- Email: [email protected]
Randomised controlled trial of conservation therapy for breast cancer: 6-year analysis of the Scottish trial
THE LANCET
Randomised controlled trial of conservation therapy for breast cancer: 6-year analysis of the Scottish trial A Patrick Forrest, Helen J Stewart, Dawn Everington, Robin J Prescott, Colin S McArdle, Adrian N Harnett, David C Smith, W David George, on behalf of Scottish Cancer Trials Breast Group
Summary Background To determine whether, when primary breast cancer is treated by local excision supported by systemic therapy appropriate to the oestrogen receptor status (ER) of the tumour, local radiotherapy can be avoided. Methods We carried out a randomised controlled trial in 585 patients aged less than 70 years with primary breast cancers of 4 cm or less in size in four specialist units and seven other hospitals in Scotland. After local excision of the tumour (1 cm margin) and an axillary lymph-node clearance or sample, all patients received systemic therapy with oral tamoxifen 20 mg daily or six 3-weekly intravenous bolus injections of cyclophosphamide 600 mg, methotrex ate 50 mg, and fluorouracil 600 mg per m2, depending upon the ER concentration in the primary tumour. Patients were then randomly allocated to postoperative radical radiotherapy (50 Gy to breast with boost to the tumour bed) or to no further local treatment. The median follow-up of living patients was 5·7 years. The primary analysis was by intention to treat but since some patients did not receive systemic therapy appropriate to their ER status, a subsidiary analysis was restricted to 464 patients in whom all details of the protocol had been observed. Findings In the primary analysis survival was equal in the radiotherapy and non-radiotherapy groups (hazard ratio [HR] 0·98, 95% CI 0·67–1·44). Event-free survival showed an advantage in the irradiated patients (HR 0·54, 95% CI 0·39–0·74), largely due to fewer loco-regional relapses (HR 0·20, 95% CI 0·12–0·33). The relapse rate in the ipsilateral breast was 24·5% in the non-irradiated group and 5·8% follow ing breast irradiation. The subsidiary analysis confirmed these findings and indicated the advantage of radiotherapy irrespective of ER concentration. There was a non-significant trend towards fewer distant metastases in the irradiated group. Interpretation After local excision of a primary breast cancer, we conclude that radiotherapy to the residual breast tissue is advisable even when selective adjuvant systemic therapy is given.
Lancet 1996; 348: 708–13 Scottish Cancer Trials Office, Medical School, University of Edinburgh, Edinburgh, UK (A Patrick Forrest FRCS, H J Stewart FRCR); Medical Statistics Unit (D Everington MSc , R J Prescott PhD); Department of Surgery, Western General Hospital, University of Edinburgh (C S McArdle FRCS); Department of Radiotherapy, Beatson Oncology Centre, Glasgow (A N Harnett FRCR); Department of Surgery, Victoria Infirmary, Glasgow (D C Smith FRCS); and Department of Surgery, Western Infirmary, University of Glasgow (W D George FRCS) Correspondence to: Sir Patrick Forrest, Scottish Cancer Trials Office, Medical School, University of Edinburgh, Teviot Place, Edinburgh EH8 9AG, UK
708
Introduction In 1982, when the need for mastectomy was being questioned, the Scottish Cancer Trials Breast Group decided to test the feasibility of mounting a randomised controlled trial to compare mastectomy with local excision and radiotherapy. It was soon apparent that the number of women prepared to accept such randomisation was limited.1 Consequently, with increasing evidence that these two forms of local treatment were equivalent,2–4 and also that adjuvant systemic therapy might protect against loco-regional relapse, it was decided to initiate a trial in women accepting breast-conserving therapy. The aim of the trial was to determine whether radiotherapy was still necessary routinely for women who had received systemic therapy appropriate to their oestrogen receptor (ER) status following local excision of their tumour.
Methods Between April 1, 1985, and Oct 2, 1991, 589 consenting patients, aged 70 years or less with invasive breast cancer of clinical size 4 cm or less, were included provided there was no fixation of the primary tumour or axillary lymph nodes. Premenopausal women with involved axillary nodes treated by conservation were not eligible because they were being entered into the Scottish adjuvant trial of chemotherapy versus ovarian ablation (trial A).5 Also excluded were those with a second tumour in the same or opposite breast, those with a history of previous invasive cancer at any site, and those with medical or other factors contraindicating adherence to the protocol. Minimum preoperative investigations were plain radiographs of the chest and pelvis. The protocol was agreed by members of the Scottish Cancer Trials Breast Group and circulated, with a request for collaboration, to all surgeons, radiotherapists, and medical oncologists working in Scotland. Ethical approval of the protocol, which provided for informed consent, was required from each participating centre. Local excision was defined as removal of the tumour with a 1 cm margin of normal breast tissue. If a preoperative needle biopsy had been done, the skin puncture site together with the needle track was to be included in the excision, as was the nipple in centrally placed tumours if there was recent associated retraction or Paget’s disease. Participants were asked to adhere to a single surgical policy for the axilla. This might be either sampling (3–4 nodes) or surgical clearance (levels I–III). Routine histopathological examination of the specimen was carried out in the related pathology laboratory. Confirmation that excision margins were clear of tumour was not a requirement for entry to the trial but, as most specialist breast units had a policy of “clear margins”, those requiring re-excision were accepted. ER concentrations were determined by the dextran-charcoal assay.6,7 The laboratories conducting most of the assays were participating in the external quality control scheme of the European Organisation for Research and Treatment of Cancer (EORTC). Results were available for 87% of the patients. As in our previous adjuvant trials, 5,8 ER levels were designated “negative” (<5 fmol/mg protein), “poor” (5–19 fmol/mg protein), or “positive with clinical relevance” (肁20 fmol/mg protein). The surgeon was responsible for initial assessment of ER data and arrangement of appropriate systemic therapy. When ER was 20 fmol/mg protein or more this was oral tamoxifen 20 mg daily
Vol 348 • September 14, 1996
THE LANCET
Radiotherapy (n=291)
Registered for trial = 589
Randomised for radiotherapy = 293
Randomised for no radiotherapy = 296
Median age at entry in year (range)
57 (28–70)
57 (28–70)
Premenopausal
78
78
Postmenopausal (%)
213 (73)
Median clinical tumour size (cm) Number (%) 聿2 cm
Available for primary analysis = 291 Lost to follow-up = 0 Median follow-up 5·14 yr
Violation of randomisation = 6 No ER available = 40 Inappropriate systemic therapy = 17
Additional analysis (all correct)= 228 (tamox ifen = 165 and CMF = 63)
Inappropriately randomised = 2
Available for primary analysis = 294 Lost to follow-up = 0 Median follow-up = 5·49 yr
Violation of randomisation = 6 No ER available = 36 Inappropriate systemic therapy = 16
Additional analysis (all correct) = 236 (tamoxifen = 166 and CMF = 70)
Figure 1: Trial profile for Scottish Conservation Trial for at least 5 years or until relapse of disease, and, when less than 20 fmol/mg protein, six 3-weekly intravenous bolus injections of cyclophosphamide 600 mg, methotrexate 50 mg, and fluorouracil 600 mg per m2 (CMF). Those without an ER assay result were to receive tamoxifen. Patients were registered with the Scottish Cancer Trials Office, (SCTO) when eligibility was confirmed. They were stratified within six subgroups according to menstrual status (within or more than 12 months from last menstrual period) and ER status as initially available (levels of 肁20, <20, or ER unknown) and the intended treatment identified from the appropriate computergenerated random list, namely postoperative irradiation to the residual ipsilateral breast or no further immediate local treatment. The correctness of stratification by ER was subsequently checked with the related laboratory. When randomised for radiotherapy, the recommended maximum megavoltage dose to the residual breast was 50 Gy in 20–25 daily fractions. A boost to the tumour bed by iridium implant of 20–30 Gy or by electron or orthovoltage external beam irradiation of 10–15 Gy was required. In those cases in which the axilla had not been surgically cleared, the axillary and supraclavicular regions were also irradiated, regardless of axillary node histology. A dose equivalent to 45 Gy in 20 fractions, with a higher dose when the sampled nodes were found to be involved, was recommended. The precise regimens of radiotherapy were those normally used in each of the five Scottish regional oncology centres. Follow-up, which included annual mammography, was by a member of the referring clinical team. Annual updates of progress were recorded by SCTO, and hospital records were inspected for consistency and completeness on report of death. Four of the 589 patients were withdrawn on account of inappropriate randomisation (two were treated by mastectomy and two were entered into trial A), leaving 585 for analysis. Relapse was classified as local (in ipsilateral breast or overlying skin), regional (involving ipsilateral axilla, supraclavicular or internal mammary nodes), or distant (all systemic sites including contralateral disease not satisfying the criteria for a second primary tumour as previously defined).5 Since all patients in this trial had breastconserving surgery, local relapse was included as an event but new primary disease at any site (including contralateral primary
Vol 348 • September 14, 1996
216 (74)
2·5
2·3
125 (43)
126 (43)
Number (%) impalpable
8 (3)
12 (4)
In-situ tumour present
111
115
In-situ tumour absent (%)
180 (62)
179 (61)
Node status (%) Positive clearance Positive sample Negative clearance Negative sample Not known
27 (9) 32 (11) 88 (30) 142 (49) 2
35 (12) 40 (14) 90 (31) 122 (42) 7
Confirmed ER status (fmol per mg protein) (%) <20 肁20 Not known
75 (26) 171 (59) 45 (15)
81 (27) 172 (59) 41 (14)
Systemic therapy given Chemotherapy Tamoxifen Median tamoxifen duration in yr (range) No adjuvant systemic therapy
75 77 213 214 4·9 (0·7–8·95) 4·9 (0·4–8·95) 3 3
Table 1: Distribution of variables breast cancer) was ignored in the calculation of event-free survival. The database was closed when information on follow-up for 1993 was available on all patients. Median follow-up for the 482 (82%) living patients was 5·7 (range 1·7–9·0) years, with all but ten patients having been followed for at least 2 years.
Statistics At initiation of the trial we were aware that large numbers would be required to detect small differences (1800 patients to have an 80% power of detecting a difference between recurrence rates of 15% and 20% at the 5% level of significance) but with the realisation that failure to achieve this would still allow use of the data in meta-analysis. If a difference of 10% existed then the number required would be about 450. Annual monitoring was planned with results reported to the trial committee only if a significant difference at the 1% level was detected. This resulted in the interim report in 19899 and subsequent closure of the trial in 1991 due to the unacceptable loco-regional relapse rate in nonirradiated patients. Analysis was carried out in the Department of Public Health Sciences of the University of Edinburgh by a member of the trial staff (DE). Kaplan-Meier survival curves were plotted for treatment groups on an intention-to-treat basis. Cox survival models incorporating time-dependent covariates were used to test the assumption of proportionality—ie, that hazard ratios (HRs) 1·0
Proportion alive
Inappropriately randomised = 2
No radiotherapy (n=294)
0·8 0·6 Watch Radiotherapy
0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation Radiotherapy 291 No radiotherapy 294
288 293
271 280
230 241
194 197
152 167
110 112
66 72
26 4 31 9
Figure 2: Kaplan-Meier curves for total survival based on allcause mortality HR=0·98, 95% CI 0·67–1·44. Numbers given below curves are those remaining at risk within compared groups for time periods analysed.
709
THE LANCET
Distribution of total sites of recurrence
Radiotherapy (n=291)
No radiotherapy (n=294)
Ipsilateral breast (no axillary node relapse) Ipsilateral breast and axillary nodes* Ipsilateral axillary nodes (no breast relapse)
14 (5) 3 1
51 (17) 21 12
Contralateral site(s) alone
3
Distant site(s) alone Distant and other sites*
2
27 (9) 12 (4)
16 (5) 42 (14)
Percentage with L/R relapse Radiotherapy (n=291) Age (yr) <50 50–59 肁60
8·0 8·7 2·7
Odds ratio (95% CI)
No radiotherapy (n=294) 31·6 29·5 25·7
0·19 (0·06–0·52) 0·23 (0·09–0·53) 0·08 (0·02–0·27)
Premenopausal
5·1
30·8
0·12 (0·03–0·39)
*Not necessarily concurrent.
Postmenopausal
6·6
27·8
0·18 (0·09–0·35)
Table 2: Distribution of confirmed sites of relapse (%) within randomised groups
Clinical size of tumour (cm) 聿2 >2 Occult
26·2 30·8 (3/12)*
0·12 (0·03–0·32) 0·18 (0·09–0·38) 0·43 (0·01–7·07)
28·8 28·4
0·16 (0·06–0·39) 0·17 (0·08–0·35)
Results 589 patients were registered (see figure 1 for trial profile). The 585 patients included were referred by 20 surgeons from 11 hospitals, including four with dedicated breast units where 487 (83%) were treated. The characteristics of the 585 patients and of their disease and treatment are recorded in table 1 from which it can be noted that threequarters of the patients had node-negative disease. Deviations from the protocol were most notably due to initial missing or incorrect information on ER status of the tumour leading to prescription of systemic therapy inappropriate to the subsequently confirmed concentration of ER in 26 patients, 13 in each randomised arm of the trial. A further 11 patients elected or were advised not to have the appropriate drug therapy. True violations of randomisation were recorded in only 12 patients, 6 in each arm, who were or were not treated by radiotherapy, contrary to allocation; in addition there were 27 patients in whom the prescribed fields of radiotherapy were less than indicated by protocol. Minor infringements of eligibility were noted in nine patients. Unless otherwise stated, all patients have been included in the analyses which were by intention to treat (ie, as randomised). Overall survival (figure 2) is similar for women in each arm of the trial (HR 0·98, 95% CI 0·67–1·44). However, the rates of distant disease were lower (but not significantly so) in those randomised for irradiation (HR 0·68, CI 0·45–1·03). Event-free survival (figure 3) indicates a significantly greater risk of relapse at any site (or death from any cause) in the non-irradiated patients (HR 0·54, 95% CI Outcome of treatment for L/R relapse
Breast-conserving treatment Treatment included mastectomy Total
L/R relapse after initial local excision and radiotherapy
L/R relapse after initial local excision without radiotherapy
Currently controlled
Not controlled
Currently controlled
Not controlled
0
8
29
21*
10
0
26
8
6·1 6·3
Node histology Negative Positive None
4·3 13·6 (0/2)*
21·2 49·3 (2/7)*
0·16 (0·06–0·41) 0·17 (0·08–0·36) 0·00 (0·00–22·46)
Node histology Negative clearance Negative sample Positive clearance Positive sample
3·4 4·9 14·8 12·5
24·4 18·9 40·0 57·5
0·13 (0·02–0·48) 0·22 (0·08–0·57) 0·26 (0·06–1·03) 0·11 (0·02–0·39)
8·1 5·0
21·7 32·9
0·32 (0·13–0·76) 0·11 (0·05–0·23)
3·1 13·6 4·3
25·0 43·8 16·0
0·09 (0·03–0·25) 0·20 (0·09–0·46) 0·23 (0·02–1·28)
In-situ cancer Present Absent ER as stratified (fmol per mg protein) 肁20 <20 Not done Available follow-up in years (total entry) Median Mean Range
5·14 7·71 0·14–8·95
5·49 7·74 0·35–8·95
*Actual numbers not %.
Table 4: Percentage with loco-regional (L/R) relapse according to variables and total follow-up for randomised groups
0·39–0·74). This is largely due to the difference in locoregional relapse rates (HR 0·20, CI 0·12–0·33, figure 4). In the irradiated arm of the trial there were 18 locoregional relapses (17 in the ipsilateral breast) and in those randomised for no radiotherapy there were 84 locoregional relapses (72 in the breast, table 2). Control of loco-regional disease following relapse has been achieved in half of initially irradiated and two-thirds of initially non-irradiated patients (table 3). As treatment of relapse included mastectomy as well as further local excision, radiotherapy, and/or systemic therapy, it resulted, for some, in failure of the conservation policy. The numbers of patients currently or at death clinically 1·0 0·8 0·6 Watch Radiotherapy
0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation 10
8
55
29
*13 of 21 had no surgical treatment for L/R recurrence.
Table 3: Control of loco-regional (L/R) relapse within randomised groups and according to whether or not breast remains conserved
710
Nodal surgery Clearance Sample
Proportion disease free
did not vary significantly with time. The HR of events together with a 95% confidence interval (CI) was estimated by fitting a Cox survival model, provided that there was no evidence of nonproportionality. An HR of 1·0 is equivalent to no difference between the randomised treatment groups, whilst an HR less than 1·0 indicates a benefit from radiotherapy. A CI that excludes 1·0 demonstrates a significant difference at the 5% level. Where odds ratios are reported, these have been calculated using EpiInfo, with exact confidence intervals based on the method of Mehta et al.10
4·0 7·6 (1/8)*
Radiotherapy 291 No radiotherapy 294
280 272
263 243
217 196
184 153
141 125
102 85
62 56
24 3 23 8
Figure 3: Kaplan-Meier survival curves for event-free survival HR=0·54, 95% CI 0·39–0·74. Numbers given below curves are those remaining at risk within compared groups for time periods analysed.
Vol 348 • September 14, 1996
Proportion free of loco-regional disease
THE LANCET
1·0
Radiotherapy
0·8
Tamoxifen (n=165) No known recurrence Alive Dead
0·6 Watch Radiotherapy
0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation Radiotherapy 291 No radiotherapy 294
284 276
267 247
222 201
188 158
147 129
107 87
65 58
26 4 24 8
Figure 4: Loco-regional disease rates
Proportion free from recurrence or death
Radiotherapy ER<20 Watch ER<20
1·0 0·8 0·6 0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation Radiotherapy ER肁20
165
164
153
128
110
83
62
33
13
3
Watch ER肁20
166
161
149
119
94
81
56
37
16
5
Radiotherapy ER<20
63
56
52
42
36
30
19
15
6
Watch ER<20
70
57
49
39
30
25
15
8
3
Figure 5: Event-free survival by results of ER assays in subgroup of patients given appropriate systemic therapy and radiotherapy according to randomisation Numbers given below curves are those remaining at risk within compared groups for time periods analysed.
free from loco-regional disease and with the breast preserved are 273 (93·8%) for the radiotherapy group and 239 (81·3%) for the observation group. In table 4 loco-regional relapse rates are related to a number of characteristics of the patients and their disease. The differences in these rates within the randomised groups consistently support the protective effect of radiotherapy. The poor prognosis associated with axillary node involvement and with low concentrations of ER is confirmed in both arms of the trial.
Additional analyses It was intended that assays of ER would be done on all tumours and that the systemic therapy to be administered would depend on this value. In the event, systemic therapy appropriate to the confirmed assay of ER was given to 472 of the 585 randomised patients of whom 464 were given radiotherapy as randomly allocated. Of those excluded, 17 incorrectly did not receive tamoxifen and 20 did not receive CMF, while 76 were excluded solely because no Trial
NSABP B-064 Toronto12 Milan13 Orebro-Uppsala14 Scottish
No of patients
1131* 837 567 381 585
Follow-up
5 yr 43 mo 39 mo 5 yr 5·7 yr
Size (cm)
聿4·0 聿4·0 聿2·5 聿2·0 聿4·0
Tamoxifen (n=166)
CMF (n=70)
137 8
41 3
113 3
34 1
Loco-regional relapse Ipsilateral breast, not nodes Breast and nodes Ipsilateral nodes, not breast
4 1 0
5 1 1
23 11 8
20 7 2
Other sites (including contralateral) No loco-regional at any time Loco-regional at some time
15 3
12 6
8 16
6 19
Table 5: Sites of recurrence of breast cancer in patients who received systemic therapy according to ER status of primary tumour
HR=0·20, 95% CI 0·12–0·33. Numbers given below curves are those remaining at risk within compared groups for time periods analysed. Radiotherapy ER肁20 Watch ER肁20
No radiotherapy CMF (n=63)
Margins
Clear Clear Wide Clear 1 cm
ER assay was performed. An additional analysis has been done in these 464 patients, of whom 228 received radiotherapy and 236 did not. The distribution of node positivity was similar to that in the total randomised groups. Event-free survival curves (figure 5) confirm the known prognostic value of ER status while indicating the advantage of postoperative radiotherapy to those patients with ER values of 20 fmol/mg protein or more and those with less than 20 fmol/mg protein, who were given tamoxifen and CMF chemotherapy, respectively. For completeness, sites of relapse for those given appropriate treatment with tamoxifen or CMF and radiotherapy according to randomisation are given in table 5.
Discussion The aim of this trial was to determine whether routine radical radiotherapy was still required when a primary cancer of the breast was treated by local excision supported by systemic therapy appropriate to the ER status of the tumour. The results indicate that, in the population of women with breast cancer studied (97% with palpable and 23% with node-positive disease), locoregional relapse, at a median follow-up of 5·7 years, was high in the non-irradiated group compared with that in those irradiated (28·6% and 6·2%, respectively). Comparative relapse rates for the ipsilateral treated breast alone were 24·5% and 5·8%, most of which were shown by mammography to be close to the initial disease site. These rates will increase with longer follow-up—ipsilateral breast recurrence at 9 years in the US National Surgical Adjuvant Breast and Bowel Project B-06 trial reached 12% in irradiated and 43% in non-irradiated breasts.11 The loco-regional relapse rate was not related to the size of the tumour or menstrual status. Not surprisingly, it was greater in those patients in the non-irradiated arm whose axillary nodes had only been sampled and in whom residual disease in remaining nodes did not receive further treatment. It was not related to the presence or absence of in-situ change, the extent of which was not recorded. % node positive
35·4 None 1·6 None 22·9
Systemic therapy
Node positive None Node positive None All
% with relapse in breast Radiotherapy
No radiotherapy
7·7 5·5 0·3 2·3 5·8
27·9 25·7 10·2 18·4 24·5
*All patients originally analysed in B-06 are included.
Table 6: 5-year relapse rates in treated breast in trials of local excision with or without radiotherapy
Vol 348 • September 14, 1996
711
THE LANCET
Although these results are similar to those of four other trials4,12–14 of radiotherapy following local excision of the tumour (table 6), the trial we now report is unique in that all patients, irrespective of node status, were given adjuvant systemic therapy by longterm tamoxifen or CMF. Adjuvant systemic therapy, by chemotherapy alone, was given in only two of the four trials and then only for patients with involved nodes. In the Milan trial this accounted only for 1·2% of patients. As premenopausal node-positive patients were not eligible for inclusion in our trial, the proportion of node-positive cases was only 23%, but the results of the overview reported by the Early Breast Cancer Trialists’ Collaborative Group indicate that both tamoxifen and chemotherapy are effective in reducing mortality in node-negative disease as well as in nodepositive disease.15 The decision to select adjuvant systemic therapy according to ER status when the trial started was questioned. The results of the trial of ovarian ablation versus chemotherapy which we have reported,5 those of the International Breast Cancer Study Group,16 and those from the overview of tamoxifen trials15 have justified this approach. We accept that the intravenous regimen of CMF may be less than now considered optimal but it was selected to ensure certainty of delivery. It is generally desirable to confirm that excision margins are histologically clear of tumour before acceptance for breast-conserving therapy. This was not a requirement in the protocol which asked only that at least 1 cm of surrounding normal breast tissue be excised. In the Milan “lumpectomy versus quadrantectomy” trial, in which tumours greater than 2·5 cm were excluded, this margin of excision gave clear histological margins in 84% of 283 specimens examined.17 From studies of shavings of the tumour bed following “lumpectomy” it would appear that with tumours of up to 4 cm in size, residual tumour may remain in 28–38% of breasts.18,19 However, it must be recognised that, as stated by Veronesi and his colleagues, pathological positivity of resection margins, although important, is far from accurate in predicting local relapse.17 Despite the uncertainty about tumour margins in our trial, it is relevant that 483 of the 585 patients were entered from four specialist breast units in which clear margins were regarded as necessary. In our trial, subsequent treatment achieved control of loco-regional relapse in more than half of those who were not initially irradiated. However, the number of patients free of loco-regional disease with both breasts preserved still favours initial irradiation. From table 3 it can be calculated that an additional 8% of patients lost their breasts through not having postoperative irradiation. As this was a trial of local therapy, the main measurable effect was control of loco-regional disease. As in other trials,12,20 survival was not compromised by the increased loco-regional relapse rates in the non-irradiated arm, being almost identical in the two arms of the trial (HR 0·98, CI 0·67–1·44). There has been an increase in the rate of distant relapse which is approaching significance, although this also has not affected survival as yet. It has been accepted that loco-regional relapse is not causative of distant spread but only an indicator of an aggressive tumour.21 Within our trial randomisation should have distributed aggressive tumours equally between the irradiated and non-irradiated arms so that the only difference between these two groups was in the local treatment given. The similar relation between loco712
regional relapse and distant relapse in the Toronto trial and in our trial supports the view that inadequate local treatment, as indicated by logo-regional relapse, may be a factor for dissemination. Equally, the difference could be due to earlier local relapse in the non-irradiated group precipitating earlier investigation for systemic disease. Longer follow-up of both these trials may resolve this. Optimal therapy for small well-differentiated tumours is currently uncertain but is being investigated in a UK trial, run under the auspices of the British Association of Surgical Oncology. There is a tendency for some surgeons to regard local excision and tamoxifen therapy as adequate treatment for these small tumours, particularly when the tumour has high or moderate ER concentrations. When our results are taken in association with those of other studies, they suggest that even when tamoxifen (or chemotherapy) is given following local excision of a tumour, a decision not to give radiotherapy should be made only when the risk of local relapse has been considered and fully explained to the patient. Scottish Cancer Trials Breast Group O Eremin, A Hutcheon, T Sarkar, Aberdeen; J Dewar, P Preece, Dundee; U Chetty, A P M Forrest, R A Hawkins, R C F Leonard, R J Prescott, A Rodger, Edinburgh; W D George (chairman), T Habeshaw, A Harnett, S Kaye, R E Leake, C S McArdle, H M M McCallum, D C Smith, Glasgow; P V Walsh, Inverness; and D Everington, C C McDonald, H J Stewart, trials office.
Acknowledgments We thank all SCTO staff, particularly C C McDonald, trial co-ordinator from 1989–93. Our thanks are also due to the patients for agreeing to be randomised, the 20 surgeons who referred them, and the many others who supported the trial in various ways, particularly the numerous general practitioners who responded to requests for additional information. The trial was administered by SCTO for the Scottish Cancer Trials Breast Group with a grant from the Medical Research Council (PG 7901641) and support from Zeneca Pharmaceuticals. We acknowledge personal support given by The Cancer Research Campaign to HJS and from the Leverhulme Trust to APMF, for secretarial assistance.
References 1
Stewart HJ, Robinson SM. Recruitment to breast cancer trials. Lancet 1984; ii: 1096–97. 2 Veronesi U, Saccozzi R, Del Vecchio M, et al. Comparing radical mastectomy with quadrantectomy, axillary dissection and radiotherapy in patients with small cancers of the breast. N Engl J Med 1981; 305: 6–11. 3 Sarrazin D, Le M, Rouesse J, et al. Conservative treatment versus mastectomy in breast cancer tumors with macroscopic diameter of 20 millimeters or less: the experience of the Institut Gustave-Roussy. Cancer 1984; 53: 1209–13. 4 Fisher B, Bauer M, Margolese R, et al. Five-year results of a randomised clinical trial comparing total mastectomy and segmental mastectomy with or without radiation in the treatment of breast cancer. N Engl J Med 1985; 132: 665–73. 5 Scottish Cancer Trials Breast Group and ICRF Breast Unit, Guy’s Hospital, London. Adjuvant ovarian ablation versus CMF chemotherapy in premenopausal women with pathological stage II breast carcinoma: the Scottish trial. Lancet 1993; 341: 1293–98. 6 Hawkins RA, Black R, Steele RJC, Dixon JMJ, Forrest APM. Oestrogen receptor concentration in primary breast cancer and axillary node metastases. Br Cancer Res Treat 1981; 1: 245–51. 7 Leake RE, Habib F. Steroid hormone receptors: assay of characterisation. In: Green B, Leake R, eds. Steroid hormones: a practical approach. Oxford: IRL Press, 1987: 67–94. 8 Breast Cancer Trials Committee. Adjuvant tamoxifen in the management of operable breast cancer: the Scottish trial. Lancet 1987; ii: 171–75. 9 Stewart HJ, Prescott RJ, Forrest APM. Conservation therapy of breast cancer. Lancet 1989; ii: 168–69. 10 Mehta CR, Patel NR, Gray R. Computing and exact confidence interval for the common odds ratio in several 2⫻2 contingency tables. J Am Stat Assoc 1985; 80: 969–73. 11 Fisher B, Anderson S, Fisher ER, et al. Significance of ipsilateral breast tumour recurrence after lumpectomy. Lancet 1991; 338: 327–31.
Vol 348 • September 14, 1996
THE LANCET 12 Clark RM, McCulloch PB, Levine MN, et al. Randomised clinical trial to assess the effectiveness of breast irradiation following lumpectomy and axillary dissection for node-negative breast cancer. J Natl Cancer Inst 1992; 84: 683–89. 13 Veronesi U, Luini A, Del Vecchio M, et al. Radiotherapy after breastpreserving surgery in women with localised cancer of the breast. N Engl J Med 1993; 328: 1587–91. 14 Liljegren G, Holmberg L, Adami H-O, et al. Sector resection with or without postoperative radiotherapy for stage I breast cancer: five-year results of a randomised trial. J Natl Cancer Inst 1994; 86: 717–22. 15 Early Breast Cancer Trialists’ Collaborative Group. Systemic treatment of early breast cancer by hormonal, cytotoxic or immune therapy: 133 randomised trials invovling 31 000 recurrences and 24 000 deaths among 75 000 women. Lancet 1992; 339: 1–15. 16 International Breast Cancer Study Group. Late effects of adjuvant oophorectomy and chemotherapy upon premenopausal breast cancer
patients. Ann Oncol 1990; 1: 30–35. 17 Veronesi V, Volterrani F, Luini A, et al. Quadrantectomy versus lumpectomy for small size breast cancer. Eur J Cancer 1990; 26: 671–73. 18 Macmillan RD, Purushotham AD, Mallon E, Ramsay G, George WD. Breast-conserving surgery and tumour bed positivity in patients with breast cancer. Br J Surg 1994; 81: 56–58. 19 van Dalen R, Bates A, Dhesi I, Bates T, Houghton J. Positive cavity biopsy does not predict early local relapse for patients treated by local excision for early breast cancer. Breast 1995; 4: 231 (abstr no 15). 20 Hayward JL. The Guy’s Hospital trials on breast conservation. In: Harris JR, Hellman S, Silen W, eds. Conservative management of breast cancer. Philadelphia: Lippincott, 1983: 77–90. 21 Veronesi U, Marubini E, Del Vecchio M, et al. Local recurrences and distant metastases after conservative breast cancer treatments: partly independent events. J Natl Cancer Inst 1995; 87: 19–27.
Double-blind randomised trial of co-amoxiclav versus placebo for persistent otitis media with effusion in general practice
F A M van Balen, R A de Melker, F W M M Touw-Otten
Summary Background The treatment of persistent otitis media with effusion (OME) remains controversial, but this condition is the commonest reason for children to require ear, nose, and throat (ENT) surgery. Trials of antibiotics are inconclusive, are often weak methodologically, and have not been done in general practice. Our aim was a trial of an antibiotic for OME in such a population. Methods 433 children, aged 6 months to 6 years, with OME from 57 general practices entered a 3-month watchful waiting period. Of 223 (52%) with persistent bilateral OME, 162 were randomised double-blind to receive co-amoxiclav suspension (20 mg/kg amoxicillin, 5 mg/kg clavulanate potassium) or matching placebo, orally three times a day for 14 days. All cases also received xylometazoline 0·25% decongestant nosedrops thrice daily. Of the 61 not randomised, 13 children were referred to an ENT surgeon and parents refused consent in 48 cases. The main outcome measures were persistent OME in both ears and in one or both ears, as assessed clinically and by tympanometry. Analysis was by intention-to-treat. Findings 79 children in the treatment group and 70 in the placebo group were analysed for efficacy. 3 withdrew in the co-amoxiclav group (2 lost to follow-up, 1 due to sideeffects); 6 withdrew in the placebo group (5 and 1, respectively) . In addition, 4 tympanograms w ere uninterpretable in the controls. Compliance was over 90% in both groups. Persistent OME in both ears and in one or both ears were found at significantly lower rates in the coamoxiclav group than in the controls at the 2-week followup: 53 vs 84% and 77 vs 93%, respectively. Odds ratios adjusted for sex, history of adenoidectomy, and upper
Prof F W M M Touw-Otten died in April, 1996 Department of General Practice, University of Utrecht, 3584 CG Utrecht, Netherlands (F A M van Balen MD, R A de Melker MD, Prof F W M M Touw-Otten PhD) Correspondence to: Dr F A M van Balen
Vol 348 • September 14, 1996
respiratory tract infection at follow-up were 0·25 (95% CI 0·11, 0·58, p=0·001) and 0·30 (0·10, 0·89, p=0·03), respectively. Parents of children in the co-amox iclav group reported significantly more side-effects than those of control children (44 vs 22%, p=0·03). Side-effects were mostly gastrointestinal and mild. Interpretation Our study in a general-practice setting confirmed the positive short-term effect of antibiotic treatment for persistent middle-ear infection. Before referral to an ENT surgeon, children with persistent OME presenting to general practitioners could be considered for such treatment, depending on the individual child and possible adverse sequelae.
Lancet 1996; 348: 713–16
Introduction Otitis media with effusion (OME) incurs great morbidity and costs in children.1 The treatment of first choice remains controversial. There is agreement on watchful waiting before active treatment, but not for how long.1–3 Both US2 and Dutch4 guidelines include watchful waiting for 3 months, after which treatment options are medical or surgical. Medical treatment could be started by the general practitioner which might lead to fewer referrals, thereby reducing the high costs of surgery.5 Decongestants and antihistamines are ineffective. 6 Treatment with corticosteroids is still under discussion.7,8 The rationale underlying antimicrobial therapy is based on the observation that pathogenic bacteria can be recovered from the middle-ear effusion in a third of all cases.9,10 The microorganisms most frequently found are Streptocococcus pneumonia, Haemophilus influenzae, and Moraxella catarrhalis.11,12 Two meta-analyses13,14 show a positive short-term effect from antibiotics. Amoxicillin, used alone or with clavulanate potassium, produced a uniform, statistically significant effect in three placebocontrolled randomised trials,15–17 but not in one other.18 Randomised studies in which the antibiotic was given for 14 days showed a positive effect, but none was carried out in general practice. 713
Randomised controlled trial of conservation therapy for breast cancer: 6-year analysis of the Scottish trial A Patrick Forrest, Helen J Stewart, Dawn Everington, Robin J Prescott, Colin S McArdle, Adrian N Harnett, David C Smith, W David George, on behalf of Scottish Cancer Trials Breast Group
Summary Background To determine whether, when primary breast cancer is treated by local excision supported by systemic therapy appropriate to the oestrogen receptor status (ER) of the tumour, local radiotherapy can be avoided. Methods We carried out a randomised controlled trial in 585 patients aged less than 70 years with primary breast cancers of 4 cm or less in size in four specialist units and seven other hospitals in Scotland. After local excision of the tumour (1 cm margin) and an axillary lymph-node clearance or sample, all patients received systemic therapy with oral tamoxifen 20 mg daily or six 3-weekly intravenous bolus injections of cyclophosphamide 600 mg, methotrex ate 50 mg, and fluorouracil 600 mg per m2, depending upon the ER concentration in the primary tumour. Patients were then randomly allocated to postoperative radical radiotherapy (50 Gy to breast with boost to the tumour bed) or to no further local treatment. The median follow-up of living patients was 5·7 years. The primary analysis was by intention to treat but since some patients did not receive systemic therapy appropriate to their ER status, a subsidiary analysis was restricted to 464 patients in whom all details of the protocol had been observed. Findings In the primary analysis survival was equal in the radiotherapy and non-radiotherapy groups (hazard ratio [HR] 0·98, 95% CI 0·67–1·44). Event-free survival showed an advantage in the irradiated patients (HR 0·54, 95% CI 0·39–0·74), largely due to fewer loco-regional relapses (HR 0·20, 95% CI 0·12–0·33). The relapse rate in the ipsilateral breast was 24·5% in the non-irradiated group and 5·8% follow ing breast irradiation. The subsidiary analysis confirmed these findings and indicated the advantage of radiotherapy irrespective of ER concentration. There was a non-significant trend towards fewer distant metastases in the irradiated group. Interpretation After local excision of a primary breast cancer, we conclude that radiotherapy to the residual breast tissue is advisable even when selective adjuvant systemic therapy is given.
Lancet 1996; 348: 708–13 Scottish Cancer Trials Office, Medical School, University of Edinburgh, Edinburgh, UK (A Patrick Forrest FRCS, H J Stewart FRCR); Medical Statistics Unit (D Everington MSc , R J Prescott PhD); Department of Surgery, Western General Hospital, University of Edinburgh (C S McArdle FRCS); Department of Radiotherapy, Beatson Oncology Centre, Glasgow (A N Harnett FRCR); Department of Surgery, Victoria Infirmary, Glasgow (D C Smith FRCS); and Department of Surgery, Western Infirmary, University of Glasgow (W D George FRCS) Correspondence to: Sir Patrick Forrest, Scottish Cancer Trials Office, Medical School, University of Edinburgh, Teviot Place, Edinburgh EH8 9AG, UK
708
Introduction In 1982, when the need for mastectomy was being questioned, the Scottish Cancer Trials Breast Group decided to test the feasibility of mounting a randomised controlled trial to compare mastectomy with local excision and radiotherapy. It was soon apparent that the number of women prepared to accept such randomisation was limited.1 Consequently, with increasing evidence that these two forms of local treatment were equivalent,2–4 and also that adjuvant systemic therapy might protect against loco-regional relapse, it was decided to initiate a trial in women accepting breast-conserving therapy. The aim of the trial was to determine whether radiotherapy was still necessary routinely for women who had received systemic therapy appropriate to their oestrogen receptor (ER) status following local excision of their tumour.
Methods Between April 1, 1985, and Oct 2, 1991, 589 consenting patients, aged 70 years or less with invasive breast cancer of clinical size 4 cm or less, were included provided there was no fixation of the primary tumour or axillary lymph nodes. Premenopausal women with involved axillary nodes treated by conservation were not eligible because they were being entered into the Scottish adjuvant trial of chemotherapy versus ovarian ablation (trial A).5 Also excluded were those with a second tumour in the same or opposite breast, those with a history of previous invasive cancer at any site, and those with medical or other factors contraindicating adherence to the protocol. Minimum preoperative investigations were plain radiographs of the chest and pelvis. The protocol was agreed by members of the Scottish Cancer Trials Breast Group and circulated, with a request for collaboration, to all surgeons, radiotherapists, and medical oncologists working in Scotland. Ethical approval of the protocol, which provided for informed consent, was required from each participating centre. Local excision was defined as removal of the tumour with a 1 cm margin of normal breast tissue. If a preoperative needle biopsy had been done, the skin puncture site together with the needle track was to be included in the excision, as was the nipple in centrally placed tumours if there was recent associated retraction or Paget’s disease. Participants were asked to adhere to a single surgical policy for the axilla. This might be either sampling (3–4 nodes) or surgical clearance (levels I–III). Routine histopathological examination of the specimen was carried out in the related pathology laboratory. Confirmation that excision margins were clear of tumour was not a requirement for entry to the trial but, as most specialist breast units had a policy of “clear margins”, those requiring re-excision were accepted. ER concentrations were determined by the dextran-charcoal assay.6,7 The laboratories conducting most of the assays were participating in the external quality control scheme of the European Organisation for Research and Treatment of Cancer (EORTC). Results were available for 87% of the patients. As in our previous adjuvant trials, 5,8 ER levels were designated “negative” (<5 fmol/mg protein), “poor” (5–19 fmol/mg protein), or “positive with clinical relevance” (肁20 fmol/mg protein). The surgeon was responsible for initial assessment of ER data and arrangement of appropriate systemic therapy. When ER was 20 fmol/mg protein or more this was oral tamoxifen 20 mg daily
Vol 348 • September 14, 1996
THE LANCET
Radiotherapy (n=291)
Registered for trial = 589
Randomised for radiotherapy = 293
Randomised for no radiotherapy = 296
Median age at entry in year (range)
57 (28–70)
57 (28–70)
Premenopausal
78
78
Postmenopausal (%)
213 (73)
Median clinical tumour size (cm) Number (%) 聿2 cm
Available for primary analysis = 291 Lost to follow-up = 0 Median follow-up 5·14 yr
Violation of randomisation = 6 No ER available = 40 Inappropriate systemic therapy = 17
Additional analysis (all correct)= 228 (tamox ifen = 165 and CMF = 63)
Inappropriately randomised = 2
Available for primary analysis = 294 Lost to follow-up = 0 Median follow-up = 5·49 yr
Violation of randomisation = 6 No ER available = 36 Inappropriate systemic therapy = 16
Additional analysis (all correct) = 236 (tamoxifen = 166 and CMF = 70)
Figure 1: Trial profile for Scottish Conservation Trial for at least 5 years or until relapse of disease, and, when less than 20 fmol/mg protein, six 3-weekly intravenous bolus injections of cyclophosphamide 600 mg, methotrexate 50 mg, and fluorouracil 600 mg per m2 (CMF). Those without an ER assay result were to receive tamoxifen. Patients were registered with the Scottish Cancer Trials Office, (SCTO) when eligibility was confirmed. They were stratified within six subgroups according to menstrual status (within or more than 12 months from last menstrual period) and ER status as initially available (levels of 肁20, <20, or ER unknown) and the intended treatment identified from the appropriate computergenerated random list, namely postoperative irradiation to the residual ipsilateral breast or no further immediate local treatment. The correctness of stratification by ER was subsequently checked with the related laboratory. When randomised for radiotherapy, the recommended maximum megavoltage dose to the residual breast was 50 Gy in 20–25 daily fractions. A boost to the tumour bed by iridium implant of 20–30 Gy or by electron or orthovoltage external beam irradiation of 10–15 Gy was required. In those cases in which the axilla had not been surgically cleared, the axillary and supraclavicular regions were also irradiated, regardless of axillary node histology. A dose equivalent to 45 Gy in 20 fractions, with a higher dose when the sampled nodes were found to be involved, was recommended. The precise regimens of radiotherapy were those normally used in each of the five Scottish regional oncology centres. Follow-up, which included annual mammography, was by a member of the referring clinical team. Annual updates of progress were recorded by SCTO, and hospital records were inspected for consistency and completeness on report of death. Four of the 589 patients were withdrawn on account of inappropriate randomisation (two were treated by mastectomy and two were entered into trial A), leaving 585 for analysis. Relapse was classified as local (in ipsilateral breast or overlying skin), regional (involving ipsilateral axilla, supraclavicular or internal mammary nodes), or distant (all systemic sites including contralateral disease not satisfying the criteria for a second primary tumour as previously defined).5 Since all patients in this trial had breastconserving surgery, local relapse was included as an event but new primary disease at any site (including contralateral primary
Vol 348 • September 14, 1996
216 (74)
2·5
2·3
125 (43)
126 (43)
Number (%) impalpable
8 (3)
12 (4)
In-situ tumour present
111
115
In-situ tumour absent (%)
180 (62)
179 (61)
Node status (%) Positive clearance Positive sample Negative clearance Negative sample Not known
27 (9) 32 (11) 88 (30) 142 (49) 2
35 (12) 40 (14) 90 (31) 122 (42) 7
Confirmed ER status (fmol per mg protein) (%) <20 肁20 Not known
75 (26) 171 (59) 45 (15)
81 (27) 172 (59) 41 (14)
Systemic therapy given Chemotherapy Tamoxifen Median tamoxifen duration in yr (range) No adjuvant systemic therapy
75 77 213 214 4·9 (0·7–8·95) 4·9 (0·4–8·95) 3 3
Table 1: Distribution of variables breast cancer) was ignored in the calculation of event-free survival. The database was closed when information on follow-up for 1993 was available on all patients. Median follow-up for the 482 (82%) living patients was 5·7 (range 1·7–9·0) years, with all but ten patients having been followed for at least 2 years.
Statistics At initiation of the trial we were aware that large numbers would be required to detect small differences (1800 patients to have an 80% power of detecting a difference between recurrence rates of 15% and 20% at the 5% level of significance) but with the realisation that failure to achieve this would still allow use of the data in meta-analysis. If a difference of 10% existed then the number required would be about 450. Annual monitoring was planned with results reported to the trial committee only if a significant difference at the 1% level was detected. This resulted in the interim report in 19899 and subsequent closure of the trial in 1991 due to the unacceptable loco-regional relapse rate in nonirradiated patients. Analysis was carried out in the Department of Public Health Sciences of the University of Edinburgh by a member of the trial staff (DE). Kaplan-Meier survival curves were plotted for treatment groups on an intention-to-treat basis. Cox survival models incorporating time-dependent covariates were used to test the assumption of proportionality—ie, that hazard ratios (HRs) 1·0
Proportion alive
Inappropriately randomised = 2
No radiotherapy (n=294)
0·8 0·6 Watch Radiotherapy
0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation Radiotherapy 291 No radiotherapy 294
288 293
271 280
230 241
194 197
152 167
110 112
66 72
26 4 31 9
Figure 2: Kaplan-Meier curves for total survival based on allcause mortality HR=0·98, 95% CI 0·67–1·44. Numbers given below curves are those remaining at risk within compared groups for time periods analysed.
709
THE LANCET
Distribution of total sites of recurrence
Radiotherapy (n=291)
No radiotherapy (n=294)
Ipsilateral breast (no axillary node relapse) Ipsilateral breast and axillary nodes* Ipsilateral axillary nodes (no breast relapse)
14 (5) 3 1
51 (17) 21 12
Contralateral site(s) alone
3
Distant site(s) alone Distant and other sites*
2
27 (9) 12 (4)
16 (5) 42 (14)
Percentage with L/R relapse Radiotherapy (n=291) Age (yr) <50 50–59 肁60
8·0 8·7 2·7
Odds ratio (95% CI)
No radiotherapy (n=294) 31·6 29·5 25·7
0·19 (0·06–0·52) 0·23 (0·09–0·53) 0·08 (0·02–0·27)
Premenopausal
5·1
30·8
0·12 (0·03–0·39)
*Not necessarily concurrent.
Postmenopausal
6·6
27·8
0·18 (0·09–0·35)
Table 2: Distribution of confirmed sites of relapse (%) within randomised groups
Clinical size of tumour (cm) 聿2 >2 Occult
26·2 30·8 (3/12)*
0·12 (0·03–0·32) 0·18 (0·09–0·38) 0·43 (0·01–7·07)
28·8 28·4
0·16 (0·06–0·39) 0·17 (0·08–0·35)
Results 589 patients were registered (see figure 1 for trial profile). The 585 patients included were referred by 20 surgeons from 11 hospitals, including four with dedicated breast units where 487 (83%) were treated. The characteristics of the 585 patients and of their disease and treatment are recorded in table 1 from which it can be noted that threequarters of the patients had node-negative disease. Deviations from the protocol were most notably due to initial missing or incorrect information on ER status of the tumour leading to prescription of systemic therapy inappropriate to the subsequently confirmed concentration of ER in 26 patients, 13 in each randomised arm of the trial. A further 11 patients elected or were advised not to have the appropriate drug therapy. True violations of randomisation were recorded in only 12 patients, 6 in each arm, who were or were not treated by radiotherapy, contrary to allocation; in addition there were 27 patients in whom the prescribed fields of radiotherapy were less than indicated by protocol. Minor infringements of eligibility were noted in nine patients. Unless otherwise stated, all patients have been included in the analyses which were by intention to treat (ie, as randomised). Overall survival (figure 2) is similar for women in each arm of the trial (HR 0·98, 95% CI 0·67–1·44). However, the rates of distant disease were lower (but not significantly so) in those randomised for irradiation (HR 0·68, CI 0·45–1·03). Event-free survival (figure 3) indicates a significantly greater risk of relapse at any site (or death from any cause) in the non-irradiated patients (HR 0·54, 95% CI Outcome of treatment for L/R relapse
Breast-conserving treatment Treatment included mastectomy Total
L/R relapse after initial local excision and radiotherapy
L/R relapse after initial local excision without radiotherapy
Currently controlled
Not controlled
Currently controlled
Not controlled
0
8
29
21*
10
0
26
8
6·1 6·3
Node histology Negative Positive None
4·3 13·6 (0/2)*
21·2 49·3 (2/7)*
0·16 (0·06–0·41) 0·17 (0·08–0·36) 0·00 (0·00–22·46)
Node histology Negative clearance Negative sample Positive clearance Positive sample
3·4 4·9 14·8 12·5
24·4 18·9 40·0 57·5
0·13 (0·02–0·48) 0·22 (0·08–0·57) 0·26 (0·06–1·03) 0·11 (0·02–0·39)
8·1 5·0
21·7 32·9
0·32 (0·13–0·76) 0·11 (0·05–0·23)
3·1 13·6 4·3
25·0 43·8 16·0
0·09 (0·03–0·25) 0·20 (0·09–0·46) 0·23 (0·02–1·28)
In-situ cancer Present Absent ER as stratified (fmol per mg protein) 肁20 <20 Not done Available follow-up in years (total entry) Median Mean Range
5·14 7·71 0·14–8·95
5·49 7·74 0·35–8·95
*Actual numbers not %.
Table 4: Percentage with loco-regional (L/R) relapse according to variables and total follow-up for randomised groups
0·39–0·74). This is largely due to the difference in locoregional relapse rates (HR 0·20, CI 0·12–0·33, figure 4). In the irradiated arm of the trial there were 18 locoregional relapses (17 in the ipsilateral breast) and in those randomised for no radiotherapy there were 84 locoregional relapses (72 in the breast, table 2). Control of loco-regional disease following relapse has been achieved in half of initially irradiated and two-thirds of initially non-irradiated patients (table 3). As treatment of relapse included mastectomy as well as further local excision, radiotherapy, and/or systemic therapy, it resulted, for some, in failure of the conservation policy. The numbers of patients currently or at death clinically 1·0 0·8 0·6 Watch Radiotherapy
0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation 10
8
55
29
*13 of 21 had no surgical treatment for L/R recurrence.
Table 3: Control of loco-regional (L/R) relapse within randomised groups and according to whether or not breast remains conserved
710
Nodal surgery Clearance Sample
Proportion disease free
did not vary significantly with time. The HR of events together with a 95% confidence interval (CI) was estimated by fitting a Cox survival model, provided that there was no evidence of nonproportionality. An HR of 1·0 is equivalent to no difference between the randomised treatment groups, whilst an HR less than 1·0 indicates a benefit from radiotherapy. A CI that excludes 1·0 demonstrates a significant difference at the 5% level. Where odds ratios are reported, these have been calculated using EpiInfo, with exact confidence intervals based on the method of Mehta et al.10
4·0 7·6 (1/8)*
Radiotherapy 291 No radiotherapy 294
280 272
263 243
217 196
184 153
141 125
102 85
62 56
24 3 23 8
Figure 3: Kaplan-Meier survival curves for event-free survival HR=0·54, 95% CI 0·39–0·74. Numbers given below curves are those remaining at risk within compared groups for time periods analysed.
Vol 348 • September 14, 1996
Proportion free of loco-regional disease
THE LANCET
1·0
Radiotherapy
0·8
Tamoxifen (n=165) No known recurrence Alive Dead
0·6 Watch Radiotherapy
0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation Radiotherapy 291 No radiotherapy 294
284 276
267 247
222 201
188 158
147 129
107 87
65 58
26 4 24 8
Figure 4: Loco-regional disease rates
Proportion free from recurrence or death
Radiotherapy ER<20 Watch ER<20
1·0 0·8 0·6 0·4 0
1
2
3
4
5
6
7
8
9
Years from randomisation Radiotherapy ER肁20
165
164
153
128
110
83
62
33
13
3
Watch ER肁20
166
161
149
119
94
81
56
37
16
5
Radiotherapy ER<20
63
56
52
42
36
30
19
15
6
Watch ER<20
70
57
49
39
30
25
15
8
3
Figure 5: Event-free survival by results of ER assays in subgroup of patients given appropriate systemic therapy and radiotherapy according to randomisation Numbers given below curves are those remaining at risk within compared groups for time periods analysed.
free from loco-regional disease and with the breast preserved are 273 (93·8%) for the radiotherapy group and 239 (81·3%) for the observation group. In table 4 loco-regional relapse rates are related to a number of characteristics of the patients and their disease. The differences in these rates within the randomised groups consistently support the protective effect of radiotherapy. The poor prognosis associated with axillary node involvement and with low concentrations of ER is confirmed in both arms of the trial.
Additional analyses It was intended that assays of ER would be done on all tumours and that the systemic therapy to be administered would depend on this value. In the event, systemic therapy appropriate to the confirmed assay of ER was given to 472 of the 585 randomised patients of whom 464 were given radiotherapy as randomly allocated. Of those excluded, 17 incorrectly did not receive tamoxifen and 20 did not receive CMF, while 76 were excluded solely because no Trial
NSABP B-064 Toronto12 Milan13 Orebro-Uppsala14 Scottish
No of patients
1131* 837 567 381 585
Follow-up
5 yr 43 mo 39 mo 5 yr 5·7 yr
Size (cm)
聿4·0 聿4·0 聿2·5 聿2·0 聿4·0
Tamoxifen (n=166)
CMF (n=70)
137 8
41 3
113 3
34 1
Loco-regional relapse Ipsilateral breast, not nodes Breast and nodes Ipsilateral nodes, not breast
4 1 0
5 1 1
23 11 8
20 7 2
Other sites (including contralateral) No loco-regional at any time Loco-regional at some time
15 3
12 6
8 16
6 19
Table 5: Sites of recurrence of breast cancer in patients who received systemic therapy according to ER status of primary tumour
HR=0·20, 95% CI 0·12–0·33. Numbers given below curves are those remaining at risk within compared groups for time periods analysed. Radiotherapy ER肁20 Watch ER肁20
No radiotherapy CMF (n=63)
Margins
Clear Clear Wide Clear 1 cm
ER assay was performed. An additional analysis has been done in these 464 patients, of whom 228 received radiotherapy and 236 did not. The distribution of node positivity was similar to that in the total randomised groups. Event-free survival curves (figure 5) confirm the known prognostic value of ER status while indicating the advantage of postoperative radiotherapy to those patients with ER values of 20 fmol/mg protein or more and those with less than 20 fmol/mg protein, who were given tamoxifen and CMF chemotherapy, respectively. For completeness, sites of relapse for those given appropriate treatment with tamoxifen or CMF and radiotherapy according to randomisation are given in table 5.
Discussion The aim of this trial was to determine whether routine radical radiotherapy was still required when a primary cancer of the breast was treated by local excision supported by systemic therapy appropriate to the ER status of the tumour. The results indicate that, in the population of women with breast cancer studied (97% with palpable and 23% with node-positive disease), locoregional relapse, at a median follow-up of 5·7 years, was high in the non-irradiated group compared with that in those irradiated (28·6% and 6·2%, respectively). Comparative relapse rates for the ipsilateral treated breast alone were 24·5% and 5·8%, most of which were shown by mammography to be close to the initial disease site. These rates will increase with longer follow-up—ipsilateral breast recurrence at 9 years in the US National Surgical Adjuvant Breast and Bowel Project B-06 trial reached 12% in irradiated and 43% in non-irradiated breasts.11 The loco-regional relapse rate was not related to the size of the tumour or menstrual status. Not surprisingly, it was greater in those patients in the non-irradiated arm whose axillary nodes had only been sampled and in whom residual disease in remaining nodes did not receive further treatment. It was not related to the presence or absence of in-situ change, the extent of which was not recorded. % node positive
35·4 None 1·6 None 22·9
Systemic therapy
Node positive None Node positive None All
% with relapse in breast Radiotherapy
No radiotherapy
7·7 5·5 0·3 2·3 5·8
27·9 25·7 10·2 18·4 24·5
*All patients originally analysed in B-06 are included.
Table 6: 5-year relapse rates in treated breast in trials of local excision with or without radiotherapy
Vol 348 • September 14, 1996
711
THE LANCET
Although these results are similar to those of four other trials4,12–14 of radiotherapy following local excision of the tumour (table 6), the trial we now report is unique in that all patients, irrespective of node status, were given adjuvant systemic therapy by longterm tamoxifen or CMF. Adjuvant systemic therapy, by chemotherapy alone, was given in only two of the four trials and then only for patients with involved nodes. In the Milan trial this accounted only for 1·2% of patients. As premenopausal node-positive patients were not eligible for inclusion in our trial, the proportion of node-positive cases was only 23%, but the results of the overview reported by the Early Breast Cancer Trialists’ Collaborative Group indicate that both tamoxifen and chemotherapy are effective in reducing mortality in node-negative disease as well as in nodepositive disease.15 The decision to select adjuvant systemic therapy according to ER status when the trial started was questioned. The results of the trial of ovarian ablation versus chemotherapy which we have reported,5 those of the International Breast Cancer Study Group,16 and those from the overview of tamoxifen trials15 have justified this approach. We accept that the intravenous regimen of CMF may be less than now considered optimal but it was selected to ensure certainty of delivery. It is generally desirable to confirm that excision margins are histologically clear of tumour before acceptance for breast-conserving therapy. This was not a requirement in the protocol which asked only that at least 1 cm of surrounding normal breast tissue be excised. In the Milan “lumpectomy versus quadrantectomy” trial, in which tumours greater than 2·5 cm were excluded, this margin of excision gave clear histological margins in 84% of 283 specimens examined.17 From studies of shavings of the tumour bed following “lumpectomy” it would appear that with tumours of up to 4 cm in size, residual tumour may remain in 28–38% of breasts.18,19 However, it must be recognised that, as stated by Veronesi and his colleagues, pathological positivity of resection margins, although important, is far from accurate in predicting local relapse.17 Despite the uncertainty about tumour margins in our trial, it is relevant that 483 of the 585 patients were entered from four specialist breast units in which clear margins were regarded as necessary. In our trial, subsequent treatment achieved control of loco-regional relapse in more than half of those who were not initially irradiated. However, the number of patients free of loco-regional disease with both breasts preserved still favours initial irradiation. From table 3 it can be calculated that an additional 8% of patients lost their breasts through not having postoperative irradiation. As this was a trial of local therapy, the main measurable effect was control of loco-regional disease. As in other trials,12,20 survival was not compromised by the increased loco-regional relapse rates in the non-irradiated arm, being almost identical in the two arms of the trial (HR 0·98, CI 0·67–1·44). There has been an increase in the rate of distant relapse which is approaching significance, although this also has not affected survival as yet. It has been accepted that loco-regional relapse is not causative of distant spread but only an indicator of an aggressive tumour.21 Within our trial randomisation should have distributed aggressive tumours equally between the irradiated and non-irradiated arms so that the only difference between these two groups was in the local treatment given. The similar relation between loco712
regional relapse and distant relapse in the Toronto trial and in our trial supports the view that inadequate local treatment, as indicated by logo-regional relapse, may be a factor for dissemination. Equally, the difference could be due to earlier local relapse in the non-irradiated group precipitating earlier investigation for systemic disease. Longer follow-up of both these trials may resolve this. Optimal therapy for small well-differentiated tumours is currently uncertain but is being investigated in a UK trial, run under the auspices of the British Association of Surgical Oncology. There is a tendency for some surgeons to regard local excision and tamoxifen therapy as adequate treatment for these small tumours, particularly when the tumour has high or moderate ER concentrations. When our results are taken in association with those of other studies, they suggest that even when tamoxifen (or chemotherapy) is given following local excision of a tumour, a decision not to give radiotherapy should be made only when the risk of local relapse has been considered and fully explained to the patient. Scottish Cancer Trials Breast Group O Eremin, A Hutcheon, T Sarkar, Aberdeen; J Dewar, P Preece, Dundee; U Chetty, A P M Forrest, R A Hawkins, R C F Leonard, R J Prescott, A Rodger, Edinburgh; W D George (chairman), T Habeshaw, A Harnett, S Kaye, R E Leake, C S McArdle, H M M McCallum, D C Smith, Glasgow; P V Walsh, Inverness; and D Everington, C C McDonald, H J Stewart, trials office.
Acknowledgments We thank all SCTO staff, particularly C C McDonald, trial co-ordinator from 1989–93. Our thanks are also due to the patients for agreeing to be randomised, the 20 surgeons who referred them, and the many others who supported the trial in various ways, particularly the numerous general practitioners who responded to requests for additional information. The trial was administered by SCTO for the Scottish Cancer Trials Breast Group with a grant from the Medical Research Council (PG 7901641) and support from Zeneca Pharmaceuticals. We acknowledge personal support given by The Cancer Research Campaign to HJS and from the Leverhulme Trust to APMF, for secretarial assistance.
References 1
Stewart HJ, Robinson SM. Recruitment to breast cancer trials. Lancet 1984; ii: 1096–97. 2 Veronesi U, Saccozzi R, Del Vecchio M, et al. Comparing radical mastectomy with quadrantectomy, axillary dissection and radiotherapy in patients with small cancers of the breast. N Engl J Med 1981; 305: 6–11. 3 Sarrazin D, Le M, Rouesse J, et al. Conservative treatment versus mastectomy in breast cancer tumors with macroscopic diameter of 20 millimeters or less: the experience of the Institut Gustave-Roussy. Cancer 1984; 53: 1209–13. 4 Fisher B, Bauer M, Margolese R, et al. Five-year results of a randomised clinical trial comparing total mastectomy and segmental mastectomy with or without radiation in the treatment of breast cancer. N Engl J Med 1985; 132: 665–73. 5 Scottish Cancer Trials Breast Group and ICRF Breast Unit, Guy’s Hospital, London. Adjuvant ovarian ablation versus CMF chemotherapy in premenopausal women with pathological stage II breast carcinoma: the Scottish trial. Lancet 1993; 341: 1293–98. 6 Hawkins RA, Black R, Steele RJC, Dixon JMJ, Forrest APM. Oestrogen receptor concentration in primary breast cancer and axillary node metastases. Br Cancer Res Treat 1981; 1: 245–51. 7 Leake RE, Habib F. Steroid hormone receptors: assay of characterisation. In: Green B, Leake R, eds. Steroid hormones: a practical approach. Oxford: IRL Press, 1987: 67–94. 8 Breast Cancer Trials Committee. Adjuvant tamoxifen in the management of operable breast cancer: the Scottish trial. Lancet 1987; ii: 171–75. 9 Stewart HJ, Prescott RJ, Forrest APM. Conservation therapy of breast cancer. Lancet 1989; ii: 168–69. 10 Mehta CR, Patel NR, Gray R. Computing and exact confidence interval for the common odds ratio in several 2⫻2 contingency tables. J Am Stat Assoc 1985; 80: 969–73. 11 Fisher B, Anderson S, Fisher ER, et al. Significance of ipsilateral breast tumour recurrence after lumpectomy. Lancet 1991; 338: 327–31.
Vol 348 • September 14, 1996
THE LANCET 12 Clark RM, McCulloch PB, Levine MN, et al. Randomised clinical trial to assess the effectiveness of breast irradiation following lumpectomy and axillary dissection for node-negative breast cancer. J Natl Cancer Inst 1992; 84: 683–89. 13 Veronesi U, Luini A, Del Vecchio M, et al. Radiotherapy after breastpreserving surgery in women with localised cancer of the breast. N Engl J Med 1993; 328: 1587–91. 14 Liljegren G, Holmberg L, Adami H-O, et al. Sector resection with or without postoperative radiotherapy for stage I breast cancer: five-year results of a randomised trial. J Natl Cancer Inst 1994; 86: 717–22. 15 Early Breast Cancer Trialists’ Collaborative Group. Systemic treatment of early breast cancer by hormonal, cytotoxic or immune therapy: 133 randomised trials invovling 31 000 recurrences and 24 000 deaths among 75 000 women. Lancet 1992; 339: 1–15. 16 International Breast Cancer Study Group. Late effects of adjuvant oophorectomy and chemotherapy upon premenopausal breast cancer
patients. Ann Oncol 1990; 1: 30–35. 17 Veronesi V, Volterrani F, Luini A, et al. Quadrantectomy versus lumpectomy for small size breast cancer. Eur J Cancer 1990; 26: 671–73. 18 Macmillan RD, Purushotham AD, Mallon E, Ramsay G, George WD. Breast-conserving surgery and tumour bed positivity in patients with breast cancer. Br J Surg 1994; 81: 56–58. 19 van Dalen R, Bates A, Dhesi I, Bates T, Houghton J. Positive cavity biopsy does not predict early local relapse for patients treated by local excision for early breast cancer. Breast 1995; 4: 231 (abstr no 15). 20 Hayward JL. The Guy’s Hospital trials on breast conservation. In: Harris JR, Hellman S, Silen W, eds. Conservative management of breast cancer. Philadelphia: Lippincott, 1983: 77–90. 21 Veronesi U, Marubini E, Del Vecchio M, et al. Local recurrences and distant metastases after conservative breast cancer treatments: partly independent events. J Natl Cancer Inst 1995; 87: 19–27.
Double-blind randomised trial of co-amoxiclav versus placebo for persistent otitis media with effusion in general practice
F A M van Balen, R A de Melker, F W M M Touw-Otten
Summary Background The treatment of persistent otitis media with effusion (OME) remains controversial, but this condition is the commonest reason for children to require ear, nose, and throat (ENT) surgery. Trials of antibiotics are inconclusive, are often weak methodologically, and have not been done in general practice. Our aim was a trial of an antibiotic for OME in such a population. Methods 433 children, aged 6 months to 6 years, with OME from 57 general practices entered a 3-month watchful waiting period. Of 223 (52%) with persistent bilateral OME, 162 were randomised double-blind to receive co-amoxiclav suspension (20 mg/kg amoxicillin, 5 mg/kg clavulanate potassium) or matching placebo, orally three times a day for 14 days. All cases also received xylometazoline 0·25% decongestant nosedrops thrice daily. Of the 61 not randomised, 13 children were referred to an ENT surgeon and parents refused consent in 48 cases. The main outcome measures were persistent OME in both ears and in one or both ears, as assessed clinically and by tympanometry. Analysis was by intention-to-treat. Findings 79 children in the treatment group and 70 in the placebo group were analysed for efficacy. 3 withdrew in the co-amoxiclav group (2 lost to follow-up, 1 due to sideeffects); 6 withdrew in the placebo group (5 and 1, respectively) . In addition, 4 tympanograms w ere uninterpretable in the controls. Compliance was over 90% in both groups. Persistent OME in both ears and in one or both ears were found at significantly lower rates in the coamoxiclav group than in the controls at the 2-week followup: 53 vs 84% and 77 vs 93%, respectively. Odds ratios adjusted for sex, history of adenoidectomy, and upper
Prof F W M M Touw-Otten died in April, 1996 Department of General Practice, University of Utrecht, 3584 CG Utrecht, Netherlands (F A M van Balen MD, R A de Melker MD, Prof F W M M Touw-Otten PhD) Correspondence to: Dr F A M van Balen
Vol 348 • September 14, 1996
respiratory tract infection at follow-up were 0·25 (95% CI 0·11, 0·58, p=0·001) and 0·30 (0·10, 0·89, p=0·03), respectively. Parents of children in the co-amox iclav group reported significantly more side-effects than those of control children (44 vs 22%, p=0·03). Side-effects were mostly gastrointestinal and mild. Interpretation Our study in a general-practice setting confirmed the positive short-term effect of antibiotic treatment for persistent middle-ear infection. Before referral to an ENT surgeon, children with persistent OME presenting to general practitioners could be considered for such treatment, depending on the individual child and possible adverse sequelae.
Lancet 1996; 348: 713–16
Introduction Otitis media with effusion (OME) incurs great morbidity and costs in children.1 The treatment of first choice remains controversial. There is agreement on watchful waiting before active treatment, but not for how long.1–3 Both US2 and Dutch4 guidelines include watchful waiting for 3 months, after which treatment options are medical or surgical. Medical treatment could be started by the general practitioner which might lead to fewer referrals, thereby reducing the high costs of surgery.5 Decongestants and antihistamines are ineffective. 6 Treatment with corticosteroids is still under discussion.7,8 The rationale underlying antimicrobial therapy is based on the observation that pathogenic bacteria can be recovered from the middle-ear effusion in a third of all cases.9,10 The microorganisms most frequently found are Streptocococcus pneumonia, Haemophilus influenzae, and Moraxella catarrhalis.11,12 Two meta-analyses13,14 show a positive short-term effect from antibiotics. Amoxicillin, used alone or with clavulanate potassium, produced a uniform, statistically significant effect in three placebocontrolled randomised trials,15–17 but not in one other.18 Randomised studies in which the antibiotic was given for 14 days showed a positive effect, but none was carried out in general practice. 713