Breast cancer survival has been reported as higher in the US than in Europe^1,2. For women diagnosed 1985–89, five-year survival was higher in each of the nine Surveillance, Epidemiology and End Results (SEER) areas than in any of the 22 European countries participating in the EUROCARE-2 study.

The first worldwide analysis of cancer survival (CONCORD³) provided a systematic comparison of survival for adults (15–99 years) diagnosed with a cancer of the breast, colon, rectum or prostate in one of 31 countries during 1990–94 and followed up to 1999. International differences in age-standardised survival were very wide, even after adjustment for differences in mortality from other causes of death. Breast cancer survival in the US and Canada was higher than in other countries, but differences between the US and most European regions were smaller than for women diagnosed during 1985–89². The largest differences were between the US and Eastern Europe.

A population-based comparison of five-year breast cancer survival among women diagnosed in 17 territories in 6 European countries during 1990–92 and in the 9 states and metropolitan areas of the US covered by the SEER programme in 1990 showed that differences were mainly attributable to stage at diagnosis and the diagnostic procedures used to determine the stage⁴.

Both the assiduity of investigation and the appropriateness of treatment by stage varied widely for women diagnosed in Europe during 1990–92 and 1996–98^5,6. Primary treatment for breast cancer also varies greatly throughout the US^7,8. Following the NIH Consensus Development Conference in 1990, which recommended breast-conserving surgery and radiotherapy instead of mastectomy for women with stage I and II breast cancer, the proportion treated with breast-conserving surgery increased steadily up to 1995⁹, but the percentage who also received radiotherapy and/or axillary lymphadenectomy declined¹⁰. Differences in protocol and calendar period make it difficult to draw firm conclusions from these studies about whether the differences in survival between Europe and the US are attributable to differences in stage, or treatment, or both.

The CONCORD protocol incorporated “high-resolution” studies designed to explain the international variations in survival for breast, colorectal and prostate cancer. The analyses involve large random samples of patients, with detailed clinical and pathological data that are not routinely abstracted by population-based cancer registries. The study reported here provides a trans-Atlantic comparison of stage, treatment and survival for women with breast cancer. The aims were to compare the stage distributions in Europe and the US; to determine whether the transatlantic differences in 5-year survival persist and, if so, to assess the extent to which they are attributable to differences in stage. We also set out to compare adherence to “standard care” for breast cancer in relation to age, stage and hormone receptor status, before widespread introduction of clinical guidelines.

Data on stage, diagnostic procedures, treatment and follow-up were collected for a representative sample of about 20,000 women aged 15–99 years who were registered with a diagnosis of breast cancer in the US or one of 12 European countries during 1996–98. A common protocol was used, based on the EUROCARE high-resolution protocols^5,6.

The European data were provided by 26 population-based cancer registries in 12 countries, 7 of which with national coverage, denoted by an asterisk (*). For some analyses, the data were grouped into four European regions defined by the United Nations (UN, http://unstats.un.org/unsd/methods/m49/m49regin.htm) - Northern Europe: Denmark*, Finland*, Iceland*, Sweden*; Western Europe: France (Bas-Rhin, Côte d’Or, Doubs, Isère, Tarn) and the Netherlands (Eindhoven, North East Netherlands); Southern Europe: Italy (Firenze, Genova, Modena, Palermo, Ragusa, Varese), Slovenia*, Spain (Basque Country, Castellon, Granada, Navarra); Eastern Europe: Estonia*, Poland (Cracow, Warsaw), Slovakia*. Estonia is classified by the UN as being in Northern Europe, but cancer survival has usually resembled that in Eastern European countries¹¹, and the data from Estonia are included here with Eastern Europe. Data from the US were provided by 7 state-wide registries: California, Colorado, Illinois, Louisiana, New York, Rhode Island and South Carolina. The US registries are part of the National Program of Cancer Registries, based at the Centers for Disease Control and Prevention.

For this study the cancer registries included in the EUROCARE-3 high-resolution study⁶ made special efforts to update the follow-up to at least five years after diagnosis for all patients. The North East Netherlands registry was not included in EUROCARE-3, but it is unusual in that it routinely collects almost all the data required for high-resolution studies on all registered cancer patients, so it was able to provide such data on virtually all women with breast cancer, not just a sample.

Most European registries provided a random sample of at least 500 women diagnosed during 1996–98, as specified by the protocol. Denmark and Sweden provided a sample of women diagnosed in 1994, and Palermo (Italy) provided data for all women diagnosed in 1999, the first year for which data were available there. The Finnish cases were a population-based sample of women diagnosed in the Tampere hospital region, which is considered representative of the whole of Finland. Despite these slight departures from protocol, these cases were retained to ensure the widest possible geographic coverage. Each of the US registries provided a random sample of at least 500 women diagnosed with breast cancer in 1997.

Anonymised, individual cancer registration records were supplied for 20,150 women diagnosed with a malignant neoplasm of the breast. In situ tumours (1,168, 5.8%) were excluded from the analyses because they were collected systematically in the US, but not in Europe. A further 20 cases (0.1%) were excluded because they did not meet the protocol (2 with benign or uncertain behaviour, 2 with the morphology of leukaemia or lymphoma, and 16 aged less than 15 or more than 100 years). In all, 18,962 women with a primary, invasive, malignant neoplasm of the breast were included in the analysis of stage and treatment. Women whose cancer was only registered from a death certificate (23; 0.1%), or of unknown vital status (18; 0.1%), or for whom the date of last known vital status preceded the date of diagnosis (32; 0.2%), were excluded from the survival analyses, which thus involved 18,889 women.

Information on stage, diagnostic examinations, treatment and follow-up was obtained by direct examination of the clinical record. Where records were incomplete, pathology reports, hospital discharge records and other sources were consulted as necessary.

Disease stage was defined according to the 4^th edition of the TNM (Tumour, Nodes, Metastasis) manual.¹² If pathological data on tumour size and lymph node status (pT and pN) were unavailable, clinical data (cT and cN) were used. Following advice from epidemiologists, pathologists and clinicians, records for which the metastatic status was unknown (MX) were considered as negative (M0), if T and N were known. Patients were grouped into six categories: early, node-negative disease (T1N0M0), larger node-negative (T2-3N0M0), node-positive (T1-3N+M0), locally advanced (T4, any N, M0), metastatic (M1) and unstaged. Within the category of early, node-negative disease, we also assessed the distribution of small tumours by size: less than 5mm (T1a), 5–10mm (T1b) and over 10mm and up to 20mm (T1c). Estrogen-receptor (ER) status was categorized as positive, negative or unknown. Age at diagnosis was categorised into four groups (15–39, 40–49, 50–69, 70–99 years) for survival analysis. Treatment comparisons were made in wider age groups: 15–49 and 50–99 years for chemotherapy and hormone therapy; 15–69 and 70–99 years for breast-conserving surgery plus radiotherapy.

Data on surgical procedures were collected in 7 categories: conservative surgery (including quadrantectomy, tumour excision, lumpectomy), simple mastectomy, any modified radical mastectomy, extended radical (Halsted) mastectomy, surgery (not otherwise specified), unknown if surgery was performed, and no surgery. When a surgical procedure was performed, axillary procedures were collected in 5 categories: for lymph-node sampling, for axillary clearance, unspecified whether for sampling or clearance, not specified if done or not, and not done. Information was also sought on sentinel lymph-node biopsy, with or without lymphadenectomy, but sentinel biopsy was very uncommon during 1996–98. Information on biopsy or needle aspiration of the breast was coded in 5 categories as either done; not done because of refusal or death, or for specified medical contraindications, or for other or unspecified reasons; or unknown if done or not. Chemotherapy, radiotherapy and hormonal therapy were coded as yes, no or unknown.

Primary treatment for early node-negative disease was dichotomised as breast-conserving surgery with radiotherapy (BCS+RT) vs. all other surgical procedures, whether or not followed by radiotherapy. Chemotherapy and endocrine treatment were dichotomised as administered vs. not administered or unknown.

We included 18,962 women with invasive primary breast cancer: 15,842 women in 26 jurisdictions in 12 European countries and 3,120 women in 7 US states (Table 1). Microscopic verification was available for 98–99% of the women in each of the US states and 94% in Europe, ranging from 79% in Estonia to 100% in the Basque country (Spain). Data were available on stage for about 90% of cases in both data sets, ranging from 78% (Ragusa, Italy) to 95% or more in 8 of the 26 European registries and from 81% (New York) to 94% (Colorado and S Carolina) in the US.

Breast cancers were generally less advanced in the US than in Europe, and the stage distribution varied less between US states than between European jurisdictions. Early node-negative tumours were more frequent in the US (39%, range 33–45%) than in Europe (32%, 16–49%). Large node-negative tumours were of similar frequency (Europe 14%, 9–22%; US 14%, 10–18%), while node-positive tumours were more common in Europe (33%, 25–44%) than the US (26%, 22–29%). Locally advanced tumours were twice as frequent in Europe (8%, 0–24%) as in the US (4%, 3–6%), but the overall frequency of metastatic tumours was similar (5–6%). The proportion of tumours with unspecified stage was slightly higher in the US (11%) than Europe (8%), but up to 18–22% in three European registries (Finland; Italy: Firenze, Ragusa), while only New York (19%) differed much from the US average. Exclusion of these registries did not substantially alter the overall stage distributions in Europe or the US (data not shown).

Lymphadenectomy was reported for 13,687 (86%) women in Europe and 2,531 (81%) in the US, but it was generally more extensive in the US, where 10 or more nodes were examined in 78% (range 76–83%) of procedures, compared with 66% (23–93%) in Europe (Table 2). Among women with early node-negative tumours, the distribution of tumour size was more favourable in the US than in Europe.

More than 90% of women received surgical treatment: 91% in Europe (from 77% in Estonia to 95% or more in 10 of 26 jurisdictions) and 96% in the US (93–97%; Table 3). Among operated women, 35% had early node-negative disease in Europe, compared with 41% in the US. Among women operated for early node-negative disease, breast-conserving surgery plus radiotherapy was received by 55% in Europe and 49% in the US, but the variability was much wider in Europe (9% in Estonia; 78–84% in four of the five French regions) than in the US (34% in S Carolina; 58% in Rhode Island). The proportion of women aged 70–99 years who received breast-conserving surgery and radiotherapy for early node-negative disease varied more between European countries and regions (4–6% in two Polish regions; 84% in Tarn) than between US states (21% in Louisiana; 47–48% in Rhode Island and California).

For early node-negative disease, and relative to Southern Europe (1,848 women, reference category), the odds of receiving both breast-conserving surgery and radiotherapy (vs. any other surgical procedure, with or without radiotherapy), adjusted for age and tumour size, were lower in the US (OR=0.80; 95%CI 0.69–0.94) and Northern Europe (OR=0.60; 0.50–0.72); much lower in Eastern Europe (OR=0.16; 0.12–0.20), and higher in Western Europe (OR=1.57; 1.36–1.81) (Table 4). The odds of receiving this treatment were significantly lower for women aged 70–99 years than for those aged 60–69 years (OR=0.48; 0.41–0.56), after adjustment for region and tumour size. Women with tumours of 5–10mm (T1b) received this treatment more than women with larger tumours (up to 20mm, T1c) (OR=1.31; 1.16–1.48).

Among women with node-positive tumours, 58% received chemotherapy in the 26 European jurisdictions, compared with 69% in the 7 US states (Table 5). Among women aged less than 50 at diagnosis, the overall proportion was similar in Europe and the US (90%), but the range was wider in Europe (54–100%) than the US (84–94%). Among older women, the proportion who received chemotherapy was higher, and varied less, in the US (60%, range 53–67%) than in Europe (46%, range 14–75%).

Overall, endocrine treatment in ER-positive tumours was slightly higher in the US (62%) than in Europe (55%). The proportion was similar in women aged 50 and over (63% in the US; 59% in Europe), but younger women received tamoxifen more often in the US (58% vs. 44%).

Overall, age-standardised net survival at five years was 81% in Europe and 84% in the US (Figure 1). Survival in Northern, Western and Southern Europe (81–84%) was similar to that in the US (84%), but it was lower in Eastern Europe (69%). Survival varied more widely between European jurisdictions (88% in Iceland to 62% in Estonia) than between US states (from 91% in Colorado to 76% in South Carolina).

Five-year age-standardised net survival was broadly similar in all European regions and the US for early, node-negative tumours (96–98%) and for large, node-negative tumours (85–90%) (Figure 2, available in web-appendix). The geographic range in survival was much wider for locally advanced disease, from 37% in Eastern Europe to 77% in Northern Europe, and 44% in the US. As with overall survival, stage-specific five-year survival was similar in Northern, Western and Southern Europe and the US. In Eastern Europe, survival for node-positive, locally advanced and metastatic tumours was lower than in other European regions or the US.

The mean excess hazard was higher in Eastern Europe than in other jurisdictions at 1 month, 6 months and 1, 3 and 5 years after diagnosis, both for all ages and in each of 5 age categories (Figure 3, available in web-appendix). The difference was most marked for women aged 70–99 years. No striking differences were found between Northern, Western, Southern Europe and the US. The high excess hazard of death in Eastern Europe was mainly confined to women with locally advanced or metastatic tumours (Figure 4, available in web-appendix).

Transatlantic differences in cancer survival have raised questions about early diagnosis and the adequacy of investigation and treatment. To our knowledge, this is the first population-based high-resolution study to use clinical data that were collected by trained abstractors from the primary medical records under a common protocol, subjected to standard quality control procedures and analysed centrally with the same statistical methods. We compared survival using clinical data on stage, diagnostic procedures and treatment. The survival differences appear likely to be related to differences in diagnosis and patterns of care shortly after diagnosis. The women were diagnosed more than 10 years ago, but most diagnostic and therapeutic approaches used at that time remain in widespread use: understanding their role in international differences in survival remains relevant.

Overall, five-year net survival was not very different in Europe (81%) and the US (84%). Differences were mainly confined to the three Eastern European countries, Estonia, Poland and Slovakia, where average five-year survival was 69%. Estonia and Slovakia are both covered by national cancer registries, and the women from those countries were thus nationally representative. Survival varied more widely between the 26 European jurisdictions than between the 7 US states.

The differences in survival between Europe and the US in the late 1990s are smaller than for women diagnosed at the beginning of the decade^3,4. In the previous high-resolution study⁴, the US data were taken from the SEER public-use data set²¹ and harmonised to the extent possible with the data collected under the EUROCARE-2 high-resolution protocol. By contrast, the data for this study were collected directly from the clinical records using a standard protocol; European coverage rose from 17 to 26 registries (11 contributed to both studies), and US coverage changed from the 5 metropolitan areas and 4 states covered by the SEER program to 7 state-wide registries in the National Program of Cancer Registries (NPCR). Survival in the 1990s was lower in the NPCR territories than the SEER areas^3,22. Finally, in the previous high-resolution study, differences in background mortality in the US were controlled with a single national life table for 1990, weighted for the proportion of Blacks, Whites and other races, whereas we were able to use state-specific life tables for each calendar year 1994–2004.

The modelling approach used to estimate net survival is a strength of this study, but it does not explain the smaller transatlantic differences than those obtained with relative survival in the previous study. We found similar patterns with all the other widely used methods for survival estimation (data not shown).

The European differences in survival were generally similar to those reported for the same countries among women diagnosed 1995–99¹¹. Survival was higher than expected in Denmark (84%): the data in this study are from eastern Denmark, greater Copenhagen and Copenhagen (Zeeland), where most of the population has undergone mammographic screening since 1991²³. In these areas, survival after mammographic diagnosis is higher than in Denmark as a whole, regardless of whether it was a screening mammography. Survival in Slovenia was lower than in other Southern European countries, and more similar to that in Eastern Europe. Variation in survival between the 7 US states was less marked than in Europe, mostly in the range 81–87%, but ranging from 91% in Colorado to 76% in South Carolina, where Blacks represent approximately 30% of the population (http://www.ipspr.sc.edu/publication/Older%20SC.pdf).

The availability of information about race in this data set would have strengthened the international survival comparisons, but information about race is not available in many European countries. Race in the US and geographical area in Europe are often considered as a proxy for socio-economic status. In future studies, it would be preferable to use life tables that are specific for race and/or socio-economic status.

Stage-specific net survival was similar in most European jurisdictions and US states. In Eastern Europe, survival from node-positive, large and metastatic tumours (N+; T4; M1) was lower than in other European regions or the US, and the proportion of metastatic tumours was also high, mainly in Estonia and Slovakia.

The mean excess hazard of death by time since diagnosis was similar in Europe and the US for women with early node-negative disease, large node-negative disease or node-positive disease, and up to five years after diagnosis. The hazard was somewhat higher in Eastern Europe for locally advanced disease, and much higher for metastatic disease, especially in the first three years after diagnosis. Adjustment for the number of examined lymph nodes, necessarily restricted to women who underwent lymphadenectomy (86%), did not modify this pattern, either overall, or within each category of stage. In other words, the geographic pattern in the mean excess hazard of death was not affected by the number of nodes examined during lymphadenectomy (data not shown). This suggests that, in contrast with the findings from the study of women diagnosed in 1990⁴, stage migration does not affect the comparison of stage-specific survival between European regions and the US. This could be because the recording of stage has become more homogeneous, or because the quality and completeness of diagnostic investigation is less variable now than previously.

The mean excess hazard of death for women with late-stage disease was very high in Eastern Europe. This suggests that fewer effective treatment options were available for these women, although higher levels of co-morbidity may have restricted therapeutic options. Hormonal treatment and adjuvant chemotherapy were used more extensively in Slovakia, Estonia and Poland than in other European countries, and not just for node-positive and estrogen-receptor positive disease. Mastectomy was often used instead of breast-conserving surgery and radiotherapy, in part because radiotherapy facilities were not always available. Total national expenditure on health was low, and this is also likely to have affected the quality of treatment⁶.

Data on stage were remarkably complete, because they were collected directly from clinical records. Complete data on stage and lymph nodes were unavailable for all but 5–11% of women in the 5 broad European regions, although for up to 20% in 3 of the 26 European registries. However, exclusion of women with unknown stage or lymph node status did not change the geographic pattern of the excess hazard of death within any of the categories for which stage was known. More complex analyses after imputation of missing values are unlikely to change this picture.

Pattern of care studies and survival have been conducted separately in Europe^5,6 and the US²⁴. Here, we could make a direct comparison between Europe and the US with data on stage at diagnosis and treatment collected and coded with the same rules.

Overall, women received breast-conserving surgery and radiotherapy for early node-negative breast cancer somewhat more often in Europe (55%) than the US (49%), but the distribution by age was similar. The lower proportion in the US is mainly determined by Louisiana (37%) and South Carolina (34%) and may be explained by the attitude of some US clinicians during the late 1990s, when radiotherapy may have been considered unnecessary after breast-conserving surgery¹⁰. Another explanation may be the paucity of radiotherapy centres and/or the distance of the nearest radiotherapy facility²⁵ in these two states.

After adjusting for age and tumour size within the category of early node-negative disease, the odds of being treated with breast-conserving surgery and radiotherapy were almost 60% higher in Western Europe than Southern Europe (reference), 20–40% lower in the US and Northern Europe, and more than 80% lower in Eastern Europe.

In Denmark, the low level of breast-conserving surgery and radiotherapy was probably related to the fact that most breast cancers were treated in local or regional hospitals (not specialist centres), rather than any lack of radiotherapy facilities, although the Danish national cancer plan of 2000 recognised the need to modernize and expand radiotherapy services. Most women receiving breast-conserving surgery also received radiotherapy, but breast-conserving surgery was hardly ever done in areas where breast cancer screening was not performed²⁶.

About 90% of women aged less than 50 years with node-positive disease received chemotherapy in both Europe and the US, in accordance with contemporary clinical protocols²⁷.

The proportion of women aged 50–99 years with positive lymph nodes who received chemotherapy was notably higher in the US (60% vs. 46%). The proportion was similar in all 7 states, but slightly lower in Illinois and Rhode Island. The finding of more active treatment for older women in the US echoes the finding for women diagnosed in 1990, and may indicate the importance of health insurance programs such as MEDICARE. The US National Institutes of Health had also recommended chemotherapy for node-positive breast cancer in 1985²⁸.

In the late 1990s, tamoxifen was recommended for estrogen-receptor positive tumours on both sides of the Atlantic^29,30, especially for women aged over 50 years. In the US, the proportion of women aged less than 50 years with ER+ tumours treated with hormonal therapy was 58%, higher than in Europe (44%).

The low proportion of women with early stage disease who receive breast-conserving surgery is correlated with total national expenditure on health⁶. The wider use of chemotherapy and hormonal treatment may reflect the fact that costs are lower than for surgery and radiotherapy. Taken with the findings of this study, this suggests that low healthcare expenditure in Eastern European countries may have had an important effect on the quality of breast cancer treatment, and on survival.

Differences in breast cancer survival between Europe and the US in the late 1990s were mainly explained by lower survival in Eastern Europe, where low healthcare expenditure may have constrained the quality of breast cancer treatment. Similarly wide variation has also been reported within the US, where non-Hispanic Black women were less likely to receive guideline-concordant treatment than non-Hispanic White women⁸.

The need for population-based data on stage and treatment is recognised by clinicians and epidemiologists. High-resolution studies still seem to be the only valid way to collect this information. More funding should be directed to help cancer registries obtain timely high-resolution data for all registered patients.