Ethnoracial disparities of TNBC in the United States

In the United States (US), racial inequalities in breast cancer incidence and mortality have been known and described for decades [4]. Mortality has consistently been observed to be higher among African American women than among non-Hispanic White women, while differences in incidence are smaller (and age-dependent; breast cancer being more common among African American women under 35 years old than among non-Hispanic White women of the same age, the opposite being observed after 40 years old). Initially, the higher mortality and shorter survival of African Americans had been attributed to their lower socioeconomic status and their difficulties in accessing diagnosis and then care, resulting in a late stage at diagnosis [5]. However, in many studies, these racial differences persisted after adjustment for socioeconomic characteristics, access to care and stage at diagnosis, suggesting a possible difference in the nature of the disease itself, in terms of aggressiveness. Indeed, since the 1990s [6], many studies have accumulated the evidence of racial disparities in terms of presence or absence of one or more tumor markers [7]. In 2020, a review of the literature reported the details of the frequency of TNBC by race from some thirty US studies (published mainly between 2007 and 2014) to conclude that it is twice as high among African American women as among non-Hispanic Whites in all age groups [8]. Although BRCA1 gene mutations are a recognized genetic risk factor for TNBC, their frequency in African American women with breast cancer is of the same magnitude, or even lower, than that in non-Hispanic (and non-Ashkenazi Jewish) white patients: approximately 1% or 2% [9,10]. Other genetic abnormalities associated with a high risk of TNBC have been discovered more frequently in African Americans [11-13]. It is worth noting that US research most often, if not always, begins from ethnoracial categories, only then (and rarely) to question the role of geographical origins. For example, it is within the population of African American women with breast cancer (identified in the 1998–2015 databases of the National Program of Cancer Registries and the US Cancer Statistics which cover 99% of the US population) that Sung et al. studied the disparities in TNBC frequency according to the country of birth [14]. They were significantly lower among African American women born in the Caribbean (21.2%) or in East Africa (11.6%) than among African American women born in the US (23.7%). The difference was not significant between the latter and Black women born in West Africa (24.1%). The authors emphasize that “nativity and geography origin in black women has rarely been examined, although considering nativity-related differences may improve our understanding of the etiologic heterogeneity of breast cancer”. In the end, their results plead for a genetic base of racial disparities in TNBC since their frequency seemed proportional to the importance of the shared genetic ancestry among the 4 groups compared.

After an initial diagnosis, in the population of women with TNBC, African American women have a 5-year survival that is 2.5 times shorter than white American women (14% versus 36%) and an excess mortality of around 30% [15]. Many reasons can explain these racial differences in prognosis within TNBC patients in the US context: higher prevalence of obesity and other comorbidities, fewer cancer screening opportunities, later diagnosis, barriers in access to cancer care in the population of African American women [16]. However, these differences remained pronounced after adjustment for all these factors, strongly suggesting, here too, a biological origin [17,18]. In fact, it has been shown that at the cellular level, TNBC tumors in African American women were more aggressive than those in White women and numerous differences in gene expression and molecular phenotype have been observed between these two groups [19]. In a perspective on the genomics of human stress, some authors consider these biological (acquired) differences as the result of a cascade of macro, meso and micro social determinants (fairly usual in the conceptual models used in social epidemiology) of social stress, having ultimately a molecular impact down to the cellular level [20].

Ethnoracial disparities of TNBC in the United Kingdom

In England, in a study limited to the North East London Cancer Network in 2005–2007 [21], the frequency of TNBC was 8% in White women, 19% in South Asians (TNBC also being more common in the Indian subcontinent [22,23]) and 25% among Black women. Such differences were also observed (even if lower) in a prospective hospital cohort of women with breast cancer aged under 40 years included between 2000 and 2008 in multiple sites in the United Kingdom (UK): 18.6% of cases among Whites versus 28.6% of cases in Black women were TNBC [24]. The English National Cancer Registry has been collecting individuals’ ethnicity with a usable completion rate (≥ 90%) since 2013. The exhaustive study of racial disparities in breast cancers is therefore quite recent (published in 2021, on 2013–2018 data) [25]. Unfortunately, its authors did not report the frequency of TNBC (perhaps because of the frequency of missing data concerning these 3 markers) but estimated that “Black African” patients were more at risk of late-stage, high-grade and/or ER negative tumors than Whites at all ages at diagnosis. A lower but significant risk was also highlighted among “Black Caribbean” women aged 53–70 at the time of diagnosis.

Another recent study, based on the same English national registry, observed a higher risk of diagnosis at a late stage of breast cancer in “African” women and, to a lesser extent, in “Caribbean” women compared to White women (including after adjustment for year of diagnosis, age, comorbidities, deprivation index of their area of residence and region) [26]. Interestingly for a European perspective, in the context of a universal national health service and the existence of an organized breast cancer screening program, the authors put forward several possible explanations for the differences observed (in addition to the TNBC ethnoracial differences and their greater aggressiveness). Indeed, as observed in some quantitative British studies, faced with cancer, women from minorities, particularly Black women, may recognize fewer cancer symptoms [27,28] and manifest higher cancer fatalism [29], resulting in delayed use of primary health care [30].

TNBC in North Africa and Sub-Saharan Africa

In Africa, a literature review and meta-analysis of 67 publications on the frequency of TNBC was published in 2018 [31]. In most countries except in North Africa, survey samples are small (a few hundred cases at best) and most often focus on the very small minority of breast cancer cases who have access to tertiary health care. In the absence of high-quality epidemiological data (population-based cancer registries only exist in 20 of the 46 African countries and are of very variable quality [32], particularly for the analysis of tumor markers [33]), these estimates provide only orders of magnitude. Estimates showed that TNBC was by far the most frequent in West Africa (45.7%, 95% CI = [38.8%− 52.8%]), comparing with East Africa (25.0%), Maghreb (i.e., Morocco, Algeria and Tunisia in Northern Africa, 23.0%), Southern Africa (20.4%) and Central Africa (14.8%) [34]. The difference in the frequency of TNBC between West and East Africa was even greater in a study which compared Ghana (53%) and Ethiopia (9%) respectively [35]. Data has only been published in 6 of the 14 French-speaking West African countries (the countries of origin of the vast majority of Sub-Saharan African immigrants in France). In Mali (n = 114) [36], Sénégal (n = 522) [37], Côte d’Ivoire (n = 335) [38], Togo (n = 117) [39], Bénin (n = 244) [40], and Guinea (n = 56) [41], respectively 46%, 47%, 43%, 38%, 32% and 26% of women consulting the capital’s referral hospital for breast cancer were TNBC cases. Estimates of the frequency of the BRCA1 mutation in Sub-Saharan Africa remain sparse and based on small sample sizes [42]. For instance, in a cohort of 434 unselected Nigerian women with breast cancer, 7% had a BRCA1 mutation [43]; this means that it cannot explain the very high frequency of TNBC estimated between 12.2% and up to 49.4% according to different regional studies [44,45].

Ethnoracial characterization: between rejection by principle and popular uses

Contrary to widespread belief, France is far from being the only European country not to do so, even if the issue of race/ethnicity is particularly sensitive here [49]. Indeed, the criminal use of the notion of “race” during the Second World War by the Nazi regime in occupied Europe meant that the notion of race was completely abandoned in all the countries of Western continental Europe after the Second World War. In 1952, Claude Lévi-Strauss wrote that “the original sin of anthropology consists of the confusion between the purely biological notion of race (assuming that […] this notion can claim objectivity) and sociological and psychological productions of human cultures” [50]. In a 1982 book, Albert Jacquard (a celebrated and popular scientist in biology and genetics in France) asserted that, for modern genetics, the notion of race was absurd since the majority of genetic variability is within geographical groups, and not between them [51]. We only quote these two scientists briefly to illustrate to what extent, in France, the notion of race and ethnicity has been historically and epistemologically discussed in terms of socially constructed groups (some French social scientists speak of “racialized” groups) and widely discredited as a biological attribute in medicine and epidemiology (with the notable exception of hemoglobinopathies and G6PD deficiencies [52]).

Contrary to the US, where history, marked by slavery, racial segregation and struggles for civil rights, has led to an institutionalization of racial categorization, France has historically had a “universalist” approach to citizenship, where the emphasis is on national unity rather than racial or ethnic differences (for a much more detailed comparison between the US and France on their visions of race and origins, respectively, French-speaking readers can refer to a reference article by F. Héran [53]). Apart from its colonies (where racial statistics were in common use, associated with extreme racist prejudices [54]), the French Republic has adopted an assimilationist paradigm that favors the integration of immigrants and their descendants without explicit recognition of ethnoracial identities [55]. The first article of the French Constitution stipulates that France “ensures the equality of all citizens without distinction of origin, race or religion” and, therefore, prohibits any counting of these categories of population. This principle is of course debatable and widely discussed in public and political debate and by French researchers, particularly in the social sciences and in the field of studies on discrimination. For example, P. Simon speaks of the “choice of ignorance” and criticizes this posture of “equality through invisibility” [56] and F. Héran calls for “stopping opposing republican principles to ethnic statistics” [57]. Also in public health and social epidemiology, French researchers regularly regret that ethnoracial characteristics cannot be easily queried, e.g., recently during the Covid-19 epidemic which particularly affected areas with a high proportion of immigrants [58,59]. Thus, if the assimilation of immigrants in France is an objective reality (as evidenced by a growing diversity of marriages or partnerships at each generation that is infinitely more frequent in France than in other Western countries [60]), at the same time experiences of discrimination are 3–4 times more frequent among immigrants or children of immigrants of Caribbean, African or North African origin [61]. A large, multipurpose, demographic survey dedicated to French population’s origins conducted in 2008 showed that class dimensions were stronger than ethnoracial ones in shaping self-perceived personal identity in France but suggested also that blackness was a crucial dimension of ethnoracial identification [62].

The legal background to collect data on “origins”

There is a general principle in French legislation which prohibits the collection of data on race or ethnicity in public and private statistics. In 2007, the French Constitutional Court reminded that the processing of personal data that directly or indirectly reveal people’s racial and ethnic origin, and the introduction of race or religion variables in administrative files, to be contrary to the French constitution, as well as the a priori definition of an ethnoracial reference system. But, at the same time, it indicated that “the processing necessary for carrying out studies on measuring the diversity of people’s origins, discrimination and integration can relate to objective data on people’s ancestry”. In the commentary to its decision, it explained that “these objective data may, for example, be based on the name, geographical origin or nationality prior to French nationality.” In addition to these “objective data”, it authorizes the collection of data based on the “feeling of belonging” to an ethnic group in specific surveys. So, for research purposes only, these data can be collected with special authorization from an ethics committee which must be argued and justified regarding the scientific objectives pursued. They are considered as “sensitive data”, in the same way as political opinions, religious affiliation or sexual orientation, whose collection can only be exceptional and strictly regulated by the French data protection authority. At no time did the Constitutional Court open the possibility of routine statistics according to ethnoracial categories but allows the detailed collection of nationalities and countries of birth over several generations in specific surveys. Although such statistical studies on detailed origins have been emerging in France over the past decade, particularly in demography and sociology, they remain rare in epidemiology and almost non-existent in clinical research. Patients’ origin may possibly (but rarely) be mentioned in their medical record when their life circumstances are briefly detailed but no doctor would systematically categorize their patients according to their race or ethnicity.

The diverse origin of the French population

In the absence of ethnoracial statistics, only estimates from unreliable sources circulate on the proportion of Black people in France: between 3.5% and 6.5% (French overseas regions (FOR) included) [63], i.e., approximately the same proportion as Black (1.98%) and Indian (2.30%) people altogether in the UK, and 2–4 times less than in the US (where 12.4% of the population is “African American”). Another estimate (just as poorly sourced but plausible [64]) postulated in 2004 that 3.5% of the French mainland population were Black people, broken down into 25% from the FOR (including the two Caribbean islands in the French West Indies), and 75% of Sub-Saharan African origin (most of whom having immigrated to France since the 1980s [65,66]).

In the Paris region in 2008, 43% of the population aged 18–50 had a direct link with immigration to mainland France over two generations, in the sense of being immigrants or descendants of immigrants, or natives from FOR or descendants from FOR: 11% are of European origin, 14% from the Maghreb, 7% from Sub-Saharan Africa, 4% from FOR, 3% from Asia, 1% from Turkey, and 3% from other regions [67]; which possibly leads to a (very rough) estimate of 11% of Black people in the Paris regional population. In France, strictly speaking, in the general population census and public statistics, “immigrants” are people living in France who were born abroad and who are not French citizens by birth. Whether or not they were subsequently naturalized as French citizens, they are still counted as immigrants. For their part, French people born abroad are not immigrants, nor are French people born in French overseas regions (such as the French Antilles) who have settled in mainland France. So, if we only count immigrants, 17.9% of the population of the Paris region are immigrant: half of which (48.3%) from Africa: 29.7% from the Maghreb and 18.6% from Sub-Saharan Africa respectively. Four countries of origin account for half of the latter: Mali (2.6%), Côte d’Ivoire (2.2%), Senegal (2.1%) and Congo (2.0%).

TNBC in the French West Indies

In the absence of ethnoracial statistics available in France, it is interesting to look at the epidemiology of breast cancer in the French West Indies, two French overseas islands where Black people are the overwhelming majority of the population. It is composed of descendants of African slaves (and more marginally of Indian forced laborers who succeeded the slaves on the plantations after the abolition of slavery in 1848) and white settlers, as well as metropolitan expatriates in the modern era. Both islands (Guadeloupe and Martinique) have a cancer registry.

In Martinique particularly, society remains very marked by socio-racial stratification which dates back to the era of slavery: “the stereotypes and prejudices attached to skin color and racial categories, which are in reality emic categories, are always extremely operational” [68]. The epidemiology of TNBC in this island of 367,000 inhabitants has only recently been published [69]: the proportion of TNBC is 20.9% (compared to 15% in mainland France) with a 5-year survival rate of 70.1% and a 1-year survival rate of only 34.1% for metastatic forms. In this paper, the race issue is furtively raised at the beginning of the discussion in a single sentence (“TNBC is a subtype with a higher prevalence in women younger than 40 years old with African or Asian ancestry or non- Hispanic Black or Hispanic women, as is the case in Caribbean populations”), then is implicit when comparing TNBC among other Caribbean islands... Not only are ethnoracial statistics not provided (since this is impossible) but the racial question is never addressed in the conclusive recommendations on the identification and management of TNBC (perhaps precisely because the racial question remains sensitive on this island). On the other hand, in 2017, in the only article on TNBC in Guadeloupe (the other French Caribbean Island, with a population of 379,000) where TNBC accounted for 14% of all cases, and were more frequent in patients under 40 (21.6% vs. 13.4%, p = 0.02), the authors not only cite a substantial bibliography on TNBC in African American women in the US but also explicitly outline the “lack of data on ethnicity and socio-economic status” in French cancer registries as the main limitation of their study [70].

The absence of ethnoracial data on cancer in France

In the absence of systematic collection of patients’ race or ethnicity, and considering the absence of details on their origin (i.e., their nationality and country of birth, and those of their parents) in any of the existing routine health information systems, it is impossible in France to easily study the biological and clinical disparities of diseases according to one or the other of these two characteristics. Notably, the 25 existing French Cancer Registries collect neither of them, nor any socio-economic characteristics of people (for example, there is no data on educational level or occupation). Social inequalities in cancer incidence are only understood indirectly, through ecological analyses which use socioeconomic statistics of people’s area of residence [71].

On the other hand, in computerized hospital administrative files (in outpatient consultation as well as in hospitalization), only patients’ gender, date and country of birth are collected systematically, along with their address and health insurance status. These files can be reconciled with medical records even if this has to be done manually.

Both the US ethnoracial and the French classification of origins have their own limits for the study of TNBC

Not only do these two classification systems make it difficult to compare inequalities in the prevalence of TNBC, but each has its own limitations.

In the US, from a biological perspective, the category “African American” actually tells us little about the ancestry of this population: genome-wide ancestry estimates of self-reported African Americans show an average proportion of 73.2% African, with notable differences across states in the US (between 64% and 81%) [72]. It would still be necessary to know which African origins we are talking about, since we saw previously [33] that the prevalence of TNBC varies considerably today between West, East, Central, and South Africa.

In France, the collection of French nationality (at birth or acquired subsequently) and country of birth in the population census makes it possible to distinguish French nationals from immigrants according to their country of birth (but not systematically in health information systems, nor in epidemiology as mentioned above). In 2025, the introduction of the parents’ place of birth (in France versus abroad, but without specifying in which country) in the population census will make it possible to distinguish French nationals born to immigrant parents. More detailed demographic surveys (the TeO surveys being a reference in France [73]) or some rare epidemiological surveys [74,75] have also collected the nationality and country of origin of parents, with special authorization from an ethics committee.

As we can see, the collection of “origins” can be more or less detailed. Given the complexity of migratory and historical links due to colonization (and also of the changing rules for acquiring French nationality, by birth or descent, over time), misclassification errors are not rare. For example, people born French abroad to French parents can be either “white” settlers or “natives.” Even in its most detailed form, this characterization tells us little (and probably even less than the racial categorization of the US) about the ancestry of people.

A statistical analysis of case series from a single Paris university hospital

Three samples of patients were selected from the total number of patients with breast cancer followed in the oncology department of Tenon Hospital (a university hospital located in the east of Paris, in a socially diverse, mainly working-class, neighborhood) between 01/01/2000 and 12/31/2017, according to their “origin” (defined here by their region of birth): 1) all the 180 patients born in Sub-Saharan Africa (SSA); 2) a random sample of 200 patients born in the Maghreb (Morocco, Algeria, Tunisia: MAG); and 3) a random control sample of 400 patients born in France (FRA). Indeed only the country of birth was available in patients’ files.

Let us emphasize from the outset the limits of this proxy for characterizing the ethnoracial origin of women as it shows possible classification bias. Of course, a “White European” woman may be born in Africa (for example to French expatriate parents), vice versa (and surely more often) a “Black” woman can be born in France (for example to Black African immigrants, to French Caribbean parents, or with an even more distant black ancestry). Furthermore, by limiting ourselves to women born in France as a proxy for “White European women”, we excluded from our sample all those who were born in another EU country (in particular in Southern EU, which account for a significant – although aging – proportion of the immigrant population in France).

Initially, the required sample size was calculated based on our primary hypothesis regarding the difference in frequency of TNBC between SSA (p1) and FRA (p2), with the following assumptions: p1 = 30%, p2 = 15%, with α = 0.05 (two-sided test) and 1-β = 0.90. At least 159 subjects per group had to be included. For greater power, we chose to include twice as many FRA as SSA. Finally, we chose to add a third (exploratory) group of breast cancer patients born in North Africa (MAG).

The study period covering a time when medical files were not or only minimally computerized, all of the paper files of the selected patients were retrieved from the department’s archives under the supervision of the department’s head oncologist, then anonymized, re-read and manually analyzed one by one. After excluding empty files and those concerning patients who only occasionally consulted for a second medical opinion but who were not followed in the department, the numbers of the 3 populations studied were 177 SSA, 153 MAG and 336 FRA women.

For each patient, the following characteristics were collected anonymously from their paper medical record: age, health insurance status, professional situation, level of deprivation of their neighborhood of residence, results (positive or negative) of ER, PR and HER2 tumor markers and its TNM stage (the TNM staging system standing for tumor, node and metastasis). Health insurance status was coded into 4 modalities: usual health insurance by Social Security (SS); specific health insurance for low income people (designated by the acronym CMUc for Universal Supplementary Health Coverage, it is granted below a certain threshold of household income: for example €810 for a single person); specific health insurance for the undocumented (designated by the acronym AME for State Medical Aid, also subject to a resource ceiling and accessible after 3 months of presence in France); and paid care. This last category concerns (non-EU) foreigners who do not live in France but can afford to travel to seek treatment in France: their care is then fully billed to them by the hospital (but may be covered in their country of origin by their own public or private health insurance). As we can see, patients’ insurance status gives several indications on the patients’ situation: poor, in an illegal situation, or foreign citizens living abroad. It should be noted that once a diagnosis of breast cancer has been made, all cancer treatments are free for all residents in France, regardless of their residency status (as is the case for all severe or chronic diseases in France).

The level of deprivation of the neighborhood of residence used the FDep [76] (for French Deprivation index), a composite indicator used in France since 1990 and updated at each population census. It was created to provide a geographic aggregate indicator of social disadvantage specifically adapted to health studies on the French population, when individual socioeconomic data are not available. It is calculated from the unemployment rate, the proportion of blue collars, the proportion of baccalaureate holders and the household median income per consumption unit in the area. Here, we used the FDep calculated in 2015, at the finest geographic level routinely available in France: the “neighborhood of residence” (or IRIS, an acronym for smallest spatial area for statistics). Indeed, in France, IRIS are the finest spatial divisions for which aggregated population census data are available. Their average population is 2000 inhabitants in the Greater Paris area, for an average surface area of 0.42 km² (i.e., a circle of 370 meters radius). They are homogeneous in terms of the type of housing and their limits are based on major cuts in the urban fabric (main roads, railways, waterways, etc.). The distribution of the FDep is conventionally divided into quintiles (the first quintile corresponding to the most advantaged neighborhoods). Each patient’s neighborhood of residence was identified based on the geocoding of their full address. Among women born in France, only 3 addresses (out of 336) were not found. Among women born in SSA, 22% came from abroad to seek cancer care at Tenon Hospital. Most declared an accommodation address with a relative in France (this accommodation address was geocoded), except 9 (out of 177) for whom no accommodation address was reported. Among women born in the Maghreb, 13% came from abroad and all except 6 declared an accommodation address with a relative in France (here too, it is this accommodation address which was geocoded, and we therefore have 6 missing data on the 153 women concerned).

The data were collected between 06/29/2016 and 08/03/2018. French law does not consider the retrospective reuse of existing data (e.g., from archived medical records) as “research involving human beings” and therefore does not require approval from a national ethics committee. Moreover, at that time, retrospective studies conducted at the Paris University Hospitals did not require special authorization 1) if they were conducted in a single care department, under the supervision and responsibility of its department head, a university professor (here, the third author of this article), and 2) if no nominative data was collected and used by researchers outside the department’s medical team. At the time the paper medical records were compiled, the retrospective reuse of anonymous data did not require formal patient consent.

Proportions between subgroups were compared using the Chi-square test and age means using the two-sided Student t-test. For multivariate analysis, two logistic regression models were fitted, estimating adjusted Odds ratios (aOR) and their 95% confidence intervals (95% CI). All statistical analyses were performed using IBM^® SPSS^® Statistics software v. 24.0 (Armonk, NY: IBM Corp, 2016). To assess multicollinearity among the explanatory variables, the Variance Inflation Factor (VIF) was calculated for each predictor. Because the proportions of missing data regarding triple-negative status were significantly different in the 3 patient samples (SSA, MAG, and FRA), a sensitivity analysis was performed, assuming that all missing values were positive or negative, respectively.

An ethnographic study on West African immigrants with breast cancer in the Paris region

To supplement the quantitative and archival research reported above, we discuss a five-year ethnographic study of cancer and sociality among West African immigrants residing in Paris and in treatment for breast cancer. This research on meanings and strategies about breast cancer among immigrants from the Senegal River Valley (Mali, Mauritania, Senegal) and northern regions of Guinea and Cote d’Ivoire living in greater Paris, France, focused on three central questions: 1) how sufferers, families, interpreters and clinicians collectively and progressively define and manage cancer; 2) how social relations are restructured around the collective experience of breast cancer; 3) how transnational meanings and strategies are continuously generated by means of cell phone and personal courier communications between those residing in Paris and in West Africa.

Its methodology has been described in detail previously [77,78]. Briefly, the research team consisted of two anthropologists: the last author of the present paper, and another (see Acknowledgements), a retired midwife trained in ethnographic methods who had established three family planning centers in northern suburbs of Paris where she also offered prenatal and postnatal care, and sexual education (her career was largely devoted to the population on whom we focused in this ethnographic study). Both researchers had previous experience working in West Africa (one of them spoke Wolof and the other Baatonu and Bambara). The team was completed by two interpreters (in Wolof, Bambara and Soninké).

The core of the ethnographic research in this study was the McGill Illness Narrative Interview (MINI) [79]. This questionnaire, translated in French in 2013, is a semi-structured interview schedule that systematically presents questions for patients regarding the history and ongoing process of their health condition. The questions encompass description of the illness symptoms, types of consultations to obtain diagnosis and therapeutic interventions, as well as prognosis, impact on everyday life, and spiritual concerns. It guided us in framing questions concerning how African women with breast cancer living in the Greater Paris area progressively define, understand, and manage cancer whether using local West African concepts or biomedical vocabulary. The MINI allows replication of questions systematically across patient populations using appropriate terminology (see Supporting Information 1 S1 File for the English version of the MINI). Data were interpreted by thematic analysis, a means of identifying patterns in interview transcripts and open-ended surveys, for the purpose of defining and naming codes and themes. The resulting collection of coded themes then constitute the basis for conclusions to the diverse questions to which informants replied.

During the 5-year project, ethnological data on breast cancer in African women data were collected between 01/06/2014 and 31/05/2019 in 7 different ways: 1) we interviewed 36 women in treatment for breast cancer who gave their oral consent, in the presence of their doctor – born in the Senegal Valley River countries (Mali, Sénégal, Mauritania), as well as Guinée and Côte d’Ivoire – at four public hospitals in Greater Paris (Tenon, where the quantitative study was conducted subsequently, Avicenne, Delafontaine, and Saint-Louis), as well as caretakers, clinicians and interpreters; 2) regular home visits were conducted between July 2014 and July 2019 for 10 women to assess kinship obligations, gender hierarchies, and moral conventions, using semi-structured interviews; 3) participant observation in four women’s immigrant associations in northern Paris suburbs allowed us to assess popular representations of breast cancer and treatment modalities; 4) we held focus group discussions on breast cancer with group leaders and members of 50 immigrant women’s associations; 5) the research team also observed oncologist-patient consultations two days weekly for two years; 6) it also accompanied patients from their home to the hospital, visited frequently (sometimes daily) hospitalized patients, and assisted them with access to social services; and 7) we held a meeting with five interpreters of Senegal River Valley languages (Bambara, Soninke, Pulaar, Wolof, Hassania) from ISM Interprétariat – a non-profit association which is the main interpreter resource used in Parisian public hospitals – to have a semi-structured discussion about occasions on which these interpreters had been asked to intervene to convey information to patients from Senegal, Mali, Mauritania, or Guinea, whether in person or by telephone.

The ethnographic results presented here are a part of a larger research project on “The Influence of Sociality in Cancer Decision Making”, funded by the National Science Foundation (grant no. 1354336). Verbal informed consent was obtained from all participants in accordance with institutional ethical guidelines. Prior to participation, the study’s objectives and procedures were explained in clear and comprehensible language. Participants provided an explicit verbal agreement to participate, which was documented in the study records, including the date and time of consent. No witnesses were present during the consent process. All data were recorded anonymously to ensure participant confidentiality. The study protocol and consent procedure were reviewed and approved by the Ethics committee at Washington University in Saint Louis (IRB no. 202403075, 18 April 2014).

Sociodemographic characteristics

The 3 groups of patients differed significantly in terms of age (the SSA being 10 years younger on average than the FRA), poverty rate (45.7% of the SSA benefited from CMUc or AME, compared to 22.3% of MAGs and only 0.6% of FRA) and type of neighborhood of residence (SSA and MAG living 2.5 times more often in the most disadvantaged neighborhoods than FRA: respectively 26.8%, 29.3% and 11.4%, p < 0.001). Regarding their professional situation, there were 2–3 times more managers or higher intellectual professions among the FRA (17.9%) than among the SSA (7.3%) or MAG (5.9%) and twice as many more intermediate professions as well. Half of the SSA and MAG had no occupation compared to a third of the FRA. (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Sociodemographic characteristics.

https://doi.org/10.1371/journal.pone.0349041.t001

Frequency of TNBC and associated factors

TNBC were more than 3 times more frequent among SSA than among FRA (Fig 1): respectively 38.2% (95% CI = [31.2%−5.8%]), and 12.2% (95% CI = [8.9%−12.2%]). The difference between MAG and FRA was not statistically significant. Note that the results of tumor markers were missing for 5.6% of SSA, 9.2% of MAG and 0.9% of FRA (p < 0.01). Sensitivity analysis showed that these differences in the number of missing data (see Supporting information 2, S1 Table in S2 File) cannot explain the discrepancies noted (see Supporting information 2, S2 Table in S2 File).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Frequency of TNBC by country of birth (% and 95% CI).

https://doi.org/10.1371/journal.pone.0349041.g001

Half of the TNBC cases were found in SSA and half of the non-TNBC cases in FRA (Table 2). Patients with TNBC were also significantly younger (by 5 years on average), more often undocumented migrants (25.8% versus 9.1%) and/or poor (33.1% versus 13.2%), more often with a low-skilled occupation or not in the labor force (65.3% versus 49.2%, p = 0.01). They also lived more often in a disadvantaged neighborhood of Greater Paris (50.0% versus 37.5%, p = 0.013).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Description of TNBC.

https://doi.org/10.1371/journal.pone.0349041.t002

In age-adjusted multivariate analysis, the patients’ region of birth was associated with TNBC (Table 3, Model 1). When the other social characteristics were added to the model, being born in Sub-Saharan Africa remained associated with TNBC as well as being an undocumented migrant in France and/or being inactive or with a low-skilled occupation (Table 3, Model 2), but patients’ neighborhood deprivation index was not. Between Model 1 and 2, the strength of the association between being born in Sub-Saharan Africa and TNBC decreased by about 30%, suggesting that some of it was related to a lower social status.

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Factors associated with TNBC (multivariate analysis).

https://doi.org/10.1371/journal.pone.0349041.t003

Sensitivity analysis showed that region of birth and being inactive or having a low-skilled occupation were always significantly associated with TNBC (see Supporting Information 2, Tables S3 and S4 in S2 File) whereas being an undocumented migrant was significantly associated with TNBC in only one of the two sensitivity models. All the covariates’ VIF values were below 1.5, suggesting the absence of multicollinearity.

Limiting our analysis to patients residing in France (excluding the 54 patients coming from abroad for treatment, i.e., the “paid care” ones) had little or no impact on the differences in frequency of TNBC according to origin, nor on the estimates in Table 3 (see Supporting Information 3, Tables S2 and S3 in S3 File).

Frequency of advanced tumors and associated factors

The frequency of stage T ≥ 3 tumors was also almost 3 times higher in SSA (39.5%, 95% CI = [32.6%−46.9%]) than in FRA (13.4%, 95% CI = [10.1%−17.3%]) (Fig 2) but they were not significantly different between MAG and FRA (no missing data regarding TNM stage). Prevalence of metastatic stages was not significantly different in the 3 populations: respectively 11.9% in SSA, 8.5% in MAG and 6.3% in FRA (p = 0.09).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Frequency of stage T ≥ 3 tumors by country of birth (% and 95% CI).

https://doi.org/10.1371/journal.pone.0349041.g002

Patients with a stage T ≥ 3 tumor were significantly younger (on average 5.4 years), with a TNBC, more often SSA, more often undocumented immigrants or foreigners coming from abroad to be treated at Tenon Hospital. (Table 4). There was a correlation between the tumor stage and the absence of markers: 34.7% of the 124 TNBC cases were at stage T ≥ 3, compared to 20.8% of the 486 cases of tumors expressing at least 1 marker (and 12, 7% of the 56 cases with missing marker results, p = 0.002).

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Description of stage T ≥ 3 tumors.

https://doi.org/10.1371/journal.pone.0349041.t004

Factors affecting tumor stage

In our dataset, differences in tumor stage can be explained by triple negative status but also by a delay in access to health care, which can be due to women’s origin, the quality of their health insurance and/or their poverty.

To further investigate the impact of these different factors, we limited our analysis to patients residing in France only; excluding the 54 women who came from abroad for treatment (it is likely that they travel to consult in Paris at a more advanced stage, but this is another question). We estimated a logistic regression model that considered age, triple negative status, origin and health insurance status (Table 5). All the covariates’ VIF values were below 1.5, suggesting the absence of multicollinearity. Its estimates showed that triple negative status and the origin of patients residing in France were both associated with a more advanced tumor stage, but not their insurance status. This last point deserves to be underlined: whether women are insured by usual health insurance (SS), by that for the poor (CMUc) or by that for undocumented immigrants (AME) did not seem to have an influence on the tumor stage at their first consultation. In the sensivity analysis, the origin of patients was the only factor associated with a more advanced tumor stage (see Supporting Information 2, Tables S3 and S4 in S2 File) and especially the origin of women residing in France were associated

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Factors associated with stage T ≥ 3.

https://doi.org/10.1371/journal.pone.0349041.t005

Women of Sub-Saharan African origin consulted for the first time at the outpatient oncology consultation at a more advanced stage than others, all things being equal regarding the triple negative status of their tumor (and their age). Nearly half of SSA women with TNBC consulted with a tumor at stage T ≥ 3 at the time of diagnosis compared to 27% of French TNBC women (Fig 3). Among non TNBC women, the prevalence of tumors at stage T ≥ 3 was also higher among SSA (37%) than among French women (13%) and this latter difference was statistically significant. Sensitivity analysis led to similar conclusions (see Supporting information 2, Table S5 in S2 File).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Comparison of the frequency of stage T ≥ 3 tumors by country of birth according to triple negative status (% and 95% CI).

https://doi.org/10.1371/journal.pone.0349041.g003

Factors explaining late access to breast cancer care

This last result can be fruitfully supplemented by certain results from ethnographic research. Indeed, certain cases drew our attention and time. One such example involved a woman in her early thirties with breast cancer with a grim prognosis, whom we visited regularly at home and hospital. She had spent a year consulting an African herbalist living in Paris. He prescribed a medication of stewed leaves to place twice daily on her sore and swollen breast. After deciding that the herbalist’s preparation was not reducing the pain, she finally decided to consult the midwife who had assisted her when she last gave birth. The midwife immediately sent her to a hospital for a mammogram and biopsy. She was subsequently referred to another hospital for chemotherapy, to which she was unresponsive. With metastases and pulmonary problems, she was hospitalized intermittently. Eventually, she returned to Mali, to her husband’s family, to undergo treatment from a ritual specialist. Her mother and her husband’s family agreed that her failure to respond to biomedical treatment was evidence of a “sent” illness, caused by a jealous relative, an etiology often suspected in cases of breast cancer. Her oncologist was supportive of her trip (intended to be a three-week visit) but expressed doubts to the team about the likelihood that she would live to return to France and indeed, she died within two weeks of arriving at the home of the healer.

The pattern evident in the case above, in which the patient waited a year for an herbalist’s treatment to work before turning to biomedical institutions, is similar to that found in other patients’ decision making. For example, Ami came to France from Guinea in response to a job offer from a man she met at work. He offered her a position in a restaurant in Paris, but when she arrived, the job proved to involve sex work. She quit but had no housing so she slept on the street until a social worker in that neighborhood stopped to speak with her and took her to stay with a friend. After settling there, she sought care at a public clinic that accepted undocumented patients to request advice for her painful breast. There, she was referred to a hospital for preliminary examination of her breast condition and then sent to another one for chemotherapy and additional treatment, as is a common process of care seeking. She was told about our team and asked us to accompany her. We found that she was terrified of her diagnosis and convinced that this disease always ends in death. Her immediate problem was that the woman who had agreed to house her required her to move out in case cancer was a contagious disease.

In yet another case, Mariam came to France from Mali to join her younger sister. She had been diagnosed with epilepsy in Mali but was well enough to work as a secretary. She had noticed breast pain and swelling but did not consult anyone for this problem, she explained, because she already had another sickness and did not want to bother her sister with this issue. “Eventually,” months after her arrival in Paris, she told her sister’s general practitioner that she had breast pain. He sent her to a hospital for a mammogram and there she was referred again to another public hospital in the region. Her breast cancer did not respond to chemotherapy and ultimately, she died.

Many patients mentioned the burden of maintaining breast symptoms secret, from the earliest manifestations: a lump, swelling, or pain in the breast. Patients considered it important to hide their concerns from even close family members (including their mother or their husband) to assure that no one in the community would become aware of their symptoms. Several women observed that « it is very important to keep this sickness (cancer) a secret because ‘Africans’ are capable of ‘doing things’ [faire des choses]”. By this they meant to engage in evil actions directed at the women concerned, for example ruin their reputation or send a curse to cause sickness. It is very important to emphasize that none of those interviewed mentioned having been informed, in France, of symptoms signifying breast cancer.

The complexity of cancer care pathways

In general, most patients followed a similar route: first, keeping this anxiety private, then consulting with an herbalist or ritual specialist, or/then getting advice from a social worker or a midwife, and finally referral to a first hospital for mammography and biopsy, and ultimately a referral to a level 3 hospital certified in oncology. Those for whom mastectomy was recommended but who were undocumented would need to transfer to yet another public hospital, the only one to perform mastectomies on undocumented women. Navigating this complex array of institutions likely constrains some patients from continuing treatment over time. Similarly, the demands of everyday life—collecting a child at day care, trying to maintain a work schedule, domestic responsibilities, traveling long distances from home to the hospital while debilitated from chemotherapy (or in one instance, living in a high-rise apartment on the 14^th floor with no functioning elevator): these obligations lead many women to skip appointments or stop treatment altogether.

What may cause breast cancer?

Among the focus groups held with group leaders and members of immigrant associations, one deserves to be mentioned here. Indeed, in an especially lively discussion, referred to by the 25 participants as a “chat group” (“une causerie”), we began by asking a general question: “what causes breast cancer?” Immediately, one woman replied, “money.” Everyone seemed to talk at once, debating whether “money” meant bills or coins (when wearing a traditional dress without pockets). The principal respondent said we (team members) did not seem to appreciate the accommodating capacity of the bra. She proceeded to demonstrate that a cell phone, money (bills and coins) but also a small radio could easily be stored in a woman’s bra. The cell phone and bills were most suspect as causes of breast cancer, given the proximity of the breast to the potentially contaminated objects. In general, the group agreed also that breast cancer, although of uncertain origin, could be a deadly disease sent by a resentful or jealous relative or friend. A co-wife was a particular risk of sending breast cancer to another wife of her husband in polygynous societies.

Given that many West African immigrant women speak only some conversational French, their capacity to comprehend complex explanations of hormone function, tumor growth/stage/grade, metastases, and treatment regimens is limited. In one case, we observed that during over a year of appointments, a particular patient never asked a question of her oncologist. Our data show that many women with breast cancer are misinformed, not informed at all, or disregard medical advice. Not only some patients mention consulting an herbalist or ritual specialist before an appointment with an oncologist as we mentioned above, but others receive packages of herbal preparations from family in West Africa. Also, consultations with healers from Africa living or visiting in Paris may continue throughout the course of hospital interventions. Often clinicians are unaware of this serial or simultaneous reliance on alternative healing remedies.

Accordingly, interpreters had difficulties translating physicians’ information that they did not understand, and/or considered inappropriate or even dangerous for the patient. For instance, in Bambara, a lingua franca for the region, women we interviewed mentioned that there are numerous words they prefer to use for “cancer” among themselves, such as “m’bon” or “bon,” (a curse); “kourou” or “kourouni” (a lump or heaviness); or “sim di”. They agreed they prefer not to use the word cancer because – even for well-educated patients – they felt it “violent,” and, as many informants told us, it is a word that “moves towards death.” They may substitute a religious discourse (for example, “this is a disease in the hands of God”) to serve as a less brutal code for delivering “bad news”. To say the word cancer, to diagnose cancer and speak of it explicitly to the patient, is a “destiny in itself.”