Background

Among females in the Caribbean, breast cancer was the leading cause of cancer deaths, and accounted for 1.4 million disability adjusted life years (DALYs) in 2013 [1–3]. Age-standardized breast cancer mortality rates in the Caribbean have shown a 37% increase to 20.6 per 100,000 since 1990; this is in contrast to the decrease seen among many industrialised countries [1, 2].

Despite this high regional burden, little is known about the social distribution of breast cancer incidence and outcomes within the Caribbean. Internationally, social inequalities in breast cancer burden and outcomes are evident, such as by race and education [4–8]. Examining whether there are differences among populations groups, and determining their basis, can guide policy towards improving outcomes.

In 2007, the Port of Spain Declaration was affirmed by Caribbean Community (CARICOM) Heads of Government, aimed at the prevention and control of non-communicable diseases (NCDs), and there is an ongoing progress evaluation of political responses to this commitment [9, 10]. The World Health Organization (WHO) Commission on the Social Determinants of Health (CSDH) has highlighted the role of health research in understanding health inequalities and inequities, and through the 2011 Rio Political Declaration, countries have committed to monitoring, understanding and addressing health inequities [11, 12]. These agreements have set the scene for efforts to understand the social drivers of chronic disease, including cancers.

To date, there has been no published systematic review of research evidence on the social determinants of breast cancer among Caribbean populations. This systematic review is guided by the analytical framework to examine social determinants of disease by the WHO CSDH [13]. This review uses a simplified version of the framework to answer the primary research question: what is the distribution, by known social determinants of health, of the risk factors, frequency, and adverse outcomes of breast cancer among female populations living in the Caribbean?

Methods

Full details of the review methodology are available in the study protocol (see Additional File 1). The protocol was guided by a previous systematic review of social determinants of diabetes [14] and an initial scoping review of the social determinants of breast cancer.

Eligibility criteria

Observational studies were sought that reported relationships between a social determinant and known risk factors for breast cancer (alcohol intake, overweight/obesity, infrequent breastfeeding, physical inactivity, dietary sugar, ionizing radiation, late age at first pregnancy, and low parity), disease frequency (incidence or prevalence), or disease outcomes (cancer stage at diagnosis, cancer grade at diagnosis, recurrence, survival, mortality). Articles written in the dominant Caribbean languages (English, Spanish, French, and Dutch) were sought from 32 Caribbean territories. Included studies drew upon samples from either the general population or from healthcare facility catchments. No age restrictions were used in determining study eligibility. Sample sizes ≤50 were excluded as unlikely to be representative of underlying populations. Risk factors were identified using three compendiums of evidence-based information: The Global Burden of Disease Consortium, UpToDate, and Cancer Epidemiology and Control [15–17]. Articles presenting risk factor data from a sample of combined genders or males only were excluded so as to more accurately represent the risk factor profile in females. The selection of social determinants was guided by the extension of the PRISMA statement for the transparent reporting of systematic reviews and meta-analyses with a focus on health equity, which recommends the “PROGRESS-Plus” checklist: place of residence, race or ethnicity (alternatively culture or language), occupation, gender, religion, education, socio-economic position (SEP), and social capital [18]. Age was not examined as a social determinant for overweight/obesity and breast cancer frequency and outcomes due to its biological associations with these variables. Reports published between January 2004 and December 2014 were considered for inclusion. This 10-year period was selected as relevant to the current situation and able to inform policy response as it is taking place within the context of a major review of regional and national policy responses in the Caribbean to NCDs [10].

Search strategy, study selection, data extraction

The databases searched were: MEDLINE (via Pubmed); EMBASE (via Ovid); SciELO; CINAHL (via EBSCO); CUMED, LILACS, and IBECS (via WHO Virtual Health Library) [19–23]. The final search was conducted in February 2015. The search strategies are detailed in a supplementary file (See Additional File 2). Search results were maintained in Endnote reference management software [24].

Study selection and data abstraction were undertaken in duplicate by two independent reviewers (CB, SH); any inconsistencies were resolved by a third reviewer (NS-G). Study selection was conducted in two stages. First, titles and abstracts were screened to identify potentially relevant articles; second, full-text screening of potentially relevant articles identified articles for inclusion in the review. If inadequate information was available for decision-making in the first stage, the article automatically progressed to full-text review. In addition to those not meeting the inclusion criteria, 10 articles were either inaccessible or awaiting publication [25–34]. With guidance by the STROBE statement on strengthening the reporting of observational studies in epidemiology and the PRISMA-Equity statement [35, 36], an electronic data abstraction form was created in the REDCap database (see Additional file 1) [37].

Risk of bias assessment

Risk of bias was assessed using a tool adapted from STROBE and Cochrane ACROBAT-NRSi guidelines (see Additional file 1) [35, 38]. Bias was assessed at the relationship level across 5 domains: confounding (was control for known and potential confounders adequate?); participant selection (is the sample representative of the target population?); missing data (is the data reasonably complete?); outcome measurement (is a social determinant/risk factor/disease endpoint appropriately measured?); selective reporting (is a relationship selectively reported?). Articles were classified as having serious, moderate, low, or unclear risk of bias. Two reviewers (CB, NS-G) made an independent judgement on the overall risk of bias of each included article, considering each domain as equally important and also the direction and magnitude of the bias from each domain. Discrepancies were discussed by the two reviewers to achieve consensus.

Synthesis of results

The review was planned as a narrative synthesis with supplementary meta-analysis if possible. Key study details were presented, followed by a description of associations between a social determinant and either a risk factor, a measure of disease frequency, or a measure of disease outcomes. The number and type of inequality relationships were summarised in an ‘evidence gap map’ – a visual tool to highlight the current evidence on the known social determinants of breast cancer in the Caribbean and a guide for focusing future research [39]. Meta-analysis of quantitative evidence was planned for inequality relationships reported by ≥2 studies with low to moderate heterogeneity and classified as having a low or moderate risk of bias [38]. Meta-analysis was not performed because of lack of sufficient evidence (number and quality) for each domain of social indicators.

Results

Summary of included studies

Thirty-four articles from 32 original studies were included from 5,190 screened citations (Fig. 1). Of these 34 articles, 23 reported on breast cancer risk factors, 9 reported on breast cancer frequency, and 3 reported on breast cancer outcomes (1 article examined both breast cancer frequency and outcomes); 10 social determinants were examined (Table 1).

Fig. 1
figure 1

Flowchart of search strategy and article selection

Full size image
Table 1 Characteristics of 34 articles describing the social distribution of breast cancer in Caribbean women [40–70, 72, 73]
Full size table

Included articles reported on studies conducted in English-speaking (Antigua, Bahamas, Barbados, British Virgin Islands, Dominica, Grenada, Guyana, Jamaica, St. Lucia, Trinidad and Tobago, United States Virgin Islands); French-speaking (Guadeloupe); Dutch-speaking (Bonaire, St. Eustatius, Saba, Suriname); and Spanish-speaking (Cuba, Puerto Rico) territories. Most studies originated in Cuba (n = 7) and Jamaica (n = 7). Across the 8 categories of social determinants, there were a total of 15 different social determinants and 14 review endpoints, leaving 189 possible inequality relationship groups that could have been reported (Fig. 2). Only 30 (16%) of these relationship groups were reported by the 34 articles, leaving 159 relationship groups (84%) without an evidence base. There were 75 inequality relationships reported: 59 on breast cancer risk factors, 13 on breast cancer frequency, and 3 on breast cancer outcomes.

Fig. 2
figure 2

Summary of 75 inequality relationships from 34 articles between a social determinant and review endpoint [40–70, 72, 73]. Legend: Age and limited breastfeeding cells do not separate youth and adult samples as the studies have combined these age groups in their samples

Full size image

Risk of bias of included studies

Of the 34 articles, 16 were classified as moderate-risk, 14 were classified as serious-risk, 1 was classified as unclear-risk, 2 were classified as moderate/serious-risk, and 1 was classified as serious/unclear-risk (Table 2). At the relationship-level, of the 75 relationships, 35 were classified as moderate-risk, 34 were classified as serious-risk, and 6 were classified as unclear-risk. Figure 3 details the proportion of relationship classifications within each of the 5 risk of bias domains. Overall, lack of adjustment for confounding was the main contributor to an increased risk of bias, followed by non-disclosure or inadequate handling of missing data.

Table 2 Risk of bias among 75 relationships from 34 included articles [40–70, 72, 73]
Full size table
Fig. 3
figure 3

The proportion of risk of bias classifications of the 75 relationships among each of the 5 risk of bias domains [40–70, 72, 73]

Full size image

Results of inequality relationships

Risk factors

Alcohol

There were 14 inequality relationships for alcohol, reported across 8 social determinants in 6 articles: age (n = 5), education (n = 2), ethnicity (n = 1), income (n = 1), marital status (n = 2), occupation (n = 1), religion (n = 1), residence (n = 1) [40–45].

All adolescent studies found that older adolescents consumed more alcohol than younger adolescents [43, 44], with less conclusive findings among adults [40, 42, 45]. Persons with higher education tended to drink more than those with less education in Barbados and Cuba [42, 45]. For example, 1.1% of elderly in Barbados with 1–6 years education versus 11.8% of persons with >12 years education consumed alcohol ≥4 days/week [42]; likewise, 4.8% (95% CI 3.8–5.7) of Cuban adults with primary level education versus 13.2% (95% CI 10.8–15.7) with university level education consumed alcohol in the past 30 days [45]. However, Cuban elderly report low frequency of consumption across all education levels (0%–1.5% consume alcohol ≥4 days/week) [42, 45]. This is in line with the one article examining residence, which reported higher frequency in overall consumption in Barbados (2.7%) as compared to Cuba (1.1%) [42]. With respect to ethnicity, more black and mestizo Cubans reported alcohol consumption (14.9%, 95% CI 12.3–17.6 and 14.7%, 95% CI 12.9, 16.5 respectively) within the past 30 days than white Cubans (8.2%, 95% CI 7.3–9.0) [45]. A large regional study found that adolescents with increased religious service attendance consumed alcohol less frequently than those who had less attendance (OR 0.50, p < 0.001) [41]. Studies examining marital status showed mixed findings; those examining income and occupation showed no association [42, 45].

Overweight/Obesity

There were 28 inequality relationships for overweight/obesity, reported across 8 social determinants in 14 articles: education (n = 8), ethnicity (n = 5), income (n = 5), marital status (n = 3), occupation (n = 2), residence (n = 2), social household structure (n = 2), and SEP (n = 1) [46–59].

Studies examining education and occupation tended towards a negative relationship [47, 50–53, 55]. All but one study (examining elderly) reported overweight/obesity to be associated with lower levels of individual education [47, 50–53], as well as maternal and paternal education [50]. Adults with lower-level occupations and children with parents working in lower-level occupations tended to be more overweight/obese than those with higher-level occupations [50, 51]. Yet reports on income showed mixed results [47, 48, 51, 53, 54], and the single study examining SEP showed higher levels of overweight/obesity among Jamaican girls of a higher family SEP (OR 1.87, 95% CI 1.0-3.4) [49]. Studies reporting on ethnicity, marital status, social household structure and residence showed mixed results.

Limited breastfeeding

There were 13 inequality relationships for breastfeeding, reported across 5 social determinants in 3 articles: age (n = 3), education (n = 3), income (n = 2), marital status (n = 3), and occupation (n = 2) [60–62].

The likelihood of breastfeeding initiation was higher among older mothers in Puerto Rico (OR 1.39, 95% CI 1.00–1.95 for 35–49 year olds), with no age differences found in Jamaica [60, 61]. Also, Puerto Rican mothers who practiced breastfeeding initiation and exclusive breastfeeding tended to be more educated than those who did not [61, 62]. Mixed results were found for marital status, income, and occupation; to note is that Jamaican mothers who were employed were less likely to exclusively breastfeed (of those employed, 21.1% exclusively breastfeed versus 31.0% nonexclusively breastfeed), while Puerto Rican mothers who were employed were more likely to initiate breastfeeding (crude OR 1.63, 95% CI 1.31–2.03; adjusted OR 1.15, 95% CI 0.89–1.48) [60, 61].

Physical inactivity

There were 4 inequality relationships for physical inactivity, reported across 2 social determinants in 4 articles: age (n = 1), ethnicity (n = 3) [40, 46, 56, 57].

In Grenada, the amount of persons participating in physical activity through walking/biking drastically decreased by 72.5% (p = <0.001) after 54 years of age; at the same time, the amount of persons participating in >10 min of leisure time per day was also found to gradually increase with age (78.1% for persons <35 years old to 83.5% for persons >64 years old, p = 0.53) [40]. The two studies examining ethnicity found that Guadeloupian Asian-Indian adults reported lower levels physical activity than their non-Asian-Indian counterparts when considering time and level of vigour of activity (physical activity level score mean 1.62 (SD 0.22) versus mean 1.74 (SD 0.34), p = <0.05) [56, 57].

Frequency & outcomes

Fewer studies examined the social determinants of the frequency and outcomes of breast cancer, than those for risk factors. There were 13 inequality relationships for breast cancer frequency, reported across 5 social determinants in 9 articles: education (n = 2), ethnicity (n = 2), marital status (n = 3), occupation (n = 1), and residence (n = 5) [63–71]. Most articles reported the number of new breast cancer cases, with 4 out of the 9 articles converting these counts to a breast cancer incidence rate. Relationships examining occupation, residence, and ethnicity showed no association. A Puerto Rico study found a higher likelihood of breast cancer among women with only primary and secondary education as compared to women with higher education (OR 3.38, 95% CI 1.5–5.7 for primary; OR 1.33, 95% CI 0.9–1.9 for secondary) [66]. Lastly, unmarried women in Puerto Rico tended to have a higher likelihood of being diagnosed with breast cancer as compared to married women (divorced OR 2.57, 95% CI 1.4–4.4; single OR 1.36, 95% CI 0.7–2.6; widow OR 2.08, 95% CI 1.1–4.0) [66], but no differences were seen in Trinidad or Barbados.

There were 3 inequality relationships for breast cancer mortality, reported across 2 social determinants in 3 articles: ethnicity (n = 1) and residence (n = 2) [72, 73]. No evidence found reporting on the other 4 breast cancer outcomes. While no associations were observed between breast cancer frequency and ethnicity, mortality from breast cancer was shown to be higher among Indian-decent compared to African-descent populations in Trinidad (OR 1.2, 95% CI 1.1–1.4) and Guyana (OR 1.3, 95% CI 1.0–1.6) [73].

Discussion

Summary of evidence

This systematic review examined the extent of evidence on the influence of social determinants of health on breast cancer risk factors, frequency, and adverse outcomes in the Caribbean. Thirty-four articles from 32 separate studies were included. With 189 possible ways of exploring the role of social determinants on breast cancer, 75 inequality relationships were reported within 30 distinct relationship groups, leaving 159 (84%) relationship groups without an evidence base. The results of this review highlight a critical evidence gap on the effects of social determinants on breast cancer among Caribbean women, with limitations in the quantity and quality of published evidence. Nearly half of the articles were classified as having serious risk of bias, mostly because of failure to adjust for important potential confounders. Furthermore, included articles reported a range of inconclusive findings for each relationship group, at least partly due to study heterogeneity and small numbers of studies available for each relationship group.

Measures of breast cancer frequency and adverse outcomes showed weak relationships with social determinants. Though, the racial disparity in breast cancer mortality between women of Indian origin and women of African origin in two different settings is worthy further investigation. The connection between breast cancer and social inequity is a not a new phenomenon. While low social status is known to place women at a higher risk of developing and dying from breast cancer [74, 75], a higher social status tends to predispose women to certain reproductive risk factors including later age at first pregnancy, lower parity and less breastfeeding [76, 77]. However, a higher SEP also affords women a higher screening rate, an earlier stage of diagnosis, and improved treatment effect and adherence, indicating a complex interchange of risk and protection [74, 75, 78]. Our depicted lack of regional evidence seems a logical result of the absence of a structured network of cancer surveillance in the Caribbean [79, 80]. Cancer registries exist in only twelve Caribbean territories, of which only four are considered high-quality [80, 81]. Challenges are wide-reaching, with limitations in resources, political will, policy and regulation, healthcare service, data quality and security, and local, regional, and international communication and collaboration [80, 81]. The PAHO Plan of Action for Cancer Prevention and Control 2008–2015 [82] has detailed areas for improvement in monitoring and surveillance and consequently, the regional Caribbean Cancer Registry Hub was conceptualized and is progressing towards implementation [81]. While this Hub is expected to greatly improve regional cancer surveillance efforts, measures of inequalities should be highlighted in its plans, with hopes to increase attention to social determinants of cancers and advance health promotion in this area.

Most results lie within the relationships between social determinants and breast cancer risk factors. Overall, Caribbean women with indicators of a lower SEP could be at a higher risk of breast cancer as they reported a higher alcohol intake (except for education), higher levels of overweight/obesity, and limited breastfeeding. The trends reported between age and education with breastfeeding is in line with evidence in other settings, with low maternal education being the strongest predictor of poor breastfeeding practices [83–86]. The inverse relationship between overweight/obesity and education and occupation is similar to what is found in other middle and higher income regions; while being overweight or obese was previously thought to be a condition of the elite, more recent transitions have occurred whereby obesity is shifting towards the persons with a lower socioeconomic standing, particularly as the country’s gross national product increases [87–89]. Typically though, alcohol consumption is found to be higher among persons of a higher SEP [90, 91]. Yet the relationship between alcohol and SEP is complex. Varying environmental factors such as alcohol availability and affordability, economic development, culture, and national alcohol policy flout the gradient typically observed whereby risk factor harm increases with decreasing SEP [90, 91]. The Caribbean is particularly vulnerable to this risk factor as its cultural norms embrace alcohol consumption as a commonplace social activity, which is further compounded by a lack of national alcohol policies [92–94]. While no relationships were reported on social capital, the inverse relationship between alcohol and religion is noteworthy. Religiosity is consistently shown to be protective from substance use by creating a positive personal identity, fostering community acceptance, and providing a coping outlet for stress [95–97]. The Caribbean touts a predominant religious identity which could confer some form of protection from alcohol’s influence on breast cancer and the wider range of NCDs afflicting the region.

Continued and standardized approaches to understanding risk factor profiles is a key element in efforts to reduce cancer risk factors, as evidenced in the WHO’s recommended STEPwise approach to Surveillance (STEPS) [98]. With relevant information on social determinants included in this instrument, it is up to Caribbean territories to fulfil their commitment to the Port-of-Spain Declaration in continuing to implement this in their ongoing efforts to reduce NCDs such as breast cancer [99].

Limitations

The review was limited by a small number of articles within each relationship group, the validity of which was further limited by their significant risk of bias. Further, few studies investigating the effects of social determinants on health have also explored the interrelationships among the social determinants themselves. The Caribbean has been considered as one region in this review, masking the possible and important country-level variations in the relative importance of social determinants. Country-level information on screening and access to treatment such as mammogram screening rates and wait times for diagnosis or treatment are important potential confounders that were not assessed. Publication bias is an important concern as no explicit searching was conducted for grey literature due to limited resources.

Conclusions

This review highlights a crucial gap in the quantity and quality of the evidence examining the social determinants of breast cancer risk factors, frequency, and outcomes. Risk factors were the main endpoints for which relationships with social determinants were reported, with implications for age, ethnicity, education, SEP, and religion. Information on frequency and outcomes were limited, but held implications on marital status and ethnicity respectively. Although the need for more research in this area is acknowledged, this effort should also include an attempt at standardizing reporting guidelines for observational studies of health inequality. Finally, the development of a validated risk of bias assessment tool is imperative for systematic reviewing of observational studies.