC. In one study there were only 6 patients who hadbeen treated with rifampicin for 2 weeks.14 In another study9 patients received rifampicin and placebo in a crossoverrandomised manner.1S These patients also received
cholestyramine at the same time, in accordance with theseverity of itching, so the anti-pruritic effect could not beattributed entirely to rifampicin.We have compared the anti-pruritic effect of rifampicin
with that of phenobarbitone over a 2-week period in patientswith PBC. We have also assessed the effects of the two drugson liver function and cholestasis.
Patients and Methods
The subjects were 22 women, mean age 49 (SD 84) years, whohad clinical, biochemical, immunological, and histological featuresof PBe. 1.2 All patients had pruritus. 17 had previously been treatedwith cholestyramine, and 7 had also been treated with pheno-barbitone, agents which were withdrawn for at least a monthbefore entry to the trial. Patients gave informed consent for the
study, which was approved by the ethical committee of the HospitalClinic i Provincial.