DIFFERENCES IN STAGE AND THERAPY FOR BREAST CANCER ACROSS EUROPE

Transcription

1 Int. J. Cancer: 93, (2001) 2001 Wiley-Liss, Inc. DIFFERENCES IN STAGE AND THERAPY FOR BREAST CANCER ACROSS EUROPE Milena SANT* and the Eurocare Working Group Division of Epidemiology, Istituto Nazionale per lo Studio ela Cura dei Tumori, Milan, Italy Publication of the International Union Against Cancer We examined variations in stage, diagnostic workup and therapy for breast cancer across Europe. Seventeen cancer registries in six European countries contributed 4,480 cases diagnosed in The clinical records of these cases were examined, and the distribution of stage, diagnostic examinations and therapy were analyzed. Stage was earliest in the French registries, followed by those of Italy and Eindhoven (Netherlands). The proportion of stage Icancers was highest in the French areas with screening in place. Estonia, the English registries and Granada (Spain) had the most advanced stage at diagnosis. Use of liver ultrasonography varied from 84% (Italian registries) to 18% (Granada). Bone scan use varied from 81% (Italian registries) to 15% (Mersey, UK). The highest proportions treated by breast-conserving surgery were in the French (57%) and English registries (63%); the lowest were in Estonia (6%) and Granada (11%). The highest proportions of Halsted mastectomies were in Italy(19%)andGranada(8%).InallcountriesexceptEngland, 90% of operations included axillary lymphadenectomy. Medical treatment only was given to 8% of (mostly advanced) cases overall. Estonia (21%) and the English registries (14%) had the highest proportions of patients given medication only. Chemotherapy was given to low proportions of nodepositive cases in the Italian (76%) and English (74%) areas; breast-conservingsurgeryforstageitumorsvariedfrom24% in Granada to 84% in England. These wide differences in breast cancer care across Europe in the early 1990s indicate aneed for continual monitoring of past treatments to help ensure application of the most effective protocols Wiley-Liss, Inc. Key words: breast cancer; population registries; care; survival; Europe The Eurocare project, which involved 45 population-based cancerregistriesin17europeancountries,revealedlargevariationsin survival for breast cancer across Europe. These variations were attributed mainly to differences in stage at diagnosis, but the data also suggested that differences in the quality of care played a role. 1,2 Studies of patterns of cancer care are usually carried out on hospital series of patients selected with respect to the general population of cancer sufferers. These studies almost always originate from specialist oncological centers and usually exclude patients treated in non-specialist structures (general hospitals) and those not treated in hospital. 3 6 Toprovide areliable picture of how cancer patients are cared for in current medical practice, it is necessary to study all the cases occurring in the population, or representative samples thereof. Only population-based cancer registries collect data on all cancer cases, and analyses deriving from them can provide indications as to the overall effectiveness of healthcare.however,theinformationondiseasestageatdiagnosis and on treatment collected routinely by most cancer registries is limited and insufficient to illuminate in detail the reasons for differences in cancer survival from one area to another. Cancer registriesarenowcollectingmoredetailedinformationoncareand clinical aspects of cancer management As part of the Eurocare project, we have now collected and analyzed population-based information from selected European cancer registries on the procedures used to diagnose and treat breast cancer. Asingle protocol was employed to abstract the clinicalinformation.thisarticlepresentstheresultsoftheanalysis and compares distribution by stage and therapeutic procedures in 17 populations from 6European countries at the beginning of the 1990s, in order to analyze trends and suggest what effects differences in care have on survival. METHODS The cancer registries participating in the study were asked to provide data on all cases of invasive breast cancer, or arepresentative sample, diagnosed in At least 200 cases per registry were required; if this minimum was not reached over 1year, the studyperiodwasprolongeduntiltherequirednumberwasreached. The clinical information was obtained by trained registry personnel who inspected the clinical records of all patients diagnosed with breast cancer and abstracted the required information onto forms specially designed for the purpose. The information was then sent to the study coordinating center, where, following data input and data consistency checks, the analyses were performed. Data collection was considered complete when, for each registry and the study period or area chosen, the number of cases collected by the trained personnel was equal to the number of incident cases in the registry records. Appendix Ashows the information recorded in the forms. In all, 4,480 breast cancer cases were analyzed. Table Ishows the number of cases by registry and indicates how cases were sampled. The registries of Calvados, Tarn, Doubs, Cô te d Or, Somme (France), Modena, Varese, Ragusa (Italy), Eindhoven (the Netherlands) and Granada (Spain) sent all cases registered during 1990 (and also in 1991 for Eindhoven, and for Ragusa, which covers asmall area in southern Italy where the incidence of breast cancer is low). The registries of Herault and Isè re (France) andmersey(uk)drewrandomsamplesfromtheir1990incidence sets. Estonia, Thames and Bas Rhin (Lower Rhine) obtained the Grant sponsor: the BIOMED Program of the European Union. The Eurocare Working Group for this article is composed of the following members: Estonia: T. Aareleid, A. Kupp (Estonia Cancer Registry); France: P. Schaffer (Bas Rhin Cancer Registry), G. Chaplain (Côte d Or Breast and Gynecologic Cancer Registry), C. Exbrayat (Isère Cancer Registry), H. Mathieu-Daude (Herault Cancer Registry), P. Grosclaude (Tarn Cancer Registry), J. Macè-Lesech (Calvados General Cancer Registry), P. Arveux (Doubs Cancer Registry), N. Raverdy (Somme Cancer Registry); Italy: L. Mangone, M. Federico (Modena Cancer Registry), A. Barchielli, E. Paci (Tuscan Cancer Registry), F. Berrino (Project Leader), G. Gatta, D. Speciale, M.R. Ruzza, E. Frassoldi, C. Mancina (Varese Cancer Registry), R. Capocaccia, A. Verdecchia (Istituto Superiore di Sanita, Rome), L. Gafa, R. Tumino, M. La Rosa (Ragusa Cancer Registry); Netherlands: J. Wilhem W. Coebergh, A. Voogd (Eindhoven Cancer Registry); Spain: C. Martinez (Granada Cancer Registry); United Kingdom: J. Bell (Thames Cancer Registry), M.P. Coleman (London School of Hygiene and Tropical Medicine), J. Youngson (Merseyside and Cheshire Cancer Registry). *Correspondenceto:DivisionofEpidemiology,IstitutoNazionaleperlo Studio ela Cura dei Tumori, Via Venezian 1, I Milan, Italy. Fax: eurocare@istitutotumori.mi.it Received 27 November 2000; Revised 21 March 2001; Accepted 30 March 2001 Published online 13 June 2001

2 BREAST CANCER CARE IN EUROPE 895 TABLE I REGISTRIES PARTICIPATING IN THE EUROCARE HIGH RESOLUTION STUDY ON BREAST CANCER, SHOWING THE TOTAL NUMBER OF CASES BY REGISTRY, THE PROCEDURE FOR CASE RECRUITMENT, AND FIVE-YEAR SURVIVAL IN AND (PRESENT STUDY PERIOD) 1 Registry Total no. of cases Study period Sampling procedure 5-year rel. surv year rel. surv Bas Rhin All cases incident in entire registry area from 1/1/90 to NA 89 1/6/90; all cases from 1/6/90 to 29/6/90 Calvados All cases incident in the considered year Doubs All cases incident in the considered year Somme All cases incident in the considered year Tarn All cases incident in the considered year NA 82 Herault Random sample taken from 500 cases incident in the NA 90 registry area in the considered year Côte d Or All cases incident in the considered year Isère Random sample from the 527 cases incident in the registry area in the considered year Modena of the 398 cases incident in the considered year (from two random samplings in two successive periods) Firenze All cases incident from 1/1/90 to 31/9/90 in the municipality of Florence only (pre-screening period) Varese All cases incident in the considered year Ragusa All cases incident in the considered years Eindhoven All cases incident in the considered year Granada All cases incident in the considered year Estonia All cases incident from 1/1/91 to 31/8/91 in northern and western Estonia Thames All cases incident in entire registry area from 15/1/ to 31/1/90 plus all cases in the North Thames area from 1/1/90 to 15/1/90 and from 1/2/90 to 29/2/90 Mersey Random sample of the 1300 cases incident in the registry area Total NA, not available. 1 Weighted age-standardized 5-year relative survival. (From Berrino F et al. 17 ) required number of cases by considering a limited part of their area or by restricting the time interval within the study year. Five-year relative survival was calculated by the Hakulinen method 15 using age- and sex- matched specific life tables of the general population pertaining to each registry area. For each registry, relative survival rates were calculated. These express the probability of cancer survival after adjustment for competing causes of death and were estimated as the ratio of the observed survival to the expected survival for the corresponding general population. Overall, 3% of cases were lost to follow-up, ranging from 0% (Calvados, Modena and Ragusa) to 11% (Doubs and Herault). The analyses of frequency and percentage distribution of variables were performed using the SPSS package. The variables were analyzed by registry, but overall figures were calculated for France, Italy and the UK, which are represented by more than one registry. For these countries, weighted percentages for the overall country were calculated for stage, therapy and diagnostic examinations performed. Weightings were the total number of cases by country in Disease stage at diagnosis was assessed by the TNM system. 16 Patients who underwent surgery were assessed by pathological pt and pn stage; for those not treated surgically the clinical TNM was used. For purposes of the analysis, five stage categories were defined: T1N0M0; T2 3N0M0 (no regional lymph node involvement); T1 3N M0 (with regional lymph node involvement); T4M0 regardless of node status; and M1 (any T, any N). As the probability of receiving a given treatment depends on age and stage at diagnosis, we used multiple regression analysis to estimate the age- and stage-adjusted variability of specific treatment patterns between registries grouped by country. RESULTS In addition to modalities of case selection, Table I shows overall 5-year relative survival by registry. The survival figures for patients diagnosed in are also presented for comparison; these were drawn from the Eurocare II survival study. The main aim of this comparison was to check the consistency of the present data with previous survival data from the same registry. The rank of overall 5-year relative survival in this study was similar to that obtained by Eurocare for The French registries had high survival, as did the Modena and Florence registries in Italy. Low survival was found in the cancer registries of Thames (UK), Granada (Spain), Ragusa (Italy) and Estonia. In most registries participating in both studies, survival was the same or slightly higher in the present study than in Eurocare II. The two figures are close, however, suggesting that no major biases occurred in sampling cases for the present study. Table II shows the distribution of cases by pathological stage or clinical stage (when the former was not available) in each of the participating registries. Breast cancer was generally diagnosed at an earlier stage in France than the other countries; however, the between-registry variation in stage distribution was marked in that country. The registries of Bas Rhin, Herault, Isère, and Côte d Or had, in order, the highest percentages of T1N0M0 tumors. Mass screening for breast cancer was in place in the first three of these registries during the study period. In the other French registries of Tarn, Calvados, Doubs and Somme, where mass screening was not in place, stage was more advanced. In Italy, breast cancer was diagnosed earliest in Modena and Florence, and at the most advanced stage in Ragusa. Only in the area of Florence is mass screening now established, but this was set up in late 1990 outside the study period. Average stage at diagnosis in Eindhoven, the only registry representing the Netherlands, was close to that in France and Italy. The difference between the two UK registries in terms of stage at diagnosis was marked; however, the high percentage (31%) of unstaged and unknown stage cases in Thames made more precise analysis impossible. In Granada, the only registry representing Spain, the percentage of early stage tumors (T1N0M0) was among the lowest; further-

3 896 SANT ET AL. TABLE II TOTAL NUMBER OF BREAST CANCER CASES BY REGISTRY AND COUNTRY AND DISTRIBUTION (%) BY STAGE AT DIAGNOSIS Registry and country Total no. of cases % Distribution by stage T1 N0M0 T2 3 N0M0 T1 3 N M0 T4 M1 Stage not available France 1 1, Bas Rhin Côte d Or Isère Herault Tarn Calvados Doubs Somme Italy 1 1, Modena Firenze Varese Ragusa NL Eindhoven Spain Granada Estonia England Thames Mersey Weighted mean. Registry and country TABLE III BREAKDOWN OF PT1 OPERATED CASES BY REGISTRY AND PATHOLOGICAL SIZE pt1 cases No. % % PT1a ( 9 mm) % PT1b (10 14 mm) %PT1c (15 19 mm) % pt1 with path. size NA 1 France Bas Rhin Côte d Or Isère Herault Tarn Calvados Doubs Somme Italy Modena Firenze NA NA NA NA Varese Ragusa NL Eindhoven Spain Granada Estonia England Thames Mersey NA NA NA NA NA, not available. 1 Weighted mean. 2 Only data from Thames available. more, node-negative tumors were larger than in France and Italy, and the percentage of T4 tumors was high. Cases from Estonia were characterized by the most advanced stage distribution at diagnosis, the lowest proportion of T1N0M0 cases, the highest proportion of T4 cases and the second highest proportion of M1 cases. The English registries (Thames and Mersey) had the highest proportion of M1 cases, and the proportion of N M0 cases was relatively low. Considering all the data, a positive relation was found between age at diagnosis and stage; in addition, unstaged tumors were more frequent in older patients. T1N0M0 cases constituted 32% of all tumors diagnosed in women up to 55 years old but were 28% of all tumors in women above 55 years. Considering all node-negative cases, tumor size increased with age. Surprisingly, node-positive tumors were more frequent in younger patients: 34% in women up to 55 years and 28% in women above 55 years. Five percent of tumors in women up to 55 years were T4 and 8% after 55 years; for M1 the corresponding figures were 5% and 7%. The proportion of very small tumors gives an indication of the intensity of diagnostic activity in a given area. Table III shows for each registry and country the number of operated cases at stage T1. The pt1 category is further subdivided as follows: up to 9 mm, mm and mm (pt1a, -b and -c, respectively). In 72% of all cases microscopically measured tumor size was available. Table III also shows the proportions of pt1 cases for which pathological size was not available and the total pt1 cases (with and without pathological size) as a proportion of all operated cases. The proportion of pt1 cases (among operated cases) was highest in the French registries, intermediate in the Netherlands and Italy, and lowest in Granada and Estonia. In England the proportion of

4 BREAST CANCER CARE IN EUROPE 897 pt1 cases among operated cases was only slightly lower than in France, Italy and Eindhoven. This, considered with the staging information in Table II, suggests that larger tumors are less likely to be operated on in England than in most other countries considered. In France, the proportion of small tumors was higher where mass screening was in place during the study period. No cases with T 9 mm were reported by the registries of Granada and Estonia. Table IV shows the diagnostic procedures carried out to establish disease extent at the time of diagnosis. More than 70% of patients underwent liver ultrasonography in the French and Italian registry areas, whereas only around 20% did so in Granada and Estonia. The two English registries differed markedly: 41% of cases in Thames and 10% in Mersey received liver ultrasound. Fewer patients received bone scan than liver scan; however, the country rankings for the performance of these two procedures were similar, and were unchanged when T1N0M0 cases (for which extensive investigations are not recommended) were excluded (data not shown). Table IV also shows the mean number of examined nodes, the percentage of patients with at least four examined nodes, and the percentages of patients with three or more metastatic nodes on pathological examination of the axilla. There was a direct relationship between the number of nodes examined and the number found metastatic. Both these figures were higher in the French than Italian registries but were highest in Granada. In the two English registries, the numbers of examined and metastatic nodes were low compared with the other countries. Table V shows the proportions of patients treated surgically and the type of surgery performed. Overall, 90% patients underwent surgery. The feasibility of breast-conserving surgery depends on the extent of the disease, and in all registries and countries breastconserving surgery was more frequent in early stage disease. This was particularly the case among the French registries, where breast-conserving surgery was performed more frequently in the registry areas characterized by earlier stage at diagnosis than those characterized by later stage at diagnosis. However, even allowing for this, variations in attitudes to breast conservation emerged. France and the UK had the highest percentages of cases undergoing breast-conserving surgery, whereas in Italy and Spain very low proportions of patients were treated conservatively. The highest percentages of Halsted mastectomies were in Italy and Spain. However, the high overall figure for Italy is mainly due to Ragusa, where considerably more mastectomies were performed than in the other Italian registry areas. It is also noteworthy that Italy and Spain had the highest percentage of T1N0M0 cases treated by Halsted mastectomy (8% in both countries), compared with less than 1% in the other countries. In all countries except England, 90% or more of the breast operations included axillary lymphadenectomy. Table VI gives a detailed breakdown of therapeutic regimens by country. Nine categories of breast treatment are shown according to type of surgery performed (breast-conserving surgery, any mastectomy) and whether or not axillary lymphadenectomy and radiotherapy were performed. The addition of systemic chemotherapy and hormonal treatment (always tamoxifen) is also indicated for each treatment category. This table shows that patterns of treatment by country vary considerably. The most common treatment was breast-conserving surgery plus axillary lymphadenectomy plus radiotherapy, which was given to 31% of patients overall; however, this figure ranged from 47% in France to 1.3% in Estonia. The second most common treatment was mastectomy (with or without axillary lymphadenectomy). This modality was frequent in Italy, Granada (Spain) and Estonia, where breastconserving surgery was less frequent. In Italy and Spain the use of radiotherapy in addition to both breast-conserving surgery and mastectomy was low compared with France, the Netherlands and England. The addition of hormonal therapy to surgery was most common in Eindhoven (the Netherlands) and in the English registries. Italy had the lowest percentages of patients given systemic adjuvant therapy. Medical treatment only, without surgery, was given to 8% of all cases. These were mainly advanced stage cases: in the T4 M1 group, 42% received medical treatment only. Estonia and England had the highest proportions of patients who received medical treatment only. Disease stage was a major determinant of treatment choice. However, the trends shown in Table VI persisted after adjustment by stage and age at diagnosis by multivariate logistic regression analysis. Taking France as the reference category, the odds ratio (OR) of not performing surgery was 5.2 in Registry and country Liver ultrasound (%) TABLE IV DIAGNOSTIC PROCEDURES 1 Bone scan (%) % patients with at least 4 examined nodes % patients with 3 or more positive nodes Mean no. of nodes examined France Bas Rhin Calvados Doubs Somme Tarn Herault Côte d Or Isère Italy Modena Firenze NA NA Varese Ragusa NL Eindhoven NA NA Spain Granada Estonia England Thames Mersey Percentages of patients receiving bone scan and liver ultrasound, with 4 or more examined nodes, 3 or more metastatic nodes and mean number of nodes examined. NA, not available. 2 Weighted mean.

5 898 SANT ET AL. TABLE V NUMBERS AND PERCENTAGES OF BREAST CANCER PATIENTS OPERATED ON, TYPE OF SURGERY PERFORMED AND PROPORTION RECEIVING AXILLARY LYMPHADENECTOMY DURING BREAST SURGERY Registry and country No. of operated cases (% of total cases) Any conservative surgery Any mastectomy (simple/modif.) Type of surgery Halsted mastectomy Not specified % of operated patients who received axillary lymphadenectomy France (92.0) Bas Rhin 197 (93.8) Calvados 278 (90.3) Doubs 207 (92.4) Somme 220 (90.2) Tarn 160 (92.0) Herault 180 (89.1) Côte d Or 234 (93.2) Isère 193 (95.5) Italy (92.1) Modena 288 (95.4) Firenze 197 (95.2) Varese 424 (90.8) Ragusa 190 (87.6) NL Eindhoven 472 (92.7) Spain Granada 158 (88.3) Estonia 167 (74.6) England (85.0) Thames 283 (83.2) Mersey 192 (87.7) Weighted mean. Registry and country (no. of cases) 2 No surgery BCS only TABLE VI BREAKDOWN OF THERAPY BY COUNTRY 1 BCS axillary BCS radiotherapy Breast treatment (%) BCS axillary radiotherapy Mastectomy only Mastectomy axillary Mastectomy radiotherapy Mastectomy axillary radiotherapy France (1,815) % Chemotherapy % Hormonal therapy Italy (986) % Chemotherapy % Hormonal therapy NL: Eindhoven (510) % Chemotherapy % Hormonal therapy Spain: Granada (179) % Chemotherapy % Hormonal therapy Estonia (224) % Chemotherapy % Hormonal therapy England (559) % Chemotherapy % Hormonal therapy Percentage of patients treated by systemic therapy added to each category of breast treatment. BCS, breast-conserving surgery; axillary, axillary lymphadenectomy. 2 Chemotherapy, systemic chemotherapy added to breast treatment (or administered alone for cases not operated); Hormonal therapy, hormonal therapy added to breast treatment (or administered alone for cases not operated). 3 Excluding Florence, for which information on adjuvant treatment was not available. Estonia (p 0.000) and 1.7 in Granada (p 0.008). Still with France as the reference category, the OR of undergoing breastconserving surgery versus mastectomy were significantly lower in Estonia (OR 0.1) and Granada (OR 0.3), but was significantly higher in England (OR 1.5, p 0.001). The OR of undergoing radiotherapy in addition to breast-conserving surgery was significantly lower than France in all countries except the Netherlands (Eindhoven). To provide indications as to the appropriateness of treatments, we considered three treatment recommendations derived from protocols widely available during the study period in the countries included in this study: (i) premenopausal patients with involved lymph nodes benefit from chemotherapy; (ii) postmenopausal patients benefit from tamoxifen (since data on menopausal status were not collected, age above and below 50 years was taken as a proxy for menopausal status and (iii) patients with small tumors (T1) can be treated by breast-conserving surgery. Table VII shows the extent of adherence to these criteria. In Italy and the UK adhesion to the first criterion was low, in that only 76% and 70%, respectively, of patients under 50 years with node-positive tumors received chemotherapy. The use of tamoxifen in patients over 50 years was lower in Italy and Eindhoven than in other countries. In the English registries and in Estonia, where tamoxifen was most frequently used, 59% of

6 BREAST CANCER CARE IN EUROPE 899 patients younger than 50 received tamoxifen as the main or adjuvant treatment, compared with 27% in France, 20% in Italy and 12% in Granada. With regard to the third criterion, patients with T1 tumors in Spain and Estonia received breast-conserving surgery less frequently than in the other countries; in Italy the percentage of T1 cases receiving such treatment was also low. DISCUSSION This study has shown that, at the beginning of the 1990s, breast cancer stage at diagnosis varied considerably across Europe. Furthermore, although in general treatment was related to disease extent and to age, regional variations in treatment within stage categories and within age groups also emerged. In some registry areas adherence to established treatment guidelines was poor. In particular, chemotherapy was given to a relatively low number of node-positive cases in Italy and in England; the use of breastconserving surgery for small tumors varied markedly from one part of Europe to another; and the Halsted mastectomy was performed in a relatively high proportion of patients with small tumors in the Italian registries and the Spanish area (Granada). Most of these geographic differences in treatment remained significant after controlling for stage and age at diagnosis. This suggests that clinical traditions or practice are as important as disease factors in the choice of treatment. The reliability of the results in this kind of study depends on the completeness of the clinical information on the clinical records. We had direct access to that information; however, the clinical records may underreport the procedures actually performed. A clinical study covering about the same period suggests that the frequency of adjuvant treatment in Italy may be higher than that found in this population-based study, but a larger proportion of the patients was included in clinical trails. 6 Underreporting of crucial information such as that on stage is in itself an indicator of poor quality of care, since clear information on stage is essential for adequate planning of therapy and also of care subsequent to first-line treatment. Most registries had to search actively for stage at diagnosis, in that it was not explicitly stated on the clinical record but had to be reconstructed from the information available on the clinical record. For most registries, the information on stage was satisfactorily complete; the only exception was Thames, where stage was unknown in 31% of cases. Among the other registries, the highest percentages of unknown stage cases were in Somme (18%), Florence, Ragusa and Doubs (12%); for the remaining registries the figures were less than 10%. The high proportion of cases with unknown stage in Thames was due to lack of information on the clinical records. This has been confirmed by another study noting that stage was available in only 50% of Thames registry cases. 8 By contrast, the completeness of information on whether or not surgery was performed, and the type of surgery, was good. The type of surgery is usually recorded routinely in all hospitals in the clinical records and is easily available to cancer registries. With regard to adjuvant therapy associated with surgery, it is possible that some underreporting occurred, especially for therapies not TABLE VII PERCENTAGE OF PATIENTS TREATED ACCORDING TO SELECTED GUIDELINES, BY COUNTRY Country % of T1N0M0 patients receiving conservative surgery % of patients over 50 yrs receiving tamoxifen %ofn patients below 50 yrs, receiving chemotherapy France Italy NL: Eindhoven Spain: Granada Estonia England requiring overnight hospital admission. However, considering the French, Italian and English registry areas, it was found that the proportions receiving different treatment regimens were fairly similar. This is an indirect indication that major underreporting did not occur. The extent to which the picture provided by the registries considered in this study is representative of countries as a whole is not known. However, we did find that interregistry variations in survival within England, France and Italy were generally smaller than intercountry variations. When a small area represents an entire country (in the present study Granada in Spain and Eindhoven in the Netherlands), the area probably does not reflect the situation in the country as a whole. For example, stage distribution in Granada is likely to be more advanced than in other Spanish areas, especially wealthier regions. 17 Estonia, which is covered by a national registry, provided a sample of cases from two-thirds of the country, which were considered to be fairly representative of the whole country. Irrespective of whether the situation in a given registry area is representative of the entire country, the identification of an area where the level of care is low constitutes a public health problem that needs to be addressed. A case in point is Ragusa, in the south of Italy, where stage was advanced and therapy inadequate compared with other Italian areas. On the other hand, where an area has higher than average levels of treatment and survival, this identifies a reachable standard that the health authorities in other areas should seek to attain. In the French and Italian areas there was considerable betweenregistry variation in stage at diagnosis, mainly due to differences in the proportions of T1N0M0 tumors. In France, this variation was related to coverage by screening programs, whose effect was to increase the detection of early stage tumors. In areas where screening was not in place, the stage distribution was more advanced. In Italy, the two regions with the highest T1N0M0 percentages were Florence and Modena. In both these areas screening was not in place in the study period; however, screening was established soon after the study period in Florence and it is possible that activity before screen was set up may have encouraged diffusion of more modern treatment protocols. In Modena, survival for most tumors is higher than in any other Italy registry area, and this is almost certainly because the standard of care is somewhat better than in the other areas. By contrast, the most advanced stages were in Ragusa, an area in which survival is low for many tumors. 18 In England, Mersey cases were characterized by earlier stage at presentation than Thames cases; this is surprising, as the published data show worse survival in Mersey than Thames. 19 However, the high percentage of unknown stage patients in the Thames registry makes comparison impossible. Data for the period from the East Anglia cancer registry show that the percentage of stage I breast cancers (corresponding to T1N0M0 in the present study) was 27%, 10 close to the weighted mean of the two English registries considered in this study. We compared the 5-year survival of the cases analyzed in this study (diagnosed in ) with those published in Eurocare II and diagnosed in The Eurocare figures are in general lower, confirming the well-documented trend for breast cancer survival to improve over time. The greatest improvements occurred in registry areas where survival was particularly low in the 1980s. 20 Stage at diagnosis correlated strongly with 5-year survival; furthermore, within stage categories, the percentage of tumors that were node negative correlated with survival. The registry areas with more than 50% patients with node-negative tumors (French and Italian areas and Eindhoven) also had the highest survival. The relation between the percentage of node-negative tumors and survival is apparently straightforward, but interpretation of the distribution of node-positive tumors is complicated by the positive relation between the number of lymph nodes examined and the number of metastatic nodes found, as shown in Table IV. In the

7 900 SANT ET AL. English areas, the percentage of node-positive tumors was low, because axillary dissection is performed less often in the UK than other countries. Thus, the low proportion of node-positive tumors in England reflects clinical practice and is not indicative of earlier stage at diagnosis. In countries where axillary dissection is more widespread, the proportion of patients with metastatic nodes was higher. Our study revealed regional differences in some of the procedures used to assess stage. When the registries were grouped by country, the mean number of lymph nodes examined pathologically following lymphadenectomy varied, with very few nodes examined in England and Estonia and a tendency for more nodes to be examined in Granada and Italy, followed by France and the Netherlands (Table IV). The frequency of investigations to detect distant metastases also varied. These investigations were most frequently and uniformly performed in the Italian and French registry areas and considerably less frequently in Granada, Estonia and Mersey. In Estonia and Granada the low use of liver ultrasound and bone scan is probably due to lack of economic resources. In other areas it is likely that these techniques are low priority for reasons other than cost. Data were not available for Eindhoven. In areas where patients were investigated less extensively and the quality of the diagnostic workup seemed lower, stage was more advanced, and patients underwent surgery less frequently. The two latter factors are related since the feasibility of surgical treatment, possibly by breast-conserving resection, is greater when stage is limited. In the French and English registries breast-conserving surgery was performed more frequently than in the other areas included in the study. However, in France breast-conserving surgery was performed more frequently in relation to limited stage, whereas in England breast-conserving surgery was performed even though stage distribution was less favorable. By contrast, the low proportion of patients who received breast-conserving surgery in Estonia and in Granada was probably related to the fact that stage at diagnosis was generally advanced, although it is also possible that breast-conserving treatment protocols had not been widely adopted in those areas. It is surprising that such a low proportion of breast cancer patients underwent breast-conserving surgery in Italy, since this technique was first developed and successfully applied in that country. 21 In fact, there were considerable differences between the Italian registry areas in the use of breast-conserving surgery. Proportionally more women underwent such surgery in Florence, because of limited stage and perhaps also because of modern treatment protocols had been adopted in relation to the imminent adoption of screening. Considerably fewer patients were treated conservatively in Varese. Even for early tumors (T1N0M0), the percentage of patients treated conservatively was low in Italy, in spite of recommendations arising from the results of extensive controlled clinical trials. It is interesting to compare data on breast conservation for other areas at about the same time: 52% received breast conservation in Ontario ( ), 12 40% in New South Wales (1990), 11 41% in US SEER registries and 38% in the US National Cancer Data Base clinical series (1992). 14 In the latter study, 32% and 33% of patients also received adjuvant radiotherapy. The relatively common use of the Halsted mastectomy in Italy and Granada found in this study and also in another study 22 requires comment. This mutilating and disabling operation should have been abandoned for all but a few breast cancer cases at the beginning of the 1990s. The continued widespread use of Halsted mastectomy in several areas may be related in part to a lack of adequate radiotherapy facilities: for patients who have to travel excessively long distances for such treatment, mastectomy might be preferred. The use of adjuvant therapy following surgery was lower in the Italian and English registry areas than in the areas covered by the French registries and Eindhoven. This may in part be due to differential loss of information on therapy administered outside the hospital. However, adhesion to the recommendations we selected as indicators of good treatment was poor in Italy, particularly in comparison with the French registries and Eindhoven. The English registry areas were characterized by highest adhesion to two of the three recommendations (breast-conserving surgery in stage I disease and use of tamoxifen in women over 50). In fact, tamoxifen was widely used in women of all ages in the English registry areas, due to the positive results of two British studies, 23,24 even though during the study period there was little overall evidence that it benefited women aged less than 50 years. 25 CONCLUSIONS This study has found that there were important differences in the quality of care for breast cancer in different areas of Europe in the early 1990s. These differences were suspected from the regional differences in breast cancer survival found in earlier populationbased studies. 17 The treatment differences can in part be attributed to the fact that treatment guidelines were not uniform across Europe during the study period, particularly as concerns the use of adjuvant systemic therapy. 26 Furthermore, in general the observed differences in survival across Europe seem related to differing stage at diagnosis, which in turn influenced choice of treatment. During the 1980s there were marked increases in survival for breast cancer in most European countries, particularly in countries with lower than average survival. 20,27 Mortality for breast cancer started to decrease during the same period, and this trend has continued over subsequent years The improvement in survival together with the decrease in mortality suggests that patients are diagnosed at increasingly more limited stages and are better treated than in the past. The results of this study have important consequences for the present since the survival and mortality figures for breast cancer today depend on the diagnoses and treatments of 10 years ago (and more). This study therefore identifies a need for continual monitoring of past treatments for breast cancer and also for wider application of the most effective treatments for this disease. ACKNOWLEDGEMENTS We are grateful to Ms. Emily Taussig for editorial assistance in the preparation of this manuscript, and to Don Ward for help with the English. APPENDIX A Variables collected by the Eurocare high-resolution study on breast cancer: Registry, country, anonymous identification code of patient Date of diagnosis, date of birth Marital status Occupation or indicator of social status Tumor site, coded according to ICD code Histotype, coded centrally according to ICDO code Inclusion in a clinical trial Pathological stage (or clinical stage if pathological not available): TNM Diameter of the tumor in mm Number of lymph nodes examined and number of metastatic nodes found Diagnostic examinations performed: Mammography Biopsy or needle aspiration Imaging of thorax, skeleton, liver Hormonal receptors

8 BREAST CANCER CARE IN EUROPE 901 Therapy: Type of surgery and whether lymphadenectomy performed If no surgery, reasons for not performing it Chemotherapy (regardless of regimen), date of beginning treatment If no chemotherapy, reasons for giving it Radiotherapy (regardless of dose and modality), date of beginning radiotherapy If no radiotherapy, reasons for not giving it Treatment with tamoxifen, date of beginning treatment If no tamoxifen, reasons for not giving it Vital status of the patient and date of last check. REFERENCES 1. Sant M, Capocaccia R, Verdecchia A, et al. Survival of women with breast cancer in Europe: variation with age, year of diagnosis and county. Int J Cancer 1998;77: Quinn MJ, Martinez-Garcia C, Berrino F, et al. Variations in survival from breast cancer in Europe by age and country, Eur J Cancer 1998;34: Kosecoff J, Kanouse D, Rogers WH, et al. Effects of the National Institutes of Health Consensus Development Program on physician practice. JAMA 1987;258: McCarthy M, Bore J. Treatment of breast cancer in two teaching hospitals: a comparison with consensus guidelines. Eur J Cancer 1991;27: Nicolucci A, Mainini F, Penna A, et al. The influence of patient characteristics on the appropriateness of surgical treatment for breast cancer patients. Ann Oncol 1993;4: Alexanian AA, Scopriglione N, Apolone G, et al. Breast cancer surgery in 30 Italian general hospitals. Eur J Surg Oncol 1993;19: Nab HW, Hop WCJ, Crommelin MA, et al. Improved prognosis of breast cancer since 1970 in south-eastern Netherlands. Br J Cancer 1994;70: Chouillet AM, Bell CMJ, Hiscox JG. Management of breast cancer in southeast England. BMJ 1994; 308: Sainsbury R, Rider L, Smith A, MacAdam A on behalf of the Yorkshire Cancer Registry. Does it matter where you live? Treatment variation for breast cancer in Yorkshire. Br J Cancer 1995;71: Stockton D, Davies T, Day N, et al. Retrospective study of reasons for improved survival in patients with breast cancer in East Anglia: earlier diagnosis or better treatment? BMJ 1997;314: Adelson P, Lim K, Churches T, et al. Surgical treatment of breast cancer in New South Wales 1991, Aust NZ J Surg 1997;67: Iscoe NA, Goel V, Wu K, et al. Variation in breast cancer surgery in Ontario. Can Med Assoc J 1994;150: Voogd AC, Van Beek MWPM, et al. on behalf of the regional Breast Cancer Study Group. Management of early breast cancer in southeast Netherlands since A population-based study. Acta Oncol 1994; 33: Mettlin CJ, Menck HR, Winchester DP, Murphy GP. A comparison of breast, colorectal, lung and prostate cancers reported to the National Cancer Data Base and the Surveillance, Epidemiology, and End Results Program. Cancer 1997;79: Hakulinen T, Abeywickrama KH. A computer program package for relative survival analysis. Comp Prog Biomed 1985;19: Spiessel B, Beahers OH, Hermanek P, et al., eds. UICC: TNM atlas classification of malignant tumours, 3rd ed. Berlin: Springer, Berrino F, Capocaccia R, Estève J, et al., eds. Survival of cancer patients in Europe: the EUROCARE-2 study. IARC Scientific Publications no Lyon: IARC, Gatta G, Buiatti E, Conti E, et al. Variations in the survival of adult cancer patients in Italy. Tumori 1997;83: Coleman MP, Babb P, Damiecki P, et al. Cancer survival trends in England and Wales : deprivation and NHS region. Series SMPS no. 61. London: The Stationery Office, Estève J, De Angelis G, Verdecchia A. Trends in cancer survival probability over the period In: Berrino F, Capocaccia R, Estève J, et al., eds. Survival of cancer patients in Europe: the EUROCARE-2 study. IARC Scientific Publications No Lyon: IARC, p Veronesi U, Saccozzi R, Del Vecchio M, et al. Comparing radical mastectomy with quadrantectomy, axillary dissection, and radiotherapy in patients with small cancers of breast. N Engl J Med 1981;305: Grilli R, Mainini F, Penna A, et al. Inappropriate Halsted mastectomy and patient volume in Italian hospitals. Am J Public Health 1993;83: Medical Research Council Scottish Trials Office. Adjuvant tamoxifen in the management of operable breast cancer. Lancet 1987;ii: Nolvadex Adjuvant Trial Organisation. Controlled trial of tamoxifen as a single adjuvant agent in the management of early breast cancer. Br J Cancer 1988;57: Early Breast Cancer Trialists Collaborative Group. Effects of adjuvant tamoxifen and of cytotoxic therapy on mortality in early breast cancer. An overview of 61 randomized trials among 28,896 women. N Engl J Med 1988;319: Clinical trials and clinical practice [editorial]. Lancet 1993;342: Brenner H, Stegmaier C, Ziegler H. Recent improvement in survival of breast cancer patients in Saarland, Germany. Br J Cancer 1998;78: Beral V, Hermon C, Reeves G, Peto R. Sudden fall in breast cancer death rates in England and Wales [letter]. Lancet 1995;345: Blanks RG, Moss SM, McGahan CE, et al. Effect of NHS breast screening programme on mortality from breast cancer in England and Wales, : comparison of observed with predicted mortality. BMJ 2000;321: La Vecchia C, Negri E, Levi F, et al. Age, cohort-of-birth, and period-of-death trends in breast cancer mortality in Europe [letter]. J Natl Cancer Inst 1997;89:732 4.