Breast-conserving surgery with or without radiotherapy vs mastectomy for ductal carcinoma in situ: French Survey experience

From March 2003 to April 2004, 77 physicians throughout France prospectively recruited 1289 ductal carcinoma in situ (DCIS) patients and collected data on diagnosis, patient and tumour characteristics, and treatments. Median age was 56 years (range, 30–84). Ductal carcinoma in situ was diagnosed by mammography in 87.6% of patients. Mastectomy, conservative surgery alone (CS) and CS with radiotherapy (CS+RT) were performed in 30.5, 7.8 and 61.7% of patients, respectively. Thus, 89% of patients treated by CS received adjuvant RT. Sentinel node biopsy (SNB) and axillary dissection (AD) were performed in 21.3 and 10.4% of patients, respectively. Hormone therapy was administered to 13.4% of the patients (80% tamoxifen). Median tumour size was 14.5 mm (6, 11 and 35 mm for CS, CS+RT and mastectomy, respectively, P<0.0001). Nuclear grade was high in 21% of patients, intermediate in 38.5% and low in 40.5%. Excision was considered complete in 92% (CS) and 88.3% (CS+RT) of patients. Oestrogen receptors were positive in 69.8% of assessed cases (31%). Treatment modalities varied widely according to region: mastectomy rate, 20–37%; adjuvant RT, 84–96%; hormone treatment, 6–34%. Our survey on current DCIS management in France has highlighted correlations between pathological features (tumour size, margin and grade) and treatment options, with several similar variations to those observed in recent UK and US studies.

In France, 49.236 new breast cancers were diagnosed in women in 2004 (Allemand et al, 2008). Between 2000 and 2004, the annual incidence rate increased by 2%. However, the precise proportion of ductal carcinoma in situ (DCIS) in France is unknown. Presently, DCIS accounts for approximately 15 -20% of all breast cancers (Baxter et al, 2004;Leonard and Swain, 2004). The French screening programme was introduced in several counties (départements) in 1989 and was progressively extended all over France within the framework of the National Cancer Plan in 2000. The national coverage was achieved in 2005. Ductal carcinoma in situ rate among abnormal screen-detected mammographies reached 13.8%. Ductal carcinoma in situ is treated by mastectomy or conservative surgery (CS) with or without radiotherapy (RT) (Morrow et al, 2002;Ceilley et al, 2004;Mokbel and Cutuli, 2006). The best option is still a matter of debate, with wide differences in practice observed across European countries and the United States (Ceilley et al, 2004;Mokbel and Cutuli, 2006). Moreover, important regional differences were also reported.
Up-to-date data on DCIS are not available for France. We therefore performed a national prospective survey of the diagnosis, histopathological features and treatment of pure DCIS in France and compared our findings with those from other countries. The methodology of this survey was similar to a study on invasive breast cancer conducted in .

MATERIALS AND METHODS
Our observational, prospective cross-sectional study was performed from March 2003 to April 2004. A total of 77 physicians took part in the study and were recruited in surgery, oncology and radiation oncology departments throughout France. Most of them participated in the first national survey on infiltrating breast cancer in . Those physicians are coordinators of a local breast unit, or general or gynaecological surgeons involved in breast cancer care. Finally, they are usually good representatives of all French regions (urban and rural) and types of sanitary structures. Indeed, patients were managed in cancer centres (49% of patients), private clinics (33%), or university or general hospitals (18%), reflecting quite well the daily current practice in our country. All patients seen in each hospital were prospectively included, but obviously with a more important recruitment in several specialized centres (7 with more than 50 cases each).
Eligibility criteria were female gender, X18 years of age, with a diagnosis of pure DCIS. Patients with microinvasive lesions, previous or synchronous contralateral breast cancer or other cancers were excluded.
We collected data on demographics (age, family history of breast cancer, menopausal status, hormonal replacement therapy (HRT)), clinical presentation, mammography findings, biopsy procedures, specimen characteristics as given by the pathology report (tumour subtype, size, grade, excision quality and hormone receptor status) and treatment (types of surgery, number of operations, adjuvant therapy (RT with or without boost and hormone therapy) (see form in Supplementary Information). All data were checked (exhaustive pathological forms, radiation doses and other treatments) by the first author in collaboration with two pathologists of the scientific committee (FP-L, EM) and the statistical team, especially regarding some cases of 'incoherencies'. The individual survey forms were made by each physician who sent them to the data management centre.
To compare two groups, we used the w 2 test for categorical variables, the t-test for continuous variables and the nonparametric Mann -Whitney test when the assumption of normality was questionable. To compare more than two groups, we used the w 2 test for categorical variables, one-way analysis of variance for continuous variables and the non-parametric Kruskal -Wallis test when the assumption of normality was questionable. Three treatment groups were compared, namely CS alone (99 patients: 7.8%), CS þ RT (797 patients: 61.7%) and mastectomy (393 patients: 30.5%). We also performed regional comparisons dividing the country in five regions, namely Paris, northwest, northeast, southeast and southwest. Finally, a multivariate analysis was included, allowing predicting the therapeutic choices, such as mastectomy vs CS as well as sentinel node biopsy (SNB) vs axillary dissection (AD) according to age, tumour size (both quantitative variables), or grade and the presence of necrosis (both qualitative variables). A stepwise procedure was used to select variables for the multiple logistic regression analysis.
In total, 816 (63.3%) patients were menopausal (median age, 50 years); 52.3% of them had undergone HRT for a median duration of 7.7 years.

Presentation at diagnosis
In 87.6% of patients, DCIS was diagnosed by mammographic abnormality. Only 12% of patients had clinical symptoms, such as palpable mass, Paget's disease, serosanguineous nipple discharge and/or retraction; such clinical symptoms were present in 9.5% of CS groups (CS and CS þ RT) compared with those in 17.3% of the mastectomy group.

Preoperative biopsy
Preoperative needle biopsy was performed in 61.8% of patients. Vacuum-assisted needle biopsy was performed in 66.4% of these patients and stereotactic core-needle biopsy in 33.6%.
Excision was considered complete (usually margins X1 mm) in 92% of CS patients and in 88.3% of CS þ RT patients. The corresponding percentages were 3 and 5.8% for doubtful excision, 4 and 2.4% for incomplete excision, with 1 and 3.5% for nonevaluable excision. Tumour-free margins were measured in 66.7% of CS patients and 70.3% of CS þ RT patients. The margin was o1 mm in 10.6% (CS) and 12.1% (CS þ RT) of patients, 1 -3 mm in 21% of patients (whether CS or CS þ RT), 4 -10 mm in 51.5% (CS) and 57.1% (CS þ RT) of patients, and 41 cm in 16.7% (CS) and 10.3% (CS þ RT) of patients. Thus, we do not have any specific analysis of multifocality and multicentricity, but we have the notion of 'residual tumour' on the specimen in case of multiple surgery, maximal tumour size and final margin status to predict the best surgical option (especially mastectomy). Indeed, the definitions of multifocality or multicentricity are often very subjective and confusing, especially in case of several little foci in the same area, but each one separated only by few millimetres.
Steroid hormone receptors were assayed in 31% of tumours. Oestrogen receptors (ER) were positive in 70% of patients and progesterone receptors in 59.5%. Hormonal replacement therapy had no influence on lesion size, architectural subtype and hormone receptor status, but was significantly associated with tumour grade. Indeed, low-grade tumours were found in 26.2% of women who had undergone HRT and in 18.3% who had not. Tumours were high grade in 34.2% (HRT) and 47.6% (no HRT) of women (P ¼ 0.0004).

Treatment
Out of 1289 patients, 896 (69.5%) underwent CS and 393 (30.5%) mastectomy. The 69.5% undergoing CS could be further classified into 7.8% who underwent just CS and 61.7% who received adjuvant RT (89%). Generally, 70% of the patients underwent one operation, 27% two and 3% three. Residual DCIS foci were found in 59% of patients undergoing a second or third operation. Mastectomy was performed in 40% of patients as a first option but in 50% after one or two lumpectomies, and in 10% after quadrantectomy. Most patients who underwent multiple operations ended up by having a mastectomy. The number of breast surgical procedures was significantly reduced if a preoperative biopsy (stereotactic core-needle biopsy and more specially vacuum-assisted needle biopsy) was performed ( Table 1). The rate of AD was also significantly reduced by the preoperative biopsy use, whereas the rate of SNB was increased (Table 2).
Tumour size, nuclear grade, necrosis and age significantly influenced treatment modalities (Table 3). Mastectomy was used to treat 41% of high-grade tumours vs 15.3% of low-grade tumours. Generally, 93% of the patients with tumours p10 mm were treated by CS (84.5% with RT) and 63.2% of patients with tumours X20 mm underwent mastectomy. Mastectomy rates were 43% for comedocarcinoma and 28% for other lesions (Po0.0001). Conservative surgery was used to treat 1.5% of comedocarcinoma and 9.5% of other lesion subtypes (Po0.0001). The multiple logistic regression analysis confirmed that the factors significantly influencing the mastectomy choice were tumour size (OR: 1.088), young age (OR: 0.975) and high or intermediate grade (Table 4).
Of 896 women who were treated by CS, 797 (89%) received adjuvant RT. The median dose administered to the whole breast was 50 Gy. A 10 Gy boost (median dose) was given to 49% of patients (48% if excision was complete and 63.1% in case of incomplete or doubtful excision (P ¼ 0.018).
Tumour size, nuclear grade, lesion subtype (comedo vs others) and treatment strongly influenced axilla treatment (Table 5). Generally, 21.3% of patients underwent SNB and 10.4% AD. The multiple logistic regression analysis showed that sentinel biopsy procedure rates were higher in case of mastectomy, comedocarcinoma subtypes, large tumour size and high-grade lesions ( Table 6). The AD use was significantly and only correlated to the mastectomy option (OR: 5.83; IC 95%: 3.68 -9.24, Po0.001).

Regional variations in practice
There are some variations among regions in histopathological diagnosis (e.g., percentage of high-grade lesions, comedocarcinoma subtype and ER assessment) as well as in treatments ( Table 7). The mastectomy rate ranged from 35 to 50% for highgrade DCIS and from 38.5 to 50% for comedocarcinoma. It varied from 20 to 37% according to the region. There were also wide regional variations in the delivery of adjuvant RT (81 -96%) and in the administration of adjuvant hormone therapy (6 -34%). Mastectomy rates also varied according to the type of institution (37.4, 32 and 19.4% in cancer centres, academic or general hospitals and private clinics, respectively, Po0.0001). The corresponding CS þ RT rates were 87.3, 81.3 and 94.2%.

DISCUSSION
Our survey was conducted to obtain comprehensive and unselected up-to-date data on diagnosis, assessment and treatment of DCIS in France. The 77 physicians who included patients are breast cancer specialists working in different types of institutions (cancer centres, clinics and hospitals), reflecting the current daily practice in our country. However, we may have a small overrepresentation in our study of large centres.
Since 2000, regional breast cancer screening programmes in France have targeted women aged from 50 to 74 years. Ductal carcinoma in situ was diagnosed by mammography in 87.6% of our patients, but only in 61.2% of patients under the age of 40 years. This 87.6% rate of mammographically detected DCIS is slightly higher than the mean 81.5% rate of infraclinical lesions found in a former review of 909 patients treated for DCIS in Bordeaux) (Barreau et al, 2005). French (Cutuli et al, 2005a, b) and European (Rutgers, 2001) guidelines also recommend preoperative diagnosis. Our study confirmed that preoperative biopsy significantly decreased the number of breast surgical procedures as well as AD and should be performed routinely (Table 1).
Our mastectomy rate is identical to that reported by two recent American and English studies (30.5%) (Katz et al, 2005;Dodwell et al, 2007). In the United States, the overall mastectomy rate decreased from 43% in 1992 to 37% in 1995 and 28% in 1999 Baxter et al, 2004;Mokbel and Cutuli, 2006). However, this rate was much higher (58%) in a large series from the Netherlands Cancer Institute (Meijnen et al, 2008). Tumour     (Fisher et al, 2001;Solin et al, 2001;Cutuli et al, 2002;Fourquet et al, 2002;Vicini and Recht, 2002;Bijker et al, 2006). These data probably explain our 50% mastectomy rates in women under the age of 40 years. In theory, there is no axillary lymph node involvement in DCIS, although a recent review has reported 1.4% involvement in 1621 patients with pure DCIS (Morrow et al, 2002;Leonard and Swain, 2004). There may, however, be a risk of associated microinvasive foci, especially when the lesion is extensive (McMasters et al, 2002;Moran et al, 2005;Tunon de Lara et al, 2008). This may be a reason why, even though axilla management is pointless, 10.4% of our patients underwent AD and 21.3% underwent SNB. The potential risk of axillary lymph node involvement was probably overestimated by several physicians. The AD practice was highly influenced by tumour size and nuclear grade (Table 5), but especially by the breast surgery type; indeed, the AD rate was 5% after CS vs 22.6% after mastectomy (Po0.001). The global AD rate recorded in an earlier study by French Cancer Centres, including 1223 patients treated from 1985 to 1996 was 51.3% (Cutuli et al, 2005a, b). Such a steep decrease (from 51.3 to 10.4% in France) has already been observed in the United States, where the global AD rate decreased from 34% in 1992 to 15% in 1999 (Po0.001), but nevertheless still remained high (30%) in 1999 in mastectomy patients (Baxter et al, 2004). In another US study , AD rate ranged from 18 to 28% across different sites. At present, AD is useless and should be replaced by SNB. In our study, SNB rate was 12.3% after CS vs 41.7% after mastectomy (Po0.001). Finally, a recent French multicentric study (Tunon de Lara et al, 2008) including 116 large DCIS found 4 positive sentinel nodes (3.4%). Thus, SNB could be proposed in patients treated by mastectomy.
For most authors, complete tumour excision is the key factor in reducing LR after CS (Silverstein et al, 1999;Kell and Morrow, 2005;MacDonald et al, 2005), but can be difficult to achieve. However, there is no clear consensus on the definition of 'complete' excision and adequate margins (Schwartz et al, 2000;Kell and Morrow, 2005). Tumour size and excision quality are factors often omitted in the reports of randomized controlled trials (Bijker et al, 2001) and retrospective studies (Solin et al, 2001). In our recent national practice guidelines (Cutuli et al, 2005a, b), we have defined complete excision as a resection with at least 3 mm circumferential margins. Approximately 76% of our patients underwent complete tumour excision according to this criterion. However, some teams consider a 2 mm margin to be adequate if the patient subsequently receives RT (Kell and Morrow, 2005).
In patients with tumour excision who did not receive RT, the 10-year LR rates ranged from 19 to 27.8% (Schwartz, 2002). In four randomized controlled trials, the LR rates after CS without RT were 32, 28, 26 and 22%, with follow-ups ranging from 52 to 120 months (Fisher et al, 2001;Hougton et al, 2003;Bijker et al, 2006;Holmberg et al, 2008). In a French study, the 7-year LR rate was 32.4% in patients not receiving RT (Cutuli et al, 2002). In two latest Dutch studies (Schouten Van Der Velden et al, 2007;Meijnen et al, 2008), the 8-and 10-year LR   after CS alone were 15.6 and 25%, respectively. Whole breast irradiation at 50 Gy after CS for DCIS halved in situ and invasive recurrence rates in earlier quoted trials (Fisher et al, 2001;Hougton et al, 2003;Bijker et al, 2006;Emdin et al, 2006;Holmberg et al, 2008) and several large retrospective studies (Solin et al, 2001;Cutuli et al, 2002;Fourquet et al, 2002;Schouten Van Der Velden et al, 2007;Meijnen et al, 2008). On the other hand, RT administration after wide excision (margins 41 cm) of small low-grade lesions is questioned (Cornfield et al, 2004;MacDonald et al, 2005;Solin et al, 2006;Buchholz et al, 2007, Silverstein et al, 2007. However, this result should be weighed up against the 12% rate observed at 5 years in a series of low-risk patients (Wong et al, 2006). On the other hand, the Van Nuys Prognostic Index (VNPI), combining tumour size, grade and margin width, was used to select 'low-risk' patients not requiring RT (Silverstein et al, 1996), but was widely criticised in two large American and English studies including 222 and 237 patients, respectively (Boland et al, 2003;MacAusland et al, 2007). A recent Italian study (Di Saverio et al, 2008) including 259 patients also used VNPI and concluded in the absence of statistically significant advantage in the CS þ RT vs the CS group. However, the CS þ RT group had several unfavourable factors widely increasing LR risk, such as lesion size 415 mm (67 vs 30%) and excision margins o1 mm (29 vs 4%). These features clearly explain the similarity of LR rates with an 'apparent' inefficiency of RT (7.5% for CS and 9.5% for CS þ RT). Despite this, other English studies still use VNPI in current practice (Gilleard et al, 2008). Our survey showed that 89% of patients receive adjuvant RT after CS. A recent study on 1140 screen-detected DCIS in the United Kingdom reported a 57% 'planned RT' after CS. The RT use was influenced by tumour size (less or more than 15 mm), the presence of comedonecrosis and, more particularly, nuclear grade, but paradoxically not by margins (Dodwell et al, 2007). Such differences may show the high confidence of France (and of many centres in the United States and partly the United Kingdom) in randomised trials and large unselected retrospective studies on RT after CS for pure DCIS. Besides, there is clear evidence that boost increases local control in invasive breast cancer, especially in women under the age of 50 years (Bartelink et al, 2001). A boost was given to 49% of our RT patients compared with 50-75% of patients in other studies (Cutuli et al, 2002;Fourquet et al, 2002;Solin et al, 2005). A recent study (Omlin et al, 2006) on 375 young women (p45 years) treated for pure DCIS showed 28 and 14% LR rates with and without a 10 Gy boost. This 50% LR reduction rate is identical to that in invasive BC. However, there are several limitations in this study, that is long recruitment period, lack of specification on margin status and types of surgery. The NSABP-24 trial (Fisher et al, 1999(Fisher et al, , 2001 found a significant reduction in ipsilateral and contralateral tumours on combining tamoxifen with CS þ RT, which was however not confirmed by the results of a randomized trial from the United Kingdom, Australia and New Zealand (Hougton et al, 2003). Tamoxifen was effective in patients with ER-positive tumours only (37% of the patients in the NSABP trial) but not in patients with negative margins, absence of necrosis and in women over 50 years (Allred et al, 2002;Fisher et al, 2002). It has some serious side effects (endometrial cancer and thromboembolic events) and various 'minor' others. In a study of 94 patients with DCIS, 20 (21%) discontinued tamoxifen because of side effects or complications (mainly hot flushes, fatigue, weight gain and gynaecological or digestive symptoms) (Yen et al, 2004). The tumours of 31% of our patients were assayed for ER and progesterone receptors. The positivity rate was 70% for ER and 60% for progesterone receptors, in line with the 64 and 57% rates, respectively, recorded in a review of 1920 patients (Baxter et al, 2004).
There were important regional differences in histopathological features as well as treatment in our study population (Table 7). The mastectomy rate ranged from 20 to 37% and the use of RT after CS from 82 to 96%. Tamoxifen use varied widely from 6% in the northwest to 34% in the northeast of France. On the other hand, the size of DCIS 'selected' for CS alone ranged from 2.5 to 8 mm. It is also important to point out that wide histopathological variations (partly unexplained) can induce various therapeutic procedures.
Physician discretion and habitual practice rather than actual differences in DCIS probably explain these variations. Indeed, in the above-mentioned UK study (Dodwell et al, 2007), among 69 breast screening units included in the study, wide variations in the use of adjuvant RT were observed, ranging from 0% in six units to 100% in two (median: 57.7%). The authors explain that such variations in the use of RT reflect the differences in the perception of the risk-benefit relationship in this treatment, and also probably some difficulties to have access to RT facilities (Dodwell and Crellin, 2006). Wide variations have also been recorded in the United States (Ceilley et al, 2004;Katz et al, 2005), especially in the mastectomy rates (range: 26-45%) and the use of RT after CS (range: 39-74%), and have been confirmed in a recent analysis of the SEER data . In a retrospective study on 727 patients with DCIS enrolled in the Ontario Breast Screening Program from 1991 to 2000 (Rakovitch et al, 2007), there were also significant regional variations in RT use after BCS, ranging from 43 to 71% regardless of tumour pathological features, surgeon and hospital volume. On the other hand, some physicians are afraid by possible RT side effects. However, with modern techniques, the 'simple' whole breast irradiation (with or without boost, but without nodal irradiation) delivering 50 Gy in 25 fractions (or equivalent dose) resulted in o1% of complications, whereas in case of invasive LR, 15-20% of the patients finally died by metastatic evolution (Mokbel and Cutuli, 2006;Silverstein et al, 1998). Moreover, in the study reporting the SEER results (Warren et al, 2005), LR rates after CS and CS þ RT were 15 and 10.7%, respectively, with a significant increase in breast cancer-related deaths when RT was omitted: 2.7 vs 0.8% (P ¼ 0.02).
In conclusion, randomised controlled trials can provide some answers, even if they are not exempt from criticism (Bijker et al, 2002). Several selection biases have been noted in DCIS trials such as inclusion of benign or microinvasive lesions and lack of compliance with study criteria (lesion size/surgical margins, quality of excision and treatment modalities) (Mokbel and Cutuli, 2006). However, there are now very consistent data confirming the importance of RT to minimise LR risk rates after CS for most DCIS, such as shown in a large study including 23 547 women from the California Cancer Registry (Innos and Horn-Ross, 2008). The future trials should identify more clearly selected patients (i.e., older than 60 years, with low-grade lesions under 10 mm and clear margin over 10 mm), in which RT can be safely omitted, and also those with aggressive DCIS requiring mastectomy.