Outcomes of Breast Cancer in Brazil Related to Health Care Coverage: A Retrospective Cohort Study

Over 52,000 women in Brazil are diagnosed annually with breast cancer and mortality has risen steadily (1, 2). These trends parallel the aging of the population and the growth of noncommunicable diseases in general (3, 4). Currently, there are few data on the clinical characteristics of patients with breast cancer and their clinical outcomes (5). It has been estimated that screening for breast cancer through the public health system of Brazil is low, with annual rates of less than 10% of the target population (5). Although national data on screening through the private sector are not available, one study has estimated that in the Brazilian State of Goiás, 48% of all exams were done through the private sector in the year of 2008 (6). The 5-year breast cancer survival rates in Brazil have been estimated to be 58%, which are low when compared with the United States (84%; ref. 5). The Brazilian breast cancer mortality-to-incidence ratio is also higher when compared with the United States (0.291 vs. 0.193, respectively; ref. 5).

The Amazona Project is a retrospective observational cohort study of 4,912 women with breast cancer from centers representing all different geographical regions of Brazil (7). The project was planned and conducted by The Brazilian Breast Cancer Study Group (GBECAM) and analyzed in collaboration with the MGH-AVON International Breast Cancer Program. The objective of this study was to assess differences in outcomes according to type of healthcare coverage, public or private. Our secondary objective was to determine whether there are differences in access to specific cancer therapies between these two groups.

Data were collected from 28 centers in 11 states through a systematic chart review of women diagnosed with breast cancer between July 2008 and January 2009. Patients were included if their medical chart was accessible, they were 18 years of age or more, had pathologic confirmation of breast cancer, and were registered at the participating centers between January 1st and December 31st of 2001 or January 1st and December 31st of 2006. Information collected included: the patient's age, race, menstrual and gestational history, family history of cancer, date of diagnosis, clinical and pathologic characteristics of breast cancer, institution where the patient received care, date and type of primary local and systemic treatments received, date of recurrences, type of recurrences, date and type of first-line treatment received for recurrences, survival, and date and cause of death. Ethnicity was categorized according to the Brazilian National Institute of Geography and Statistics (4). Patients registered in 2001 were staged clinically according to AJCC/UICC TNM 5th edition, whereas those from 2006 were staged with AJCC/UICC TNM 6th edition (8, 9). For those patients who did not have a clinical stage determined but had available information on tumor, lymph nodes, and metastasis, clinical stage was determined according to AJCC/UICC TNM 6th edition. Participating centers were of three types: public, which attend almost exclusively to public healthcare patients; private, which attend to private healthcare patients; and philanthropic, which have a mixed population of public and private healthcare patients. For the purpose of this analysis, patients from public and private centers were considered to have public and private health coverage, respectively. Individual data on type of health insurance was retrieved for patients from philanthropic institutions. Initial chart review included 4,912 patients, 4,905 of which had valid survival data. Seven hundred and twenty-five patients were originally from private and 1,068 patients from public institutions. Of the 3,112 patients from philanthropic institutions, we were able to categorize by further chart review 262 as privately insured and 1,087 as publicly insured, for a total of 3,142 patients included in this analysis. For hormone receptor data, we considered positive those patients that had positivity stated in records or that received hormonal therapy; patients who had neither were considered as hormone receptor negative. Ethics committees of all participating centers approved the study and the study was conducted according to the Declaration of Helsinki.

Comparisons of demographics, disease characteristics, and treatment regimens between those with public and private coverage were analyzed using the χ² test. Overall survival (OS) was calculated from the time of diagnosis until death from any cause. Disease-free survival (DFS) was calculated as the time from initial diagnosis to the earlier of either breast cancer recurrence or death. Date of breast cancer recurrence was available for stage 0–III patients, but only date of death was available for stage IV patients. Postrelapse survival (PRS) was calculated from the time of first recurrence of breast cancer to death. OS, DFS, and PRS plots according to type of health coverage and stage at diagnosis were calculated using Kaplan–Meier methodology. Comparisons were made using the log-rank test. Effect of insurance type on DFS and OS were analyzed using Cox regression models with adjustment for patient characteristics (age and race) and stage at presentation (tumor size and nodal status). For PRS, the Cox regression model was adjusted for age, race, type of relapse (locoregional vs. distant), and type of health coverage. All P values are two sided and considered significant when P < 0.05. All analyses were performed using SAS version 9.2.

Funding sources had no role in study design, data collection, analysis, and interpretation, writing of the report, and decision to submit to publication.

Of the 3,142 patients included in this analysis, 2,155 (69%) had public healthcare coverage and 987 (31%) had private coverage (Table 1). At a median follow-up time of 28 months from diagnosis, 439 patients had progressed and 141 had died. Of these deaths, 128 were attributed to breast cancer, 1 to treatment toxicity, 9 to nondisease-related causes but with evidence of breast cancer, and 3 due to nondisease related causes without evidence of disease.

Table 1 provides the characteristics of all subjects shown by type of healthcare coverage. Median age at diagnosis was 53 years (range 15–94) for all patients and not significantly different between the groups (P = 0.09). The majority of women were postmenopausal (62%). Overall, the majority of subjects were Caucasian (60%). Ethnicity was significantly different according to the type of healthcare coverage (P < 0.001), with public healthcare having more patients of mixed ethnicity and those with private coverage more frequently being Caucasians; however, there were a substantial number of subjects with unknown ethnicity in both groups. Ninety-three percent of patients had invasive breast cancer and 83% of these were ductal. Clinical stage at diagnosis was different according to the type of healthcare (P < 0.001): only 15% of patients with public coverage had stage 0–I disease compared with 33% of those with private coverage, whereas 33% of the public patients had stage III compared with only 16% of private patients (Table 1). Sixty-two percent of patients had hormone receptor-positive disease and distribution was similar between the groups (P = 0.75). HER2 testing was only analyzed in patients who were diagnosed in 2006 because it was not routinely tested in 2001. Of 1,797 patients, only 1,063 (59%) were tested (Table 1) and there was no difference in tumor HER2 expression according to insurance type (P = 0.91).

Type of therapy received after diagnosis is shown in Table 2. Overall, the proportion of patients undergoing any breast surgery was slightly higher among private patients (P = 0.003). The majority of public patients had a mastectomy (68%), whereas mastectomy and lumpectomy were equally distributed among the private patients. Mastectomy was more common in public healthcare patients in all local clinical disease stages when compared with private patients (P = 0.012 for stage 0–I; P < 0.001 for stage II; P = 0.008 for stage III; Supplementary Table S1). Sentinel lymph node biopsy was less frequent in public patients than among private patients (15% vs. 26%, respectively; P < 0.001). When analyzed according to stage, this difference was only significant among patients with stage 0–I disease (P = 0.003; Supplementary Table S1). Overall adjuvant radiotherapy was given at similar rates in both insurance types (P = 0.582), but when analyzed by stage, private patients with stage 0–I disease tended to have more adjuvant radiotherapy (P = 0.003; Supplementary Table S1).

More public patients received chemotherapy than private patients (70% vs. 66%, respectively; P = 0.015). Among patients receiving chemotherapy, most received anthracycline-based regimens (66.2% of public and 56.3% of private). Taxanes were routinely available only for patients diagnosed in 2006 and overall use of these agents was generally low and not significantly different between patients of the two types of health coverage when analyzed by stage (Supplementary Table S1). Analysis of trastuzumab use included HER2-positive patients from 2006 only and, of these, 9% of public patients compared with 53% of private patients received this drug (P < 0.001; Table 2). There was a marginally significant difference between groups in the use of endocrine therapy for patients with hormone receptor-positive tumors (P < 0.052), and tamoxifen was the most frequently used drug in both groups and for all the stages. Use of adjuvant aromatase inhibitors alone or in sequence after tamoxifen was analyzed only for patients diagnosed in 2006 because in 2001, this adjuvant therapy was not in routine use. Private patients received more adjuvant aromatase inhibitors than public patients (40% vs. 12%, respectively; P < 0.001) and within all early stages of disease (Supplementary Table S1).

Among patients who received any treatment, the time from diagnosis to first treatment was significantly longer for public patients (public median 1 month vs. private median <1 month; P < 0.001). Similarly the time from diagnosis to surgery (public median 2 months vs. private median <1 month; P < 0.001) and time from diagnosis to chemotherapy (public median 2 months vs. private median 1 month; P < 0.001) were significantly shorter for private patients. There was a marginally significant difference in time from surgery to postoperative radiation therapy among patients who underwent both surgery and radiation therapy (P = 0.0962) in favor of private patients.

Overall, public patients had a shorter DFS compared with private patients (P < 0.001; Fig. 1A). When DFS was analyzed by clinical stage, there was no difference between private and publicly insured patients with stage 0–II (P = 0.89; Fig. 1B), but public patients with stage III–IV disease had significantly shorter DFS (P = 0.002; Fig. 1C).

Public patients had worse OS than private patients (P < 0.001; Fig. 2A). When analyzed by clinical stage, patients with stage 0–II disease did not have significantly different survival outcomes (P = 0.176; Fig. 2B), but those with stage III–IV disease with public coverage experienced significantly shorter survival (P = 0.008) than those in the private sector (Fig. 2C).

When analyzing survival after any relapse (locoregional or distant), patients with private coverage had significantly longer PRS than public patients (P < 0.001; Fig. 3). At 3 years, more than 80% of private patients were alive as compared with less than 60% of those with public healthcare.

In a minimally adjusted multivariate model (Supplementary Table S2), adjusting for insurance, age, and race, we found that private insurance was associated with an improved DFS [HR 0.72; 95% confidence interval (CI), 0.55–0.94; P = 0.017] and OS (HR 0.39; 95% CI, 0.21–0.73; P = 0.003). When expanding the multivariate Cox regression model to include data on clinical stage at diagnosis insurance was not independently associated with improved DFS (HR 0.87; P = 0.332; Table 3). In this expanded model, the association between insurance type and OS was marginally significant (HR 0.53; P = 0.052). Insurance type, however, was independently associated with PRS, before (HR 0.38; 95% CI, 0.20–0.73; P = 0.004) and after adjustment for stage at diagnosis and patients with private coverage had 45% lower risk of death after relapse than those in the public sector in the expanded model (HR 0.55; P < 0.001; Table 3).

To our knowledge, this is the first large, multicenter retrospective analysis of the impact of medical health coverage on outcomes of patients with breast cancer in Brazil. Many factors such as screening, pathologic characteristics of tumors, and access to medical care can influence outcomes of patients with breast cancer (10–14). Of these, socioeconomic factors, specifically health coverage type, have been associated with stage at diagnosis and survival from breast cancer (15–19).

Our results have demonstrated important differences in clinical stage at diagnosis in patients from the public versus private centers. Patients with public healthcare had about half the proportion of stage 0–I breast cancer but were diagnosed with stage III breast cancer about twice as often as patients with private coverage. It is important to note that availability of imaging at diagnosis is probably different between public and private patients, and stage IV disease might be underestimated. Because there was no difference in other characteristics that might suggest a difference in tumor subtypes, these data strongly suggest that dependence on public coverage is an important factor leading to more advanced disease at diagnosis. Ethnicity has been a factor associated with stage of disease presentation (19). In our cohort, there were more patients with mixed ethnicity in the public healthcare group, whereas the privately covered patients were predominantly Caucasians. Unfortunately, the proportion of patients recorded as “unknown ethnicity” was high in both groups, and thus did not allow us to draw any firm conclusions. A possible explanation for the differences in disease stage at diagnosis might be access to healthcare. Although we did not have information on access to screening mammography and time from breast cancer suspicion to diagnosis in this cohort, a previous publication by our group has shown that access to screening and delays to diagnosis are major issues in public healthcare sector patients in Brazil (5). Other factors such as public awareness and education might also contribute to delays in diagnosis (5).

The slightly higher rates of breast surgery seen in privately covered patients were not unexpected because of more advanced stage of public patients, potentially precluding them from having curative surgery. Higher rates of mastectomy in public patients were seen in all stages and the reasons for this should be studied further. Publicly insured patients also had lower rates of sentinel lymph node biopsy, although the precise reasons for this are not clear from our study. There were no clinically significant differences in access to radiotherapy between insurance types, and the higher rates of adjuvant radiotherapy for stage 0–I patients with private coverage are not surprising given more frequent conservative surgery.

Chemotherapy was given slightly more frequently to public patients and again this might reflect more advanced stage at diagnosis. Type of chemotherapy regimen was different between the groups with the use of taxanes being more common in those with private coverage, although overall use was uncommon for both groups in this cohort. Although the type of adjuvant chemotherapy has been shown to influence recurrence and survival, we did not have enough events in this cohort to see a difference (20). There was no difference in use of endocrine therapy for hormone receptor-positive tumors between the groups. Aromatase inhibitors were prescribed more frequently for private patients, but again we had very few events to be able to analyze differences in outcome due to this therapy. Although HER2 testing was performed in only about one third of all patients, there was a clear difference between trastuzumab use and type of health coverage, with less than 10% of those publicly insured receiving this therapy. Median times from diagnosis to any therapy in both groups were within the 12-week period generally considered ideal (21), and although there was a statistically significant difference in this time in favor of the privately covered patients, this difference was not likely of clinical relevance.

Univariate analysis showed that DFS of patients treated through the public system was generally worse than those with private coverage. In the minimally adjusted multivariate analysis insurance maintained association with improved DFS, which was not shown after adjusting for T and N stage. This strongly suggests that stage at presentation is what is driving the difference seen in univariate analysis.

OS was also worse for public healthcare patients in univariate analysis, and this difference can be attributed to death from breast cancer because this was the most common cause of death in the cohort. Again, the minimally adjusted multivariate analysis initially showed association with insurance type and OS, which disappeared when including the stage at diagnosis. This again suggests that later stage at diagnosis is what is related to the higher mortality rates seen in publicly insured patients.

Interestingly, PRS showed a large difference between those with public versus private coverage, with the first group doing considerably worse. This difference persisted even after adjusting for demographic and clinical factors. Reasons for this are not clear. One possible explanation is that privately insured patients might have earlier diagnosis of recurrent disease due to easier access to clinical consultation and imaging. This might cause a lead-time bias, where those privately insured have an apparent longer survival after diagnosis of recurrence. We cannot, on the other hand, exclude that differences in postrelapse therapy might in some way influence these findings.

A potential pitfall of this study is that more than 1,800 patients from the original dataset were excluded because of lack of information on healthcare coverage. Although we still maintained statistical power for our main analysis, subgroup analyses might have been underpowered. In addition, there could be a chance that our cohort was not representative of the population as a whole. We also had relatively short follow-up and a longer observation period with more events might impact these results. Another issue is that for some prognostic variables, there was a high percentage of missing information. It is, however, our assumption that “missingness” occurred randomly, and thus we assumed it did not impact significantly on our results. Despite these caveats, we found no differences between groups in distribution of several other characteristics and prognostic factors known to influence prognosis of breast cancer and even with short follow-up, we were able to detect differences in outcomes.

An important consideration when interpreting these results is that all participating centers attending to publicly covered patients are national centers of excellence and several are academic institutions. Access to treatment, resources, and outcomes at these centers is probably better than for the breast cancer population treated through the general public healthcare system. It is possible that the actual difference in outcomes between publicly and privately insured might be greater than we report.

In conclusion, Brazilian patients with breast cancer with public health coverage present with more advanced stage breast cancer, and in this cohort differences in DFS and OS seem to be mostly related to this factor. The difference in risk of death from breast cancer after recurrence between the groups needs further investigation. Although we have identified disparities in access to therapies such as conservative breast surgery, sentinel lymph node biopsy procedures, and use of taxanes, aromatase inhibitors, and trastuzumab, in this cohort we could not determine with certainty that this was responsible for differences in clinical outcomes. Access to therapy among public and private patients should be an issue for further investigation. Government and society should make efforts to diagnose publicly insured patients earlier and to improve their access to what is currently considered standard therapy for early breast cancer. Avoidance of stage III–IV disease with clinical downstaging would allow for significant decrease in suffering and save the costs associated with end-of-life care.

C.H. Barrios has commercial research support from Roche and GSK, has honoraria from speakers' bureau from Roche, and is a consultant/advisory board member of Roche. No potential conflicts of interest were disclosed by the other authors.

Conception and design: P.E.R. Liedke, D.M. Finkelstein, C.H. Barrios, J. Bines, S.D. Simon, P.E. Goss

Development of methodology: P.E.R. Liedke, D.M. Finkelstein, C.H. Barrios, J. Bines, S.D. Simon, P.E. Goss

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): C.H. Barrios, J. Bines, S.D. Simon

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): P.E.R. Liedke, D.M. Finkelstein, J. Szymonifka, C.H. Barrios, Y. Chavarri-Guerra, J. Bines, S.D. Simon, P.E. Goss

Writing, review, and/or revision of the manuscript: P.E.R. Liedke, D.M. Finkelstein, J. Szymonifka, C.H. Barrios, Y. Chavarri-Guerra, J. Bines, C. Vasconcelos, S.D. Simon, P.E. Goss

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.M. Finkelstein

Study supervision: C.H. Barrios, J. Bines, S.D. Simon, P.E. Goss

The authors thank the members of GBECAM who contributed to this study: João Soares Nunes, Ercio Ferreira Gomes, Fernando Chalu Pacheco, Andréa J. S. Gomes, José Getúlio Segalla, Susanne Crocamo, Daniel Luiz Gimenes, Brigitte Marie H. R. Adam Van Eyll, Geraldo Silva Queiroz, Giuliano Santos Borges, Lissandra Dal Lago, Ruffo de Freitas Jr, Sérgio Jobim de Azevedo, Carlos A. Sampaio P. Filho, Gilson Luchezi Delgado, Daniela Dornelles Rosa, Yeni Verônica N. do Nascimento, Nivaldo Farias Vieira, Hélio Pinczowski, Martha Reinisch Perdicaris, Clarissa Mathias, Jéferson José da Fonseca Vinholes, Manuela Zereu, Elicie Lins Svirski, Jorge Sabbaga.

This work was supported by Roche Brazil through a grant provided to GBECAM. P.E.R. Liedke, J. Symonifka, D. Finkelsein, Y. Chavarri-Guerra, and P.E. Goss are supported through the MGH–AVON International Breast Cancer Program, funded by the Avon Foundation, NY. C.H. Barrios, J. Bines, C. Vasconcelos, and S.D. Simon received grant support through GBECAM.

Planning, conducting, analyzing, and reporting the results of this study were the sole responsibility of the authors.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.